/* edge-build: 4c77d0a3043c */
/* ============================================================
   EDGE - graphite terminal, repaletted 2026-08-05 on the user's direction:
   off the red-orange accent and violet fade, toward a quieter professional
   scheme. Color budget: ONE accent (cyan, shared with the value pole, as v1
   shared orange), one fade (burnt orange), amber reserved for injury risk,
   and position hues demoted to ghost chips. Everything else is graphite.

   Measured per constitution 17 (OKLab dE x100 on the cell surface):
     value #1fa9c9 vs fade #b25a12 - 27.8 normal, 21.8 worst gated CVD
     fade vs TE gold 17.1 / value vs RB green 16.7 (same-row, clear the floor)
     fade vs risk amber 25.8
   Accepted weak pair, documented: value vs WR blue 10.1 normal / 9.5 CVD -
   cross-channel (signed text vs lettered chip), stronger than the accepted
   TE/RB 4.3. Signs (+/-) always accompany color.
   ============================================================ */
/* ------------------------------------------------------------------
   TYPE, SELF-HOSTED. Three families, 124KB total, served from this origin.

   NOT a Google Fonts <link>. This app's service worker exists on the premise
   that draft-day wifi dies and the board should not, and a render-blocking
   stylesheet on a third-party origin is the one asset that promise cannot
   cover - the fonts would fall back exactly when the room is worst, which is
   the hour the design matters most. Self-hosted, they are shell assets like
   app.js: cached on install, versioned by the deploy hash, offline forever.

   Subset to LATIN ONLY, checked rather than assumed: zero of the 862 players
   in players.json carry a non-ASCII character, so latin-ext was 121KB of
   coverage for nothing. Newsreader and IBM Plex Sans are variable fonts and
   ship as one file each across their whole weight range.
   ------------------------------------------------------------------ */
@font-face{font-family:"Newsreader";font-style:normal;font-weight:300 600;
  font-display:swap;src:url(fonts/newsreader.woff2) format("woff2")}
/* THE RANGE OPENS DOWNWARD TO 300 AND COSTS NOTHING. The Lions skin's body
   copy is a light weight - 15px/1.85 at 300 is most of why that design reads
   as editorial rather than as an interface - and this was declared 400 600,
   which CLAMPS a request for 300 to 400 and would have rendered the whole
   skin a step too heavy with no error anywhere.

   Measured before it was claimed, in the browser, against this exact file:
   'Handgloves 0123' at 40px advances 313.17px at wght 300 and 315.88 at 400,
   and font-variation-settings:'wght' 300 gives the identical 313.17 - so the
   axis is real and reaches 300, rather than the browser synthesising it. The
   shipped woff2 is unchanged; only what this declaration admits to. */
@font-face{font-family:"IBM Plex Sans";font-style:normal;font-weight:300 600;
  font-display:swap;src:url(fonts/ibm-plex-sans.woff2) format("woff2")}
@font-face{font-family:"IBM Plex Mono";font-style:normal;font-weight:400;
  font-display:swap;src:url(fonts/ibm-plex-mono-400.woff2) format("woff2")}
@font-face{font-family:"IBM Plex Mono";font-style:normal;font-weight:500;
  font-display:swap;src:url(fonts/ibm-plex-mono-500.woff2) format("woff2")}
/* PRIMETIME, scoped to the player card. One extra family and 10KB, which is a
   real cost paid deliberately: the card is a broadcast graphic, not a panel,
   and the oblique Archivo Black over a team colour IS the idea. Nothing
   outside .pcard is allowed to use it, so the rest of the app keeps one voice.
   Only the weight the template actually sets is shipped, which is 400.

   THE CONDENSED FACE IS GONE, not merely unreferenced. Three rules beside this
   one declared `font-family:var(--font-label)`, and var() is not valid in a
   descriptor - the whole rule is dropped at parse time, in every engine. So
   Barlow Condensed never rendered a pixel of this app: the declarations and
   `--font-label:"Newsreader"` landed in the SAME commit, and `document.fonts`
   on production listed Newsreader, IBM Plex and Archivo Black and nothing
   else. Every comment in this repo saying the condensed face is "scoped to
   .pcard" was describing the mock, never the app.

   Naming it literally would have been the wrong repair. `--font-label` is the
   editorial serif by a decision this app has since CALIBRATED AGAINST - the
   schedule grid's 13px/18px proportions and the board's name abbreviation are
   both measured in the serif (both 2026-08-08) - so the family would have
   loaded with nothing entitled to ask for it. The 44KB went to the
   service worker's precache on every install, and `sheetDocument()` inlines
   every face this stylesheet names as base64, so it also went into every
   exported cheat sheet at ~59KB. Dead weight is cheapest to carry and cheapest
   to delete when it is dead in one place; it was dead in three. */
@font-face{font-family:"Archivo Black";font-style:normal;font-weight:400;
  font-display:swap;src:url(fonts/archivo-black.woff2) format("woff2")}
/* THE POSTER FACE, scoped to the Lions skin the same way Archivo Black is
   scoped to the player card: one family, one job, and nothing outside the
   skin may reach for it. 35KB, latin only, taken from his own mock's bundle
   rather than from Google - this app's service worker exists on the premise
   that draft-day wifi dies, and a font on a third-party origin is the one
   asset that promise cannot cover. Variable across the range, so the 700 the
   design sets and the 400 it does not are one file. */
@font-face{font-family:"Big Shoulders Display";font-style:normal;font-weight:400 700;
  font-display:swap;src:url(fonts/big-shoulders-display.woff2) format("woff2")}

/* THE GRAPHITE PALETTE, AND THE TWO ISLANDS THAT WEAR IT ANYWHERE.
   2026-08-11 (6th): `.pcd` and `.hz-ban` were added to this selector when the
   board joined the lit views, and the addition IS the design decision.

   An EDGE card is a dark object. Its fill is #0b0d14 - a literal, chosen for
   that reason, that no skin has ever moved - while the 37 rules that paint
   inside it read --text-primary, --positive, --pos-rb, --accent and a dozen
   more, every one of which a skin does move. Half of a matched set was themed
   and the other half was not. Light the view and 397 player names turned
   #131c22 on near-black and left the page; the position chips lost a third of
   their contrast; the foot's value cells went Honolulu blue on graphite at
   3.94:1. Nothing errored and nothing shifted. It was found by measuring, and
   only after a preview was built to be thrown away.

   THE RULING IS THAT A CARD IS A GRAPHITE OBJECT, NOT A THEMED ONE, and this
   is the smallest mechanism that can hold it: the card declares the whole
   graphite palette ON ITSELF, so every rule inside it resolves against the
   ground it was actually written for - including the thirty-odd that nobody
   has audited and the ones written next year. Custom properties resolve per
   element, so a `.pcd` inside a `.lions` view simply answers its own
   declaration and hands it to its descendants.

   WHY NOT NAME THE INKS ONE BY ONE. That was built first: --card-ink,
   --card-ground and the rest, swapped in at the five sites the two known
   defects lived at. It fixed those five and left thirty-two, and the reason
   is instructive - a hand-maintained list of "the tokens a card happens to
   use" is a second copy of the palette, and this repo spent the morning of
   the same day on four hand-copied cache tags that agreed with each other
   about everything except the one input that mattered. Agreement between
   duplicates is not a property you keep, it is one you re-earn on every edit.

   WHAT IS DELIBERATELY *NOT* AN ISLAND: the KPI chip row above the cards. A
   chip's lit state is the one thing the ground has to have a say in - a
   near-white cap is the brightest object in the row on graphite and a hole in
   the row on paper - so those five rules stay themed, through the --chip-*
   tokens further down. And `.pcd-bandhd`, the tier band header, is a sibling
   of the grid rather than a child of a card: it sits on the page, it follows
   the page, and a pattern-matched version of this rule would have dragged it
   onto the graphite ink it must never wear. */
:root, .pcd, .hz-ban{
  /* SURFACES - near-flat, per the Ice Editorial template. Its cards carry no
     fill at all: a hairline rule on the page ground, no raised planes and no
     glow. That reads as editorial rather than as a HUD, and it is the single
     biggest reason the template looks calmer than what it replaces.

     Kept, against the template: a row hover and a selected tint. A 400-row
     table you navigate with j/k needs to answer "which row am I on", and the
     template has no dense table in it to have solved that problem. */
  /* THE GRID IS THE BASE FOR EVERY DARK DISPLAY, 2026-08-11 (6th), his
     call: the intro card, Momentum's top half and the Injuries top half
     become the same material as the EDGE card grid. That is the app's
     densest dark surface and the one he singled out on the lit board;
     the page and the header sat two points off it for a reason nobody
     ever wrote down. */
  --bg-app:#0b0d14;
  --bg-header:#0b0d14;
  --bg-surface:#0e1216;
  /* THE CONTROL BAR'S PLATE, and it is the THIRD instance of a trap this
     repo has now booked twice: the three rules that paint it named
     --bg-panel with a #0f1215 fallback, and --bg-panel was declared NOWHERE.
     The fallback was the rule, so a stylesheet audit searching for the token
     found nothing at all three sites. It surfaced the moment the board was
     lit, as a graphite plate wearing the white keycaps the skin had correctly
     handed it - the keys followed the ground and the thing beneath them could
     not. Declared here at the value all three sites already resolved to, so
     graphite does not move by a channel, and the fallbacks are gone from the
     call sites so it cannot happen a fourth time. */
  --bg-panel:#0f1215;
  --bg-surface-raised:#11161a;
  --bg-row:transparent;
  --bg-row-hover:#12181c;
  --bg-selected:rgba(22,131,196,.10);

  /* borders - the template's hairline, which is doing all the structural work
     now that surfaces are flat */
  --border-subtle:rgba(140,180,200,.12);
  --border-strong:rgba(140,180,200,.22);
  /* A THIRD, LIGHTER RULE, from his 7a/6a: the hairline BETWEEN ROWS is a
     step quieter than the one that bounds a panel, so a 25-row list stops
     reading as 25 stacked panels. */
  --border-row:rgba(140,180,200,.08);
  /* An UNFILLED pip is an outline and nothing else, so its edge has to be
     heavier than any of the three rules above or it disappears. */
  --pip-edge:rgba(140,180,200,.28);

  /* text */
  --text-primary:#eef6fa;
  --text-secondary:#8ea2ae;
  --text-muted:#5d707c;
  /* RETUNED 2026-08-08 to the value his 7a/6a mocks use for this step (#3c4b53,
     297 occurrences across the five builds; #4a5a64 appears in none of them).
     This is the "present but explicitly not information" step - a zero delta,
     an unplayed bye, a provenance note, the word "certain", a keyboard hint -
     and the design's answer for it is one stop darker than the app's guess was.
     Both ramps have four steps and this is the fourth in each, so it is the
     same slot repriced, not a fifth token. */
  --text-dim:#3c4b53;
  /* A FIFTH STEP, AND IT IS NOT TEXT YOU READ. The quiet board's unmade picks
     carry their number and nothing else, at a value below every step above:
     96 of 150 cells wear it, and at --text-dim they still read as a list. It
     was the literal #2b3941 at one site; named here because a skin has to be
     able to move it and because "below the last readable step" is a real slot
     in the ramp, not an accident of that one rule. */
  --text-ghost:#2b3941;

  /* THE INK ON A LIT SURFACE. Wherever the accent stops being a line and
     becomes a FILL - the on-state of a key, the one cell on the clock, the
     lead rank disc - whatever sits on it needs the opposite ink, and #04222b
     was that ink written out at nine sites. It is not --bg-app: it is chosen
     against the accent, and the moment the accent moves (as it does in a
     skin) it has to move with it or the fill goes unreadable. */
  --on-accent:#04222b;
  --on-accent-soft:rgba(4,34,43,.55);

  /* ══ THE KPI CHIP ROW ══ 2026-08-11 (6th), and it is the ONE part of the
     board's dark furniture that is deliberately NOT a graphite island.

     Four chips sit above the cards: `8/8 SOURCES`, `389 RANKED`, an outlier
     count, and the room the prices were read from. His second reported defect
     was that they half-ported - the first went light while the rest kept
     their graphite pills - and the note in the handoff was "decide the chip
     treatment; today it reads as a mistake rather than a design".

     THE DECISION. The row is four chips in the cards' own language, so its
     ground and its inks are graphite like theirs. But a chip has a state a
     card does not - one of them is LIT - and lit is the single thing whose
     meaning genuinely depends on the page it is read against. A near-white
     cap is the brightest object in the row on graphite; on #eef1f3 it has
     barely more contrast with the page than the page has with itself, so it
     stops reading as the loudest chip and starts reading as a gap where a
     chip should be. That is why these are tokens and the card is an island:
     the card asks nothing of the ground, and the chip asks it exactly one
     question.

     WHAT THE ROW SAYS, now that it has to say it on two grounds: a FILLED
     chip reports a number this board computed, and the OUTLINED one names the
     room those numbers came from. Three findings and a setting. That
     distinction was always in the markup and was never legible, because on
     graphite an unfilled chip and a #161b21 chip are nearly the same object.
     On paper they finally read as two kinds of thing - which is the one place
     this seam made the design better rather than only different.

     Every value here is byte-identical to the literal or token it replaced,
     so the port is a provable identity on graphite: the ground these rules
     were written for does not move by a single channel. --card-ink-soft and
     --card-ink-dim are steps 2 and 3 of the graphite text ramp above, named
     so a chip can reach them from inside a lit view. */
  --chip-ground:#161b21;
  --card-ink-soft:#8ea2ae;
  --card-ink-dim:#5d707c;
  /* Face and ink are TWO tokens rather than one because they have to move
     together or the fill goes unreadable - the same reason --on-accent exists
     a few lines above, and the same failure it was extracted to prevent. */
  --chip-on-face:#eef6fa;
  --chip-on-ink:#0b0d14;
  /* The warn chip carries the app's negative, which every skin reprices; on a
     #2a1215 pill the skin's softer red is the wrong half of that pair. */
  --chip-warn-face:#2a1215;
  --chip-warn-ink:#ff2244;
  /* An outlined chip's edge is drawn against the PAGE rather than against a
     chip, so this one is genuinely themed - it is the only edge in the row
     with nothing of its own behind it. */
  --chip-ghost-edge:rgba(140,180,200,.12);

  /* THE TRACK a meter fills, and it was TWO VALUES four percent apart -
     rgba(...,.10) under Best Available's survival bar, .14 under the pick
     path's and the roster rail's. His mocks draw one track colour for all
     three, which is the right reading: they are the same object. Unified on
     the .14, the majority value and the one the design's own grey is nearer. */
  --track:rgba(140,180,200,.14);
  /* the whisper that says a row is FILLED rather than open */
  --row-tint:rgba(140,180,200,.03);

  /* ICE EDITORIAL PALETTE, adopted whole from the template on his direction.
     Accent is now INTERACTIVE ONLY - the value pole moves to its own mint,
     which is the template's structure and a real separation the old scheme
     did not have: "you can click this" and "this is good" stop sharing a hue.

     THE CONTRAST FLOOR IS OVERRULED HERE, deliberately and by his call after
     the measurements were put in front of him. Constitution 17 sets an OKLab
     dE x100 floor of 15 between any two semantics that can appear together.
     Measured on this palette (surface #0e1216):

       positive #7fe0c8 vs accent  #5fd4f0     7.8   UNDER, by half
       negative #e0796b vs risk    #e0b46b    14.7   UNDER, marginally
       negative         vs QB pink #e0447c    11.7   UNDER (cross-channel)
       positive vs negative                   26.7   clears
       negative vs accent                     27.2   clears

     Recorded rather than quietly dropped, because the first pair is the real
     cost: mint and ice-cyan are close enough that a rising delta and a
     clickable control read as the same family at a glance. What keeps this
     survivable is that NOTHING in this app carries meaning by hue alone -
     every delta ships a sign, every position mark ships its letter, every
     severity ships its code (constitution 18). Colour is the second channel
     here, never the only one. See DECISIONS 2026-08-06. */
  --accent:#1683c4;
  --accent-hover:#2b8fcb;
  --accent-soft:rgba(22,131,196,.14);
  --accent-border:rgba(22,131,196,.62);
  /* THE ONE GLOW IN THIS APP, around the cell on the clock, so it is findable
     from across the room. A glow is a dark-theme device - light spilling off a
     lit object onto a dark ground - and on a white page it has to become a
     drop instead, which is the whole reason it is a token and not a literal. */
  --accent-glow:rgba(22,131,196,.34);

  /* VALUE IS THE ACCENT AGAIN, and fade is a hard red - which is what the
     EDGE Cards template does and what this app did before the Ice palette.
     The mint was a SEPARATE hue for "good" sitting 7.8 from the cyan for
     "clickable", close enough that the hero made them read as one family. The
     answer is not more distance between them; it is to stop pretending they
     are different questions. One hue means "good / act here".

     What that buys, measured on #0e1216: value vs fade 26.7 -> 39.3, fade vs
     the injury amber 14.7 -> 25.8, fade vs the TE gold 12.2 -> 22.5. Two of
     the three overruled pairs clear the floor now and are asserted as real
     checks again. The one that remains is fade vs the QB pink at 8.7, which
     is cross-channel - signed delta text against a lettered position chip -
     and carried by sign and letter as constitution 18 requires. */
  --positive:#1683c4;
  --positive-ink:#2b8fcb;
  --negative:#ff2244;
  --negative-ink:#ff5c72;
  --negative-soft:rgba(255,34,68,.10);
  --warning:#e0b46b;
  --warning-soft:rgba(224,180,107,.08);
  --neutral:#6e828f;

  /* availability risk (injury) - still a RESERVED STATUS slot, not a series
     hue, and still carried by FORM (filled = out, outlined = questionable)
     plus the letter code rather than by colour alone. Retuned to the
     template's softer amber; the 14.7 against the fade coral is part of the
     overruled set above. */
  --risk:#e0b46b;
  --risk-soft:rgba(224,180,107,.12);
  --risk-border:rgba(224,180,107,.5);

  /* position marks - four hues and a gray, rendered as GHOST chips (tinted
     background, colored letter) rather than solid fills. K and DST share ONE
     neutral gray on purpose: they are streamed positions, excluded from every
     signal, and giving them hues spent color budget on the two positions that
     matter least. The letters tell them apart (constitution 18 - letters are
     mandatory on every position mark and may never be removed). K's old teal
     also sat 4.7 from the new cyan accent, so it had to move regardless. */
  /* WR MOVED OFF THE VALUE POLE'S STEP, 2026-08-11 (6th), and it is the
     one thing the single-brand-blue ruling forced. WR has always been
     blue and so, now, is `value`: measured against the suite's own
     Machado CVD metric the pair fell to 6.0, under the documented
     secondary-encoding band of 8, and a WR chip beside a blue +22 is
     two meanings in one hue.
     HUE CANNOT FIX IT. Violet was tried first and scores WORSE (3.2):
     under protan and deutan simulation violet collapses toward blue.
     A search over the whole space for a colour clearing 9 against
     value, fade, QB, RB and TE returned 206 answers and every good one
     separates by LIGHTNESS. So WR stays the blue it has always been
     and moves several steps up it: dE 6.0 -> 19.2, and 9.90:1 on the
     grid where it is read. His call to make the move: "we can pivot the
     WR blue to a different color if you want." */
  --pos-qb:#e0447c; --pos-rb:#2fae57; --pos-wr:#80c0f0;
  --pos-te:#d29a24; --pos-k:#77838c;  --pos-dst:#77838c;
  --pos-ink:#0c0d0e;

  /* draft grid cells (v2 handoff surfaces; the specular treatment is gone) */
  --bg-cell:#11161a;
  --bg-cell-hover:#171e23;

  /* TYPE - three families, each with one job.

     --font-display  Newsreader, a serif, for MASTHEADS AND DISPLAY NUMERALS
                     only: page titles, the brand, the big figure in a stat
                     tile. This is the template's own rule - it sets 26-64px
                     numbers in the serif and never uses it below that.
     --font-ui       IBM Plex Sans, for anything that is a sentence.
     --font-num      IBM Plex Mono, for the data grid, deltas and caps labels.

     The mono is the part that is not merely cosmetic. This board is a
     17-column table read by scanning DOWN a column, and a proportional face
     was making 10 / 100 / 1000 sit at three different optical positions in
     the same column. Mono digits are one advance wide by construction. */
  --font-display:"Newsreader",Georgia,"Times New Roman",serif;
  /* THE POSTER FACE, and it is a FOURTH job rather than a fourth family here:
     a page title and a clock numeral are set at 34-56px and read as signage,
     where the pick path's 22px claim is a SENTENCE that happens to be large.
     The graphite theme answers both with the serif, which is why one token
     covered it; the Lions skin answers them with two different faces, so the
     jobs have to be named apart before either can move. Defaulting to
     --font-display keeps this theme byte-identical. */
  --font-poster:var(--font-display);
  /* THE LABEL FACE, behind one token so it is a single line to change back.
     Barlow Condensed was doing the small-caps labels on the cards, the hero
     chips and the Primetime panel; set to the editorial serif to see the whole
     app speak in one voice. Newsreader tops out at 600, so the 700/800 weights
     the condensed face carried are clamped rather than silently synthesised
     into a fake bold. */
  --font-label:"Newsreader",Georgia,"Times New Roman",serif;
  --font-ui:"IBM Plex Sans",-apple-system,BlinkMacSystemFont,"Segoe UI",Helvetica,Arial,sans-serif;
  --font-num:"IBM Plex Mono",ui-monospace,SFMono-Regular,Menlo,monospace;
  --fs-xs:10px; --fs-sm:11px; --fs-md:12px; --fs-lg:13px; --fs-xl:16px; --fs-hero:26px;
  --track-caps:.1em; --track-mono:.18em;

  /* space + radius */
  --sp-1:4px; --sp-2:8px; --sp-3:12px; --sp-4:16px; --sp-5:24px;
  --r-sm:4px; --r-md:6px; --r-lg:11px;

  /* row metrics */
  --row-h:36px;
  --cell-x:10px;
}

/* ============================================================ BASE */
/* The `hidden` ATTRIBUTE only sets display:none through the UA stylesheet, so
   ANY author rule that sets display beats it. `.draftbanner{display:flex}` did
   exactly that: the board's draft bar reported hidden:true and still occupied
   18px as an empty bordered strip above the filters, with no draft running.
   One global rule closes the whole class of bug rather than the one instance. */
[hidden]{display:none !important}

*{box-sizing:border-box;margin:0}
html{background:var(--bg-app);scrollbar-color:var(--border-strong) var(--bg-app)}
body{font:var(--fs-md)/1.45 var(--font-ui);color:var(--text-primary);background:var(--bg-app)}
a{color:var(--accent)}
a:hover{color:var(--accent-hover)}
button{font:inherit;cursor:pointer}
.num{font-family:var(--font-num);font-variant-numeric:tabular-nums}
/* Every small caps label in the app is the template's mono micro-label. */
.caps{font-family:var(--font-num);font-size:var(--fs-xs);letter-spacing:var(--track-mono);
  text-transform:uppercase;color:var(--text-muted);font-weight:500}
.wrap{max-width:1440px;margin:0 auto;padding:0 var(--sp-4)}
.view{display:none}
.view.active{display:block}
:focus-visible{outline:2px solid var(--accent-border);outline-offset:1px}
.dim{color:var(--text-muted)}

/* delta text - always signed; color redundant to the sign */
.d-up{color:var(--positive-ink)}
.d-dn{color:var(--negative-ink)}
.d-0{color:var(--neutral)}

/* ---------------------------------------------- ADP CHANGE OVER TIME
   GREEN = the ADP number went UP. RED = it went DOWN. Keyed to the NUMBER and
   nothing else, on the user's call (2026-08-06), so the arrow, the sign and
   the hue are three drawings of one fact and cannot contradict each other.

   This is a SEPARATE axis from the cyan/orange pair, and deliberately so.
   Cyan/orange answers "which side of the market price does this source sit
   on", where the favourable direction INVERTS for opinion sources - an expert
   rank of +5 means they are out on him. Painting that green because the number
   is bigger would revive the FP polarity bug this project shipped once
   already. Two different questions, two different palettes, both stated in the
   board legend.

   Red/green is the one pair that fails for red-green colour blindness, which
   is why it is never the only channel here: every use carries a signed number
   and an arrow glyph. Same rule as the position chips (constitution 18). */
:root{ --adp-up:#3ddc97; --adp-dn:#e0645a; }
/* The tape's inks, as R,G,B triplets because they ride inside rgba() with a
   size-scaled alpha. Tokens, not literals in the renderer: each ground
   chooses its own - graphite keeps the cyan/red the tape shipped with, and
   the Lions plate re-inks them from its own palette below. */
:root{ --tape-up:22,131,196; --tape-dn:255,34,68; --tape-0:140,180,200; }
.adp-up{color:var(--adp-up)}
.adp-dn{color:var(--adp-dn)}
/* a delta that clears no bar is data, not signal - it stays in the margins */
.d-quiet{color:var(--text-muted)}

/* inline keyboard hints (prep drawer, queue copy) */
.kbd{display:inline-block;background:var(--bg-surface-raised);border:1px solid var(--border-strong);
  border-radius:3px;padding:1px 5px;font:600 var(--fs-xs)/1.3 var(--font-ui);color:var(--text-secondary)}

/* ---------------------------------------------------------- THE ALERT
   NINE SITES EMITTED THIS AND NOTHING STYLED IT. `.alert`, `.alert-fade`,
   `.alert-info` and `.tag` appear nowhere else in this file, so every
   degradation this app is careful enough to announce - offline, rows failed,
   no sim, no join, preseason, panel failed, and the fatal "could not load
   players.json" - rendered as one run-on line of body text with its own
   label welded to the front of the sentence: "OFFLINE Could not reach ESPN's
   scoreboard". Found while porting Live, where out of season that message is
   the ENTIRE content of the tab.

   Built on the advisory idiom this app already has (a tinted band, a rule
   down its edge, a caps tag beside a sentence) rather than a new object, and
   entirely on tokens, so it reads correctly in both palettes. The variant
   colours the rule and the tag only: the message is a sentence, and a
   sentence in coloured text is harder to read for the sake of a fact the tag
   beside it already states. */
.alert{display:flex;align-items:baseline;gap:var(--sp-3);
  padding:var(--sp-3) var(--sp-4);border-radius:var(--r-md);
  background:var(--bg-surface);box-shadow:inset 3px 0 0 var(--neutral);
  font:400 var(--fs-lg)/1.6 var(--font-ui);color:var(--text-secondary)}
.alert .tag{flex:none;font:500 var(--fs-xs)/1.6 var(--font-num);
  letter-spacing:var(--track-mono);text-transform:uppercase;color:var(--text-muted)}
.alert .body{min-width:0}
/* A DEGRADATION IS NOT AN ERROR. Something the app wanted is not here, the
   rest of the app still works, and the reserved caution slot says exactly
   that - the same amber the injury tab uses for availability. */
.alert-fade{background:var(--risk-soft);box-shadow:inset 3px 0 0 var(--risk)}
.alert-fade .tag{color:var(--risk)}
/* Nothing is wrong; there is simply nothing to show yet. */
.alert-info{background:var(--accent-soft);box-shadow:inset 3px 0 0 var(--accent)}
.alert-info .tag{color:var(--accent)}
/* The one fatal case: the board itself did not load. */
.alert.run{background:var(--negative-soft);box-shadow:inset 3px 0 0 var(--negative);
  color:var(--text-primary)}

/* ========================================================== THE LIONS SKIN
   His "Lions Draft Room" mock, 2026-08-11, ported to the three surfaces he
   named: the Draft Room, the Live tab, and the Injuries page from its
   detailed tracker down. Everything else in the app stays graphite.

   IT IS A SKIN AND NOT A SECOND STYLESHEET. Every component this touches was
   already reading tokens - the 7a ledger, the 6a rail, the quiet board, the
   key system - so what a light editorial palette needs from the app is a
   second set of VALUES for the names that already exist, not a second set of
   rules. The block below is that set; the rules under it are only the places
   where the design says something the graphite theme had no name for (a dark
   band inside a light page, a field behind the board, a poster face).

   TWO PALETTES IN ONE APP IS HIS CALL, made with the seam in front of him.
   The seam is drawn deliberately rather than hidden: the section rail stays
   dark against both, the topbar follows whichever view is on screen, and the
   injury tracker becomes a light CARD on the dark page rather than a light
   bleed, which is the mock's own idiom (its whole page is a 12px-radius card
   on a grey ground).

   WHAT THE MOCK DID NOT GOVERN, and why:

   - THE MONO STAYS. The mock has no monospace in it; its numerals are the
     sans with tabular-nums. Both answers solve the same problem - a column
     read downward needs one advance per digit - and this app's answer is
     already argued twice in DECISIONS and measured against a 17-column board.
     What is taken from the mock is its SIZE and TRACKING, not its family.
     Same ruling as the 7a port made for Barlow Condensed.
   - A RUN IS STILL RED. The mock draws the recent-picks run marker and its
     header chip in the accent. Here a run means the room is taking the
     position you need, which is why .pl-edge.on and .pl-run are --negative,
     and that is a meaning rather than a surface. Recolouring it accent would
     make the app's one "this is against you" signal look like its "act here"
     signal, which is the exact confusion the 2026-08-06 repalette existed to
     undo.
   - THE FIFTH TEXT STEP FOLDS INTO THE FOURTH. Graphite needs --text-ghost
     below --text-dim because 96 unmade picks at --text-dim still read as a
     list. The mock prints those same numbers, the word "certain" and every
     empty roster slot at ONE value (#b6bec3), so on this ground the two slots
     are the same colour. Taken from the design rather than invented.        */
.lions,body.skin-lions .topbar{
  /* SURFACES. The mock's cards carry no border at all - separation is the
     fill against the #eef1f3 ground - so the hairlines below are for the
     rules INSIDE a card (a header, a row), not around it. */
  --bg-app:#eef1f3;
  --bg-header:#fff;
  --bg-surface:#fff;
  /* On paper the control bar is a surface like any other. It is a page
     element and not one of the graphite islands - the reader acts on it, and
     what it holds are keys the skin already lights. */
  --bg-panel:var(--bg-surface);
  --bg-surface-raised:#fff;
  --bg-row:transparent;
  --bg-row-hover:#f4f7f8;
  --bg-selected:#e7f2f9;
  --bg-cell:#fff;
  --bg-cell-hover:#f4f7f8;
  /* SUNKEN, which the graphite theme has no use for: on a dark ground a
     recessed thing is darker than the page, and there is nothing darker here
     to be. The team headers across the top of the board are the only object
     that wants it. Skin-local, and only skin-scoped rules may name it. */
  --bg-sunken:#e3e8ea;
  /* THE ONE DARK OBJECT IN A LIGHT PAGE: the clock band, which the mock
     draws as a strip of turf under stadium lighting. Same navy as the rail,
     so the two dark things in the design agree. */
  --turf:#0f1a22;

  --border-subtle:#e2e7ea;
  --border-strong:#d3d9dc;
  --border-row:#edf0f2;
  --pip-edge:#c3ccd1;

  --text-primary:#131c22;
  --text-secondary:#55636b;
  /* LIFTED A STEP, 2026-08-11 (6th): "I need some contrast here. Can barely
     see it." Every panel caption on a lit view - BEST AVAILABLE, TARGET QUEUE,
     YOUR ROSTER, RECENT PICKS - is a `.caps` in this ink, and at #8a949a it
     measured 3.10:1 on white and 2.73:1 on the paper ground. That is under the
     floor for 11px tracked capitals by a wide margin; it was picked as a
     "muted" step against the mock's own greys without anyone measuring what it
     would be asked to carry, which turned out to be every section header in
     the draft room. #6b757c is 4.71 and 4.15 and still reads as the quiet step
     - the ramp keeps its shape, it just stops being decorative. */
  --text-muted:#6b757c;
  --text-dim:#b6bec3;
  --text-ghost:#b6bec3;

  /* HONOLULU BLUE, and it is the accent, the value pole and the on-state -
     one hue doing what cyan does in graphite. --accent-hover DARKENS rather
     than lightens: on a white ground the brighter direction is toward the
     page, so the graphite move would make a hovered control recede. */
  --accent:#1683c4;
  --accent-hover:#0f6ba3;
  --accent-soft:#e7f2f9;
  --accent-border:#a9cfe6;
  /* NOT A GLOW ANY MORE. Light spilling off a lit object only exists against
     a dark ground; on white the same job is done by a drop, so the shadow
     that uses this is re-shaped below as well as recoloured. */
  --accent-glow:rgba(22,131,196,.42);

  --positive:#1683c4;
  --positive-ink:#1683c4;
  --negative:#d92d3f;
  --negative-ink:#d92d3f;
  --negative-soft:rgba(217,45,63,.08);
  --warning:#9a6d10;
  --warning-soft:#faf1de;
  --neutral:#8a949a;
  --risk:#9a6d10;
  --risk-soft:#faf1de;
  --risk-border:#e2cfa0;
  --adp-up:#1e7a3c;
  --adp-dn:#c0392b;
  /* The tape re-inked for paper: the same triplets as the plate's own accent,
     negative ink and neutral, so a wash and the solid it fades from are one
     statement. */
  --tape-up:0,118,182; --tape-dn:217,45,63; --tape-0:138,148,154;

  /* POSITION MARKS, the mock's own inks. Ghost chips still: a tinted ground
     under a coloured letter, mixed from these at 14% exactly as in graphite,
     which lands within a point or two of the mock's #e6f5ec / #e7f0fb /
     #fbe7ef / #faf1de. K and DST keep sharing one grey. */
  /* The same collision on paper, resolved the same way in the other
     direction: on a light ground a chip separates by going DARKER.
     dE 8.1 -> 20.0, and 8.84:1 on the page. */
  --pos-qb:#c0245c; --pos-rb:#1e7a3c; --pos-wr:#2b3f7a;
  --pos-te:#9a6d10; --pos-k:#6b757b;  --pos-dst:#6b757b;
  /* the ink ON a solid position or accent fill, which is now the light one */
  --pos-ink:#fff;
  --on-accent:#fff;
  --on-accent-soft:rgba(255,255,255,.72);

  --track:#dfe4e7;
  --row-tint:#f6f8f9;

  /* THE KEYS FLATTEN. His button system is a physical keycap - lit lip, hard
     bottom edge, 2px press - and that object is made of light falling on a
     dark material. The mock's controls are flat white rectangles with a
     hairline; the press survives as a fill change and a shadow that inverts
     inward, so the control still moves under the finger. */
  --key-face:#fff;
  --key-face-hover:#f4f7f8;
  --key-face-down:#e8edef;
  --key-edge:#d3d9dc;
  --key-edge-hover:#b6c1c7;
  --key-ink:#48545b;
  --key-ink-hover:#131c22;
  --key-lift:0 1px 1px rgba(15,26,34,.05);
  --key-down:inset 0 2px 3px rgba(15,26,34,.14);
  /* THE LIT KEY IS AN OUTLINE ON THIS GROUND, not a filled slab. His call,
     2026-08-11 (6th): "put the blue button highlights on the border of the
     boxes so they stand out."

     Two reasons it is the right call and not just a preference. A solid blue
     cap fights the KPI chip row directly above it - that row's whole job is to
     have exactly ONE lit blue object in it, and seven more of them a hundred
     pixels down flattens the page's loudest signal into wallpaper. And on
     paper the box itself is the object: a filled cap hides the box, an outline
     is the box, which is literally what he asked for.

     Done here rather than in a new rule, so the key system stays ONE system -
     every control still reads --key-on-face, --key-on-edge and --key-on-ink,
     and only the values move. The face stays the page's own white so a lit key
     and an unlit key are the same material at different weights; the edge and
     the ink carry the state. The edge is doubled to 2px through --key-on-lift,
     since a 1px hairline in the accent reads as a focus ring rather than as a
     selection. */
  --key-on-face:#fff;
  --key-on-face-hover:#f2f8fc;
  --key-on-face-down:#e7f2f9;
  --key-on-edge:var(--accent);
  --key-on-ink:var(--accent);
  /* The second blue line that makes the edge read as chosen rather than
     focused, plus the same soft drop every key on this ground carries. */
  --key-on-lift:inset 0 0 0 1px var(--key-on-edge),0 1px 2px rgba(0,118,182,.22);
  --key-on-down:inset 0 0 0 1px var(--key-on-edge),inset 0 2px 4px rgba(0,118,182,.18);

  /* THE KPI ROW ON PAPER, and it is the whole of what this skin says about
     the board's dark furniture. The cards and the hero banners are graphite
     ISLANDS - they declare the graphite palette on themselves and ask this
     block for nothing at all - so the three declarations below are the entire
     seam between a light page and 397 dark objects standing on it.

     The lit chip goes Honolulu blue with white ink rather than keeping the
     near-white cap, which on #eef1f3 has barely more contrast with the page
     than the page has with itself and reads as a gap in the row instead of
     the loudest thing in it. Blue-on-white is already what every on-state in
     this skin is - the key system's --key-on-face, the YOU column, the cell
     on the clock - so the row joins a sentence the page was speaking rather
     than inventing a fifth one of its own.

     The warn chip is NOT here, and that is the same decision read the other
     way. Its pill is dark on both grounds, so handing it this skin's softer
     red would put one ground's ink on the other ground's fill - #d92d3f on
     #2a1215, measurably worse than the pairing it replaced - which is exactly
     the half-port this block exists to end. A pair moves together or it does
     not move. */
  --chip-on-face:var(--accent);
  --chip-on-ink:#fff;
  /* Graphite's ghost edge is a translucent light, which over paper is nothing
     at all. The skin's own hairline-against-a-light-ground already exists as
     --pip-edge, for exactly this reason: an outline is all an unfilled thing
     has, so it is the one thing it cannot afford to lose. */
  --chip-ghost-edge:#c3ccd1;

  /* TYPE. The poster face is the identity; --font-display drops to the sans
     because this design has no serif in it and the slot's other jobs (the
     pick path's claim, a stat figure) are sentences and numbers at size, not
     signage. The mono keeps its own token and its own job. */
  --font-poster:"Big Shoulders Display","Barlow Condensed",Impact,sans-serif;
  --font-display:var(--font-ui);
  --font-label:var(--font-ui);

  /* GEOMETRY. The mock rounds cards at 10, cells and controls at 8. */
  --r-sm:8px; --r-md:8px; --r-lg:10px;
}

/* ------------------------------------------------------------ THE GROUND */
/* A light view has to paint its own ground: the document's background is the
   graphite one and always will be, since three tabs of six are lit. The
   min-height keeps the ground under a short page - Live out of season is a
   masthead and one line - so the page does not end in a dark band. */
.view.lions{background:var(--bg-app);color:var(--text-primary);
  min-height:calc(100vh - 52px)}
/* THE TRICOLOUR IS GONE, site wide. His call, 2026-08-11 (6th): "drop this
   blue silver blue silver site wide."

   It came from the mock and it was decorative from the first day - a dashed
   blue-and-grey rule under the bar, carrying nothing. What changed is the
   company it kept: with one brand blue and a lit chip row a few pixels below
   it, a second blue element that means NOTHING competes with the one that
   means "this is the lit state". Removing it is the cheapest thing this page
   can do for the signals that are left.

   Left as a comment rather than a silent deletion because the rule is in the
   mock, and the next person to open that file will wonder why it is not here. */
/* THE INJURY TRACKER IS A CARD, not a bleed: it is the bottom half of a page
   whose top half stays graphite, and the mock's own page is exactly this
   object - a radiused plate on a ground, with its content inset. */
.lions-block{background:var(--bg-app);color:var(--text-primary);
  border-radius:12px;box-shadow:0 3px 20px rgba(15,26,34,.16);
  padding:26px 30px 30px;margin:var(--sp-5) 0}
.lions-block>:first-child{margin-top:0}
/* On a phone the plate's desktop inset stacks on the wrap's 8px and the
   method card's own padding - three walls before a word. The plate keeps its
   idiom and gives back the width; scoped by WIDTH, never by dr-full. */
@media (max-width:640px){
  .lions-block{padding:16px 12px 18px;margin:var(--sp-4) 0}
}

/* The bar belongs to the view under it, so it changes with the tab. The rail
   does not: it is the one object that stands against both palettes, which is
   what makes the seam read as a design rather than as a mistake. */
body.skin-lions .topbar{border-bottom-color:var(--border-strong)}
/* THE SKIN PAINTS THE BODY, NOT JUST THE FURNITURE (round 11). Every lions
   view painted its own light ground while the page body behind it stayed
   graphite, so any margin collapsing through a view's top edge opened a
   black band between the topbar and the content - the Season tab wore a
   24px one and a critic read it as broken. The literal matches the skin's
   own --bg-app above; it cannot be var(--bg-app) because the token is
   re-valued on .lions and .topbar, and body is neither. */
body.skin-lions{background:#eef1f3}

/* ---------------------------------------------------------- THE MASTHEAD */
/* 52px uppercase in the poster face, on one line with the kicker beside it. */
.lions .mh-title{font-weight:700;font-size:clamp(34px,4vw,52px);line-height:.95;
  letter-spacing:.01em;text-transform:uppercase}
.lions .mh-dek{font-size:15px;font-weight:300;line-height:1.85;margin-top:14px}
/* The rule the mock draws between the dek and the first row of controls. It
   hangs off the control row rather than the masthead because the masthead is
   78ch wide and this rule spans the column. */
.lions .draft-setup,.lions .live-head{margin-top:24px;padding-top:22px;
  border-top:1px solid var(--border-strong)}

/* ------------------------------------------------------------- THE CARDS */
/* No border, because the fill against the ground already separates them, and
   a hairline on top of that reads as a box drawn around a box. */
.lions .panel,.lions .pickpath,.lions .method,.lions .board-scroll,
.lions .keepers,.lions .rptpanel,.lions .game{border:none;border-radius:var(--r-lg)}
.lions .pickpath{padding:24px 26px 22px}
/* Rows inside a card keep the quietest rule the palette has; two of them were
   written as literals below the token floor and vanish on white. */
.lions .pl{border-bottom-color:var(--border-row)}
.lions .dg-cell.filled{border-top-color:transparent;border-right-color:transparent;
  border-bottom-color:transparent}
/* the unchanged third of the injury split bar, a literal graphite */
.lions .ms-0{background:var(--track)}
/* THE SEVERITY BAR, whose three bands were written straight rather than in
   tokens: a stale share drawn at #222c31 is a black bar across a white card,
   and the questionable band's amber at 55% goes to peach on this ground. The
   band that means "we do not know" is the one that must NOT look like a
   reading, so it takes the track grey - the same colour every empty meter in
   this skin already uses for "nothing here". */
.lions .is-stale{background:var(--track)}
.lions .is-q{background:color-mix(in srgb,var(--risk) 38%,transparent)}

/* ------------------------------------------------------- THE CLOCK BAND */
/* The one dark object in the page, and the mock's whole hero: a strip of turf
   with the pick number set 56px in the poster face. The stripes are the
   mowing pattern, at 4.5% - visible as texture, never as lines you count.

   THE ONLY min-width QUERY IN THIS FILE, and it is here for a reason worth
   the exception. Below 640 the phone's full-screen mode LIFTS this same card
   into its fixed bar - `background:none`, 16px type, laid between the X and
   the pick count - and it is a light bar. These rules set white ink for a
   near-black band, so on the phone the clock read white-on-#eef1f3 and the
   pick number vanished into the bar, on top of the count. Found by shooting
   375 rather than by reading the cascade.

   Scoped by WIDTH and not by `body:not(.dr-full)`, which was the first fix
   and was wrong: dr-full is on the body at every width once a draft is
   running, so excluding it would have taken the turf off the desktop hero
   the moment the draft it exists for started. The band is a desktop object;
   the phone bar is its own design, and it only wants the palette. */
@media (min-width:641px){
  .lions .clock-card{align-items:flex-end;gap:24px;padding:22px 26px;
    background:var(--turf);border:none;border-radius:var(--r-lg);
    background-image:repeating-linear-gradient(90deg,
      rgba(255,255,255,.045) 0 16px,transparent 16px 32px);
    margin-bottom:var(--sp-4)}
  .lions .clock-time{font:700 56px/.9 var(--font-poster);letter-spacing:.02em;color:#fff}
  /* WHITE EVEN WHEN IT IS YOUR TURN, which the mock is right about: the
     accent is a mid blue and this ground is near-black, so the one state that
     must never be missed would be the LEAST legible thing on the band. The
     turn is carried where the mock carries it - the team reads YOU and the
     countdown reads now - and by the lit cell on the board. */
  .lions .clock-time.you{color:#fff}
  .lions .clock-card .caps{color:#b0b7bc}
  .lions .clock-team{font:400 17px/1.2 var(--font-ui);letter-spacing:.04em;color:#fff}
  .lions .clock-until .num{font:400 17px/1.2 var(--font-ui);color:#fff}
}

/* -------------------------------------------------------------- THE FIELD
   Yard lines behind the board, keyed to the SAME measurement the YOU rail is
   drawn from: syncGridMetrics() writes the team column's real pitch and the
   round gutter's width, so the lines land on the column boundaries at any
   league size instead of on a guessed cadence. The fallbacks are the CSS
   track's own numbers, so it is right before the first measurement too.

   At 8% on a #eef1f3 ground this is texture, not information - it carries
   nothing, and every cell that has something to say paints over it.

   THE HASH MARKS ARE NOT REPRODUCED, and that is a measurement rather than a
   preference. The mock's field is three rounds tall, so its two rows of
   hashes sit a third and two thirds down a short plate and read as field
   markings. This grid is fifteen rounds in a 62vh scroller: the same two rows
   land beside round 4 and round 11 for no reason either number can give, and
   on screen they read as two stray rules across the board. The vertical lines
   survive because they are keyed to something real - the columns. */
.lions .draft-grid{--dg-track:107px;--dg-gutter:72px}
.lions .draft-grid::before{content:"";position:absolute;inset:-10px 0;z-index:0;
  pointer-events:none;opacity:.08;
  background-image:
    repeating-linear-gradient(90deg,var(--turf) 0 2px,transparent 2px var(--dg-track));
  background-position:var(--dg-gutter) 0}

/* --------------------------------------------------------- THE BOARD GRID */
.lions .dg-team{background:var(--bg-sunken);border-color:transparent;
  border-radius:var(--r-md)}
.lions .dg-team.you{background:var(--accent);color:#fff}
/* The lit cell keeps being the only lit thing, but a glow becomes a drop - and
   on paper the inverted plate needs the accent ring to stay attached to the
   rest of the app's "this is live" vocabulary. */
.lions .dg-cell.now{box-shadow:0 0 0 2px var(--accent),0 6px 18px rgba(15,26,34,.28)}
/* :not(.now) IS LOAD-BEARING, and it was found by looking rather than by
   reasoning: `.lions .dg-cell.mine` (0,3,0) out-specifies `.dg-cell.now`
   (0,2,0), so tinting your own column painted the accent-soft wash straight
   over the one lit cell on the board - the cell went pale blue and empty, and
   the only thing on screen that says where you are stopped saying it. Your
   column is a channel; the cell on the clock is not part of it. */
/* AND SO IS THE REST OF THIS CHAIN, found the same way when the ghost board
   filled the other 130 cells (2026-08-15). `.lions .dg-cell.mine` (0,3,0)
   out-specifies every position rule on the board (0,2,0), so the accent wash
   was painting over the position colour of every cell in YOUR OWN COLUMN -
   the one column where "how many backs have I taken" is the question. It was
   invisible while your column held six picks against 144 empties; a full
   board made it fifteen colourless cells in a coloured grid.
   The channel tint is for cells with nothing else to say. A cell carrying a
   player carries his position, and the column still reads as a column because
   `.dg-you-rail` draws one outline around the whole thing - which was always
   the argument for that rail. */
.lions .dg-cell.mine:not(.now):not(.filled):not(.keeper):not(.ghost){
  background:var(--accent-soft)}

/* -------------------------------------------------------- THE PICK PATH */
/* The lead row is a tinted plate with the accent as a rule down its edge,
   which is the mock's mark for "this is the one" in three separate panels. */
.lions .pp-lrow.lead{background:var(--accent-soft);border-color:transparent;
  box-shadow:inset 3px 0 0 var(--accent)}
.lions .pp-rank{background:var(--track);border-color:transparent}
.lions .pp-lrow.lead .pp-rank{background:var(--accent)}
.lions .pp-vclaim{font-weight:300;font-size:25px;line-height:1.35}
.lions .pp-vclaim b{font-weight:400}
.lions .pp-cand b{font-weight:400}

/* ----------------------------------------------------- THE BEST AVAILABLE */
/* The mock's rows are taller and lighter than the graphite table's: 400 at
   15px, because on white a 600 is a shout where on graphite it was the only
   way to be read at all. */
.lions .ba{padding:13px var(--sp-4)}
.lions .ba .nm{font-weight:400;font-size:15px}
/* THE PILL IS A DESKTOP SHAPE, and the suite is what said so: at 375 and 414
   the phone gives these position tabs a 44px thumb target, and `.lions
   .batab` (0,2,0) out-specifies the phone rule that grants it even from
   inside the media query - so the mock's 30px pill quietly reimposed itself
   on nine controls a thumb has to hit. Second time this port made that
   mistake; the clock band above made it first. */
@media (min-width:641px){
  .lions .batab{height:30px;padding:0 16px;border-radius:20px}
}
/* THE MIDDLE BAND GOES GREY, which is the mock's whole ladder for this
   column: one colour for the players who will not last, and grey for
   everyone else. Graphite runs an amber step between them, and on white that
   amber at .75 is a muddy olive that reads as a third meaning rather than as
   a middle. The percentage beside the bar is what carries the number; the
   bar only has to say "urgent or not". */
.lions .meter.mid i{background:var(--neutral);opacity:1}
.lions .slot .who,.lions .pl-nm,.lions .qrow .nm{font-weight:400}

/* ============================================================ TOPBAR
   THE BAR WRAPS AT EVERY WIDTH, NOT AT A BREAKPOINT. brand + tabs + meta only
   fit on one line above ~1160px, so a nowrap row was a promise the layout
   could not keep: below that the meta strip ran off the right edge and took
   the whole document's scrollWidth with it - 451px of scroll at a 375px
   viewport, and still broken at 768. Hand-tuning breakpoints against measured
   string widths would only relocate the bug, because every string in the strip
   is dynamic (the countdown, the ADP stamp, the room name).

   So the row wraps whenever it must, at any width, for any content, and the
   height is a floor rather than a fixed 52px. .hide-m below is then purely a
   question of taste - what a phone is worth showing - and can never again be
   the only thing standing between the page and a horizontal scrollbar. */
/* ══════════════════ THE SECTION RAIL ══════════════════
   Fixed rather than a flex column with its own scroller, which is how the
   template does it. The template has no 1325px table and no sticky sort
   header; making the main column an overflow container here would have
   re-parented `position:sticky` onto it and taken the board's column headers
   with it. A fixed rail and a body inset changes nothing about how the
   document scrolls. */
/* ── 2c THREE PLATES ── his chosen direction, 2026-08-11, ported from
   design/edge-rail-2c.html, whose own THE PATCH block is written against these
   selectors. The rail's ground stops being one flat black: the brand and the
   foot are RAISED PLATES with a lit top edge, and the nav between them is a
   RECESSED WELL. The active row rises back to plate level.

   VALUE STRUCTURE ONLY - no gradient, no texture, no hue - and that is the
   whole reason this was the direction to take rather than a nicer colour. The
   rail is the one object that has to stand against BOTH grounds; DECISIONS
   records that as why it stays graphite while the topbar follows the view. A
   rail that gained a tint would be picking a side. Depth is the single
   property a paper page and a graphite page both read the same way, so the
   rail says "I am the frame" in a language neither ground argues with.

   The active mark moves from the graphite theme's cyan to Lions silver for
   exactly that reason: it was importing an accent from a view that is not
   necessarily the one on screen.

   THE TOKENS ARE RAIL-LOCAL, declared on .sidenav rather than in :root. They
   describe one object's interior lighting - a lip, a cut, a well - and nothing
   else in this app has an interior to light. In the palette they would invite
   a second user, and a second user is what turns a value into a compromise. */
.sidenav{--rail-well:#07090a;   --rail-plate:#141a1d;
  --rail-lip:rgba(255,255,255,.06);   --rail-cut:rgba(0,0,0,.7);
  --rail-mark:#b0b7bc;   --rail-ink:#7f8d95;
  --rail-ink-hover:#eef2f4;   --rail-rule:rgba(176,183,188,.1);
  position:fixed;left:0;top:0;bottom:0;width:212px;z-index:30;
  display:flex;flex-direction:column;
  /* PADDING ZERO, and it is structural rather than cosmetic: a plate has to
     reach the rail's own edges or it is a card lying on a black ground instead
     of a surface the rail was cut from. The horizontal padding moves into each
     child, which is why every rule below carries its own. */
  padding:0;
  background:var(--rail-well);border-right:1px solid var(--border-subtle);
  transition:transform .18s ease}
/* The 212px body inset stays exactly where it already was. The mock's patch
   restates it, and a second copy is how two numbers that must agree begin to
   disagree - so this is the existing line, not a new one. */
body{padding-left:212px}

/* PLATE ONE. */
.sn-brand{padding:26px 22px 22px;background:var(--rail-plate);border-bottom:none;
  box-shadow:inset 0 1px 0 var(--rail-lip),0 1px 0 var(--rail-cut)}
.sn-brand .brand-mark{display:block;font:500 27px/1 var(--font-display);
  font-style:normal;letter-spacing:.04em;color:#f6f9fa}
/* The terminal cursor after the wordmark, as ::after so index.html does not
   change - the same reason the caption's rule below is one.

   AND IT IS THE ONE HUED THING IN THE RAIL, his call 2026-08-11 (6th): "just
   the hero logo... to match the intro." That is a real distinction rather than
   an exception to 2c's no-hue rule: the cursor is part of the WORDMARK, and a
   wordmark carries the brand. Everything the rail says about NAVIGATION - the
   active row, the plates, the caption rules - stays hueless, which is what
   kept the rail able to stand against both grounds. The logo is not
   navigation. */
.sn-brand .brand-mark::after{content:"_";color:var(--accent)}
.sn-brand .brand-sub{display:block;margin-top:4px;font-family:var(--font-num);
  font-size:9px;letter-spacing:.24em;text-transform:uppercase;color:#818e95}

/* THE WELL between the two plates. */
.sn-scroll{flex:1;overflow-y:auto;padding:24px 0;display:flex;
  flex-direction:column;gap:24px;box-shadow:inset 0 3px 8px rgba(0,0,0,.55)}
/* The caption's rule runs all the way to the rail's edge, which is what gives
   the well its horizon. */
.sn-cap{display:flex;align-items:center;gap:10px;padding:0 22px 9px;
  font-family:var(--font-num);font-size:9px;letter-spacing:.22em;
  text-transform:uppercase;color:#5b666d}
.sn-cap::after{content:"";flex:1;height:1px;background:var(--rail-rule)}

/* PLATE TWO. */
.sn-foot{margin:0;padding:16px 18px 18px;background:var(--rail-plate);
  border-top:none;
  box-shadow:inset 0 1px 0 var(--rail-lip),0 -1px 0 var(--rail-cut)}
/* The one control on the foot plate, and it stays INSIDE the key system.
   The mock draws it as a flat slab, and taking that literally is what the
   first version of this did - it replaced `.key`'s face, edge and shadow
   outright, which cost the button its gradient and its press and broke the
   app-wide assertion that every button here is one physical key. That
   assertion is not decoration: the key system is defined once precisely so
   there is one test rather than one per control.

   So the rail RETUNES the tokens instead of overriding the rule. The key
   keeps its lip, its hard edge and its press from the system; only the
   material is the rail's, which is the same move the plates make and the same
   argument that keeps the rail graphite in the first place - the --key-*
   tokens are the PAGE's and light with whatever view is on screen, and this
   plate belongs to neither ground.

   The gradient is centred on the mock's #1b2226 rather than replacing it, so
   the slab it draws is what you see and the cap is still machined. */
.sn-foot{--key-face:linear-gradient(#232b30,#161c20);
  --key-face-hover:linear-gradient(#2b343a,#1c2429);
  --key-face-down:#141a1e;
  --key-edge:rgba(176,183,188,.28);
  --key-edge-hover:var(--rail-mark);
  --key-ink:#e2e8ea;
  --key-ink-hover:#fff}
.sn-foot .key{width:100%;height:auto;padding:8px 10px;font:500 12px/1.4 var(--font-ui)}

/* The rail is a LIST here, so the tab rules below - built for a horizontal bar
   with an underline - are restated rather than inherited. */
.sidenav .tabs{display:block}
.sidenav .tab{display:block;width:100%;text-align:left;min-height:0;
  padding:8px 20px;border:none;border-left:2px solid transparent;border-radius:0;
  font:400 13px/1.5 var(--font-ui);letter-spacing:0;text-transform:none;
  background:none;color:var(--rail-ink)}
.sidenav .tab:hover{color:var(--rail-ink-hover);background:none}
/* THE ACTIVE ROW RISES to plate level. No tint and no blue: the mark is the
   silver bar and the lit lip, which is the same object as the two plates - so
   "where you are" is told in the rail's own language instead of borrowing an
   accent from a view that may not be the one you are looking at. */
.sidenav .tab.active{font-weight:500;color:#fff;
  border-left-color:var(--rail-mark);background:var(--rail-plate);
  box-shadow:inset 0 1px 0 var(--rail-lip)}

/* The rail hides on demand and below 1100px, where 212px of chrome costs the
   board more than the navigation is worth. */
.sn-toggle{appearance:none;width:30px;height:30px;padding:0;cursor:pointer;
  background:none;border:1px solid var(--border-subtle);border-radius:var(--r-sm);
  color:var(--text-secondary);display:grid;place-items:center;margin-right:var(--sp-3)}
.sn-toggle:hover{border-color:var(--accent-border);color:var(--accent)}
.sn-toggle span,.sn-toggle span::before,.sn-toggle span::after{
  display:block;width:13px;height:1.5px;background:currentColor;content:""}
.sn-toggle span{position:relative}
.sn-toggle span::before{position:absolute;top:-4.5px}
.sn-toggle span::after{position:absolute;top:4.5px}
body.sn-off{padding-left:0}
body.sn-off .sidenav{transform:translateX(-100%)}
@media (max-width:1100px){
  body{padding-left:0}
  .sidenav{transform:translateX(-100%)}
  body.sn-open{padding-left:0}
  body.sn-open .sidenav{transform:none;box-shadow:0 0 0 100vmax rgba(0,0,0,.5)}
  /* As an overlay the rail needs a WAY OUT it did not have: the toggle sat
     UNDER the open drawer (z 30) at 30px, so the only exit was navigating
     somewhere. The toggle now rides above the drawer at thumb size and its
     bars cross into an X, and app.js closes on any tap outside the rail. */
  .sn-toggle{width:44px;height:44px;position:relative;z-index:31}
  body.sn-open .sn-toggle span{background:transparent}
  body.sn-open .sn-toggle span::before{top:0;transform:rotate(45deg)}
  body.sn-open .sn-toggle span::after{top:0;transform:rotate(-45deg)}
}
@media (prefers-reduced-motion:reduce){.sidenav{transition:none}}

/* ══════════════════ EDITORIAL MASTHEAD ══════════════════
   The template's signature: a large serif title in SENTENCE CASE, a mono
   kicker in caps beside it, and a dek that says what the page is for.

   Sentence case matters more than it sounds. The old headers were
   `text-transform:uppercase` on a UI sans, which is a control-panel
   convention; a serif set in sentence case is what makes the page read as
   written rather than as instrumentation, and it is the whole difference
   between this template and the six neon variants beside it.

   clamp() rather than the template's flat 64px: 64px is a landing-page size,
   and this masthead sits above a working tool that people open on a laptop. */
.masthead{margin:var(--sp-5) 0 var(--sp-4);max-width:78ch}
.mh-title{display:inline;margin:0;font:400 clamp(30px,3.4vw,46px)/1.05 var(--font-poster);
  color:var(--text-primary);letter-spacing:-.005em}
.mh-kicker{margin-left:var(--sp-3);font-family:var(--font-num);font-size:10px;
  letter-spacing:.26em;text-transform:uppercase;color:var(--accent);
  white-space:nowrap}
.mh-dek{margin:12px 0 0;max-width:60ch;font-size:14px;line-height:1.55;
  color:var(--text-secondary);text-wrap:pretty}
@media (max-width:620px){
  .mh-kicker{display:block;margin:6px 0 0}
  .mh-dek{font-size:13px}
}

.topbar{background:var(--bg-header);border-bottom:1px solid var(--border-subtle);position:sticky;top:0;z-index:10}
/* row-gap is deliberately tighter than column-gap: the columns are separate
   regions of the bar, the rows are the same bar folded. */
.topbar-inner{display:flex;align-items:center;flex-wrap:wrap;gap:var(--sp-3) var(--sp-5);min-height:52px}
.brand{display:flex;gap:var(--sp-3);padding-right:var(--sp-5);border-right:1px solid var(--border-subtle);height:52px;align-items:center}
.brand-mark{font:800 italic 19px/1 var(--font-ui);letter-spacing:.04em;color:var(--text-primary)}
.brand-mark::after{content:"_";font-style:normal;color:var(--accent)}
.brand-sub{font-size:var(--fs-xs);letter-spacing:var(--track-caps);text-transform:uppercase;color:var(--text-muted)}
/* stretch, not height:100% - the bar's height is now a min-height, and a
   percentage height against an auto-height parent resolves to auto, which
   would drop the active tab's 3px underline up under the text. */
/* Six tabs now - Momentum and Injuries joined - so on a phone the nav wraps
   like the meta strip does (same invariant as the topbar fix): a width-tuned
   rule would be a guess about dynamic text. */
.tabs{display:flex;flex-wrap:wrap;gap:0 var(--sp-2);align-self:stretch;min-width:0}
.tab{appearance:none;background:none;border:none;border-bottom:3px solid transparent;color:var(--text-muted);
  font:600 var(--fs-sm)/1 var(--font-ui);letter-spacing:var(--track-caps);text-transform:uppercase;
  padding:0 var(--sp-3);cursor:pointer;height:auto;min-height:48px}
.tab:hover{color:var(--text-secondary)}
.tab.active{color:var(--accent);border-bottom-color:var(--accent)}
.topbar-meta{margin-left:auto;display:flex;align-items:center;flex-wrap:wrap;gap:0;min-width:0}
/* The format chip became a <label> when it stopped being a literal and became
   a control, so these rules name it explicitly - a `>span` selector silently
   dropped its divider and its padding, and it sat flush against the stamp.
   The dividers hang off the RIGHT of each chip, closed by a left border on the
   first. Same four rules on screen as the old border-left-on-everything, but
   the chips that drop out on a phone are always a PREFIX of the strip
   (stamp, then format) - so a right-hand divider can never be left stranded at
   the head of the strip, and the room selector, which is always last and never
   hides, is the one chip that draws no divider at all. */
.topbar-meta>span,.topbar-meta>.fmt{padding:0 var(--sp-4);border-right:1px solid var(--border-subtle);
  color:var(--text-secondary);font-size:var(--fs-sm);font-weight:600;letter-spacing:.04em;
  white-space:nowrap}
.topbar-meta>span:first-child,.topbar-meta>.fmt:first-child{border-left:1px solid var(--border-subtle)}
.topbar-meta>span:last-child,.topbar-meta>.fmt:last-child{border-right:none}
.sync-dot{display:inline-block;width:6px;height:6px;border-radius:50%;background:var(--accent);margin-right:7px;vertical-align:1px}
.roomsel{display:flex;align-items:center;gap:var(--sp-2);min-width:0}
.roomsel select{min-width:0}
/* off-format room badge: --risk is the reserved caution slot (see its token
   comment); this is a state warning, not a series color */
.off-badge{color:var(--risk);border:1px solid var(--risk-border);
  border-radius:var(--r-sm);padding:3px 6px;font-weight:800;cursor:help}
select{background:var(--bg-surface-raised);color:var(--text-primary);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:9px 8px;font:600 var(--fs-sm)/1 var(--font-ui);min-height:36px}
select:focus{outline:none;border-color:var(--accent)}

/* ============================================================ STAT STRIP
   Proportions ported from the Claude-design template - taller tiles, a 40px
   headline, mono tracked-out labels, and a graphic in every tile. The
   GRAPHICS are not ported: the template's are decorative curves implying
   months of history, and this board has three days of it. Each one here is a
   reading of something real (see renderStats). */
.stat-strip{display:grid;grid-template-columns:repeat(4,1fr);gap:var(--sp-3);margin:var(--sp-4) 0}
/* THE SCORING RESTATEMENT. A hairline rule rather than a banner: it is on
   screen at all times, so it must read as a caption on the board and never
   compete with the verdict. The scoring name takes the accent - it is the one
   word that changed - and everything else stays secondary. `.fmtnote b` is the
   measured consequence of the last switch, which is the part that answers "did
   that do anything". No team colour and no gold: both are spoken for below the
   fold, and this sits above it. */
.fmtnote{margin:var(--sp-3) 0 0;padding:9px 12px;border:1px solid var(--border-subtle);
  border-left:2px solid var(--accent);border-radius:var(--r-sm);
  background:var(--bg-surface);color:var(--text-secondary);
  font-size:12px;line-height:1.5}
.fmtnote .sc{color:var(--accent);font-weight:700;letter-spacing:.06em;text-transform:uppercase}
.fmtnote b{color:var(--text-primary);font-weight:600}
.fmtnote .off{color:var(--text-muted)}
/* THE ROOM SHORTCUT, ON THE FIRST SCREEN. Deliberately built as a CAPTION
   WITH CONTROLS IN IT and not as a second control bar: the ctlbar below is
   the bar, and two stacked bars would tell the reader the board has two of
   them. So it takes the fmtnote's material - hairline box, accent left edge,
   12px secondary ink - and differs only by holding chips, which is the same
   argument the scoring restatement makes one line down. */
.quickbar{display:flex;align-items:center;flex-wrap:wrap;gap:var(--sp-2);
  margin:var(--sp-3) 0 0;padding:8px 12px;
  border:1px solid var(--border-subtle);border-left:2px solid var(--accent);
  border-radius:var(--r-sm);background:var(--bg-surface)}
.quickbar-cap{color:var(--text-secondary)}
/* The note takes the rest of the line and wraps inside it rather than pushing
   the chips around: `min-width:0` is what lets a flex item narrower than its
   text exist at all, and without it the sentence forces the strip wider than
   the board at 375. */
.quickbar-note{flex:1 1 220px;min-width:0;color:var(--text-muted);
  font-size:12px;line-height:1.5}
@media (max-width:640px){
  /* Same ruling the board dek already carries at this width: read-once
     instruction stands down so the first player row comes up sooner. The
     chips still say what they do, and the Compare key still states the
     resulting sentence. */
  .quickbar-note{display:none}
}
.stat{position:relative;overflow:hidden;border:1px solid var(--border-subtle);
  border-radius:var(--r-lg);padding:18px 20px;min-height:142px;
  box-shadow:0 1px 2px rgba(0,0,0,.35);
  /* A fine noise wash plus a corner glow, so the tile reads as a surface
     rather than a flat swatch. Both are inline SVG/gradients - no asset, and
     nothing for the CSP to block. */
  background:
    url('data:image/svg+xml,%3Csvg xmlns=%22http://www.w3.org/2000/svg%22 width=%22120%22 height=%22120%22%3E%3Cfilter id=%22n%22%3E%3CfeTurbulence type=%22fractalNoise%22 baseFrequency=%220.9%22 numOctaves=%222%22/%3E%3C/filter%3E%3Crect width=%22120%22 height=%22120%22 filter=%22url(%23n)%22 opacity=%220.035%22/%3E%3C/svg%3E'),
    radial-gradient(80% 70% at 100% 0%,rgba(31,169,201,.06),transparent 60%),
    var(--bg-surface-raised)}
.stat.lead{box-shadow:inset 3px 0 0 var(--accent),0 1px 2px rgba(0,0,0,.35)}
.stat .caps{display:block;margin-bottom:var(--sp-2);font-size:10.5px;letter-spacing:.19em}
/* DISPLAY NUMERALS ARE SERIF, which is the template's own rule - it sets
   every 26-64px figure in Newsreader and never uses it smaller. The grid
   below stays mono; a serif column of ADPs would be unscannable. */
.stat-val{display:block;font:400 44px/1.05 var(--font-display);font-variant-numeric:tabular-nums;
  position:relative;z-index:2;margin-top:var(--sp-2)}
.stat-val.name{font-family:var(--font-display);font-weight:500;font-size:25px;line-height:1.2;white-space:nowrap;overflow:hidden;
  text-overflow:ellipsis;margin-top:14px}
.stat-val .of{color:var(--accent)}
.stat-note{font-size:var(--fs-lg);color:var(--text-secondary);margin-top:9px;position:relative;z-index:2;
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
/* The tile with the ring reserves its column, so the sentence cannot run
   through the dots. A width, not a padding - see .stat-sub. */
.stat.has-ring .stat-note{width:64%}
/* The caption ends where the spark begins. A width, not a padding: padding
   keeps the text clear but leaves the BOX spanning the tile, which is
   invisible on screen and means nothing can assert the two do not overlap. */
.stat-sub{font-size:var(--fs-xs);color:var(--text-muted);margin-top:5px;position:relative;z-index:2;
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis;width:53%}
.stat-val .d-up,.stat-val .d-dn{font-size:22px;margin-left:var(--sp-2)}

/* ROTATING HERO TILES. The arrows sit against the caps label rather than over
   the name, so nothing moves under the pointer when the tile advances, and the
   counter is between them because "3/26" is the thing that makes the arrows
   look like arrows rather than decoration.

   The body is a fixed min-height: the tiles cycle through players whose note
   and sparkline wrap differently, and without it the whole strip changed
   height every six seconds and shoved the board down the page. */
.stat.hero .hero-nav{position:absolute;top:14px;right:16px;z-index:3;
  display:flex;align-items:center;gap:6px}
.stat.hero .hero-at{min-width:34px;text-align:center;
  font:600 var(--fs-xs)/1 var(--font-num);color:var(--text-muted);
  font-variant-numeric:tabular-nums}
.stat.hero .hero-step{display:grid;place-items:center;width:20px;height:20px;
  padding:0;border:1px solid var(--border-subtle);border-radius:5px;
  background:transparent;color:var(--text-secondary);cursor:pointer;
  font:600 13px/1 var(--font-ui)}
.stat.hero .hero-step:hover{border-color:var(--accent);color:var(--accent)}
.stat.hero .hero-step:focus-visible{outline:2px solid var(--accent);outline-offset:1px}
.stat.hero .hero-body{min-height:96px}
.stat.hero .stat-val.name[data-key]{cursor:pointer}
.stat.hero .stat-val.name[data-key]:hover{color:var(--accent)}
/* The worth is the quantity the tiles are RANKED by, so it is on screen beside
   the pick delta rather than left implicit. Quieter than the delta: it is the
   reason for the ordering, not the headline. */
.stat.hero .hero-worth{margin-left:var(--sp-2);font:600 var(--fs-sm)/1 var(--font-num);
  color:var(--text-muted);font-variant-numeric:tabular-nums}

/* One pip per source, lit where that source publishes numbers for this board.
   This is what the headline "8/8" counts, drawn. */
.stat-pips{position:absolute;right:20px;bottom:18px;display:flex;gap:4px;z-index:2}
.stat-pips i{width:5px;height:20px;border-radius:2px;background:var(--border-strong)}
.stat-pips i.on{background:var(--accent);box-shadow:0 0 8px rgba(31,169,201,.45)}

/* The ring is the headline's fraction, drawn - dotted, per the template. */
/* TOP-RIGHT and fully contained, as the template has it. Vertically centred
   it collided with the note line and ran off the tile's right edge - the ring
   was clipped and the sentence underneath it ran through the dots. */
.stat-ring{position:absolute;right:16px;top:12px;width:106px;height:106px;z-index:1;opacity:.95}
.stat-ring .track{fill:none;stroke:var(--border-strong);stroke-width:2;stroke-dasharray:.9 2.4}
.stat-ring .arc{fill:none;stroke:var(--accent);stroke-width:3;stroke-dasharray:.9 2.4;
  stroke-linecap:round}

/* The price line. Bottom-right, behind the text, clipped by the tile. It
   plots the PICK NUMBER directly - bigger sits higher - so a green line
   always climbs and a red one always falls, matching the arrow and the sign
   beside it. The caption states both endpoints regardless. */
/* Sits in the tile's bottom-right QUADRANT, not across its whole floor. At
   62% wide and 64px tall it ran straight through the caption that explains
   which way the axis points - the one line the chart cannot be read without. */
/* Sized to the template's proportions. At 46% x 38px the line was a hairline
   ruled across the corner; the area fill underneath is what makes it read as
   part of the tile. */
.stat-spark{position:absolute;right:0;bottom:0;width:52%;height:62px;z-index:1;
  pointer-events:none}
/* Same green-up / red-down rule as every other ADP change on the board.
   The gradients are driven from CSS so the two poles share one set of shapes:
   `g0/g1` fill the area top-to-bottom, `s0/s1/s2` fade the stroke in from the
   left so the line emerges rather than starting at a hard edge. */
.stat-spark .ln{fill:none;stroke-width:1.8;vector-effect:non-scaling-stroke}
/* The dashed bridge over a snapshot gap (round 11): dashed is this app's one
   vocabulary for "not a fact", so the reader sees the market went unread
   between two points instead of a chart that broke. Toned explicitly per
   class exactly as the dot is - this file has no currentColor convention to
   borrow - and no area is ever filled under a bridge. */
.stat-spark .lngap{fill:none;stroke-width:1.4;opacity:.5;
  stroke-dasharray:3 4;vector-effect:non-scaling-stroke}
.stat-spark.up .lngap{stroke:var(--adp-up)}
.stat-spark.dn .lngap{stroke:var(--adp-dn)}
.stat-spark.flat .lngap{stroke:var(--neutral)}
.stat-spark .ar{stroke:none}
.stat-spark .g1,.stat-spark .s0{stop-opacity:0}
.stat-spark .g0{stop-opacity:.42}
.stat-spark .s1{stop-opacity:.55}
.stat-spark .s2{stop-opacity:1}
.stat-spark.up .g0,.stat-spark.up .g1{stop-color:var(--adp-up)}
.stat-spark.up .s0,.stat-spark.up .s1,.stat-spark.up .s2{stop-color:var(--adp-up)}
.stat-spark.dn .g0,.stat-spark.dn .g1{stop-color:var(--adp-dn)}
.stat-spark.dn .s0,.stat-spark.dn .s1,.stat-spark.dn .s2{stop-color:var(--adp-dn)}
.stat-spark.flat .g0,.stat-spark.flat .g1{stop-color:var(--neutral)}
.stat-spark.flat .s0,.stat-spark.flat .s1,.stat-spark.flat .s2{stop-color:var(--neutral)}
.stat-spark.up .dot{fill:var(--adp-up)}
.stat-spark.dn .dot{fill:var(--adp-dn)}
.stat-spark.flat .dot{fill:var(--neutral)}

@media (max-width:1100px){
  .stat-strip{grid-template-columns:repeat(2,1fr)}
}
@media (max-width:620px){
  .stat-strip{grid-template-columns:1fr}
}

/* ==================================================== THE KEY, from his
   "Board button system". Every button in this app is one physical object: a
   gradient face, a lit top lip, a hard 2px bottom edge and a soft drop - and
   it PRESSES, moving 2px down into an inset shadow when you push it.

   DEFINED ONCE AND ADOPTED BY NAME. The app had four unrelated button styles
   (.fchip, .segbar .seg, .batab, .primary) reached from 37 markup sites, so
   the choice was one grouped selector here or 37 edits and a fifth style the
   next time. The class names stay, because they carry MEANING - a filter chip
   and a segment are different controls - and only the surface is shared.

   The face colours are his literals, named here so the system has one home.
   Nothing in the existing palette was close enough to reuse: these are a
   physical material, not a semantic. */
:root{
  --key-face:linear-gradient(#2c383f,#171f24);
  --key-face-hover:linear-gradient(#3a4952,#1e282e);
  --key-face-down:#12191e;
  --key-edge:rgba(160,200,220,.42);
  --key-edge-hover:rgba(180,220,240,.62);
  --key-ink:#cfe0e8;
  /* Hover lifts the ink to the brightest thing the theme has, and the on-state
     drops it to the ink that reads on a lit face. Both were literals (#fff and
     #04222b) - which is the same as saying the key system assumed a dark
     theme, since on a white face both are the wrong way round. */
  --key-ink-hover:#fff;
  --key-on-ink:var(--on-accent);
  /* THE PRESS. Lip, hard edge, soft drop - and the inset the face falls into
     when it is pushed, which is the whole reason a key reads as a key. */
  --key-lift:inset 0 1px 0 rgba(255,255,255,.16),0 2px 0 rgba(0,0,0,.75),0 3px 6px rgba(0,0,0,.5);
  --key-down:inset 0 3px 5px rgba(0,0,0,.8);
  /* ON: the accent as a lit keycap with dark ink, not a tinted outline. */
  --key-on-face:linear-gradient(#2b8fcb,#0f6ba3);
  --key-on-face-hover:linear-gradient(#3a9bd6,#1478b3);
  --key-on-face-down:linear-gradient(#1478b3,#0b5c8c);
  --key-on-edge:rgba(60,150,210,.95);
  --key-on-lift:inset 0 1px 0 rgba(255,255,255,.55),0 2px 0 rgba(0,0,0,.65),0 3px 7px rgba(0,0,0,.5);
  --key-on-down:inset 0 3px 5px rgba(0,0,0,.35);
}
/* `.intro-dest` LEFT THIS GROUP with his 3A. It was adopted into the button
   system deliberately - the intro taught a new reader what a pressable thing
   looks like here - and 3A gives the cover its own 2px-ruled panels instead.
   Removing it from the selector rather than overriding it is the point: the
   member that stayed behind kept `height:32px`, which clamped every
   destination to 32px and made the whole cover render on top of itself. */
.key,.fchip,.segbar .seg,.batab,button.primary{
  display:inline-flex;align-items:center;justify-content:center;
  height:32px;padding:0 14px;
  font:500 var(--fs-lg)/1 var(--font-ui);color:var(--key-ink);
  background:var(--key-face);border:1px solid var(--key-edge);
  border-radius:var(--r-sm);box-shadow:var(--key-lift);cursor:pointer;
  transition:background .12s,border-color .12s,color .12s}
.key:hover,.fchip:hover,.segbar .seg:hover,.batab:hover,button.primary:hover{
  background:var(--key-face-hover);border-color:var(--key-edge-hover);color:var(--key-ink-hover)}
.key:active,.fchip:active,.segbar .seg:active,.batab:active,button.primary:active{
  background:var(--key-face-down);box-shadow:var(--key-down);
  transform:translateY(2px);border-color:rgba(160,200,220,.3)}
.key:focus-visible,.fchip:focus-visible,.segbar .seg:focus-visible,
.batab:focus-visible,button.primary:focus-visible{
  outline:2px solid var(--accent);outline-offset:2px}
.key:disabled,.fchip:disabled,.segbar .seg:disabled,.batab:disabled,button.primary:disabled{
  opacity:.4;cursor:default;transform:none;
  box-shadow:inset 0 1px 0 rgba(255,255,255,.08),0 2px 0 rgba(0,0,0,.5)}
.key:disabled:hover,.fchip:disabled:hover,.segbar .seg:disabled:hover,
.batab:disabled:hover,button.primary:disabled:hover{
  background:var(--key-face);border-color:var(--key-edge);color:var(--key-ink)}

/* ON. One state name per control, all landing on the same lit face. */
.key.on,.fchip.active,.segbar .seg.on,.batab.active,button.primary{
  font-weight:600;color:var(--key-on-ink);background:var(--key-on-face);
  border-color:var(--key-on-edge);box-shadow:var(--key-on-lift)}
.key.on:hover,.fchip.active:hover,.segbar .seg.on:hover,.batab.active:hover,
button.primary:hover{background:var(--key-on-face-hover);color:var(--key-on-ink)}
.key.on:active,.fchip.active:active,.segbar .seg.on:active,.batab.active:active,
button.primary:active{background:var(--key-on-face-down);
  box-shadow:var(--key-on-down);transform:translateY(2px)}

/* MONO, for the counted chips: a shorter key in the data face. */
.key.mono,.batab{gap:8px;height:28px;padding:0 10px;
  font:400 var(--fs-md)/1 var(--font-num);letter-spacing:.06em}
.key.icon{width:34px;padding:0;font-size:11px}
.key.accent{color:var(--accent);background:linear-gradient(#1d3a45,#0f1c23);
  border-color:rgba(22,131,196,.55);
  box-shadow:inset 0 1px 0 rgba(255,255,255,.14),0 2px 0 rgba(0,0,0,.75),0 3px 6px rgba(0,0,0,.5)}
.key.accent:hover{background:linear-gradient(#275060,#152c36);
  border-color:rgba(140,235,255,.8);color:#c9f4ff}
.key.accent:active{background:#101d23;box-shadow:var(--key-down);transform:translateY(2px)}

/* ================================================== START HERE (intro)
   The one screen in this app that assumes you know nothing. Everything else
   is a dense instrument; this is the sentence that makes the instrument
   legible, and it is shown once. */
/* HIS 3A, "Eventually treatment - drifting photo field, 2px rules, one accent".
   The card becomes a COVER: a full-bleed drifting colour field under a heavy
   scrim, a 96px serif nameplate, and destinations that are 2px-ruled panels
   rather than the app's key primitive. That last point supersedes the note this
   block used to carry - the intro taught the button language, and 3A gives the
   cover its own. Every literal below was read off the mock's own inline styles;
   --accent already IS its rgb(95,212,240). */
/* Centered by AUTO MARGINS, not by place-items: a centered grid item taller
   than the viewport overflows in BOTH directions, and only the bottom half is
   scrollable - on a phone the card is ~1460px tall and its first 320px (the
   name of the product, the first destination) were unreachable at any scroll
   position. Auto margins collapse to zero when there is no free space, which
   is exactly "center when it fits, start when it does not". */
.intro{position:fixed;inset:0;z-index:60;display:grid;justify-items:center;
  padding:var(--sp-4);background:rgba(4,7,9,.72);backdrop-filter:blur(3px);
  overflow:auto}
.intro-card{position:relative;overflow:hidden;width:min(1280px,100%);
  margin-block:auto;
  background:#000;border:1px solid rgba(140,180,200,.14);
  border-radius:var(--r-lg);box-shadow:0 24px 60px -20px rgba(0,0,0,.9)}
/* THE FIELD. Three colour plates that pan and cross-fade. The mock ships three
   1920x900 JPEGs totalling 417KB for what are, measured, smooth low-frequency
   washes - so each is carried here as a 16x8 WebP the browser scales back up.
   Mean error under 1% per channel against the original, at 114 bytes apiece,
   and the page never asks the network for a decoration. Same reason the stat
   tiles draw their noise inline: nothing here for the CSP to block. */
.intro-field{position:absolute;inset:0;opacity:.375;pointer-events:none}
.intro-fx{position:absolute;inset:0;width:150%;background-position:center;
  background-size:cover;animation:introPan 45s linear infinite alternate,
  introFade 36s linear infinite}
.intro-fx:nth-child(1){background-image:url("data:image/webp;base64,UklGRmoAAABXRUJQVlA4IF4AAAAQAgCdASoQAAgAAMAWJbACdLoAApFVl8lYAP77mrN+//arWYP3slRxKyhenNNXoWKazzNPSyPwmjoMcjSQQrUv7f/LatE8kfz+2fDx/9fYyw+VYOq6//Gke4NrQAAA");animation-delay:0s,0s}
.intro-fx:nth-child(2){background-image:url("data:image/webp;base64,UklGRmgAAABXRUJQVlA4IFwAAAAwAgCdASoQAAgAAMAWJbACdLoAAzsZETXAMAD++1u394W1nxy0678LzPBc/j396R2Mh06pVRtJYBxGY/8qS/xEHer/EQf+x1x7qb8v/M5vfm+QJ+z/+mB/4JAAAA==");animation-delay:0s,-12s}
.intro-fx:nth-child(3){background-image:url("data:image/webp;base64,UklGRl4AAABXRUJQVlA4IFIAAAAQAgCdASoQAAgAAMAWJQBOgCHgTnEWhEQAAP7YAeYF8FIpvshHxzXSnkuCbL0WelepKGOQj8/I1ysD5+8EoW0Ftr+4QOxljbxs4Zr1QOxOcAAA");animation-delay:0s,-24s}
@keyframes introPan{from{transform:translateX(0)}to{transform:translateX(-25%)}}
@keyframes introFade{0%,22%{opacity:1}39%,94%{opacity:0}100%{opacity:1}}
/* The scrim is what makes 96px white type legible over a moving field. */
.intro-scrim{position:absolute;inset:0;pointer-events:none;
  background:linear-gradient(rgba(6,9,11,.72),rgba(6,9,11,.86))}
.intro-body{position:relative;z-index:2;padding:64px 72px 56px;
  display:flex;flex-direction:column;justify-content:center;gap:26px}
.intro-hd{display:flex;align-items:flex-end;gap:14px}
/* The nameplate. 96px is the mock's, not a token - the type scale here tops out
   at 26px because nothing else in the app is a cover. */
.intro-t{margin:0;font:400 96px/.9 var(--font-display);letter-spacing:-.02em;
  color:#fff}
.intro-bar{width:78px;height:8px;background:var(--accent);margin-bottom:11px}
/* THE KICKER IS QUIET, AND THE MARK CARRIES THE HUE. His note, 2026-08-11
   (6th): "drop the blue draft intelligence in the intro or add it to the grid.
   There's asymmetry right now." There was - the intro painted BOTH the bar and
   the words blue while the rail's identical lockup paints only the cursor,
   leaving the same two words grey. Two versions of one wordmark disagreeing
   about which half is the brand.

   Resolved toward the rail, because that is the version his earlier ruling
   already settled: "just the hero logo." The MARK is the logo - the rail's
   underscore, the intro's bar - and it takes the blue. "Draft intelligence" is
   a caption in both places, so it reads as one in both places. */
.intro-k{margin-bottom:14px;font:500 var(--fs-sm)/1 var(--font-num);
  letter-spacing:.24em;text-transform:uppercase;color:#818e95}
.intro-lead{margin:0;font:400 34px/1.3 var(--font-ui);letter-spacing:-.01em;
  color:rgba(255,255,255,.75);text-wrap:pretty}
.intro-lead b{color:#fff;font-weight:700}
.intro-dests{display:flex;flex-direction:column;gap:14px}
/* THE LEAD PANEL. 01 is the only destination that gets the accent, a filled
   START HERE and the full width - the design's way of saying where to begin,
   and the reason the other five can be quiet. */
.intro-dest{display:flex;align-items:center;gap:22px;padding:20px 26px;
  border:2px solid var(--accent);border-radius:var(--r-md);
  background:rgba(22,131,196,.16);text-align:left;color:var(--text-primary);
  width:100%;cursor:pointer;font:inherit}
.intro-n{font:500 var(--fs-md)/1 var(--font-num);letter-spacing:.16em;
  color:var(--accent)}
.intro-dest .intro-col{display:flex;flex-direction:column;gap:6px;flex:1}
.intro-dest .intro-line{display:flex;align-items:baseline;gap:16px}
.intro-dt{font:400 34px/1 var(--font-display);color:#fff}
.intro-dk{font:500 var(--fs-xs)/1 var(--font-num);letter-spacing:.14em;
  text-transform:uppercase;color:var(--accent)}
.intro-dd{margin:0;max-width:62ch;font:400 15px/1.6 var(--font-ui);
  color:rgba(255,255,255,.8);text-wrap:pretty}
.intro-go{align-self:center;font:700 var(--fs-md)/1 var(--font-ui);
  letter-spacing:.06em;text-transform:uppercase;color:var(--on-accent);
  background:var(--accent);border-radius:var(--r-md);padding:14px 20px;
  white-space:nowrap}
/* THE OTHER FIVE. auto-fit rather than the mock's repeat(5,1fr): the row is
   built from the rail, so a seventh tab must not overflow a hardcoded track
   count - and at the design's 1280 the two render identically. */
.intro-grid{display:grid;gap:12px;
  grid-template-columns:repeat(var(--intro-n,5),minmax(0,1fr))}
.intro-grid .intro-dest{display:flex;flex-direction:column;align-items:stretch;
  gap:8px;padding:16px 16px 18px;border:2px solid rgba(255,255,255,.2);
  background:rgba(255,255,255,.04)}
.intro-grid .intro-n{font-weight:400;letter-spacing:normal;
  color:rgba(255,255,255,.45)}
.intro-grid .intro-dt{font-size:24px;line-height:1.1}
.intro-grid .intro-dd{margin-top:2px;font-size:var(--fs-lg);line-height:1.6;
  color:rgba(255,255,255,.66)}
.intro-dest:hover{border-color:var(--accent)}
.intro-dest:focus-visible{outline:2px solid var(--accent);outline-offset:2px}
.intro-foot{display:flex;align-items:center;gap:14px}
.intro-btn{border:2px solid rgba(255,255,255,.35);border-radius:var(--r-md);
  background:transparent;color:rgba(255,255,255,.85);
  font:700 var(--fs-lg)/1 var(--font-ui);letter-spacing:.06em;
  text-transform:uppercase;padding:14px 22px;cursor:pointer;
  transition:border-color .12s,background .12s,color .12s}
.intro-btn.on{border-color:var(--accent);background:var(--accent);color:var(--on-accent)}
.intro-btn:hover{border-color:var(--accent)}
.intro-btn.on:hover{background:#7fe0f6;border-color:#7fe0f6}
.intro-btn:focus-visible{outline:2px solid var(--accent);outline-offset:2px}
.intro-fmts{margin-left:auto;font:400 var(--fs-sm)/1 var(--font-num);
  letter-spacing:.18em;text-transform:uppercase;color:rgba(255,255,255,.45)}
@media (max-width:900px){
  .intro-body{padding:40px 32px 36px;gap:20px}
  .intro-t{font-size:64px}
  .intro-lead{font-size:24px}
  .intro-dest{flex-direction:column;align-items:flex-start;gap:12px}
  .intro-dt{font-size:26px}
  /* Five 150px tracks do not fit a tablet; the count is a design value at the
     width it was drawn for, not at every width. */
  .intro-grid{grid-template-columns:repeat(2,minmax(0,1fr))}
}
@media (max-width:640px){
  .intro{padding:0}
  .intro-card{border-radius:0;min-height:100%}
  .intro-body{padding:28px 20px 24px}
  .intro-t{font-size:48px}
  .intro-bar{width:52px;height:6px;margin-bottom:8px}
  .intro-lead{font-size:20px}
  .intro-foot{flex-wrap:wrap}
  .intro-fmts{margin-left:0;flex-basis:100%}
  .intro-grid{grid-template-columns:1fr}
}
/* A field that drifts for the whole session is motion nobody asked for. */
@media (prefers-reduced-motion:reduce){
  .intro-fx{animation:none}
  .intro-fx:nth-child(2),.intro-fx:nth-child(3){opacity:0}
}

/* ============================================================ FILTERS */
.filters{display:flex;align-items:center;gap:var(--sp-2);flex-wrap:wrap;margin-bottom:var(--sp-3)}
.fdiv{width:1px;height:20px;background:var(--border-subtle);margin:0 var(--sp-1)}
.search{position:relative;margin-left:auto}
.search .lens{position:absolute;left:10px;top:50%;transform:translateY(-62%);width:9px;height:9px;
  border:1.5px solid var(--text-muted);border-radius:50%;pointer-events:none}
.search .lens::after{content:"";position:absolute;left:8px;top:9px;width:5px;height:1.5px;
  background:var(--text-muted);transform:rotate(45deg)}
.fsearch{background:var(--bg-surface);border:1px solid var(--border-subtle);border-radius:var(--r-md);
  color:var(--text-primary);font:var(--fs-sm)/1 var(--font-ui);padding:10px 10px 10px 28px;width:240px;min-height:36px}
.fsearch::placeholder{color:var(--text-muted)}
.fsearch:focus{outline:none;border-color:var(--accent)}

/* ============================================================ BOARD TABLE */
.board-scroll{overflow-x:auto;border:1px solid var(--border-subtle);border-radius:var(--r-md);background:var(--bg-surface)}
table.board{border-collapse:collapse;width:100%;min-width:1180px}
.board thead th{position:sticky;top:0;background:var(--bg-surface-raised);color:var(--text-muted);
  font:600 var(--fs-xs)/1 var(--font-ui);letter-spacing:var(--track-caps);text-transform:uppercase;
  text-align:right;padding:12px var(--cell-x);border-bottom:1px solid var(--border-strong);white-space:nowrap;z-index:2;height:36px}
.board thead th.tl{text-align:left}
.board thead th .star{color:var(--accent);font-size:var(--fs-sm)}
.board thead th.sortable{cursor:pointer;user-select:none}
/* sort direction is drawn AND announced (aria-sort); colour never carries it alone */
.board thead th .sc{margin-left:4px;font-size:8px;color:var(--accent)}
.board thead th.sortable:focus-visible{outline:2px solid var(--accent-border);outline-offset:-2px}
.board thead th.sortable:hover{color:var(--text-secondary)}
.board thead th.sorted{color:var(--accent)}
.board tbody td{height:var(--row-h);padding:0 var(--cell-x);background:var(--bg-row);
  border-bottom:1px solid var(--border-subtle);text-align:right;white-space:nowrap;
  font-variant-numeric:tabular-nums}
.board tbody td.tl{text-align:left}
.board tbody tr:hover td{background:var(--bg-row-hover)}
.adp{color:var(--text-secondary);font-weight:600}
.tier{color:var(--text-muted)}
/* The table name column, same rule as the cards - the grid around it stays
   mono, this one column is prose. */
.player{font-family:var(--font-display);font-weight:500;font-size:14px;
  letter-spacing:-.003em}
.player .team{font-size:var(--fs-xs);color:var(--text-muted);font-weight:400;margin-left:6px}
.bye{color:var(--text-muted)}
.sos{color:var(--text-secondary)}
.auction{color:var(--text-secondary)}

/* Separator rows (round + tier).
   SPECIFICITY MATTERS HERE, and it silently did not for two builds: the
   original rule was written as `.round-sep td` (0,1,1), which LOSES to
   `.board tbody td` (0,1,2) above. So `text-align:left` never applied, the
   label rendered right-aligned inside a 1180px cell at x~1049, and it sat
   off-screen at any viewport narrower than the table - i.e. exactly the
   narrow screens the comment said it was protecting. Invisible at 1440px,
   which is why it survived. Selectors below are scoped through .board tbody
   so they actually win.
   The label is additionally STICKY, which is what "stays visible when the
   table is horizontally scrolled" always meant: left-aligning only helps at
   scrollLeft 0. */
.board tbody tr.round-sep td,
.board tbody tr.tier-sep td{text-align:left;background:var(--bg-app);
  white-space:nowrap;border-bottom:1px solid var(--border-subtle)}
.board tbody tr.round-sep td{height:24px;padding:0 var(--cell-x)}
.board tbody tr.tier-sep td{height:auto;padding:7px var(--cell-x)}
.board tbody tr.round-sep:hover td,
.board tbody tr.tier-sep:hover td{background:var(--bg-app)}
.sep-in{position:sticky;left:0;display:inline-block;
  font:700 var(--fs-xs)/1.5 var(--font-num);letter-spacing:var(--track-caps);
  color:var(--text-muted)}
.round-sep .rn{color:var(--accent)}

/* site cells */
.site{font-family:var(--font-num);font-variant-numeric:tabular-nums;font-weight:700;
  white-space:nowrap}
/* Same grammar as the rank column: annotation in a RESERVED left slot, value
   flush right. Reserved is what makes it work - a slot that appears only when
   there is a delta reproduces the drift it was meant to remove. */
.site .dl-slot{display:inline-block;min-width:26px;text-align:right;
  vertical-align:baseline}
.site .dl{font-size:var(--fs-xs);font-weight:600}
.site .site-v{display:inline-block;min-width:2.2em;text-align:right;
  margin-left:6px;vertical-align:baseline}

/* signal badges */
.sigs{display:flex;gap:4px;justify-content:flex-end}
.sig{font:700 9px/1 var(--font-num);letter-spacing:.07em;padding:3px 6px;border-radius:var(--r-sm);
  border:1px solid var(--border-strong);color:var(--text-secondary);background:rgba(0,0,0,.25)}
.sig-val{border-color:var(--accent-border);color:var(--positive-ink);background:var(--accent-soft)}
.sig-fade{border-color:color-mix(in srgb,var(--negative) 55%,transparent);color:var(--negative-ink);background:var(--negative-soft)}
/* THE LIVE FEED'S TWO VARIANTS, and they live here with their siblings rather
   than down in the event-row block, because a chip variant belongs to the chip
   component and a second home for `.sig` is how the four of them drift apart.
   Both are built entirely from tokens that already ship.

   AMBER FOR `CORRECTED` AND `NO PLAY` IS THE CORRECT RESERVATION, not a
   convenient spare colour. The comment above `.alert-fade` already rules that
   A DEGRADATION IS NOT AN ERROR and spends --risk on exactly that, and the
   palette header reserves amber for risk. A correction is not the app failing;
   it is the feed telling the truth later, which is the same class of event.
   Red here would say Edge got it wrong, and Edge did not.

   `.sig-hold` IS THE ONLY CHIP WITH NO FILL. FORMING is the absence of a
   finished answer, so its chip is the absence of a fill - the one state on
   this row that is defined by what has not arrived yet reads that way at a
   glance, before the word is read. */
.sig-warn{border-color:var(--risk-border);color:var(--risk);background:var(--risk-soft)}
.sig-hold{border-color:var(--border-strong);color:var(--text-muted);background:none}

/* flagged rows - value (orange) / fade (raspberry); text stays neutral */
/* strong-signal rows: a 3px bar, no row wash. 39 washed rows was a wall of
   tint; the bar plus the one colored delta is the same information at a
   fraction of the ink. */
tr.glow-val td:first-child{box-shadow:inset 3px 0 0 var(--positive)}
tr.glow-fade td:first-child{box-shadow:inset 3px 0 0 var(--negative)}

/* ============================================================ CARD BOARD
   The same board, drawn as cards. Layout ported from the Claude-design
   template; PALETTE DELIBERATELY NOT PORTED. That template accents on
   #ff8b2e, which is the orange this app was moved OFF on 2026-08-05 (see
   DECISIONS, "the quiet terminal"), and whose replacement pair has measured
   contrast floors asserted in the suite. Re-introducing it would have
   reverted a decision and broken those assertions to buy a hue.

   So the card grid reads in the shipped scheme: graphite surfaces, ONE cyan
   accent, burnt orange reserved for the fade pole. The template's own player
   profile is already cyan, so this is also the more internally consistent
   half of the two designs.

   The team-tinted photo glow is likewise dropped in favour of tinting by the
   VALUE / FADE pole. It costs no new data (no 32-team color table to keep
   correct), it stays inside the color budget, and it makes the grid scannable
   by the one thing the app is for rather than by which team a player is on. */
.board-cards{padding:var(--sp-1) 0 var(--sp-4)}
/* THE TEMPLATE'S CARD IS ~430px AND THE OLD ONE WAS 270. A rank rail, a
   headshot and an Archivo Black name need width the five-across grid never
   had - at 1400 it left 120px of body for a name wanting 256. auto-fill with
   a 300px floor gives three or four across depending on the viewport, which
   costs density and buys the design its proportions. */
.pcd-grid{display:grid;gap:var(--sp-3);margin-bottom:var(--sp-3);
  grid-template-columns:repeat(auto-fill,minmax(268px,1fr))}
/* The featured row is the top THREE, so it is three columns, not an auto-fill
   that wrapped the third card onto a row by itself. */
/* THE TEMPLATE'S FEATURED ROW: one card spanning two columns with two normal
   cards beside it, not three equal ones. Four tracks, and the lead card takes
   two of them. */
/* THREE FEATURED CARDS AT ONE SIZE, and the rest smaller beneath them. The
   single span-2 hero read as one enormous card rather than as a top row. */
.pcd-grid.feat{grid-template-columns:repeat(3,minmax(0,1fr))}
/* Only the WIDE card gets the large treatment. Applying it to all three
   featured cards left the two quarter-width ones with a 62px rail, a 124px
   headshot and 84px of room for a name wanting 116. */
.board-cards .empty{padding:var(--sp-5);text-align:center;color:var(--text-muted);
  border:1px solid var(--border-subtle);border-radius:var(--r-lg);background:var(--bg-surface)}

.pcd{position:relative;overflow:hidden;display:flex;flex-direction:column;
  border:1px solid var(--border-subtle);border-radius:var(--r-lg);
  background:var(--bg-surface);cursor:pointer;
  transition:border-color .12s ease,transform .12s ease,box-shadow .12s ease}
.pcd:hover{border-color:var(--border-strong);transform:translateY(-1px);
  box-shadow:0 8px 24px rgba(0,0,0,.45)}
.pcd:focus-visible{outline:2px solid var(--accent-border);outline-offset:1px}

/* The photo well - featured cards only (see cardFace). The two gradients are
   what keep the name legible over it: one fades the photo into the surface
   horizontally, one lifts the foot strip off it.
   `.nophoto` is set by the img's own onerror, so a player with no headshot
   collapses to a clean text card rather than reserving space for nothing. */
/* The absolutely-positioned photo that used to live here is GONE, superseded
   by the flex column at the foot of this file. It was still winning: a new
   rule that sets width and clip-path does not clear `position:absolute`, so
   the headshot left the flow, the body stretched the full card width, and the
   foot's SIGNAL cell rendered underneath the picture. Reported as "the signal
   bleeds over into the picture" - the third time today an old rule outlived
   the design it belonged to. */

/* A featured body reserves the photo's column as right padding. Without it the
   name runs under the headshot and the gradient has to be relied on to keep it
   readable, which it cannot do at every name length. */
.pcd-body{position:relative;padding:var(--sp-3) var(--sp-3) var(--sp-2);flex:1;min-width:0}

.pcd-top{display:flex;align-items:center;gap:var(--sp-2);min-height:18px}
.pcd-rank{font:700 var(--fs-lg)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-secondary)}
.pcd-rd,.pcd-tier{font:700 9px/1 var(--font-num);letter-spacing:.06em;padding:3px 6px;
  border-radius:var(--r-sm);border:1px solid var(--border-strong);color:var(--text-muted)}
.pcd-rd{border-color:transparent;background:var(--bg-surface-raised);color:var(--text-secondary)}

/* The serif card-name rules that lived here are gone, superseded by the EDGE
   Cards design at the foot of this file. They were not merely redundant:
   `.pcd.feat .pcd-name` carries two classes and beat the new single-class
   rule on SPECIFICITY no matter how far down the file the replacement sat, so
   two featured names rendered at 24px in a 100px box. Deleting the old design
   is part of shipping the new one. */
/* A FEATURED NAME GETS ITS OWN LINE once the cards go two-up.
   The photo is absolutely positioned, so the name ROW is only 178px of a
   336px card, and inside that the logo, team mark and injury chip take 73px
   at fixed width - leaving 98px for a name that wants 155. Measured on
   Jahmyr Gibbs, Bijan Robinson and Ja'Marr Chase, which is to say on the
   three cards this layout exists to feature.
   Wrapping the marks to a second line hands the name the full row and it is
   the shape the rank cards on the two command centres already use: name,
   then its metadata beneath. Only below 1280px, because at three-up there is
   room for the original single line and no reason to spend a row. */
/* Sizes tuned by measurement, not by eye: at 1400px the truncation count runs
   85 / 37 / 13 / 10 / 4 for 20 / 18 / 17 / 16 / 15px. None of those is zero,
   which is what said the problem was the ROW and not the type. With the marks
   moved out, 19px clears every name on the board. */
/* The name is the one thing that must never be clipped mid-word, so it takes
   the ellipsis and the team mark rides after it at fixed width. Targeting a
   class, not :first-child - the first ELEMENT child is the team logo, so the
   selector that looked right shortened the logo and left the name overflowing. */
.pcd-name .nm{overflow:hidden;text-overflow:ellipsis;white-space:nowrap;min-width:0}
.pcd-logo{width:16px;height:16px;object-fit:contain;flex:none;opacity:.9}
.pcd-team{font:600 var(--fs-sm)/1 var(--font-num);letter-spacing:.05em;
  color:var(--text-muted);flex:none}

.pcd-meta{display:flex;align-items:center;gap:var(--sp-3);margin-top:var(--sp-3);flex-wrap:wrap}
.pcd-kv{display:flex;align-items:baseline;gap:5px}
.pcd-kv .k{font-size:var(--fs-xs);letter-spacing:var(--track-caps);text-transform:uppercase;
  color:var(--text-muted);font-weight:600}
.pcd-kv b{font:700 var(--fs-lg)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-primary)}
/* Positional rank, as one token ("RB2") rather than a label and a number -
   "RB 2" read as a second, unexplained stat rather than as his RB rank. */
.pcd-pr{font:700 var(--fs-xs)/1 var(--font-num);letter-spacing:.05em;color:var(--text-secondary);
  border:1px solid var(--border-subtle);border-radius:var(--r-sm);padding:4px 6px}
.pcd.feat .pcd-kv b{font-size:20px}
.pcd.feat .pcd-meta{gap:var(--sp-5);margin-top:var(--sp-4)}

/* The stat foot. Same five facts on a featured card, four on a small one -
   ADP drops from the foot on small cards because it is already in the meta
   line above; repeating it would spend a quarter of the strip saying it twice. */
/* A DETERMINISTIC grid, and getting here took three tries worth recording.
   The foot carries four facts in ~250px and every one of them is
   data-dependent: "PROJ +12" is five times the width of the "-" that sits
   there when nobody disagrees, and "+161.3" is twice "+92".

   1. `auto` tracks sized to content and overflowed - READ and SIGNAL printed
      on top of each other, but ONLY on the cards that had a signal to show.
      The layout looked perfect on every empty card and broke on the
      interesting ones.
   2. Wrapping flex fixed the overlap and introduced a worse problem: whether
      a card wrapped depended on how many digits its PAR happened to have, so
      feet sat at different heights across one row of the grid and the whole
      board read as misaligned.

   So SIGNAL gets its own row on the narrow cards, always, whatever the data.
   A consistently two-line foot costs ~18px of card height and cannot be
   knocked out of alignment by a number getting longer. Featured cards are
   twice as wide and keep all five on one line with room to spare. */
.pcd-foot{position:relative;display:grid;align-items:start;
  grid-template-columns:auto auto 1fr;gap:8px var(--sp-3);
  padding:9px var(--sp-3);border-top:1px solid var(--border-subtle);
  background:rgba(0,0,0,.28)}
.pcd.feat .pcd-foot{grid-template-columns:auto auto auto auto 1fr;
  padding:11px var(--sp-4);gap:var(--sp-5)}
.pcd-foot .c{min-width:0}
.pcd-foot .k{display:block;font-size:var(--fs-xs);letter-spacing:var(--track-caps);
  text-transform:uppercase;color:var(--text-muted);font-weight:600}
/* min-height pins the row against its own tallest tenant. The EDGE cell holds
   a bordered chip when a source disagrees and a bare "-" when none does, and
   those differ by 5px - which showed up as feet sitting at two different
   heights along one row of the grid, i.e. the board looked misaligned exactly
   where it had something to report. */
.pcd-foot .v{display:flex;align-items:center;min-height:19px;margin-top:3px;
  font:700 var(--fs-lg)/1.1 var(--font-num);
  font-variant-numeric:tabular-nums;white-space:nowrap}
.pcd-foot .v.sm{font-size:var(--fs-sm)}
/* On the two-row foot the bar is the full width of the card, so the count
   sits at the far right where it reads as the bar's total. */
.pcd-foot .v.sig{display:flex;align-items:center;gap:6px}
.pcd:not(.feat) .pcd-foot .v.sig{justify-content:flex-start}
.pcd:not(.feat) .pcd-foot .v.sig small{margin-left:auto}
.pcd-foot .v.sig small{font:600 var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}

/* The lead-signal chip. Same rule and same two classes as the table's .sig,
   so a player's chip reads identically in both layouts. */
.pcd-sig{display:inline-block;font:700 9px/1 var(--font-num);letter-spacing:.07em;
  padding:4px 6px;border-radius:var(--r-sm);border:1px solid var(--border-strong);
  color:var(--text-secondary);background:rgba(0,0,0,.25)}
.pcd-sig.val{border-color:var(--accent-border);color:var(--positive-ink);background:var(--accent-soft)}
.pcd-sig.fade{border-color:color-mix(in srgb,var(--negative) 55%,transparent);
  color:var(--negative-ink);background:var(--negative-soft)}

/* The READ tag. Words, always - the color is the redundant channel, exactly
   as the signed deltas are on the table. */
.pcd-foot .v.tag{font:800 var(--fs-sm)/1.1 var(--font-num);letter-spacing:.07em}
.pcd-foot .v.tag.val{color:var(--positive-ink)}
.pcd-foot .v.tag.fade{color:var(--negative-ink)}
.pcd-foot .v.tag.weak{opacity:.72}
.pcd-foot .v.tag.none{color:var(--text-muted)}

/* Segment bar: how many sources deviate in the lead direction, out of how many
   priced him. A count, so the unfilled segments carry meaning too and stay
   visible rather than fading to the background. */
.pcd-bars{display:inline-flex;gap:2px}
.pcd-bars i{width:4px;height:11px;border-radius:2px;background:var(--border-strong)}
.pcd-bars i.on{background:var(--neutral)}
.pcd.glow-val .pcd-bars i.on{background:var(--positive)}
.pcd.glow-fade .pcd-bars i.on{background:var(--negative)}

/* Strong-signal cards: a 3px edge bar and a breath of tint in the photo well,
   the card-sized version of the table's inset row bar. */
.pcd.glow-val{box-shadow:inset 3px 0 0 var(--positive)}
.pcd.glow-fade{box-shadow:inset 3px 0 0 var(--negative)}
.pcd.glow-val:hover{box-shadow:inset 3px 0 0 var(--positive),0 8px 24px rgba(0,0,0,.45)}
.pcd.glow-fade:hover{box-shadow:inset 3px 0 0 var(--negative),0 8px 24px rgba(0,0,0,.45)}
.pcd.glow-val .pcd-photo::before,.pcd.glow-fade .pcd-photo::before{content:"";position:absolute;
  inset:-20%;z-index:1}
.pcd.glow-val .pcd-photo::before{background:radial-gradient(52% 52% at 62% 40%,var(--accent-soft),transparent 70%)}
.pcd.glow-fade .pcd-photo::before{background:radial-gradient(52% 52% at 62% 40%,var(--negative-soft),transparent 70%)}

/* Already gone - same statement the struck table row makes, in card form. */
.pcd.gone{opacity:.42}
.pcd.gone .pcd-name{text-decoration:line-through;text-decoration-color:var(--text-muted)}
/* The j/k cursor, so the keyboard works identically in both layouts. */
.pcd.cursor{border-color:var(--border-strong);background:var(--bg-row-hover)}
.pcd.cursor::after{content:"";position:absolute;inset:0;border-radius:var(--r-lg);
  box-shadow:inset 0 0 0 1px var(--accent-border);pointer-events:none}

/* Tier bands in card view - the same header the table's <tr> carries. */
.pcd-band{margin-bottom:var(--sp-4)}
.pcd-bandhd{font:700 var(--fs-xs)/1.5 var(--font-num);letter-spacing:var(--track-caps);
  color:var(--text-muted);padding:var(--sp-2) 0;margin-bottom:var(--sp-1);
  border-bottom:1px solid var(--border-subtle)}
.pcd-bandhd .rn{color:var(--accent)}

/* The layout switch. Deliberately the LOUDEST control in the filter row: it
   changes what the whole board looks like, which is a bigger move than any
   chip beside it, and a user who lands in the card grid needs the way back to
   the table to be obvious rather than discovered. */
/* THE SEGMENTED KEY. One object with one drop shadow, divided - not a row of
   separate keys, which is what makes it read as a single control with a
   position rather than as two buttons that happen to be adjacent. */
.segbar{display:inline-flex;border-radius:var(--r-sm);
  box-shadow:0 2px 0 rgba(0,0,0,.75),0 3px 6px rgba(0,0,0,.5)}
.segbar .seg{padding:0 16px;border-radius:0;white-space:nowrap;
  box-shadow:inset 0 1px 0 rgba(255,255,255,.16)}
/* Restated after the shared ON rule. Not because the cascade needed it - it
   did not, and a long hunt for the rule "beating" it found nothing because
   nothing was: the headless preview was serving a STALE COMPUTED STYLE while
   an identical probe element built in the same document resolved correctly.
   Third time this session that the preview lied about a value it had already
   recomputed. Kept anyway, because a segment's ink is worth stating where the
   segment is defined rather than inferring it from a shared list. */
/* THE INK IS THE KEY SYSTEM'S, not --on-accent. Those two were the same
   value on both grounds until the lit key became an OUTLINE on paper -
   at which point --key-on-ink went blue and --on-accent stayed white,
   and this rule painted a white word on the white face it had just been
   given. A control that restates a value "for clarity" is a control that
   stops tracking it; the whole point of the token is that one edit moves
   every key. */
.segbar .seg.on{box-shadow:inset 0 1px 0 rgba(255,255,255,.55);color:var(--key-on-ink);font-weight:600}
.segbar .seg:first-child{border-radius:var(--r-sm) 0 0 var(--r-sm)}
.segbar .seg:last-child{border-radius:0 var(--r-sm) var(--r-sm) 0;border-left:none}
.segbar .seg:active{transform:translateY(2px)}

/* Card-view sort control. Lives beside the filter chips and only exists in
   card mode, because in table mode the column headers ARE this control. */
.fsort{display:flex;align-items:center;gap:var(--sp-2)}
.fsort select{min-height:32px;padding:6px 8px}
.fsort select:disabled{opacity:.45;cursor:not-allowed}
.fdir{background:var(--bg-surface-raised);color:var(--accent);
  border:1px solid var(--border-subtle);border-radius:var(--r-md);
  min-height:32px;min-width:30px;padding:0 7px;font-size:9px;line-height:1}
.fdir:hover{border-color:var(--border-strong)}
.fdir:disabled{opacity:.45;cursor:not-allowed;color:var(--text-muted)}

@media (max-width:1280px){
  .pcd-grid{grid-template-columns:repeat(auto-fill,minmax(250px,1fr))}
  .pcd-grid.feat{grid-template-columns:repeat(2,minmax(0,1fr))}
  .pcd-grid.feat .pcd.lead{grid-column:span 2}
}
/* THE FEATURED FOOT LAYS FIVE CELLS ACROSS, and that needs about 380px of
   card. Below ~820px the featured grid is already two-up, so each card falls
   under it - and the last column is `1fr`, which happily collapses to ZERO
   while the signal meter inside keeps its own width and draws straight through
   the READ tag beside it. Measured: at a 716px viewport the fifth column
   computes to 0px and READ/SIGNAL share the same x.

   `1fr` shrinking below its contents is the trap; a fifth `auto` would have
   overflowed the card instead, which is the same defect wearing a scrollbar.
   So the row drops to four cells across with the signal on its own row - the
   deterministic two-row foot every non-featured card already uses. */
@media (max-width:820px){
  .pcd.feat .pcd-foot{grid-template-columns:repeat(4,auto);
    padding:9px var(--sp-3);gap:8px var(--sp-3)}
  .pcd.feat .pcd-foot .c.wide{grid-column:1/-1;justify-self:stretch}
}
@media (max-width:980px){
  .pcd-grid,.pcd-grid.feat{grid-template-columns:repeat(auto-fill,minmax(280px,1fr))}
  .pcd-grid.feat .pcd.lead{grid-column:auto}
}
@media (max-width:620px){
  .pcd-grid,.pcd-grid.feat{grid-template-columns:1fr}
}

/* position badges - GHOST chips: a 14% tint and a colored letter. The letter
   carries identity (constitution 18); the hue is a secondary cue and no longer
   competes with the signal layer for the eye. Solid fills were the single
   largest source of ink on the board. */
.pos{display:inline-block;min-width:32px;text-align:center;font:800 var(--fs-xs)/1 var(--font-num);
  letter-spacing:.03em;color:var(--text-secondary);
  background:color-mix(in srgb,var(--neutral) 14%,transparent);
  padding:4px 5px;border-radius:var(--r-sm)}
.pos-QB{color:var(--pos-qb);background:color-mix(in srgb,var(--pos-qb) 14%,transparent)}
.pos-RB{color:var(--pos-rb);background:color-mix(in srgb,var(--pos-rb) 14%,transparent)}
.pos-WR{color:var(--pos-wr);background:color-mix(in srgb,var(--pos-wr) 14%,transparent)}
.pos-TE{color:var(--pos-te);background:color-mix(in srgb,var(--pos-te) 14%,transparent)}
.pos-K{color:var(--pos-k);background:color-mix(in srgb,var(--pos-k) 14%,transparent)}
.pos-DST{color:var(--pos-dst);background:color-mix(in srgb,var(--pos-dst) 14%,transparent)}

/* ---------------------------------------------- injury / availability risk
   Severity rides on FORM, never on a second hue (see --risk). The letter code
   is always present, so the chip is never color-alone. A STALE designation
   gets no status color at all: we do not vouch for it, and a muted chip says
   that more honestly than a colored one with a caveat in a tooltip. */
.inj{display:inline-block;font:800 9px/1 var(--font-num);letter-spacing:.06em;
  padding:3px 5px;border-radius:var(--r-sm);margin-left:6px;vertical-align:1px;
  border:1px solid var(--risk-border);color:var(--risk);background:transparent}
.inj-out{background:var(--risk);color:var(--pos-ink);border-color:var(--risk)}
.inj-stale{border-color:var(--border-strong);color:var(--text-muted);background:transparent}
.ba .inj,.slot .inj{margin-left:5px}

/* day-over-day ADP movement; signed pick delta, same polarity as site cells */
.mv{font:700 9px/1 var(--font-num);margin-left:5px;vertical-align:1px}

/* THE NUMBER HOLDS THE LINE.
   The movement badge used to trail the rank, so a column of 1..12 sat at a
   different x on every row that happened to have moved - the numbers stopped
   reading as a column. The badge now lives in a RESERVED slot to the LEFT of
   the rank: reserved so an unmoved row indents identically to a moved one,
   and left so the ranks stay flush against the column's right edge. The slot
   costs no width overall - the column already had to fit a badge. */
td.adp{white-space:nowrap}
td.adp .mv-slot{display:inline-block;min-width:40px;text-align:right;
  vertical-align:baseline}
td.adp .mv-slot .mv{margin-left:0}
td.adp .adp-n{display:inline-block;min-width:2.4em;text-align:right;
  margin-left:6px;vertical-align:baseline}

/* queue / avoid markers on the board */
.mk{display:inline-block;font:800 9px/1 var(--font-num);margin-left:6px;vertical-align:1px}
.mk-q{color:var(--accent)}
.mk-x{color:var(--negative-ink)}

/* keyboard cursor row (j/k) */
tr.cursor td{background:var(--bg-row-hover);box-shadow:inset 0 0 0 1px var(--border-strong)}
tr.cursor td:first-child{box-shadow:inset 3px 0 0 var(--text-secondary),inset 0 0 0 1px var(--border-strong)}
/* The name is a real <button> so it is keyboard-reachable and announced as
   interactive, but it must look like text everywhere it appears - board rows,
   best-available list, type-ahead. Scoping this reset to .player left default
   button chrome around every name in the draft room. */
.pname{cursor:pointer;background:none;border:none;color:inherit;font:inherit;
  text-align:left;letter-spacing:inherit;
  /* VISION P5 wants every interactive target >= 36px. The name is inline text
     inside a data row, so it grows its HIT BOX with padding and pulls the box
     back with a matching negative margin: the target is row-height without the
     text moving a pixel. */
  padding:10px 0;margin:-10px 0;display:inline-block;line-height:16px}
.pname:hover{color:var(--accent);text-decoration:underline}

/* ============================================================ TIER VIEW */
.tier-sep .rn{color:var(--accent);font-size:var(--fs-sm)}
.tcs{margin-left:var(--sp-3)}
.tc{display:inline-flex;align-items:center;gap:4px;margin-right:var(--sp-2);color:var(--text-secondary)}
.tc .pos{min-width:26px;padding:2px 4px;font-size:9px}
.tec{margin-left:var(--sp-3);text-transform:none;letter-spacing:0;font-weight:400}
.cliff{margin-left:var(--sp-3);color:var(--risk);border:1px solid var(--risk-border);
  border-radius:var(--r-sm);padding:2px 6px}
tr.gone td{opacity:.42}
tr.gone .player .pname{text-decoration:line-through}
th.sortable.disabled{opacity:.4;cursor:default}

/* ======================================================== PICK PATH (MC) */
.pickpath{background:var(--bg-surface);border:1px solid var(--border-strong);
  border-radius:var(--r-md);padding:var(--sp-3) var(--sp-4);margin-bottom:var(--sp-3)}
.pp-hd{display:flex;align-items:baseline;justify-content:space-between;margin-bottom:var(--sp-3)}
.pp-run{color:var(--text-secondary);font-size:var(--fs-sm);padding:var(--sp-2) 0}
.pp-verdict{background:var(--accent-soft);border-left:3px solid var(--accent);
  border-radius:var(--r-sm);padding:9px var(--sp-3);margin-bottom:var(--sp-3);
  font-size:var(--fs-sm);line-height:1.55;color:var(--text-secondary)}
.pp-verdict b{color:var(--text-primary);font-weight:600}
.pp-grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(210px,1fr));gap:var(--sp-3)}
.pp-card{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3)}
.pp-take{font-size:var(--fs-sm);display:flex;align-items:center;gap:6px;flex-wrap:wrap}
.pp-take b{font-weight:600}
.pp-take .dim{font-size:var(--fs-xs)}
.pp-sub{display:block;margin:var(--sp-2) 0 6px}
.pp-row{display:grid;grid-template-columns:40px 34px 1fr 30px;gap:var(--sp-2);
  align-items:center;margin-bottom:5px}
.pp-pct{font:700 var(--fs-sm)/1 var(--font-num);font-variant-numeric:tabular-nums;text-align:right}

/* ══════════════════ PICK PATH - the ranked ledger (his 5a) ═════════════════
   Values taken off the reference mock's own inline styles rather than
   approximated: the first pass rebuilt the STRUCTURE and left it wearing the
   app's table styling, which is why it read as a spreadsheet and the design
   reads as a ledger. Rows are CARDS, 8px apart, and the lead one is lit.

   One column set, shared by the header and every row, so a heading can never
   drift off its column. */
/* THE NAME COLUMN WAS COLLAPSING TO NOTHING. The fixed tracks plus gaps sum
   to 944px and min-width was 940, so `minmax(0,1fr)` resolved to ~0 and every
   candidate rendered as "J." - the ellipsis doing exactly what it was told.
   The floor now leaves the name a real 240px, so the row scrolls before the
   name truncates. A minimum that equals the sum of everything else is not a
   minimum, it is a guarantee of zero. */
.pp-ledger{--pp-cols:44px minmax(240px,1fr) 104px 104px 112px 300px 168px;
  overflow-x:auto}
.pp-lhead,.pp-lrow{display:grid;grid-template-columns:var(--pp-cols);
  gap:14px;align-items:center;min-width:1120px}
.pp-lhead{padding:0 18px 8px;font:400 10px/1 var(--font-num);
  letter-spacing:.16em;text-transform:uppercase;color:var(--text-muted)}
.pp-lrows{display:flex;flex-direction:column;gap:8px;min-width:1120px}
.pp-lrow{padding:16px 18px;border-radius:6px;
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle)}
/* THE LEAD ROW IS THE ONLY LIT ONE - same rule as the board's on-the-clock
   cell. A ranking where every row is emphasised contains no ranking. */
.pp-lrow.lead{background:color-mix(in srgb,var(--accent) 7%,transparent);
  border-color:color-mix(in srgb,var(--accent) 45%,transparent)}
/* Adds nothing to the lineup: still an answer to "what else is there", just
   not a recommendation. Receding beats hiding. */
.pp-lrow.dead{opacity:.62}
/* The rank is a disc, filled on the leader and outlined on the rest, so the
   ordering is legible before any number is read. */
.pp-rank{display:flex;align-items:center;justify-content:center;
  width:34px;height:34px;border-radius:50%;font:400 15px/1 var(--font-num);
  border:1px solid var(--border-strong);color:var(--text-secondary)}
.pp-lrow.lead .pp-rank{background:var(--accent);border-color:var(--accent);
  color:var(--on-accent)}
.pp-cand{display:flex;align-items:center;gap:10px;min-width:0}
/* 19px, from the mock. The name is the thing being chosen between; at the
   table's 11px it was the same size as its own metadata. */
.pp-cand b{font:600 19px/1.35 var(--font-ui);white-space:nowrap;
  overflow:hidden;text-overflow:ellipsis;font-weight:600}
.pp-num{font:400 13px/1 var(--font-num);color:var(--text-secondary);
  font-variant-numeric:tabular-nums}
/* A reach is a PRICE, and the mock prices it in warning amber rather than
   leaving it the same grey as the numbers that cost nothing. */
.pp-reachv{color:var(--risk)}
.pp-thinwrap{display:flex;align-items:center;min-width:0}
/* Values off the mock. The percentages sit in a fixed 44px so they form a
   COLUMN across the four rows - the point of a ledger - and the position is
   coloured letters at a fixed 24px rather than a badge, whose padding differs
   per position and would break that alignment. */
.pp-thin{display:flex;align-items:center;gap:10px;flex:1;min-width:0}
.pp-tpos{flex:none;width:24px;font:700 12px/1 var(--font-label);
  letter-spacing:.14em}
.pp-pct{flex:none;width:44px;font:400 15px/1 var(--font-num);color:var(--accent);
  font-variant-numeric:tabular-nums}
.pp-bar{flex:1;min-width:24px;height:5px;border-radius:3px;
  background:var(--track);overflow:hidden;display:block}
.pp-bar>i{display:block;height:100%;background:var(--accent);border-radius:3px}
.pp-tier{flex:none;font:400 10px/1 var(--font-num);color:var(--text-muted)}
.pp-rest{font:400 11px/1 var(--font-num);color:var(--text-muted);
  white-space:nowrap;font-variant-numeric:tabular-nums}
.pp-lfoot{margin-top:18px;padding-top:14px;
  border-top:1px solid var(--border-subtle);
  font-size:var(--fs-sm);color:var(--text-muted);line-height:1.55}
/* CERTAINTY STOPS DRAWING BARS (his rule). A meter pinned at 100% is a bar
   that says nothing, once per row - the word carries it. */
.pp-certain{font:700 9px/1 var(--font-label);letter-spacing:.14em;
  text-transform:uppercase;color:var(--text-muted)}

/* THE VERDICT, at the mock's own scale: a 44px figure beside a 22px claim in
   the display serif, with the qualifier small and grey under it. It had been a
   21px inline number inside a body-copy sentence, which is why the panel read
   as a paragraph where the design reads as a headline. */
.pp-verdict5{display:flex;align-items:baseline;gap:16px;margin:18px 0 20px}
.pp-fig{flex:none;font:400 44px/1 var(--font-display);color:var(--accent)}
.pp-vbody{flex:1;min-width:0;display:block}
.pp-vclaim{display:block;font:400 22px/1.25 var(--font-display);
  max-width:1000px;text-wrap:pretty}
.pp-vclaim b{font-weight:600}
.pp-vsub{display:block;margin-top:6px;font:400 12px/1.5 var(--font-ui);
  color:var(--text-secondary)}
@media (max-width:700px){
  .pp-verdict5{gap:12px}
  .pp-fig{font-size:32px}
  .pp-vclaim{font-size:18px}
}
.pp-med{font-size:var(--fs-xs);color:var(--text-muted);text-align:right}
.pp-note{margin-top:var(--sp-3);padding-top:var(--sp-2);border-top:1px solid var(--border-subtle);
  font-size:var(--fs-xs);color:var(--text-muted);line-height:1.6}
.pp-note b{color:var(--text-secondary);font-weight:600}

/* ==================================================== OPPONENT NEEDS */
.needs{padding:var(--sp-3) var(--sp-4)}
.nd-lead{font-size:var(--fs-sm);color:var(--text-secondary);margin-bottom:var(--sp-3);line-height:1.5}
.nd-lead b{color:var(--text-primary);font-weight:600}
.nd-row{display:grid;grid-template-columns:40px 38px 1fr 22px;gap:var(--sp-2);align-items:center;margin-bottom:5px}
.nd-u{font:700 var(--fs-xs)/1 var(--font-num);text-align:right;color:var(--risk)}
.nd-n{font:700 var(--fs-xs)/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:right;color:var(--text-secondary)}
.nd-note{margin-top:var(--sp-3);font-size:var(--fs-xs);color:var(--text-muted);line-height:1.5}

/* ============================================================ QUEUE PANEL */
.queue{list-style:none;padding:var(--sp-1) 0}
.jq-block{padding:var(--sp-3) var(--sp-4) var(--sp-3);border-bottom:1px solid var(--border-subtle)}
.jq-block>.caps{display:block;margin-bottom:var(--sp-2)}
.jq{display:grid;grid-template-columns:42px 40px 1fr;gap:var(--sp-2);align-items:center;margin-bottom:5px}
.jqp{font-size:var(--fs-xs);color:var(--text-muted);font-variant-numeric:tabular-nums}
.jqv{font:700 var(--fs-sm)/1 var(--font-num);font-variant-numeric:tabular-nums;text-align:right}
.jq-warn{margin-top:var(--sp-2);padding:7px var(--sp-2);border-left:3px solid var(--risk);
  background:var(--risk-soft);border-radius:var(--r-sm);font-size:var(--fs-xs);
  color:var(--text-secondary);line-height:1.5}
.jq-warn b{color:var(--text-primary);font-weight:600}
.jq-note{margin-top:var(--sp-2);font-size:var(--fs-xs);color:var(--text-muted);line-height:1.5}
.qrow{display:grid;grid-template-columns:30px 40px minmax(0,1fr) 38px 22px;gap:var(--sp-2);
  align-items:center;padding:6px var(--sp-4);border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.qrow:last-child{border-bottom:none}
.qrow:hover{background:var(--bg-row-hover)}
.qrow .rk{font-size:var(--fs-xs);color:var(--text-muted);text-align:right}
.qrow .nm{font-weight:600;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.qrow .nm .team{font-size:var(--fs-xs);color:var(--text-muted);font-weight:400;margin-left:5px}
.qrow .qs{font-size:var(--fs-xs);text-align:right;color:var(--text-secondary)}
.qrow.gone{opacity:.5}
.qrow.gone .nm{text-decoration:line-through}
.qx{background:none;border:none;color:var(--text-muted);font-size:14px;line-height:1;
  padding:2px 4px;border-radius:var(--r-sm);min-width:36px;min-height:36px}
.qx:hover{color:var(--negative-ink);background:var(--negative-soft)}

/* ============================================================ KEEPERS */
.keepers{background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3);margin-top:var(--sp-3)}
.kp-form{display:flex;gap:var(--sp-2);align-items:center;flex-wrap:wrap}
#kpPlayer{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);color:var(--text-primary);font:var(--fs-sm)/1 var(--font-ui);
  padding:8px 10px;min-width:190px}
#kpPlayer:focus{outline:none;border-color:var(--accent)}
/* League setup rows: one per slot type, +/- then count then name - the same
   reading order as the Sleeper panel it is modeled on. Buttons are real
   buttons at 34px so a thumb can hit them; the count is a fixed-width cell
   so a two-digit bench does not shove the labels out of column. */
.lg-row{display:flex;align-items:center;gap:var(--sp-2);padding:5px 0;
  border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.lg-row:last-of-type{border-bottom:none}
.lg-btn{width:34px;height:34px;border-radius:var(--r-md);cursor:pointer;
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  color:var(--text-primary);font:600 var(--fs-md)/1 var(--font-ui)}
.lg-btn:hover:not(:disabled){border-color:var(--accent)}
.lg-btn:disabled{opacity:.4;cursor:default}
.lg-n{min-width:2ch;text-align:right;font:600 var(--fs-md)/1 var(--font-num);
  color:var(--text-primary)}
.lg-name{color:var(--text-muted)}
.lg-foot{margin-top:var(--sp-2);color:var(--text-muted)}
.kp-list{list-style:none;margin-top:var(--sp-3)}
.kp-list li{display:flex;align-items:center;gap:var(--sp-2);padding:5px 0;
  border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.kp-list li:last-child{border-bottom:none}
.kp-list li .num{color:var(--accent);font-weight:700;width:46px}
.kp-list li .qx{margin-left:auto}

/* ---- board setup: the grid IS the keeper editor before a draft starts ---- */
.draft-grid.prep .dg-cell.open-slot{cursor:pointer}
.draft-grid.prep .dg-cell.open-slot:hover{background:var(--bg-cell-hover);
  border-color:var(--accent-border)}
.dg-plus{position:absolute;left:50%;bottom:6px;transform:translateX(-50%);
  font:600 var(--fs-lg)/1 var(--font-ui);color:var(--border-strong)}
.draft-grid.prep .dg-cell.open-slot:hover .dg-plus{color:var(--accent)}
.draft-grid.prep .dg-cell.keeper.editable{cursor:pointer}
.draft-grid.prep .dg-cell.keeper.editable:hover{border-color:var(--negative)}
/* the in-cell editor: one field, because the cell already carries team+round */
.dg-cell.editing{overflow:visible;z-index:5;border-color:var(--accent)}
.dg-in{width:100%;margin-top:4px;background:var(--bg-app);color:var(--text-primary);
  border:1px solid var(--accent-border);border-radius:var(--r-sm);
  font:600 var(--fs-sm)/1 var(--font-ui);padding:5px 6px}
.dg-in:focus{outline:none;border-color:var(--accent)}
.dg-sug{position:absolute;left:0;right:0;top:100%;margin-top:3px;z-index:6;
  list-style:none;background:var(--bg-surface-raised);border:1px solid var(--border-strong);
  border-radius:var(--r-sm);box-shadow:0 12px 28px rgba(0,0,0,.55);max-height:190px;
  overflow-y:auto;min-width:190px}
.dg-sug li{padding:6px 8px;font-size:var(--fs-sm);cursor:pointer;white-space:nowrap}
.dg-sug li.on{background:var(--accent-soft)}
.dg-sug li:hover{background:var(--bg-row-hover)}
.dg-sug .dim{margin-left:6px}
/* team headers are buttons now: claim your row by clicking your column */
button.dg-team{width:100%;background:var(--bg-surface-raised);cursor:pointer}
button.dg-team:hover{color:var(--text-primary);border-color:var(--border-strong)}
button.dg-team.you:hover{color:var(--accent)}

/* saved boards */
.kp-boards{list-style:none;margin:var(--sp-3) 0}
.kp-boards li{display:flex;align-items:center;gap:var(--sp-2);padding:6px 0;
  border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.kp-boards li:last-child{border-bottom:none}
.kp-boards li .dim{flex:1}
.kp-boards li.empty{color:var(--text-muted);display:block;line-height:1.5}
#kpBoardName{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);color:var(--text-primary);font:var(--fs-sm)/1 var(--font-ui);
  padding:8px 10px;min-width:190px}
#kpBoardName:focus{outline:none;border-color:var(--accent)}

/* ============================================================ FOLLOW MODE */
#flUrl{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);color:var(--text-primary);font:var(--fs-sm)/1 var(--font-ui);
  padding:8px 10px;min-width:320px;flex:1}
#flUrl:focus{outline:none;border-color:var(--accent)}
.fl-state{display:flex;align-items:center;gap:var(--sp-3);flex-wrap:wrap;
  margin-top:var(--sp-3);font-size:var(--fs-sm)}
.fl-id{font:600 var(--fs-xs)/1 var(--font-num);color:var(--text-secondary)}
.fl-ok{color:var(--text-primary)}
/* a failure states itself in the fade pole, which is the one colour on this
   board that is never also a value signal */
.fl-bad{color:var(--negative)}
.fl-live{color:var(--accent);font-weight:700;letter-spacing:.08em}

/* a sparse column with nothing in it yet says so instead of reading broken */
th.col-empty .sos-hint{color:var(--text-muted);font-style:italic}

/* ============================================================ WEEKLY OPS */
.ops-tabs{display:flex;gap:var(--sp-2);padding:var(--sp-3) var(--sp-4) 0;flex-wrap:wrap}
#opsBody{padding:var(--sp-3) var(--sp-4) var(--sp-4)}
.ops-list{list-style:none}
.ops-row .lab{width:38px;flex:none}
.ops-row .nm{flex:1;min-width:110px;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.ops-row .nm .team{margin-left:6px;font-size:var(--fs-xs);color:var(--text-muted)}
.ops-row .num{font-variant-numeric:tabular-nums;font-weight:700;min-width:52px;text-align:right}
.ops-gain.up{color:var(--positive-ink)}
.ops-faab{min-width:190px;font-variant-numeric:tabular-nums}
.ops-faab .dim{margin-left:6px}
.ops-trend{min-width:88px;text-align:right;color:var(--text-muted);font-size:var(--fs-xs)}
.ops-opp{min-width:64px}
.ops-empty{color:var(--text-muted);font-size:var(--fs-sm);line-height:1.6;padding:var(--sp-2) 0}
.ops-verdict.d-up{border-color:var(--accent-border)}
.ops-verdict.d-dn{border-color:color-mix(in srgb,var(--negative) 55%,transparent)}
.ops-verdict .dim{display:block;margin-top:4px}
.ops-grid .pc-kv{flex:1;min-width:120px}
.trade-chips{display:flex;flex-wrap:wrap;gap:6px;margin:var(--sp-2) 0}
.trade-in{width:100%;background:var(--bg-surface-raised);color:var(--text-primary);
  border:1px solid var(--border-subtle);border-radius:var(--r-md);
  font:var(--fs-sm)/1 var(--font-ui);padding:8px 10px}
.trade-in:focus{outline:none;border-color:var(--accent)}
@media (max-width:820px){
  .ops-faab{min-width:0}
  .ops-trend{display:none}
}

/* ---- player card: schedule strip ---- */
.sch{display:grid;grid-template-columns:repeat(9,1fr);gap:3px;margin-top:var(--sp-2)}
.sch-c{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-sm);padding:4px 2px;text-align:center;min-height:34px}
.sch-c .wk{display:block;font:600 9px/1 var(--font-num);color:var(--text-muted)}
.sch-c .op{display:block;margin-top:3px;font:700 var(--fs-xs)/1 var(--font-num);
  color:var(--text-secondary);white-space:nowrap}
.sch-c.bye{border-style:dashed}
.sch-c.bye .op{color:var(--text-muted)}
.sch-c.none .op{color:var(--border-strong)}
/* the fantasy playoffs are the weeks that decide it */
.sch-c.po{border-color:var(--accent-border);background:var(--accent-soft)}
.sch-c.po .wk{color:var(--accent)}

/* ---- player card: usage ---- */
.use{display:flex;gap:4px;margin-top:var(--sp-2)}
.use-c{flex:1;background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-sm);padding:5px 2px;text-align:center}
.use-c .wk{display:block;font:600 9px/1 var(--font-num);color:var(--text-muted)}
.use-c .pct{display:block;margin-top:3px;font:700 var(--fs-lg)/1 var(--font-num);
  font-variant-numeric:tabular-nums}
.use-c .pct small{font-size:9px;color:var(--text-muted);margin-left:1px}
.use-c.none{border-style:dashed}
.use-c.none .pct{color:var(--text-muted)}
.use-delta{margin-top:6px;font-size:var(--fs-sm);color:var(--text-secondary)}
.use-delta.up{color:var(--positive-ink);font-weight:600}
.use-delta.dn{color:var(--negative-ink);font-weight:600}
.ops-spike{color:var(--positive-ink);font-weight:700}

/* ---- player card: news ---- */
.nws{list-style:none;margin-top:var(--sp-2)}
.nws li{padding:6px 0;border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.nws li:last-child{border-bottom:none}
.nws a{color:var(--text-primary);text-decoration:none}
.nws a:hover{color:var(--accent);text-decoration:underline}
.nws .when{display:block;margin-top:2px;font:600 var(--fs-xs)/1 var(--font-num);
  color:var(--text-muted)}

/* a reach, stated in picks and points; blocked = the market says you can wait */
.pp-reach{margin-left:6px;font:600 var(--fs-xs)/1 var(--font-num);
  color:var(--negative-ink);white-space:nowrap}
.pp-reach.blocked{color:var(--text-muted)}
.pp-card.blocked{opacity:.72}

/* the recommended card leads and says so */
.pp-card.lead{border-color:var(--accent-border);background:var(--accent-soft)}
.pp-card.lead .pp-take b{color:var(--accent)}

/* the runway warning: outranks the value verdict, because value advice is
   only valid while there are picks to spare */
.pp-runway{border:1px solid var(--risk-border);background:var(--risk-soft);
  border-radius:var(--r-md);padding:8px var(--sp-3);margin-bottom:var(--sp-3);
  font-size:var(--fs-sm);line-height:1.5}
.pp-runway b{color:var(--risk)}
.pp-runway.forced{border-color:var(--negative);background:var(--negative-soft)}
.pp-runway.forced b{color:var(--negative-ink)}
/* a starter slot nobody drafted, priced at the best free agent */
.rc-hole{color:var(--negative-ink)}
.rc-cap{font-size:var(--fs-xs);color:var(--text-muted)}

/* ============================================================ REPORT CARD */
.rc-hd{display:flex;align-items:flex-start;justify-content:space-between;gap:var(--sp-3)}
.rc-title{font:700 var(--fs-lg)/1.2 var(--font-ui);color:var(--text-primary)}
.rc-sub{font-size:var(--fs-xs);color:var(--text-muted);margin-top:2px}
.rc-empty{color:var(--text-muted);font-size:var(--fs-sm)}
.rc-room{display:flex;gap:var(--sp-4);flex-wrap:wrap;margin:var(--sp-3) 0;
  padding:var(--sp-3) 0;border-top:1px solid var(--border-subtle);
  border-bottom:1px solid var(--border-subtle)}
.rc-room-item{display:flex;flex-direction:column;gap:2px;font-size:var(--fs-sm)}
.rc-list{list-style:none}
.rc-row{display:grid;gap:var(--sp-3);align-items:center;padding:7px 0;
  grid-template-columns:34px minmax(140px,1.2fr) 84px 1fr 1fr 1.1fr;
  border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.rc-row:last-child{border-bottom:none}
.rc-row.mine{background:var(--bg-cell-hover)}
.rc-row.mine .rc-team{color:var(--accent)}
.rc-grade{font:800 var(--fs-lg)/1 var(--font-num);text-align:center;
  color:var(--text-primary)}
.rc-team{display:flex;flex-direction:column;gap:2px}
.rc-str{text-align:right;font-weight:700;font-variant-numeric:tabular-nums}
.rc-str .caps{margin-left:4px;color:var(--text-muted);font-weight:600}
.rc-pick{color:var(--text-secondary);font-size:var(--fs-xs)}
.rc-flags{display:flex;gap:5px;flex-wrap:wrap;justify-content:flex-end}
.rc-flag{font-size:var(--fs-xs);color:var(--text-muted);
  border:1px solid var(--border-subtle);border-radius:var(--r-sm);padding:1px 6px}
/* value polarity, same two poles as the whole board */
.rc-up{color:var(--positive);font-weight:700}
.rc-down{color:var(--negative);font-weight:700}
.rc-basis{margin-top:var(--sp-3);font-size:var(--fs-xs);color:var(--text-muted);
  line-height:1.5}
@media (max-width:900px){
  .rc-row{grid-template-columns:30px 1fr 74px;}
  .rc-pick,.rc-flags{display:none}
}

/* ============================================================ TYPE-AHEAD */
.ta{position:relative;margin-top:var(--sp-3)}
.ta-in{display:flex;align-items:center;gap:var(--sp-3);background:var(--bg-surface-raised);
  border:1px solid var(--border-strong);border-radius:var(--r-md);padding:0 var(--sp-3)}
.ta-in:focus-within{border-color:var(--accent)}
.ta-slash{white-space:nowrap;color:var(--text-secondary)}
.ta-slash b{color:var(--accent);font-weight:700}
#taInput{flex:1;background:none;border:none;color:var(--text-primary);
  font:600 var(--fs-lg)/1 var(--font-ui);padding:13px 0;min-width:0}
#taInput:focus{outline:none}
#taInput::placeholder{color:var(--text-muted);font-weight:400}
.ta-hint{white-space:nowrap}
.ta-list{list-style:none;position:absolute;left:0;right:0;top:100%;z-index:9;
  background:var(--bg-surface-raised);border:1px solid var(--border-strong);
  border-top:none;border-radius:0 0 var(--r-md) var(--r-md);max-height:320px;overflow-y:auto;
  box-shadow:0 12px 28px rgba(0,0,0,.55)}
.ta-list:empty{display:none}
.ta-list li{display:grid;grid-template-columns:38px 40px minmax(0,1fr) auto;gap:var(--sp-2);
  align-items:center;padding:8px var(--sp-3);border-bottom:1px solid var(--border-subtle);
  cursor:pointer;font-size:var(--fs-sm)}
.ta-list li:last-child{border-bottom:none}
.ta-list li.on{background:var(--accent-soft);box-shadow:inset 3px 0 0 var(--accent)}
.ta-list li .rk{font:var(--fs-sm)/1 var(--font-num);color:var(--text-muted);text-align:right}
.ta-list li .nm{font-weight:600;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.ta-list li .nm .team{font-size:var(--fs-xs);color:var(--text-muted);font-weight:400;margin-left:6px}
.ta-list li .meta{font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted);text-align:right;white-space:nowrap}

/* ============================================================ PLAYER CARD */
.pcard-scrim{position:fixed;inset:0;background:rgba(0,0,0,.5);z-index:20;border:none}
/* WIDER, AND THE WIDTH IS WHAT BUYS THE LAYOUT. At 430px every section had to
   be a full-width band, so the card was one long column and reading the
   sources against the survival curve meant scrolling between them. At 760 the
   sections pair up (.pc-cols below) and the whole verdict-to-schedule story
   fits in about one screen, which is the difference between a reference
   document and something usable on the clock. */
/* A FLOATING PANEL, not a flush drawer. Ported from the design template,
   which insets the card, rounds it, and rings it in the accent - so it reads
   as an object laid over the board rather than a wall the board slid under.
   The tint is the app's cyan, not the template's #3ee0f7 (see the palette
   decision); the gradient is what keeps a 2,600px scroll from reading as a
   flat slab. */
.pcard{position:fixed;top:10px;right:10px;bottom:10px;width:min(760px,calc(100vw - 20px));
  z-index:21;border-radius:16px;
  border:1px solid color-mix(in srgb,var(--accent) 30%,transparent);
  background:linear-gradient(160deg,#101a1f 0%,#0b1114 55%,#090e11 100%);
  box-shadow:0 0 60px color-mix(in srgb,var(--accent) 14%,transparent),
             0 24px 60px rgba(0,0,0,.6);
  overflow-y:auto;overscroll-behavior:contain;
  /* The pinned bar sits on top of the last section rather than after it. */
  scroll-padding-bottom:72px}

/* Two columns of sections, PACKED - multi-column flow rather than a grid.
   Each .pc-sec becomes its own bordered panel; `.full` spans both for the ones
   that are the point of the card (verdict), are controls (your call), or are
   footnotes (provenance).

   A grid was the obvious choice and it was wrong: these sections differ in
   height by an order of magnitude (measured on one player - Market price 126px
   beside Rest of season 444px, Schedule 192px beside the WHY decomposition at
   1200px), and grid rows are as tall as their tallest cell. That left voids of
   300 to 1000px down the middle of the card. Column flow packs instead, so
   there are none, and `column-span:all` still gives the three sections that
   need the full width exactly that.

   The trade is reading order: column-major (down, then across) instead of
   row-major. Acceptable here because these panels are independent readings of
   one player rather than a narrative, and the two that ARE sequenced - the
   verdict at the top and the actions at the bottom - span all columns. */
.pc-cols{column-count:2;column-gap:var(--sp-3);padding:var(--sp-3)}
.pc-cols>.pc-sec{border:1px solid var(--border-subtle);border-radius:var(--r-lg);
  background:var(--bg-surface-raised);padding:var(--sp-3) var(--sp-4);
  /* inline-block + width:100% is what makes break-inside:avoid reliable across
     engines; on its own the property is advisory in several of them. */
  display:inline-block;width:100%;break-inside:avoid;-webkit-column-break-inside:avoid;
  margin:0 0 var(--sp-3)}
/* display:block is REQUIRED here, not cosmetic: column-span only applies to
   in-flow block-level boxes, so the inline-block above (which the other
   sections need for break-inside to hold) silently cancels it and the verdict
   flows into column one like any other panel. */
.pc-cols>.pc-sec.full{display:block;column-span:all;-webkit-column-span:all}
/* Below this the panel is most of a phone screen and two columns of source
   rows stop being readable, so it degrades to the single column it used to be. */
@media (max-width:820px){ .pc-cols{column-count:1} }
.pc-hd{position:sticky;top:0;z-index:3;
  background:linear-gradient(180deg,#121d23 0%,#0e161a 100%);
  border-bottom:1px solid color-mix(in srgb,var(--accent) 20%,transparent);
  border-radius:16px 16px 0 0;padding:var(--sp-4)}
/* Position and full club name, the line the template leads with - the chips
   below carry the same facts in short form, this one is for recognition. */
.pc-team{font-size:var(--fs-lg);color:var(--text-secondary);margin-top:4px}

/* The pinned action bar. */
.pc-bar{position:sticky;bottom:0;z-index:3;display:flex;gap:var(--sp-2);
  padding:var(--sp-3) var(--sp-4);
  background:linear-gradient(180deg,rgba(9,14,17,.82),#0a1013 40%);
  backdrop-filter:blur(8px);
  border-top:1px solid color-mix(in srgb,var(--accent) 20%,transparent);
  border-radius:0 0 16px 16px}
.pc-act{flex:1;min-height:40px;padding:10px var(--sp-3);border-radius:9px;
  border:1px solid var(--border-strong);background:var(--bg-surface-raised);
  color:var(--text-secondary);font:600 var(--fs-lg)/1 var(--font-ui);
  display:flex;align-items:center;justify-content:center;white-space:nowrap}
.pc-act:hover{color:var(--text-primary);border-color:var(--accent-border)}
.pc-act.on{border-color:var(--accent-border);color:var(--accent);background:var(--accent-soft)}
.pc-act.on-x{border-color:color-mix(in srgb,var(--negative) 55%,transparent);
  color:var(--negative-ink);background:var(--negative-soft)}
.pc-act.primary{flex:1.4;border:none;font-weight:700;color:var(--on-accent);
  background:linear-gradient(180deg,var(--accent-hover),var(--accent))}
.pc-act.primary:hover{filter:brightness(1.08)}
/* Not a button: there is nothing to draft into. Says why on hover rather than
   rendering a dead control. */
.pc-act.off{cursor:default;color:var(--text-muted);background:transparent;
  border-style:dashed}

/* The verdict, with a pole mark - same two poles as the READ tag. */
.pc-vwrap{display:flex;gap:var(--sp-3);align-items:flex-start}
.pc-vmark{flex:none;width:34px;height:34px;border-radius:50%;display:flex;
  align-items:center;justify-content:center;font-size:11px;
  border:1px solid var(--border-strong);color:var(--text-muted)}
.pc-vmark.val{border-color:var(--accent-border);color:var(--positive-ink);background:var(--accent-soft)}
.pc-vmark.fade{border-color:color-mix(in srgb,var(--negative) 55%,transparent);
  color:var(--negative-ink);background:var(--negative-soft)}
.pc-vmark.weak{opacity:.7}
/* TOP LEFT, WHERE A WAY BACK LIVES, and carrying its own ground so it does
   not depend on what the hero happens to be painted with - every club colour
   and a photograph pass under this thing. z-index clears the hero's scrim and
   the portrait, both of which sit at 1. */
.pc-close{position:absolute;top:var(--sp-3);left:var(--sp-3);z-index:4;
  display:inline-flex;align-items:center;gap:5px;
  background:rgba(6,9,11,.62);border:1px solid rgba(255,255,255,.24);
  color:#fff;font:600 12px/1 var(--font-label);letter-spacing:.12em;
  text-transform:uppercase;padding:9px 13px 9px 10px;border-radius:999px;
  backdrop-filter:blur(6px);-webkit-backdrop-filter:blur(6px)}
.pc-close:hover{background:rgba(6,9,11,.82);border-color:rgba(255,255,255,.45);
  color:#fff}
.pc-back-i{font:700 15px/1 var(--font-ui);margin-top:-1px}
.pc-id{display:flex;align-items:center;gap:var(--sp-3);padding-left:34px}
.pc-who{min-width:0;flex:1}
/* The face is data (recognition at a glance), not decoration - so it is small,
   squared to the grid, and slightly desaturated to sit inside the terminal
   palette rather than on top of it. */
.pc-face{width:96px;height:96px;flex:none;border-radius:var(--r-lg);
  border:1px solid var(--border-subtle);background:var(--bg-surface);
  object-fit:cover;object-position:center top;filter:saturate(.85)}
.pc-face.logo{object-fit:contain;padding:14px}
.pc-name{font:500 30px/1.15 var(--font-display);letter-spacing:-.005em}
.pc-bio{font-size:var(--fs-xs);color:var(--text-muted);margin-top:3px}
.pc-band{display:flex;margin-top:var(--sp-3);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);background:var(--bg-surface);overflow:hidden}
.pc-band .b{flex:1 1 0;min-width:0;padding:6px 10px;border-left:1px solid var(--border-subtle)}
.pc-band .b:first-child{border-left:none}
.pc-band .bk{display:block;font-size:var(--fs-xs);letter-spacing:var(--track-caps);
  text-transform:uppercase;color:var(--text-muted);font-weight:600}
.pc-band .bv{display:block;font:700 var(--fs-lg)/1.2 var(--font-num);
  font-variant-numeric:tabular-nums;margin-top:2px;white-space:nowrap}
.pc-band .bn{display:block;font-size:var(--fs-xs);color:var(--text-muted);white-space:nowrap}
/* The identity facts as chips rather than a run-on line - at 760px there is
   room for them to be separate objects, and each one is a different fact. */
.pc-sub{display:flex;align-items:center;gap:7px;margin-top:var(--sp-3);flex-wrap:wrap}
.pc-sub .f{font-size:var(--fs-sm);color:var(--text-secondary);
  border:1px solid var(--border-subtle);border-radius:var(--r-sm);padding:4px 9px}
.pc-sub .f b{color:var(--text-primary);font-weight:600}
/* The READ tag, same words and same two poles as the board card's. */
.pc-sub .f.tag{font:800 var(--fs-xs)/1.4 var(--font-num);letter-spacing:.07em}
.pc-sub .f.tag.val{color:var(--positive-ink);border-color:var(--accent-border);
  background:var(--accent-soft)}
.pc-sub .f.tag.fade{color:var(--negative-ink);background:var(--negative-soft);
  border-color:color-mix(in srgb,var(--negative) 55%,transparent)}
.pc-sub .f.tag.weak{opacity:.78}
.pc-sub .f.tag.none{color:var(--text-muted)}
.pc-sec{padding:var(--sp-4);border-bottom:1px solid var(--border-subtle)}
.pc-sec>.caps{display:block;margin-bottom:var(--sp-2)}
.pc-verdict{font:600 var(--fs-lg)/1.5 var(--font-ui);color:var(--text-primary)}
.pc-verdict .lede{color:var(--accent)}
.pc-verdict.fade .lede{color:var(--negative-ink)}
.pc-why{font-size:var(--fs-sm);color:var(--text-secondary);margin-top:var(--sp-2);line-height:1.55}

/* five-source strip */
.pc-src{display:grid;grid-template-columns:44px 46px 1fr 46px;gap:var(--sp-2);align-items:center;
  padding:6px 0;border-bottom:1px solid var(--border-subtle);font-size:var(--fs-sm)}
.pc-src:last-child{border-bottom:none}
.pc-src .lab{font:700 var(--fs-xs)/1 var(--font-num);letter-spacing:.06em;color:var(--text-secondary)}
.pc-src .lab .off{color:var(--text-muted);font-weight:400}
.pc-src .v{font:700 var(--fs-sm)/1 var(--font-num);font-variant-numeric:tabular-nums;text-align:right}
.pc-src .bar{position:relative;height:14px}
.pc-src .bar .mid{position:absolute;left:50%;top:0;bottom:0;width:1px;background:var(--border-strong)}
.pc-src .bar i{position:absolute;top:4px;height:6px;border-radius:3px}
.pc-src .bar i.up{background:var(--positive)}
.pc-src .bar i.dn{background:var(--negative)}
/* Static, in its own grid cell - see srcRow. Absolute positioning against the
   bar is what let a clamped bar push this on top of the value. */
.pc-src .d{font:700 var(--fs-xs)/1 var(--font-num);text-align:right;
  font-variant-numeric:tabular-nums}

/* production + meta grids */
.pc-grid{display:grid;grid-template-columns:1fr 1fr;gap:var(--sp-3)}
.pc-kv{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3)}
.pc-kv .k{font-size:var(--fs-xs);letter-spacing:var(--track-caps);text-transform:uppercase;color:var(--text-muted);font-weight:600}
.pc-kv .v{font:700 var(--fs-xl)/1.1 var(--font-num);font-variant-numeric:tabular-nums;margin-top:3px}
.pc-kv .v small{font-size:var(--fs-sm);color:var(--text-muted);font-weight:600}
.pc-kv .n{font-size:var(--fs-xs);color:var(--text-muted);margin-top:2px}

/* survival curve */
.pc-curve{width:100%;height:118px;display:block}
.pc-curve .axis{stroke:var(--border-subtle);stroke-width:1}
.pc-curve .grid{stroke:var(--border-subtle);stroke-width:1;stroke-dasharray:2 3;opacity:.6}
.pc-curve .line{fill:none;stroke:var(--accent);stroke-width:2}
.pc-curve .area{fill:var(--accent-soft)}
.pc-curve .mine{stroke:var(--text-secondary);stroke-width:1;stroke-dasharray:3 3}
.pc-curve .dot{fill:var(--accent)}
.pc-curve text{fill:var(--text-muted);font:9px/1 var(--font-num)}
.pc-pks{font:600 var(--fs-sm)/1.4 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-secondary);margin-top:4px}
.pc-pks b{color:var(--accent);font-weight:700}

/* actions */
.pc-actions{display:flex;gap:var(--sp-2);flex-wrap:wrap}
.pc-actions .fchip{min-height:36px}
.pc-actions .fchip.on{background:var(--accent-soft);color:var(--accent);border-color:var(--accent-border)}
.pc-actions .fchip.on-x{background:var(--negative-soft);color:var(--negative-ink);border-color:rgba(194,68,127,.55)}
.pc-note{width:100%;margin-top:var(--sp-3);background:var(--bg-surface-raised);
  border:1px solid var(--border-subtle);border-radius:var(--r-md);color:var(--text-primary);
  font:var(--fs-sm)/1.5 var(--font-ui);padding:8px 10px;resize:vertical;min-height:56px}
.pc-note:focus{outline:none;border-color:var(--accent)}
.pc-prov{font-size:var(--fs-xs);color:var(--text-muted);line-height:1.6}

@media (max-width:640px){
  /* On a phone the panel goes back to filling the screen: a 10px inset and a
     16px radius spend real estate to decorate an edge nobody can see past. */
  .pcard{inset:0;width:100vw;border-radius:0;border-left:none}
  .pc-hd,.pc-bar{border-radius:0}
  .ta-list li{grid-template-columns:34px 40px minmax(0,1fr)}
  .ta-list li .meta{display:none}
}

/* ============================================================ DRAFT ROOM */
/* Eleven controls in a row need 1321px; wraps for the same reason .topbar-inner
   does, and was found by the same scrollWidth sweep. */
.draft-setup{display:flex;align-items:center;flex-wrap:wrap;gap:var(--sp-3);margin-top:var(--sp-4)}
/* Same gap as the row, so nothing moves on a bar that already fits; the point
   is only that a field is one unit when the row folds. min-width:0 all the way
   down is for the far end: alone on its line a field cannot wrap any further,
   so without it the widest one (Profile, 279px) sets a floor the row cannot go
   below and it is the field itself that overflows a 320px phone. */
.ds-field{display:flex;align-items:center;gap:var(--sp-3);min-width:0}
.ds-field>select,.ds-field>input{min-width:0}
button.primary{background:var(--accent);color:var(--pos-ink);border:none;border-radius:var(--r-md);
  padding:9px 16px;font:700 var(--fs-sm)/1 var(--font-ui);min-height:36px}
button.primary:hover{background:var(--accent-hover)}
.room{display:grid;grid-template-columns:1fr 300px;gap:var(--sp-4)}
.room-main{min-width:0}
.panel{background:var(--bg-surface);border:1px solid var(--border-subtle);border-radius:var(--r-md)}
.panel-hd{display:flex;align-items:baseline;justify-content:space-between;
  padding:var(--sp-3) var(--sp-4);border-bottom:1px solid var(--border-subtle)}

/* clock card */
.clock-card{display:flex;align-items:center;gap:var(--sp-4);background:var(--bg-surface-raised);
  border:1px solid var(--border-strong);border-radius:var(--r-md);padding:var(--sp-3) var(--sp-4);margin-bottom:var(--sp-3)}
.clock-time{font:800 32px/1 var(--font-num);font-variant-numeric:tabular-nums}
.clock-time.you{color:var(--accent)}
.clock-meta .caps{display:block}
.clock-team{font:600 var(--fs-xl)/1.2 var(--font-ui)}
.clock-until{margin-left:auto;text-align:right}
.clock-until .num{font-size:var(--fs-xl);font-weight:700}

/* ======================================================= THE STATE RAIL ====
 *
 * The room's state and the room's one action, at the top of the room. The card
 * inside it is the same .clock-card that used to sit under the board - this
 * wrapper adds the action row and the three faces, and takes NO styling
 * responsibility away from the card, so the phone's full-screen mode (which
 * lifts .clock-card out to a fixed bar) is untouched by all of it.
 */
/* The topbar is `position:sticky`, so anything scrollIntoView lands at the top
   of the document lands UNDER it. Leaving full screen scrolls to this rail,
   and without the margin the state card came back half-hidden behind the
   stamp - measured, not guessed: 56px of a 96px card. */
.drr{margin-bottom:var(--sp-3);scroll-margin-top:64px}
.drr .clock-card{margin-bottom:0;border-radius:var(--r-md) var(--r-md) 0 0}
.drr-act{display:flex;align-items:center;gap:var(--sp-2);flex-wrap:wrap;
  padding:var(--sp-2) var(--sp-4);background:var(--bg-surface);
  border:1px solid var(--border-strong);border-top:0;
  border-radius:0 0 var(--r-md) var(--r-md)}
.drr-act .ds-field{margin:0}
.drr-meta{margin-left:auto;color:var(--text-muted);font-variant-numeric:tabular-nums}
/* The drawer door exists at one width, because the drawer does. Desktop shows
   the whole setup row and has nothing to open. */
#btnDrSetup{display:none}
/* Once a draft exists the seat and the start are settled, so the row stops
   offering them. It does not disappear: the drawer door is how every other
   control on this page stays reachable. */
.drr.live .drr-slot,.drr.fin .drr-slot,
.drr.live #btnStartDraft,.drr.fin #btnStartDraft{display:none}
/* WAS: `@media (max-width:640px){ .drr.pre .clock-time.pre{...quiet grey} }`,
   on the argument that a pre-draft numeral is a fact and not a state. The
   argument went with the face it described. `.pre` now wears the live card's
   anatomy - the room is ON pick 1.01 before anybody presses Start, and it
   ticks - so its numeral is the same state the live one is and is drawn the
   same way. The class it was keyed to is not emitted any more, and a rule that
   can no longer match anything is worse than no rule: it reads as an
   explanation for behaviour that is not happening. */

/* THE TURN, INVERTED. His reference frame spends its loudest ink on the cell
   that is on the clock, and the skin already owns the plate to spend it with:
   --turf, the dark band the design draws under stadium lighting. Same object
   as the board's own on-the-clock cell, one scale up. */
.drr.you .clock-card{background:var(--turf,#0f1a22);border-color:var(--turf,#0f1a22)}
.drr.you .clock-time,.drr.you .clock-team,.drr.you .clock-until .num{color:#fff}
.drr.you .clock-card .caps{color:rgba(255,255,255,.62)}
.drr.you .clock-time.you{color:#fff}
.clock-count{margin-left:auto;font:800 26px/1 var(--font-num);
  font-variant-numeric:tabular-nums;color:var(--text-primary);
  letter-spacing:.02em}
.drr.you .clock-count{color:#fff}
.clock-count.low{color:var(--negative)}
.drr.you .clock-count.low{color:#ff8a94}
/* The rail's clock is the LAST thing before "your pick in", so the until block
   stops claiming the auto margin when a countdown is present. */
.clock-count + .clock-until{margin-left:var(--sp-4)}

/* ==================================================== THE MISPRICE BOARD ===
 *
 * Sleeper's board answers "who is next". This one answers "who is wrong", and
 * the row is drawn as the arbitrage it is: two dots on ONE shared pick scale,
 * ours and the market's, with the gap between them as a bar. Shared, because a
 * per-row scale draws a four-pick gap and a twenty-pick gap the same width and
 * the comparison is the entire point.
 */
.dr-edge{margin-bottom:var(--sp-3);background:var(--bg-surface);
  border:1px solid var(--border-strong);border-radius:var(--r-md);
  overflow:hidden}
.dre-hd{display:flex;align-items:baseline;gap:8px;flex-wrap:nowrap;
  padding:var(--sp-3) var(--sp-4) 6px}
.dre-t{color:var(--text-primary);font-weight:700;letter-spacing:.12em;
  white-space:nowrap}
.dre-live{color:var(--accent);white-space:nowrap;font-size:9px;
  padding:3px 6px;border-radius:3px;background:var(--accent-soft)}
.dre-n{margin-left:auto;color:var(--text-muted);white-space:nowrap;font-size:9px}
.dre-n b{color:var(--text-primary);font-variant-numeric:tabular-nums}
.dre-dek{margin:0;padding:0 var(--sp-4) var(--sp-3);
  font:var(--fs-xs)/1.45 var(--font-ui);color:var(--text-muted);
  border-bottom:1px solid var(--border-subtle)}
.dre-dek b{color:var(--text-secondary);font-weight:700}
.dre-list{list-style:none;margin:0;padding:0}
.dre{padding:10px var(--sp-4) 12px;border-bottom:1px solid var(--border-row);
  /* THE READ IS AN EDGE, not a wash: a coloured fill behind six rows in a row
     turns the panel into a heat map of itself and nothing leads. */
  border-left:3px solid transparent}
.dre:last-child{border-bottom:none}
.dre.val.strong{border-left-color:var(--positive)}
.dre.fade.strong{border-left-color:var(--negative)}
.dre-top{display:flex;align-items:center;gap:8px;min-width:0}
.dre-nm{border:0;background:none;padding:0;cursor:pointer;text-align:left;
  font:700 14px/1.25 var(--font-ui);color:var(--text-primary);
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap;min-width:0}
.dre-nm:hover{color:var(--accent)}
.dre-tb{font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted);
  white-space:nowrap}
.dre-tag{margin-left:auto;font:700 9px/1 var(--font-num);letter-spacing:.14em;
  padding:4px 6px;border-radius:3px;white-space:nowrap}
.dre.val .dre-tag{color:var(--positive);background:var(--accent-soft)}
.dre.fade .dre-tag{color:var(--negative);background:var(--negative-soft)}
/* A gap that has NOT cleared the bar is still worth seeing and must not be
   narrated as a call, so its tag drops to the quiet ink. Same rule the player
   card's AT PRICE follows. */
.dre:not(.strong) .dre-tag{color:var(--text-muted);background:var(--row-tint)}

.dre-bot{display:flex;align-items:center;gap:10px;margin-top:9px}
/* THE READING, in fixed positions. Both numbers the bar is drawn from, printed
   where they cannot move: a label that follows a dot collides with the other
   label at exactly the gap sizes that are most common. */
.dre-read{display:flex;flex-direction:column;gap:3px;flex:0 0 auto;
  font:700 9px/1 var(--font-num);letter-spacing:.06em;text-transform:uppercase;
  font-variant-numeric:tabular-nums;min-width:52px}
.dre-read .e{color:var(--accent)}
.dre-read .m{color:var(--text-secondary)}
/* THE DIVERGING BAR. Centre is OUR number; right is the market letting him
   fall past it, left is the market taking him ahead of it. Direction is the
   claim and length is its size, on one denominator shared by every row. */
.dre-scale{position:relative;flex:1 1 auto;min-width:70px;height:22px}
.dre-track{position:absolute;left:0;right:0;top:10px;height:2px;
  background:var(--track);border-radius:1px}
.dre-mid{position:absolute;left:50%;top:3px;width:1px;height:16px;
  margin-left:-.5px;background:var(--accent)}
.dre-bar{position:absolute;top:7px;height:8px;border-radius:2px}
.dre.val .dre-bar{background:var(--positive)}
.dre.fade .dre-bar{background:var(--negative)}
/* A gap that has not cleared its bar is drawn, and drawn quietly. It is a
   fact; it is not a call. */
.dre:not(.strong) .dre-bar{opacity:.34}
.dre-gap{font:800 15px/1 var(--font-num);font-variant-numeric:tabular-nums;
  min-width:34px;text-align:right}
.dre.val .dre-gap{color:var(--positive)}
.dre.fade .dre-gap{color:var(--negative)}
/* THE BAR STOPS GROWING WITH THE WINDOW. On a 1440 the row is 1,640px wide and
   an unbounded flex track drew a 600px bar with 400px of white between the
   reading and the number it belongs to. A comparison you have to travel to is
   a comparison nobody makes. Everything the bar says is said inside 380px;
   past that the row's right half goes to the survival read and the action. */
@media (min-width:641px){
  .dre-scale{flex:0 1 380px}
  .dre-surv{margin-left:auto}
}
/* AT PRICE is not a call, so its number stops being coloured like one. The
   gap is still printed - it is a fact - it just stops being narrated. Same
   ruling the player card's tag already makes. */
.dre:not(.strong) .dre-gap{color:var(--text-muted)}
.dre-surv{font:var(--fs-xs)/1.2 var(--font-num);color:var(--text-muted);
  white-space:nowrap;min-width:78px;text-align:right}
.dre-surv.low{color:var(--negative)}
/* `.dre .dre-act`, and the extra specificity is load-bearing rather than
   habit: the draft button wears `.draftbtn` too, so that it goes through the
   one delegated handler every other draft control goes through - and
   `.draftbtn` is declared LATER in this file as a bare text link. At equal
   weight the later rule wins and the pill rendered as grey text. */
.dre .dre-act{min-height:34px;padding:0 12px;border-radius:17px;cursor:pointer;
  font:700 10px/1 var(--font-ui);letter-spacing:.1em;text-transform:uppercase;
  border:1px solid var(--accent-border);background:var(--accent-soft);
  color:var(--accent);white-space:nowrap;text-align:center}
.dre .dre-act:hover{background:var(--accent);color:var(--on-accent)}
.dre .dre-act.qbtn.on{background:var(--accent);color:var(--on-accent);
  border-color:var(--accent)}

/* ---------------------------------------------------- BEST AVAILABLE (7a)
   His 7a, ported 2026-08-08. Eight labelled columns where there were six
   unlabelled ones, and the cell that was carrying two different numbers (an
   ADP and a signed delta against it) is now two cells that say which is which.

   THE TRACKS ARE PROPORTIONAL, NOT COPIED. The mock is a 1000px card; this
   panel measures 720px at a 1280 viewport, so the mock's fixed widths would
   have left the player column 46px and `text-overflow:ellipsis` would have
   rendered every candidate as "J." - the exact failure the 5a port hit. Each
   fixed track is sized to the content it actually holds (a three-digit rank, a
   position chip, "CIN - bye 6", "certain" beside a bar, the draft hint) and the
   player column keeps a real 150px floor so the row scrolls before it
   collapses. A minimum equal to the sum of everything else is not a minimum. */
.ba-head{display:grid;grid-template-columns:var(--ba-cols);gap:10px;
  padding:9px var(--sp-4);border-bottom:1px solid var(--border-subtle);
  font:400 9px/1 var(--font-num);letter-spacing:.16em;text-transform:uppercase;
  color:var(--text-muted)}
.ba-head .r{text-align:right}
.ba-list{list-style:none;padding:0}
:root{--ba-cols:34px 30px minmax(150px,1fr) 80px 44px 46px 124px 72px}
.ba{display:grid;grid-template-columns:var(--ba-cols);
  gap:10px;align-items:center;
  padding:10px var(--sp-4);border-bottom:1px solid var(--border-row)}
.ba:last-child{border-bottom:none}
.ba:hover{background:var(--bg-row-hover)}
.ba .rk{font:var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums;color:var(--text-muted);text-align:right}
.ba .nm{font:600 14px/1.35 var(--font-ui);min-width:0;
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.ba .tb{font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted);
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.ba .mkt{font:var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:right;color:var(--text-secondary)}
/* The Room delta takes the POLES, not the inks. The mock sets these at
   --positive / --negative rather than the lighter .d-up/.d-dn pair the board
   uses, and it is right for this column: the ledger is scanned down, so the
   sign wants the strongest step the palette has. Own classes rather than
   .d-up/.d-dn, so the board's own deltas are untouched - and because `.ba
   .room` would out-specify them anyway, which is how this first shipped grey. */
.ba .room{font:var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:right;color:var(--text-muted)}
.ba .room.up{color:var(--positive)}
.ba .room.dn{color:var(--negative)}
/* The word this number carries on a touch screen. Off everywhere else: on a
   pointer machine the column has a header, a legend and this cell's own
   tooltip, and a word inside a numeric column would break the one thing a
   column of numbers is for. See the mode's own rule for where it comes back. */
.ba .room-w{display:none}
.surv{display:flex;align-items:center;gap:9px}
.surv .pct{font:var(--fs-sm)/1 var(--font-num);font-variant-numeric:tabular-nums;
  width:48px;color:var(--accent)}
/* A SETTLED QUESTION IS NOT A HIGH SCORE. The phrase and a flat dim rail at
   BOTH poles, because a full bar in a column of full bars reads as "best"
   when it means "not a question" - and an empty one reads as "worst". */
/* AND IT IS A SENTENCE, SO IT IS SET AS ONE. `gone by 2.12` in the tabular
   numeric face is a measurement that has lost its digits; in the UI face at
   the row's smallest step it is what it actually is - a short statement about
   a named future pick, in the quietest ink the column has. The width is auto
   rather than the 48px the percentage holds: the phrase is 60px at 10px in
   Chrome and the percentage is 26px, so a shared track would either clip the
   phrase or push the meter off the row. The meter keeps whatever is left,
   which for a settled question is a flat dim rail carrying no reading. */
.surv .pct.sure{color:var(--text-muted);font:600 10px/1.15 var(--font-ui);
  font-variant-numeric:normal;letter-spacing:0;width:auto;flex:0 0 auto;
  white-space:nowrap}
.meter{flex:1;height:4px;background:var(--track);border-radius:2px;overflow:hidden}
.meter i{display:block;height:100%;background:var(--accent);border-radius:2px}
.meter.mid i{background:var(--warning);opacity:.75}
.meter.low i{background:var(--negative)}
/* THE BAND NAMES ITS OWN POLARITY. `mid`/`low` above are keyed to a COVERAGE
   percentage (the queue's joint survival), where the low end is the bad end.
   The gone-by column runs the other way, so it takes its own pair rather than
   borrowing names that would put the alarm on the calm rows. Same two inks:
   the palette is unchanged, only which end of the scale wears them. */
.meter.warm i{background:var(--warning);opacity:.75}
.meter.hot i{background:var(--negative)}
.meter.certain i{background:var(--text-dim)}

/* ------------------------------------------------------- the pool's search
   862 players reachable only by scrolling a list that stops at twenty-five
   live rows. The reference ships a magnifier on this row and this panel had
   nothing; it is the one control the comparison found simply missing. */
.ba-search{display:flex;align-items:center;gap:6px;flex:1 1 150px;
  min-width:0;max-width:280px;height:30px;padding:0 4px 0 9px;
  border-radius:15px;background:var(--bg-surface);
  border:1px solid var(--border-strong)}
/* The title and the search pill met at zero pixels at 390 - space-between
   distributes leftovers, it does not guarantee any (round 9). */
#baPanelHd{column-gap:12px}
.ba-search:focus-within{border-color:var(--accent)}
.ba-mag{flex:none;width:13px;height:13px;color:var(--text-muted)}
.ba-search input{flex:1 1 auto;min-width:0;background:none;border:0;padding:0;
  font:400 12px/1 var(--font-ui);color:var(--text-primary)}
.ba-search input::placeholder{color:var(--text-muted);opacity:1}
.ba-search input:focus{outline:none}
.ba-clear{flex:none;width:24px;height:24px;border:0;background:none;padding:0;
  color:var(--text-muted);font:400 16px/1 var(--font-ui);cursor:pointer}
.ba-clear:hover{color:var(--text-primary)}
/* The mock draws this as a keyboard hint rather than a button, and the footer
   says why: this panel is driven with j/k and Enter. It stays a real button so
   the phone and the mouse keep an affordance the keyboard does not need. */
.draftbtn{background:none;border:none;padding:0;text-align:right;
  font:var(--fs-xs)/1 var(--font-num);letter-spacing:.12em;text-transform:uppercase;
  color:var(--text-dim)}
/* The enter glyph is a keyboard hint wearing a hint's mark (round 11). It
   was bare label text at every width, promising a phone a key it does not
   have; marked as a <kbd> the medium can decide - same one-render-path law
   as the row's direction word. */
.draftbtn .dbk{font:inherit;color:var(--text-dim);margin-right:1px}
@media (hover:none){.draftbtn .dbk{display:none}}
.draftbtn:hover{color:var(--accent)}
.draftbtn.off{color:var(--text-dim);cursor:default}
.draftbtn.off:hover{color:var(--text-dim)}

/* the fills-first rule, surfaced while browsing */
.ba-advisory{display:flex;align-items:center;gap:10px;padding:9px var(--sp-4);
  background:var(--warning-soft);border-bottom:1px solid var(--border-row)}
.ba-advisory .adv-n{font:var(--fs-xs)/1 var(--font-num);letter-spacing:.14em;
  color:var(--risk);white-space:nowrap}
.ba-advisory span+span{font:400 var(--fs-sm)/1.4 var(--font-ui);color:var(--text-secondary)}
/* SENTENCE CASE, BECAUSE IT IS A SENTENCE (round 9). This footer is two
   written explanations, and the mono-caps micro-label treatment stretched
   over 40 words reads as a terminal disclaimer - the round-8 first-timer's
   "dense abbreviations" charge, self-inflicted on the one element that
   exists to explain them. Labels shout; explanations talk. */
.ba-legend{display:flex;gap:20px;padding:var(--sp-3) var(--sp-4);
  font:400 var(--fs-sm)/1.5 var(--font-ui);letter-spacing:0;
  color:var(--text-muted)}
.ba-legend .k{margin-left:auto;text-align:right;white-space:nowrap}
/* j/k and Enter are a keyboard's affordances; on a screen with no keyboard
   the hint is three cryptic glyphs explaining controls that do not exist. */
@media (hover:none){.ba-legend .k{display:none}}

/* ------------------------------------------------------- ROSTER RAIL (6a)
   Twelve rows reading "- open -" became five short ones and a bench count. A
   filled slot is taller and lit; an empty one is a thin line whose POSITION
   CODE carries the accent, so the rail reads as progress rather than as a
   form. The label face stays the mono: the mock sets these codes in Barlow
   Condensed, but --font-label is the editorial serif in this app now and a
   three-letter code in a fixed 38px column is exactly what a mono column is
   for. Its 11px size and .16em tracking are the mock's. */
/* The mock's rail card is 400px; this aside is 300, which leaves 266 for a
   header that wants 274 - so "slot 8" broke across two lines with the 8 alone
   on the second. Both halves are nowrap and the row is allowed to wrap, so the
   count drops INTACT to its own line when it cannot fit beside the label
   (it cannot, at "slot 10") rather than a phrase breaking mid-word. The count
   keeps the mock's own .14em, which is looser than the .18em on the label and
   buys back six of the eight pixels. */
.rail-hd{display:flex;align-items:baseline;justify-content:space-between;
  gap:var(--sp-2);flex-wrap:wrap}
.rail-hd>*{white-space:nowrap}
.rail-count{margin-left:auto;color:var(--accent);letter-spacing:.14em}
.rail-meter{height:4px;margin:0 var(--sp-4) 0;border-radius:2px;
  background:var(--track);overflow:hidden}
.rail-meter i{display:block;height:100%;background:var(--accent);border-radius:2px}
.roster,.picklog{list-style:none;padding:var(--sp-1) 0}
.roster{padding-top:var(--sp-2)}
.slot{display:flex;align-items:center;gap:10px;border-bottom:1px solid var(--border-row)}
.slot:last-child{border-bottom:none}
.slot.filled{padding:11px var(--sp-4);background:var(--row-tint)}
.slot.open{padding:7px var(--sp-4);background:transparent}
.slot .lab{width:38px;flex:none;font:700 var(--fs-sm)/1 var(--font-num);
  letter-spacing:.16em;color:var(--text-muted)}
.slot.open .lab{color:var(--accent)}
.slot .who{font:600 14px/1.35 var(--font-ui);min-width:0;
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.slot.open .who{font:400 var(--fs-lg)/1.35 var(--font-ui);color:var(--text-dim)}
.slot .team{font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}
.slot .byetag{margin-left:auto;font:var(--fs-xs)/1 var(--font-num);color:var(--risk)}
.slot .rd{font:var(--fs-xs)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-dim)}
.bench-line{display:flex;align-items:center;gap:9px;padding:11px var(--sp-4);
  border-bottom:1px solid var(--border-row)}
.bpips{display:flex;gap:5px}
.bp{width:9px;height:9px;border-radius:50%;border:1px solid var(--pip-edge)}
.bp.on{background:var(--accent);border-color:var(--accent)}
.bench-line .bn{margin-left:auto;font:var(--fs-xs)/1 var(--font-num);color:var(--text-dim)}
.byeclash-line{display:flex;gap:9px;padding:11px var(--sp-4);background:var(--warning-soft)}
.byeclash-line .bw{font:var(--fs-xs)/1.4 var(--font-num);letter-spacing:.14em;
  color:var(--risk);white-space:nowrap}
.byeclash-line span+span{font:400 var(--fs-sm)/1.4 var(--font-ui);color:var(--text-secondary)}
/* ------------------------------------------------------ RECENT PICKS (7c)
   Grouped by round, and the run in progress marked as a rule down the edge.
   The round separator is a label plus a hairline that eats the remaining
   width, so the group reads as a band rather than as another row. */
.pl-run{color:var(--negative);letter-spacing:.14em}
.pl-rd{display:flex;align-items:center;gap:10px;padding:12px var(--sp-4) 6px}
.pl-rd span{font:400 9px/1 var(--font-num);letter-spacing:var(--track-mono);
  text-transform:uppercase;color:var(--text-dim)}
.pl-rd i{flex:1;height:1px;background:var(--track)}
.pl{display:flex;align-items:center;gap:10px;padding:8px var(--sp-4);
  border-bottom:1px solid rgba(140,180,200,.06)}
.pl:last-child{border-bottom:none}
/* Always present, transparent when the pick is not in the run - so a run
   starting does not shift every row sideways by three pixels. */
.pl-edge{align-self:stretch;width:3px;flex:none;border-radius:2px;background:transparent}
.pl-edge.on{background:var(--negative)}
.pl-pk{width:30px;flex:none;font:var(--fs-xs)/1 var(--font-num);
  font-variant-numeric:tabular-nums;color:var(--text-muted)}
.pl-who{flex:1;min-width:0}
.pl-nm{display:block;font:600 var(--fs-lg)/1.35 var(--font-ui);color:var(--text-primary);
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.pl-by{display:block;margin-top:3px;font:400 9px/1 var(--font-num);color:var(--text-dim);
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
/* COLOURED LETTERS, not a chip - the same call the 5a thinnest-slot cell made.
   A chip's padding differs per position and would break the column the team
   codes form on the right. Constitution 18 still holds: the letters ARE the
   position, colour only agrees with them. */
.pl-pos{flex:none;font:700 var(--fs-sm)/1 var(--font-num);letter-spacing:.14em}
.pl-pos.pos-QB{color:var(--pos-qb)} .pl-pos.pos-RB{color:var(--pos-rb)}
.pl-pos.pos-WR{color:var(--pos-wr)} .pl-pos.pos-TE{color:var(--pos-te)}
.pl-pos.pos-K{color:var(--pos-k)}   .pl-pos.pos-DST{color:var(--pos-dst)}
.pl-tm{width:26px;flex:none;text-align:right;font:400 9px/1 var(--font-num);
  color:var(--text-muted)}
.pl-legend{padding:11px var(--sp-4);font:400 var(--fs-sm)/1.5 var(--font-ui);
  color:var(--text-muted)}

/* --------------------------------------------------- THE MARKET TAPE (9a)
   The rework of what he called "the visually unappealing momentum 2nd half".
   Two additions: a split bar that answers "quiet week or loud one" before you
   read a row, and a per-night tape that turns one 23-spot number into the four
   nights it happened on. */
.mkt-split{margin:var(--sp-3) 0 var(--sp-4)}
.ms-bar{display:flex;height:9px;border-radius:3px;overflow:hidden}
.ms-bar i{display:block;height:100%}
.ms-dn{background:var(--negative)} .ms-up{background:var(--accent)}
.ms-0{background:#222c31}
.ms-key{display:flex;justify-content:space-between;margin-top:6px;
  font:400 var(--fs-xs)/1 var(--font-num);letter-spacing:.14em;
  text-transform:uppercase;color:var(--text-muted)}
.ms-t-dn{color:var(--negative)} .ms-t-up{color:var(--accent)}

.tape-hd{font-weight:400}
.tapecell{padding:6px 8px}
.tape{display:flex;gap:4px;min-width:150px}
/* Equal-width cells: a night is a night, so the tape may not imply that one
   was longer than another by being wider. */
.tp{flex:1 1 0;min-width:0;height:26px;display:flex;align-items:center;
  justify-content:center;border-radius:3px;
  font:400 var(--fs-sm)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-primary)}
/* A GAP IS DRAWN AS A GAP. Market history is never interpolated, and the one
   place that would be tempting to fudge is a cell that looks empty. */
.tp.gap{background:rgba(140,180,200,.05);color:var(--text-dim)}
/* The row's headline number, in the display serif per the template's own rule
   that big numerals are the serif's job. */
.board td.overall{font:400 19px/1.2 var(--font-display);font-variant-numeric:tabular-nums}
.board td.overall.adp-up{color:var(--positive)}
.board td.overall.adp-dn{color:var(--negative)}
.board td.overall.d-0{color:var(--text-muted)}
/* NOT `.moved` - that class already exists in this file as the ADP-change
   strip's flex container, and reusing it here turned every table row into a
   flex box and wrapped the whole leaderboard into cards. */
.board tr.mv-row td{background:rgba(140,180,200,.02)}
@media (max-width:900px){
  /* The tape is the first thing to go on a narrow screen: it is the shape of a
     move, and Overall and Overnight still carry the move itself. */
  .tapecell,.tape-hd{display:none}
}

/* --------------------------------------------------- INJURY TRACKER (11a)
   The band split is the idea here, and it is not a visual one: a designation
   on a player nobody drafts was sitting in the same list, at the same weight,
   as one on a player you are about to take in the third. Both are still shown;
   the unpriced ones are named, dimmed and put below. */
.is-out{background:var(--negative)}
.is-q{background:rgba(224,180,107,.55)}
.is-stale{background:#222c31}
.is-t-out{color:var(--negative-ink)} .is-t-q{color:var(--risk)}
.board tr.inj-band td{padding:9px var(--cell-x) 7px;text-align:left;
  font:400 9px/1 var(--font-num);letter-spacing:var(--track-mono);
  text-transform:uppercase;color:var(--text-muted);background:var(--row-tint)}
.board tr.inj-band.on td{background:var(--accent-soft);color:var(--accent)}
.board tr.inj-band:hover td{background:rgba(140,180,200,.04)}
.board tr.inj-band.on:hover td{background:var(--accent-soft)}
/* Dimmed, not hidden. The row is still readable and still opens the card. */
.board tr.unpriced{opacity:.72}
/* The 3px position edge. It is the only place on this table the position hue
   appears at full strength, so a scan down the left tells you which positions
   the injuries are landing on without reading a single chip. */
.board td.pedge,.board th.pedge{width:3px;padding:0}
.board td.pedge{background-clip:content-box}
.board td.pedge.pos-QB{background:var(--pos-qb)} .board td.pedge.pos-RB{background:var(--pos-rb)}
.board td.pedge.pos-WR{background:var(--pos-wr)} .board td.pedge.pos-TE{background:var(--pos-te)}
.board td.pedge.pos-K{background:var(--pos-k)}   .board td.pedge.pos-DST{background:var(--pos-dst)}
.board tr.unpriced td.pedge{opacity:.5}
/* FRESHNESS, read off the layer's own staleAfterDays so the bar empties at
   exactly the moment the badge beside it starts saying "stale". */
/* ------------------------------------------------- ROSTER SHAPE (12a)
   His footnote is the whole rule: the count and its chip are ONE object, so
   "WR 7 TE 1" cannot be read as "WR, 7 TE". The bar says the same fact as a
   shape, which is the part you can read without counting. */
/* Its own class rather than `.sh-sub`, which is a flex ROW - it laid the count,
   the bar and the key out side by side and gave the bar zero width, so the
   whole shape was invisible while every computed value on it was correct. */
.rs-sum{margin-top:5px;font-size:var(--fs-sm);color:var(--text-secondary)}
.rs-count{display:flex;align-items:baseline;gap:var(--sp-3);flex-wrap:wrap}
.rs-big{font:400 44px/1 var(--font-display);font-variant-numeric:tabular-nums}
.rs-else{color:var(--text-muted);letter-spacing:.14em}
.rs-bar{display:flex;gap:2px;height:12px;margin-top:var(--sp-4);
  border-radius:3px;overflow:hidden}
.rs-bar i{display:block;height:100%}
/* 62% of the hue over the page ground: strong enough to name the position,
   quiet enough that six of them side by side stay a chart and not a flag. */
.rs-bar .pos-QB{background:color-mix(in srgb,var(--pos-qb) 62%,var(--bg-app))}
.rs-bar .pos-RB{background:color-mix(in srgb,var(--pos-rb) 62%,var(--bg-app))}
.rs-bar .pos-WR{background:color-mix(in srgb,var(--pos-wr) 62%,var(--bg-app))}
.rs-bar .pos-TE{background:color-mix(in srgb,var(--pos-te) 62%,var(--bg-app))}
.rs-bar .pos-K{background:color-mix(in srgb,var(--pos-k) 62%,var(--bg-app))}
.rs-bar .pos-DST{background:color-mix(in srgb,var(--pos-dst) 62%,var(--bg-app))}
.rs-key{display:flex;flex-wrap:wrap;gap:var(--sp-3) 26px;margin-top:var(--sp-3)}
.rs-item{display:inline-flex;align-items:center;gap:9px}
.rs-n{font:400 20px/1 var(--font-display);font-variant-numeric:tabular-nums}
.rs-n.pos-QB{color:var(--pos-qb)} .rs-n.pos-RB{color:var(--pos-rb)}
.rs-n.pos-WR{color:var(--pos-wr)} .rs-n.pos-TE{color:var(--pos-te)}
.rs-n.pos-K{color:var(--pos-k)}   .rs-n.pos-DST{color:var(--pos-dst)}

/* ---------------------------------------------------- WAIVER BRIEF (12b)
   The verdict is a band, the gain is a bar, and the drops sit beside the adds
   instead of below them where they scrolled off. The explanatory phrase that
   used to repeat on every row is the group's footnote now. */
.wb-band{display:flex;align-items:baseline;gap:var(--sp-4);margin-bottom:18px}
.wb-top{font:400 40px/1 var(--font-display);font-variant-numeric:tabular-nums;
  color:var(--accent);white-space:nowrap}
.wb-lead{font:400 20px/1.3 var(--font-display)}
.wb-lead b{font-weight:600}
.wb-sub{margin-top:5px;font:400 var(--fs-md)/1.5 var(--font-ui);color:var(--text-secondary)}
.wb-cols{display:grid;grid-template-columns:1.35fr 1fr;gap:18px;align-items:start}
.wb-panel{border:1px solid var(--border-subtle);border-radius:8px;overflow:hidden}
/* THE BAR IS FIXED SO THE NAME IS THE ONLY THING THAT FLEXES. This was
   `minmax(60px,1fr)`, a second flexible track, which makes grid hand the bar
   exactly as much of the leftover as the name: 124px each at 1400, and 70px
   each at 1000, where ten names were cut. Slack belongs to the subject. */
:root{--wb-cols:3px 40px minmax(0,1fr) 46px 70px 100px 64px 64px}
.wb-hd{display:grid;grid-template-columns:var(--wb-cols);gap:12px;
  padding:10px 14px 10px 0;border-bottom:1px solid rgba(140,180,200,.14);
  background:var(--bg-surface);font:400 9px/1 var(--font-num);
  letter-spacing:.16em;text-transform:uppercase;color:var(--text-muted)}
.wb-hd .r{text-align:right}
.wb-dhd{display:flex;align-items:baseline;justify-content:space-between;
  padding:10px 14px;border-bottom:1px solid rgba(140,180,200,.14);background:var(--bg-surface)}
.wb-list{list-style:none;padding:0}
.wb-row{display:grid;grid-template-columns:var(--wb-cols);gap:12px;align-items:center;
  padding:9px 14px 9px 0;border-bottom:1px solid var(--border-row)}
.wb-row:last-child{border-bottom:none}
.wb-row.drop{grid-template-columns:3px 40px minmax(0,1fr) 46px 70px 90px;gap:10px;
  padding:10px 14px 10px 0}
/* A STREAMER IS DIMMED, NOT LABELLED. Ten rows each carrying the word would be
   the same repetition the band was built to remove; the group says it once. */
.wb-row.stream{opacity:.66}
.wb-row .nm{font:600 14px/1.35 var(--font-ui);min-width:0;
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.wb-row .tm{font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}
.wb-g{text-align:right;font:400 15px/1.2 var(--font-display);
  font-variant-numeric:tabular-nums;color:var(--text-muted)}
.wb-g.up{color:var(--accent)} .wb-g.dn{color:var(--negative-ink);font-size:17px}
.wb-bar{display:flex;height:10px}
.wb-bar i{display:block;height:10px;border-radius:2px;background:var(--accent)}
.wb-row.stream .wb-bar i{background:var(--text-dim)}
.wb-bar.dn{justify-content:flex-end}
.wb-bar.dn i{background:var(--negative);margin-left:auto}
.wb-faab,.wb-tr{text-align:right;font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}
.wb-tr{text-align:right;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.wb-foot{padding:11px 14px;font:400 var(--fs-sm)/1.55 var(--font-ui);color:var(--text-muted)}
.wb-row .pedge{align-self:stretch;border-radius:2px}
.wb-row .pedge.pos-QB{background:var(--pos-qb)} .wb-row .pedge.pos-RB{background:var(--pos-rb)}
.wb-row .pedge.pos-WR{background:var(--pos-wr)} .wb-row .pedge.pos-TE{background:var(--pos-te)}
.wb-row .pedge.pos-K{background:var(--pos-k)}   .wb-row .pedge.pos-DST{background:var(--pos-dst)}
.wb-row.stream .pedge{opacity:.45}

/* THE BAR SURVIVES ONLY WHERE THE PANEL CAN AFFORD IT, and between 900 and
   1399 it cannot. This band is a laptop window that is not maximised, or a
   tablet in landscape, and it was the WORST width in the app - worse than any
   phone - because the brief still laid out all eight columns inside a panel
   that is barely half the viewport. At 1000 the name and the bar took 70px
   each and ten ordinary names were cut. Nothing here is a phone rule reused;
   the phone simply reached this conclusion first. From 1400 up the bar comes
   back at its fixed 100px, which the panel has the room for - measured, not
   chosen: 1300 still starves the names (wb 91, cc 88) and 1400 does not
   (wb 148, cc 134), so the boundary sits on the first width that clears. */
@media (max-width:1399px){
  :root{--wb-cols:3px 40px minmax(0,1fr) 46px 70px 64px}
  .wb-hd span:nth-child(6),.wb-hd span:nth-child(8),.wb-bar,.wb-tr{display:none}
  .wb-row.drop{grid-template-columns:3px 40px minmax(0,1fr) 46px 70px}
}

@media (max-width:900px){
  /* One column, adds first: the drops are a second opinion, not the answer. */
  .wb-cols{grid-template-columns:1fr}
  .wb-band{flex-wrap:wrap}
  /* The bar and the trend are already gone by here. What a phone needs on top
     of that is narrower chrome: at 402px the name computed to 40.5px against
     the 140px "Jaxon Smith-Njigba" needs, and every name on the brief was cut
     mid-word - too narrow for even the ellipsis to draw. The gain number that
     stands where the bar was says what the bar said. */
  :root{--wb-cols:3px 30px minmax(0,1fr) 30px 56px 46px}
  .wb-hd,.wb-row{gap:8px}
  .wb-row.drop{grid-template-columns:3px 30px minmax(0,1fr) 30px 56px;gap:8px}
}

/* -------------------------------------------------- SUNDAY LINEUP (12e)
   The projection reads as a bar per slot, and close calls sit against ONE
   shaded noise band rather than each carrying the verdict in words. The old
   brief said "coin flip - inside the noise, your call" seven times on a real
   roster; the band says it once, as geometry. */
.wb-cols.lu-cols{grid-template-columns:1.15fr 1fr}
.wb-cols.tr-cols{grid-template-columns:1fr 1fr}
/* 120px of bar against a name that flexed to 62px at 1000. Same rule: the
   drawing gets a fixed, modest track and the name takes the slack. */
:root{--lu-cols:3px 52px 40px minmax(0,1fr) 100px 64px 62px}
.lu-hd{display:grid;grid-template-columns:var(--lu-cols);gap:12px;
  padding:10px 14px 10px 0;border-bottom:1px solid rgba(140,180,200,.14);
  background:var(--bg-surface);font:400 9px/1 var(--font-num);
  letter-spacing:.16em;text-transform:uppercase;color:var(--text-muted)}
.lu-hd .r{text-align:right}
.lu-row{display:grid;grid-template-columns:var(--lu-cols);gap:12px;align-items:center;
  padding:9px 14px 9px 0;border-bottom:1px solid var(--border-row)}
.lu-row:last-child{border-bottom:none}
.lu-row .nm{font:600 14px/1.35 var(--font-ui);min-width:0;
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.lu-slot{font:var(--fs-xs)/1 var(--font-num);letter-spacing:.14em;color:var(--text-muted)}
.lu-bar{display:flex;height:8px}
.lu-bar i{display:block;height:8px;border-radius:2px}
.lu-bar .pos-QB{background:color-mix(in srgb,var(--pos-qb) 62%,var(--bg-app))}
.lu-bar .pos-RB{background:color-mix(in srgb,var(--pos-rb) 62%,var(--bg-app))}
.lu-bar .pos-WR{background:color-mix(in srgb,var(--pos-wr) 62%,var(--bg-app))}
.lu-bar .pos-TE{background:color-mix(in srgb,var(--pos-te) 62%,var(--bg-app))}
.lu-bar .pos-K{background:color-mix(in srgb,var(--pos-k) 62%,var(--bg-app))}
.lu-bar .pos-DST{background:color-mix(in srgb,var(--pos-dst) 62%,var(--bg-app))}
.lu-opp{text-align:right;font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}
.lu-pts{text-align:right;font:400 16px/1.2 var(--font-display);
  font-variant-numeric:tabular-nums}
.lu-row .pedge{align-self:stretch;border-radius:2px}
.lu-row .pedge.pos-QB{background:var(--pos-qb)} .lu-row .pedge.pos-RB{background:var(--pos-rb)}
.lu-row .pedge.pos-WR{background:var(--pos-wr)} .lu-row .pedge.pos-TE{background:var(--pos-te)}
.lu-row .pedge.pos-K{background:var(--pos-k)}   .lu-row .pedge.pos-DST{background:var(--pos-dst)}

/* THE BAND, drawn ONCE across the whole panel. `.cc-grid` is an overlay that
   shares the row grid, so the strip lands in the bar column at exactly the
   noise threshold and every row is measured against the same line. */
.cc-wrap{position:relative;padding:10px 14px 0}
/* The bar was the second `1fr` here too, so it matched the name at 127px each
   at 1400. Fixed, and the band it carries is unaffected: the overlay reads the
   same track list, so the shading still lands exactly on this column. */
:root{--cc-cols:minmax(0,1fr) 56px 120px 132px}
.cc-grid{position:absolute;inset:10px 14px 0;display:grid;
  grid-template-columns:var(--cc-cols);gap:14px;pointer-events:none}
.cc-track{position:relative}
.cc-noise{position:absolute;top:0;bottom:0;left:0;background:rgba(224,180,107,.08);
  border-right:1px dashed rgba(224,180,107,.4)}
.cc-row{display:grid;grid-template-columns:var(--cc-cols);gap:14px;align-items:center;
  padding:9px 0;border-bottom:1px solid var(--border-row)}
.cc-row:last-child{border-bottom:none}
.cc-row .nm{font:600 var(--fs-lg)/1.35 var(--font-ui);min-width:0;
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.cc-m{text-align:right;font:400 15px/1.2 var(--font-display);
  font-variant-numeric:tabular-nums;color:var(--accent)}
.cc-bar{display:flex;min-width:0}
.cc-bar i{display:block;height:8px;border-radius:2px;background:var(--accent)}
/* Inside the band: amber, and HATCHED - a second channel, so the call does not
   rest on a colour that a shaded strip is already carrying. */
.cc-row.flip .cc-m{color:var(--risk)}
.cc-row.flip .cc-bar i{background:repeating-linear-gradient(115deg,
  var(--risk) 0,var(--risk) 3px,rgba(224,180,107,.35) 3px,rgba(224,180,107,.35) 7px)}
.cc-over{font:400 var(--fs-sm)/1.35 var(--font-ui);color:var(--text-secondary);
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}

/* ------------------------------------------------------- TRADE LAB (12f)
   Filled rather than empty: a side is a list of players carrying their own
   points, because the question is "is this fair" and a name with no number
   cannot answer it. */
.tr-hd{margin-bottom:11px}
.tr-side{padding:14px}
.tr-list{display:flex;flex-direction:column;gap:8px}
.tr-row{display:grid;grid-template-columns:34px minmax(0,1fr) 42px 52px 20px;
  gap:10px;align-items:center;padding:10px 12px;border-radius:6px;
  border:1px solid var(--border-subtle)}
.tr-row.pos-QB{background:color-mix(in srgb,var(--pos-qb) 9%,var(--bg-app))}
.tr-row.pos-RB{background:color-mix(in srgb,var(--pos-rb) 9%,var(--bg-app))}
.tr-row.pos-WR{background:color-mix(in srgb,var(--pos-wr) 9%,var(--bg-app))}
.tr-row.pos-TE{background:color-mix(in srgb,var(--pos-te) 9%,var(--bg-app))}
.tr-row.pos-K{background:color-mix(in srgb,var(--pos-k) 9%,var(--bg-app))}
.tr-row.pos-DST{background:color-mix(in srgb,var(--pos-dst) 9%,var(--bg-app))}
.tr-row .nm{font:600 var(--fs-lg)/1.35 var(--font-ui);min-width:0;
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.tr-p,.tr-r{text-align:right;font:var(--fs-sm)/1 var(--font-num);
  font-variant-numeric:tabular-nums}
.tr-r{color:var(--text-muted)}
.tr-side .trade-in{margin-top:8px;width:100%}

/* THE WEEKS THE AVERAGE HIDES. A diverging column per remaining week against a
   zero line: "gains 2.9 a week" cannot quietly contain a week it loses. */
.tw-panel{margin-top:18px}
.tw-chart{display:flex;gap:6px;padding:16px 14px 12px}
.tw-wk{flex:1;display:flex;flex-direction:column;align-items:center;min-width:0}
.tw-up{width:100%;height:46px;display:flex;flex-direction:column;justify-content:flex-end}
.tw-dn{width:100%;height:46px;display:flex;flex-direction:column;justify-content:flex-start}
.tw-up i,.tw-dn i{display:block;width:100%;background:var(--accent)}
.tw-up i{border-radius:3px 3px 0 0}
.tw-dn i{border-radius:0 0 3px 3px;background:var(--negative)}
.tw-zero{width:100%;height:1px;background:rgba(140,180,200,.22)}
.tw-n{margin-top:6px;font:var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}
.tw-wk.dn .tw-n{color:var(--negative-ink)}

/* The same band, for the same reason as the brief above. At 1000 the lineup
   name flexed to 62px while its bar held a fixed 120 - the drawing was twice
   the subject - and the close-call names took 83px each beside an 83px bar. */
@media (max-width:1399px){
  :root{--lu-cols:3px 52px 40px minmax(0,1fr) 64px 62px}
  .lu-hd span:nth-child(5),.lu-row>*:nth-child(5){display:none}

  /* The close-call bar STAYS at every width - the shaded band is drawn down
     that column and a row ending inside it IS the coin flip - so this row
     wraps instead: "over <name>" takes its own full-width line and returns its
     fixed 132px to the names. The overlay's trailing spacer goes with it, or
     the band would be sized by the first of two implicit rows and stop short
     of the row's full height. */
  :root{--cc-cols:minmax(0,1fr) 56px 120px}
  .cc-grid>span:nth-child(4){display:none}
  .cc-over{grid-column:1/-1}
}

@media (max-width:900px){
  .wb-cols.lu-cols,.wb-cols.tr-cols{grid-template-columns:1fr}
  /* THE NUMBER SURVIVES, NOT THE BAR. This view used to hide Pts on a phone
     and keep the projection bar, which is the precise column yielding to the
     drawn one. Inverted above: the bar goes and the number stays, and it costs
     less width than the thing it replaces. The bar is hidden by POSITION
     rather than by class, because an unfilled slot renders a bare <span> in
     that cell and not a .lu-bar - hiding the class alone left empty rows one
     cell longer than the template, in an implicit seventh column. */
  :root{--lu-cols:3px 36px 30px minmax(0,1fr) 46px 52px}
  .lu-hd,.lu-row{gap:8px}

  :root{--cc-cols:minmax(0,1fr) 46px 96px}
  .cc-grid,.cc-row{gap:10px}
}

/* THE SMALL-PHONE CLASS - 375 (SE, mini) through 393 (15/16), against the 402
   of a 16 Pro. The templates above clear every ordinary name at 402 and come
   up ~30px short at 375, so one more column yields: the team code, exactly as
   the phone draft cell already drops it. The name is a button and the card it
   opens states the team, so nothing here is the only place that fact lives.
   THIS BLOCK MUST FOLLOW EVERY 900px BLOCK ABOVE. Both match at 375 and both
   set the same custom properties at the same specificity, so source order is
   the only thing deciding them: sitting above the lineup block, its --lu-cols
   was simply overwritten by the wider phone value and the rules here read as
   if they had never been written. */
@media (max-width:400px){
  :root{--wb-cols:3px 30px minmax(0,1fr) 56px 46px}
  .wb-hd span:nth-child(4),.wb-row .tm{display:none}
  .wb-row.drop{grid-template-columns:3px 30px minmax(0,1fr) 56px}
  :root{--lu-cols:3px 30px 30px minmax(0,1fr) 44px 50px}
  .lu-hd,.lu-row{gap:6px}
}

/* ------------------------------------------------------ BYE TIMELINE (12d)
   The season as one axis, so a CLEAR week is visible instead of missing. The
   old grid drew only the weeks that hurt, which meant the most useful cell on
   the chart - the week where a stack is safe - could only be inferred from a
   gap in the numbering. */
.by-line{display:flex;align-items:flex-end;gap:5px;height:96px;margin:22px 0 6px}
.by-wk{flex:1;display:flex;flex-direction:column;justify-content:flex-end;
  align-items:center;gap:6px;min-width:0}
.by-wk i{display:block;width:100%;border-radius:3px 3px 0 0;
  background:rgba(22,131,196,.6)}
.by-n{font:var(--fs-sm)/1 var(--font-num);color:var(--text-secondary)}
.by-w{font:var(--fs-xs)/1 var(--font-num);letter-spacing:.1em;color:var(--text-secondary)}
/* A clear week is a stub and a dim number: present, and visibly nothing. */
.by-wk.clear i{background:var(--border-subtle)}
.by-wk.clear .by-w{color:#2f3f47}
/* Amber is about STARTERS, not bodies. Two bench players out is not the same
   fact as one starter out, and the old panel counted only bodies. */
.by-wk.clash i{background:var(--risk)}
.by-wk.clash .by-w{color:var(--risk)}

/* The survival sentence, with the room selector inside it. A caps label that
   contains a control needs the control to sit ON the baseline, not above it. */
.survline{display:inline-flex;align-items:center;gap:6px;text-transform:none;
  letter-spacing:normal;color:var(--text-muted)}
.survline select{background:var(--bg-surface-raised);color:var(--text-secondary);
  border:1px solid var(--border-subtle);border-radius:var(--r-sm);
  font:var(--fs-xs)/1 var(--font-num);letter-spacing:.06em;padding:3px 6px}
.survline select:hover{border-color:var(--border-strong);color:var(--text-primary)}
.survline select:focus-visible{outline:2px solid var(--accent);outline-offset:1px}

.fresh{display:inline-block;min-width:64px;text-align:right}
.fr-n{display:block;font-variant-numeric:tabular-nums}
.fr-t{display:block;margin-top:6px;height:3px;border-radius:2px;
  background:var(--track);overflow:hidden}
.fr-t i{display:block;height:100%;border-radius:2px;background:var(--accent)}
.empty{color:var(--text-muted);font-size:var(--fs-sm);padding:10px var(--sp-4)}

/* ============================================================ SEASON */
.season-head{display:flex;align-items:center;gap:var(--sp-3);margin-top:var(--sp-4);
  background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3) var(--sp-4)}
.sh-main{flex:1;min-width:0}
.sh-sub{margin-top:5px;font-size:var(--fs-sm);color:var(--text-secondary);
  display:flex;align-items:center;gap:5px;flex-wrap:wrap}
.sh-sub .pos{margin-right:2px}
.rosterbox{background:var(--bg-surface);border:1px solid var(--border-strong);
  border-radius:var(--r-md);padding:var(--sp-4);margin-top:var(--sp-3)}
.rosterbox .caps{display:block;margin-bottom:var(--sp-2)}
.rosterbox textarea{width:100%;background:var(--bg-surface-raised);
  border:1px solid var(--border-subtle);border-radius:var(--r-md);
  color:var(--text-primary);font:var(--fs-sm)/1.6 var(--font-ui);
  padding:10px;resize:vertical;margin-bottom:var(--sp-3)}
.rosterbox textarea:focus{outline:none;border-color:var(--accent)}
/* 16px IS A THRESHOLD, NOT A TYPE CHOICE. iOS Safari force-zooms any focused
   field whose text is under 16px, and the viewport meta sets no maximum-scale,
   so tapping this 11px textarea zoomed the whole document and LEFT it zoomed
   after blur: every label, the save button and the status line then ran off
   the right edge. That reads as cut-off text with no layout defect behind it.
   Emulated Chrome at 402px cannot reproduce this - only real WebKit does. */
@media (max-width:760px){
  .rosterbox textarea{font-size:16px}
}
.rb-actions{display:flex;align-items:center;gap:var(--sp-2)}
.rb-unmatched{list-style:none;margin-top:var(--sp-3);font-size:var(--fs-sm);color:var(--negative-ink)}
.rb-unmatched .caps{color:var(--text-muted);margin-bottom:var(--sp-1)}
.season-grid{display:grid;grid-template-columns:1fr 1fr;gap:var(--sp-4);margin-top:var(--sp-4)}

/* waiver radar */
.waiver-tabs{display:flex;gap:var(--sp-2);padding:var(--sp-3) var(--sp-4) 0}
.wv-list{list-style:none;padding:0}
/* HIS 12c. Adds moves to its own right-aligned column beside the position, the
   velocity gets a bar, and Status only speaks when the row is NOT the default. */
/* 150px of velocity against a 112px name, in a panel that is only half the
   season grid: the widest name needed 119px and four were clipped at 1400.
   The bar reads the same at 100 - it is a proportion, not a measurement. */
:root{--wv-radar:34px 74px minmax(0,1fr) 100px 46px 62px}
.wv-hd2{display:grid;grid-template-columns:var(--wv-radar);gap:12px;
  padding:10px var(--sp-4);border-bottom:1px solid rgba(140,180,200,.14);
  background:var(--bg-surface);font:400 9px/1 var(--font-num);
  letter-spacing:.16em;text-transform:uppercase;color:var(--text-muted)}
.wv-hd2 .r{text-align:right}
.wv{display:grid;grid-template-columns:var(--wv-radar);gap:12px;
  align-items:center;padding:8px var(--sp-4);border-bottom:1px solid var(--border-row);
  font-size:var(--fs-sm)}
.wv:last-child{border-bottom:none}
.wv:hover{background:var(--bg-row-hover)}
.wv.off{opacity:.5}
/* The runaway outlier: kept in place, dimmed, and its bar goes grey to say it
   is off this chart's scale rather than the top of it. */
.wv.far{opacity:.72;background:var(--row-tint)}
.wv-vel{display:flex;height:9px}
.wv-vel i{display:block;height:9px;border-radius:2px}
.wv-vel .pos-QB{background:color-mix(in srgb,var(--pos-qb) 70%,var(--bg-app))}
.wv-vel .pos-RB{background:color-mix(in srgb,var(--pos-rb) 70%,var(--bg-app))}
.wv-vel .pos-WR{background:color-mix(in srgb,var(--pos-wr) 70%,var(--bg-app))}
.wv-vel .pos-TE{background:color-mix(in srgb,var(--pos-te) 70%,var(--bg-app))}
.wv-vel .pos-K{background:color-mix(in srgb,var(--pos-k) 70%,var(--bg-app))}
.wv-vel .pos-DST{background:color-mix(in srgb,var(--pos-dst) 70%,var(--bg-app))}
.wv.far .wv-vel i{background:var(--text-muted)}
.wv-n{font:var(--fs-lg)/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:right;color:var(--accent)}
/* This panel is HALF the season grid until 1000px, so it runs out of room
   earlier than anything else on the view: at 1200 the name resolved to 62px
   even with the bar trimmed to 100. The velocity goes for the whole band, and
   the count it was scaled from is still in the column beside it. */
@media (max-width:1399px){
  :root{--wv-radar:34px 74px minmax(0,1fr) 46px 62px}
  .wv-hd2 span:nth-child(4),.wv-vel{display:none}
}
@media (max-width:760px){
  /* The status track is the only fixed column here that is EMPTY on most rows
     by design - the badge speaks only when the row is not free-and-ranked - so
     56px plus a 12px gutter was reserved on every row for a signal 22 of 22
     rows did not carry. It cannot simply collapse: each row is its own grid,
     so an auto track would let RANK jitter between rows that have a badge and
     rows that do not. Tightened instead, gutters included, which is enough
     because the names were missing by 3 to 19px, not by a column. */
  :root{--wv-radar:30px 58px minmax(0,1fr) 30px 54px}
  .wv-hd2,.wv{gap:8px}
}
.wv-nm{font-weight:600;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.wv-nm .team{font-size:var(--fs-xs);color:var(--text-muted);font-weight:400;margin-left:5px}
.wv-rk{font-size:var(--fs-xs);color:var(--text-muted);text-align:right}
.wv-tag{font:700 9px/1 var(--font-num);letter-spacing:.06em;padding:3px 5px;
  border-radius:var(--r-sm);border:1px solid var(--border-strong);
  color:var(--text-muted);text-align:center}
.wv-tag.free{border-color:var(--accent-border);color:var(--accent);background:var(--accent-soft)}
.wv-tag.mine{border-color:var(--border-strong);color:var(--text-secondary)}
.wv-note{padding:var(--sp-3) var(--sp-4);border-top:1px solid var(--border-subtle);
  font-size:var(--fs-xs);color:var(--text-muted);line-height:1.6}
.wv-note b{color:var(--text-secondary);font-weight:600}

/* bye matrix - the timeline lives with the 12d block above; the card grid
   it replaced is gone rather than left behind as dead rules. */
.bye-note{padding:var(--sp-3) var(--sp-4);border-top:1px solid var(--border-subtle);
  font-size:var(--fs-xs);color:var(--text-muted);line-height:1.6}
.bye-note b{color:var(--text-primary);font-weight:600}

/* live: my players inside a game card */
.game.has-mine{border-color:var(--accent-border)}
.game-mine{margin-top:var(--sp-2);padding-top:var(--sp-2);border-top:1px solid var(--border-subtle)}
.game-mine .caps{display:block;margin-bottom:5px}
.gm-row{display:flex;align-items:center;gap:6px;font-size:var(--fs-sm);padding:2px 0}
.gm-st{margin-left:auto;font-size:var(--fs-xs);color:var(--text-secondary);white-space:nowrap}

@media (max-width:1000px){ .season-grid{grid-template-columns:1fr} }

/* ============================================================ LIVE */
/* ================================================= THE GAME DAY SKIN =====
   HIS WHIPAROUND MOCK, design/live-whiparound-1b.html, executed as drawn.
   A THIRD NAMED SKIN, not a restyle, and that is what the mock actually is
   once its values are read rather than glanced at. The app's own palette is a
   blue graphite - #0b0d14 ground, #eef6fa ink, borders in rgba(140,180,200,x)
   - and every colour in the mock is neutral or warm: #050506 ground, #f4f5f7
   ink, #1e2025 borders, and a broadcast red the app has nowhere else. That is
   not a page wearing the house palette differently. It is a different palette,
   the way `.lions` is and the way `.ws-sheet` is, and this file already has
   the mechanism for exactly that: a skin class retunes the tokens locally and
   every rule underneath keeps reading the same token names. `.ws-sheet` says
   so in its own comment ("the sheet RETUNES the token locally, the same move
   the rail makes with the key tokens").
   So nothing below this block is rewritten to carry a literal. The scoreboard,
   the feed and the row go on reading --bg-card and --text-muted; they simply
   resolve to the mock's values inside the live view. That is what makes the
   port reversible, and what stops the mock's red leaking into the six other
   surfaces that share those rules.
   THE RED IS ITS OWN TOKEN AND NOT --accent. --accent is the Lions blue and
   it means "this is the lit state" across the whole app; a red that also meant
   that would be a second answer to one question. --gd-red is a BRAND surface
   (the masthead, the yours-tile, the FOLLOWING chip) and --gd-live is the
   liveness signal (the pulse dot, the running clock). Two reds, two jobs,
   because the mock uses them for two jobs. */
/* THE SKIN CARRIES ITS OWN GROUND, exactly as `.view.lions` does. Retuning the
   tokens alone left `#panel-live` with no background rule at all - the ground
   it had been standing on belonged to `.view.lions`, and dropping that class
   dropped the paint with it. The browser probe read the panel back as
   `rgba(0, 0, 0, 0)`, which is the honest description of a surface that
   defines a colour and never applies it. */
.view.gameday{background:var(--bg-app);color:var(--text-primary);
  min-height:calc(100vh - 52px)}
.gameday{
  --bg-app:#050506;
  --bg-card:#0a0a0c;
  --bg-surface:#111216;
  --bg-surface-raised:#111216;
  --bg-panel:#0e0f12;
  --bg-row:transparent;
  --bg-row-hover:#16171b;
  --border-subtle:#1e2025;
  --border-row:#17181c;
  --border-strong:#26282e;
  --text-primary:#f4f5f7;
  --text-secondary:#dfe2e6;
  --text-muted:#8b929e;
  --text-dim:#4a505b;
  /* The mock's own two reds, named for what they do rather than for what
     they are. #d50a0a is the masthead and the ownership tile; #ff2e3e is the
     LIVE dot and the running clock, and it is the brighter of the two
     because it is always small type on a dark ground. */
  --gd-red:#d50a0a;
  --gd-live:#ff2e3e;
  --gd-tile:#16171b;
  /* THE POSTER FACE IS THE HEADLINE FACE HERE. The lions skin already scopes
     Big Shoulders to --font-poster; the mock sets every headline, team code
     and score in it, so the skin points --font-display at it too and the row
     headline rule below needs no change at all. */
  --font-poster:"Big Shoulders Display","Arial Narrow",sans-serif;
  --font-display:"Big Shoulders Display","Arial Narrow",sans-serif;
}
/* ---------------------------------------------------- THE PLATE (masthead)
   The mock's red bar. `.masthead` keeps its three children because the
   Start-here cover is BUILT from them (introDests reads .mh-title, .mh-kicker
   and .mh-dek and then filters to panels that have a dek), so the title and
   the dek are hidden with the clip idiom rather than removed - a screen
   reader still reads the sentence, the cover still finds it, and the plate
   draws what the mock draws. */
.gameday .gd-plate{position:relative;display:flex;align-items:stretch;
  border-radius:var(--r-md) var(--r-md) 0 0;margin-bottom:0;padding:0;
  background:linear-gradient(105deg,#8f0505,#d50a0a 55%,#a10707);border:0}
.gameday .gd-plate .mh-title,
.gameday .gd-plate .mh-dek{position:absolute;width:1px;height:1px;padding:0;
  margin:-1px;overflow:hidden;clip:rect(0 0 0 0);white-space:nowrap;border:0}
/* THE BIG WORD IS THE KICKER, which already says "Game day". Archivo Black
   over a colour field, obliqued - the idiom this app already owns and the
   only reason that face is in the bundle. */
.gameday .gd-plate .mh-kicker{display:block;padding:20px 26px 16px;
  font:400 clamp(28px,4.4vw,40px)/1 "Archivo Black",sans-serif;color:#fff;
  letter-spacing:-.01em;text-transform:uppercase;transform:skewX(-9deg);
  transform-origin:left bottom}
.gameday .gd-ears{margin-left:auto;display:flex;flex-direction:column;
  justify-content:center;padding:0 26px;text-align:right;gap:4px}
.gameday .gd-ears .caps{font:700 11px/1.8 var(--font-num);letter-spacing:.2em;
  color:rgba(255,255,255,.85)}
.gameday .gd-ears .caps+.caps{color:rgba(255,255,255,.6)}

/* ------------------------------------------------------- THE BOARD (scores)
   Four compact cards across the top. `auto-fit` and not the mock's literal
   `repeat(4,1fr)`: the mock is drawn at 880px and says nothing about a phone,
   and a fixed four-track grid at 375px is four 80px cards - which is not the
   design at a smaller size, it is the design broken. minmax keeps the mock's
   four-up wherever four fit and reflows below that, which is also what keeps
   the 44px tap-target floor reachable on the two phone widths the suite
   measures. */
.gameday .gd-board{grid-template-columns:repeat(auto-fit,minmax(168px,1fr));
  gap:10px;margin-top:var(--sp-3)}
/* THE RED RULE IS THE THEME AND ITS BRIGHTNESS IS THE SIGNAL. His call,
   2026-08-21: "I like the red top. It's redzone themed."
   The first cut put full #d50a0a on every card, which on a preseason Thursday
   drew fourteen SCHEDULED games wearing the exact colour that means "this is
   happening now" - red everywhere is red meaning nothing. The second cut took
   red off everything but a live game, which killed the theme to save the
   signal. Neither trade was necessary, because the two jobs live on different
   axes: the HUE carries the theme and the LUMINANCE carries the state.
   So the red drains as a game runs its course. A scheduled game wears the
   deep maroon that is the mock's own gradient end - banked, anticipatory, the
   board is red before kickoff. A live game burns at full #d50a0a beside a lit
   dot. A final has spent it and goes to the neutral rule the mock draws.
   Three states, one colour family, and the brightest thing on the board is
   always the game that is actually moving. */
.gameday .gd-board .game{background:var(--bg-surface);border:0;
  border-top:3px solid #8f0505;border-radius:4px;padding:12px 14px}
.gameday .gd-board .game.is-live{border-top-color:var(--gd-red)}
/* A FINAL IS A DIFFERENT KIND OF CARD, not a dimmer one: its own ground, its
   own rule colour, and its losing side muted. The mock draws the distinction
   twice over because "this game is still moving" is the whole question a
   scoreboard answers at a glance. */
.gameday .gd-board .game.is-final{background:var(--bg-panel);
  border-top-color:var(--border-strong);color:var(--text-muted)}
/* THE LIVE CARD IS THE ONE THING ON THE BOARD THAT MOVES, and the dot beside
   its label is the only animation in this skin. Held to the app's existing
   reduced-motion rule rather than given a new one. */
.gameday .gd-board .game.is-live .game-status .live::before{
  animation:gd-pulse 2s ease-in-out infinite}
@keyframes gd-pulse{0%,100%{opacity:1}50%{opacity:.35}}
@media (prefers-reduced-motion:reduce){
  .gameday .gd-board .game.is-live .game-status .live::before{animation:none}
}
.gameday .gd-board .game-status{margin-bottom:10px;align-items:center}
.gameday .gd-board .game-status .live{color:var(--gd-live);
  font:700 10px/1 var(--font-num)}
.gameday .gd-board .game-status .live::before{background:var(--gd-live)}
.gameday .gd-board .game-status .fin,
.gameday .gd-board .game-status .sched{font:700 10px/1 var(--font-num)}
.gameday .gd-board .game-status .qtr{font:700 10px/1 var(--font-num);
  color:var(--text-muted)}
.gameday .gd-board .game-row{padding:0;margin-top:4px;align-items:baseline}
.gameday .gd-board .game-row:first-of-type{margin-top:10px}
.gameday .gd-board .game-row .tm{font:700 22px/1 var(--font-poster);
  letter-spacing:.01em}
.gameday .gd-board .game-row .sc{font:700 28px/1 var(--font-poster)}
.gameday .gd-board .game-row.losing .tm,
.gameday .gd-board .game-row.losing .sc{color:var(--text-muted)}
/* The record and the my-players block are not in the mock's card and would
   double its height. They stay in the DOM (the data is real and the card is
   the same renderer the wide layout uses) and stand down on this skin. */
.gameday .gd-board .game-row .rec{display:none}
/* The broadcast-and-venue note is not on the mock's card and it is half the
   card's height. It stays in the DOM - the data is real and every other
   layout still prints it - and stands down here, because a board whose whole
   job is four scores read at a glance cannot spend two lines per card on the
   stadium name. */
.gameday .gd-board .game-note{display:none}

/* ------------------------------------------------------- THE FOLLOWING BAR */
.gameday .gd-follow{display:flex;align-items:center;gap:14px;flex-wrap:wrap;
  margin-top:var(--sp-3);background:var(--bg-surface);
  border:1px solid var(--border-subtle);border-radius:4px;padding:10px 16px}
.gameday .gd-follow .gd-chip{background:var(--gd-red);color:#fff;
  font:700 10px/1 var(--font-num);letter-spacing:.16em;padding:4px 8px;
  border-radius:2px}
.gameday .gd-follow .gd-game{font:700 18px/1 var(--font-poster);
  letter-spacing:.04em;color:var(--text-primary)}
.gameday .gd-follow .gd-clock{font:500 12px/1 var(--font-num);color:var(--gd-live)}
.gameday .gd-follow .gd-you{margin-left:auto;font:500 11px/1 var(--font-num);
  color:var(--text-muted)}

.live-head{display:flex;align-items:center;gap:var(--sp-3);margin-top:var(--sp-4)}
/* ============================== THE WIRE =================================
   HIS BROADSHEET MOCK, 2026-08-14, executed as drawn: a newspaper front
   page on one white sheet - ears, Newsreader nameplate, double rule, desk
   index, lead story with its figure band, ruled desk columns, the trending
   index, the market in brief - and the Ledger sheet below with everything,
   graded. Tokens only; the hexes in the handoff are the tokens' own values,
   for verification, never for use. */
.ws-wrap{max-width:1240px}
/* The masthead the intro reads, never printed - the nameplate is the page's
   own masthead. display:none still answers textContent, which is all
   introDests asks of it. */
.ws-unprinted{display:none}
/* The rethink arrow on a filled draft cell: quiet until the cell is hovered,
   always present on touch (no hover to reveal it with). */
/* `align-self:center` IS A ROW-HEIGHT DECLARATION. `.dg-top` aligns on the
   BASELINE, and a 13px glyph inside 10px of padding sits its baseline 10px
   down - far enough below the name's that the line box came out 26px against
   the name band's own 24px. Every row holding a mock pick was therefore 43.5px
   against 41.5px for a row of pure forecast, which is the round-1 defect again
   at a twentieth of the size and was invisible underneath it. Centred, the
   button takes the band's height instead of setting it. Measured after: 41.5px
   on every round of a 47-pick mock board at 375. */
/* AND IT IS THE LAST THING TAKING WIDTH FROM THE NAME. With the +/- mark moved
   off this row, the mock undo is the only item left sharing it, and measured at
   1280 it was costing 21.6px of a 90px box - 13px of glyph, 16px of padding,
   less 8px of negative margin - which put eight of seventy-two mock picks back
   into an ellipsis ("Bi. Robinson" needs 70px in 68.4). 12px inside 10px of
   padding, pulled 7px right, is 15px: the name gets 75px and only the two
   double-barrelled surnames the pool's own fit contract already documents as
   its tail stay over. The tap target is unchanged - the 10px of vertical
   padding is what carries it and none of that is touched. */
.dg-swap{background:none;border:none;cursor:pointer;align-self:center;
  padding:10px 5px;margin:-10px -7px -10px auto;
  font-size:12px;line-height:1;color:var(--text-muted);
  opacity:0;transition:opacity .12s}
.dg-cell:hover .dg-swap,.dg-swap:focus-visible{opacity:1}
.dg-swap:hover{color:var(--accent)}
@media (hover:none){.dg-swap{opacity:.55}}
.ws-sheet{background:var(--bg-surface);border-radius:var(--r-lg);
  box-shadow:0 3px 20px rgba(15,26,34,.16);padding:44px 52px 40px;
  color:var(--text-primary);font:300 15px/1.6 var(--font-display);
  /* The lions skin drops --font-display to the sans (line ~700); the
     broadsheet deliberately extends Newsreader into the lit view - the
     handoff's own note - so the sheet RETUNES the token locally, the same
     move the rail makes with the key tokens. */
  --font-display:"Newsreader",Georgia,"Times New Roman",serif}
.ws-second{margin-top:40px;padding:40px 52px 36px}
.ws-ears{display:flex;justify-content:space-between;gap:20px;
  padding-bottom:10px;border-bottom:1px solid var(--text-primary);
  font:400 10.5px/1 var(--font-num);letter-spacing:.14em;
  text-transform:uppercase;color:var(--text-secondary);flex-wrap:wrap}
.ws-nameplate{text-align:center;padding:30px 0 24px}
.ws-nameplate h1{margin:0;font:500 clamp(48px,7.4vw,92px)/.95 var(--font-display);
  letter-spacing:-.015em}
.ws-tagline{margin-top:14px;font:500 10.5px/1 var(--font-num);
  letter-spacing:.22em;text-transform:uppercase;color:var(--text-muted)}
.ws-rule2 i{display:block;height:3px;background:var(--text-primary)}
.ws-rule2 i+i{height:1px;margin-top:3px}
/* The desk index IS the filter set, re-skinned: counts in dim ink, the
   active desk's count in the accent. Thumb-sized targets on every width. */
.ws-desks{display:flex;justify-content:center;align-items:center;
  padding:2px 0;border-bottom:1px solid var(--border-strong);flex-wrap:wrap}
.ws-desk{background:none;border:none;cursor:pointer;min-height:44px;
  padding:0 18px;border-left:1px solid var(--border-strong);
  font:500 10.5px/1 var(--font-num);letter-spacing:.14em;
  text-transform:uppercase;color:var(--text-primary)}
.ws-desk:first-child{border-left:none}
.ws-desk b{font-weight:400;color:var(--text-dim)}
.ws-desk.active b{color:var(--accent);font-weight:500}
.ws-refresh{margin-left:10px}
/* ---- the front page ---- */
.ws-front{display:grid;
  grid-template-columns:minmax(0,2.05fr) minmax(0,1fr) minmax(0,1fr);
  margin-top:28px}
.ws-lead{padding-right:30px;min-width:0}
.ws-col{border-left:1px solid var(--border-strong);padding:0 26px;min-width:0}
.ws-col:last-child{padding-right:0}
.ws-kicker{font:500 10.5px/1 var(--font-num);letter-spacing:.18em;
  text-transform:uppercase;color:var(--accent)}
.ws-lead h2{margin:14px 0 0;font:500 clamp(28px,3.6vw,46px)/1.04 var(--font-display);
  letter-spacing:-.01em;text-wrap:balance}
.ws-headline{color:inherit;text-decoration:none}
a.ws-headline:hover{text-decoration:underline}
.ws-standfirst{margin:16px 0 0;font:300 18px/1.6 var(--font-display);
  color:var(--text-secondary);text-wrap:pretty}
.ws-lead-player{display:flex;align-items:center;gap:10px;margin:18px 0 0;
  font:400 11px/1 var(--font-num);letter-spacing:.1em;
  text-transform:uppercase;color:var(--text-muted)}
.ws-lead-player .pname{font-weight:500;color:var(--text-primary)}
.ws-figband{display:flex;align-items:center;gap:26px;margin:22px 0 0;
  padding:20px 0;border-top:1px solid var(--text-primary);
  border-bottom:1px solid var(--border-row)}
.ws-fig{font:500 58px/1 var(--font-num);color:var(--accent);
  letter-spacing:-.03em}
.ws-fig-word{font-size:clamp(26px,3vw,40px)}
.ws-figcap{margin-top:8px;font:500 9.5px/1 var(--font-num);
  letter-spacing:.16em;text-transform:uppercase;color:var(--text-muted)}
.lions .ws-figdiv{width:1px;align-self:stretch;background:var(--bg-sunken)}
.ws-facts{font:400 13px/1.9 var(--font-num);color:var(--text-secondary)}
.ws-facts b{color:var(--text-primary);font-weight:400}
.ws-bodytext{margin:18px 0 0;columns:2;column-gap:28px;
  column-rule:1px solid var(--border-row);font:300 15px/1.75 var(--font-display);
  color:var(--text-primary);text-align:justify;text-wrap:pretty}
.ws-crossref{margin-top:16px;font:400 11px/1.6 var(--font-num);
  color:var(--text-muted)}
/* ---- desk columns ---- */
.ws-deskhd{display:flex;justify-content:space-between;gap:8px;
  padding-bottom:10px;border-bottom:2px solid var(--text-primary);
  font:500 11px/1 var(--font-num);letter-spacing:.18em;text-transform:uppercase}
.ws-deskn{font-weight:400;font-size:10.5px}
.ws-deskn.neg{color:var(--negative)}
.ws-deskn.warn{color:var(--warning)}
.ws-item{padding:16px 0;border-bottom:1px solid var(--border-row)}
.ws-topline{display:flex;align-items:baseline;gap:8px;
  font:400 10.5px/1 var(--font-num);letter-spacing:.08em;
  text-transform:uppercase;color:var(--text-muted)}
.ws-topline .pname,.ws-trow .pname{font:inherit;font-weight:500;
  color:var(--text-primary);background:none;border:none;padding:0;
  cursor:pointer;text-transform:inherit;letter-spacing:inherit}
.ws-topline .pname:hover,.ws-trow .pname:hover{color:var(--accent)}
.ws-sp{flex:1}
.ws-item-day{color:var(--text-dim);text-transform:none}
.ws-item-fig{font:500 15px/1 var(--font-num)}
.ws-item-fig.up{color:var(--adp-up)}
.ws-item-fig.dn{color:var(--adp-dn)}
.ws-item-fig.neg{color:var(--negative)}
.ws-item-hl{margin-top:8px;font:500 19px/1.25 var(--font-display);
  letter-spacing:-.005em;text-wrap:balance}
.ws-item-dt{margin:6px 0 0;font:300 13.5px/1.65 var(--font-display);
  color:var(--text-secondary);text-wrap:pretty}
/* the position ink chip (lead player line only) + the bare-ink mark the
   rest of the page uses, both keyed to the board's own position inks */
.ws-pos{display:inline-flex;align-items:center;justify-content:center;
  min-width:28px;height:19px;padding:0 6px;border-radius:4px;
  font:500 10px/1 var(--font-num);letter-spacing:.06em}
.ws-posb{font:500 10px/1 var(--font-num);letter-spacing:.06em;
  background:none !important}
.pos-ink-QB{color:var(--pos-qb);background:color-mix(in srgb,var(--pos-qb) 14%,transparent)}
.pos-ink-RB{color:var(--pos-rb);background:color-mix(in srgb,var(--pos-rb) 14%,transparent)}
.pos-ink-WR{color:var(--pos-wr);background:color-mix(in srgb,var(--pos-wr) 14%,transparent)}
.pos-ink-TE{color:var(--pos-te);background:color-mix(in srgb,var(--pos-te) 14%,transparent)}
.pos-ink-K,.pos-ink-DST{color:var(--pos-k,var(--text-muted));
  background:color-mix(in srgb,var(--text-muted) 14%,transparent)}
/* ---- trending index ---- */
/* The weekday on an archive chip: present, subordinate, never competing with
   the score it qualifies. */
.pfarch-day{color:var(--text-dim);font-weight:400;letter-spacing:.08em;
  margin-right:3px}
.fchip.active .pfarch-day{color:inherit;opacity:.7}

/* THE NEWS DESK. Rules rather than a box, because the trending index below
   it already carries a box and two boxed columns side by side turn a front
   page into a dashboard. A headline is set in the display face - it is the
   only text on the page written by a person. */
.ws-newsbox{margin-top:22px;border-top:2px solid var(--text-primary);padding-top:10px}
.ws-nrow{padding:11px 0;border-bottom:1px solid var(--border-row)}
.ws-nrow:last-child{border-bottom:none}
.ws-nhl{display:block;font:400 15px/1.3 var(--font-display);
  color:var(--text-primary);text-decoration:none}
.ws-nhl:hover{color:var(--accent);text-decoration:underline}
.ws-nmeta{margin-top:5px;font:400 10.5px/1 var(--font-num);
  letter-spacing:.06em;text-transform:uppercase;color:var(--text-dim)}
.ws-nmeta .pname{font:inherit;color:var(--text-muted);background:none;
  border:none;padding:0;cursor:pointer}
.ws-nmeta .pname:hover{color:var(--accent)}
.ws-trendbox{margin-top:20px;border:1px solid var(--text-primary);
  border-radius:var(--r-md);padding:16px 18px 12px}
.ws-trendhd{display:flex;justify-content:space-between;gap:8px;
  padding-bottom:9px;border-bottom:1px solid var(--text-primary);
  font:500 10.5px/1 var(--font-num);letter-spacing:.18em;
  text-transform:uppercase}
.ws-trendhd span+span{color:var(--text-muted);font-weight:400}
.ws-trow{display:flex;align-items:baseline;gap:8px;padding:7px 0;
  border-bottom:1px solid var(--border-row);font:400 12px/1 var(--font-num)}
.ws-trow:last-child{border-bottom:none}
.ws-leader{flex:1;border-bottom:1px dotted var(--border-strong)}
.ws-trow .ws-item-fig{font-size:12px}
.ws-tdir{color:var(--text-dim);font-size:10px;text-transform:uppercase}
/* ---- the market, in brief ---- */
.ws-brief{margin-top:30px;border-top:2px solid var(--text-primary);
  padding-top:12px}
.ws-briefhd{display:flex;justify-content:space-between;gap:10px;
  font:500 11px/1 var(--font-num);letter-spacing:.18em;text-transform:uppercase}
.ws-briefhd span+span{color:var(--text-muted);font-weight:400;letter-spacing:.08em}
.ws-brief-grid{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));
  margin-top:6px}
.ws-brief-cell{padding:14px 22px 6px 0;border-right:1px solid var(--border-row);
  margin-right:22px;min-width:0}
.ws-brief-cell:last-child{border-right:none;margin-right:0}
/* ---- provenance ---- */
.ws-prov{margin-top:26px;padding-top:12px;border-top:1px solid var(--border-strong);
  display:flex;gap:22px;flex-wrap:wrap;font:400 10.5px/1.8 var(--font-num);
  color:var(--text-muted)}
.wire-empty{padding:20px 2px;color:var(--text-muted);
  font:300 15px/1.6 var(--font-display)}
/* ---- the Ledger ---- */
.lg-head{display:flex;align-items:flex-end;justify-content:space-between;
  gap:24px;padding:26px 0 20px;flex-wrap:wrap}
.lg-head h1{margin:0;font:500 clamp(36px,4.6vw,56px)/.95 var(--font-display);
  letter-spacing:-.012em}
.lg-legend{display:flex;gap:14px;align-items:center;flex-wrap:wrap;
  font:400 11px/1 var(--font-num);color:var(--text-secondary)}
.lg-legend span{display:inline-flex;align-items:center;gap:6px}
.lg-grade{display:inline-flex;align-items:center;justify-content:center;
  width:18px;height:18px;border-radius:4px;align-self:center;
  font:500 10px/1 var(--font-num)}
.lg-a{background:var(--text-primary);border:1px solid var(--text-primary);
  color:var(--bg-surface)}
.lg-b{background:var(--bg-surface);border:1px solid var(--text-primary);
  color:var(--text-primary)}
.lg-c{background:var(--bg-surface);border:1px solid var(--border-strong);
  color:var(--text-muted)}
.lg-row{display:grid;grid-template-columns:18px 46px 30px minmax(0,1fr) auto 62px;
  gap:12px;align-items:baseline;padding:13px 0;
  border-bottom:1px solid var(--border-row)}
.lg-kind{font:700 9.5px/1 var(--font-num);letter-spacing:.1em}
.lg-kind.negative{color:var(--negative)}
.lg-kind.accent{color:var(--accent)}
.lg-kind.warning{color:var(--warning)}
.lg-kind.positive{color:var(--adp-up)}
.lg-kind:not(.negative):not(.accent):not(.warning):not(.positive){
  color:var(--text-muted)}
.lg-body{min-width:0}
.lg-body .pname{font:500 16px/1.35 var(--font-display);letter-spacing:-.005em;
  background:none;border:none;padding:0;cursor:pointer;color:var(--text-primary)}
.lg-body .pname:hover{color:var(--accent)}
.lg-team{font:400 11px/1 var(--font-num);color:var(--text-dim);margin:0 7px 0 4px}
.lg-hl{font:400 15px/1.35 var(--font-display);color:var(--text-primary);
  text-decoration:none}
a.lg-hl:hover{text-decoration:underline}
.lg-dt{display:block;margin-top:3px;font:300 13.5px/1.55 var(--font-display);
  color:var(--text-muted);text-wrap:pretty}
.lg-fig{text-align:right}
.lg-fig .ws-item-fig{font-size:14px}
.lg-day{font:400 11px/1 var(--font-num);color:var(--text-dim);text-align:right}
/* Inline names and headline links get thumb-sized HIT AREAS without moving
   the print: padding grows the target past 44px, the negative margin gives
   the space back. The 414px sweep enforces this on every wire target. */
.ws-topline .pname,.ws-trow .pname,.ws-lead-player .pname,.ws-nmeta .pname{
  display:inline-block;padding:17px 0;margin:-17px 0}
.lg-body .pname,a.lg-hl,a.ws-headline{
  display:inline-block;padding:14px 0;margin:-14px 0}
/* The news headline keeps `display:block` - it is a full-width link and
   inline-block would shrink-wrap it - and takes the same padding/negative
   margin trade for its hit box. */
a.ws-nhl{padding:14px 0;margin:-14px 0}
/* ---- the sheet folds on a phone: columns stack, rules rotate ---- */
@media (max-width:900px){
  .ws-sheet{padding:26px 20px 24px}
  .ws-front{grid-template-columns:1fr}
  .ws-lead{padding-right:0}
  .ws-col{border-left:none;padding:22px 0 0;margin-top:22px;
    border-top:1px solid var(--border-strong)}
  .ws-bodytext{columns:1}
  .ws-brief-grid{grid-template-columns:1fr;gap:0}
  .ws-brief-cell{border-right:none;margin-right:0;
    border-bottom:1px solid var(--border-row)}
  .lg-row{grid-template-columns:18px 38px minmax(0,1fr) auto;gap:10px}
  /* ws-posb, not ws-pos: the ledger's mark is the bare-ink one. When this
     selector went stale the badge stayed visible, five cells fought over
     four tracks, and auto-placement jammed "RB" flush against the name. */
  .lg-row .ws-pos,.lg-row .ws-posb,.lg-day{display:none}
}
/* ============================ COMPARE TRAY ================================
   The queue, side by side. A table in the board's own idiom: hairline rows,
   tabular numerals, the winning cell inked with the accent - one mark, not
   a heatmap. */
.compare{margin:0 0 var(--sp-3);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);background:var(--bg-surface);overflow-x:auto}
.cmp-table{border-collapse:collapse;width:100%;font-size:13px}
.cmp-table th,.cmp-table td{padding:7px 12px;text-align:left;
  border-bottom:1px solid var(--border-subtle)}
.cmp-table thead th{font:600 12.5px/1.3 inherit}
.cmp-table thead .dim{display:block;font-size:10.5px}
.cmp-table tbody th{font:600 10.5px/1.4 var(--font-label);
  letter-spacing:.1em;text-transform:uppercase;color:var(--text-muted);
  white-space:nowrap}
.cmp-table td.num{font-variant-numeric:tabular-nums}
.cmp-table td.cmp-best{color:var(--accent);font-weight:700}
.cmp-table tr:last-child th,.cmp-table tr:last-child td{border-bottom:none}
.cmp-empty{padding:14px;color:var(--text-muted);font-size:13.5px}
.cmp-prov{padding:8px 12px 10px;color:var(--text-muted);font-size:11px}
@media (max-width:640px){
  /* The phone stacks headline under the name; the date yields (it is in the
     detail sentence for injuries, and "today" is the wire's default). */
  .wire-row{grid-template-columns:44px auto 1fr;
    grid-template-areas:"tag pos name" "tag hl hl" "tag dt dt"}
  .wire-day{display:none}
}
.scores{display:grid;grid-template-columns:repeat(auto-fill,minmax(280px,1fr));gap:var(--sp-3);margin-top:var(--sp-4)}
.scores .alert{grid-column:1/-1}
.scores.finals-collapsed .game-extra{display:none}
.live-results-more{grid-column:1/-1;justify-self:start;min-height:38px}
.game{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);border-radius:var(--r-lg);padding:var(--sp-3) var(--sp-4)}
.game-status{display:flex;justify-content:space-between;align-items:baseline;margin-bottom:var(--sp-2)}
.game-status .live{color:var(--accent);font:700 var(--fs-xs)/1 var(--font-num);letter-spacing:var(--track-caps)}
.game-status .live::before{content:"";display:inline-block;width:6px;height:6px;border-radius:50%;background:var(--accent);margin-right:5px}
.game-status .fin,.game-status .sched{color:var(--text-muted);font:600 var(--fs-xs)/1 var(--font-num);letter-spacing:var(--track-caps)}
.game-status .qtr{font:var(--fs-sm)/1 var(--font-num);font-variant-numeric:tabular-nums;color:var(--text-secondary)}
.game-row{display:flex;align-items:baseline;gap:var(--sp-2);padding:3px 0}
.game-row .tm{font-weight:600}
.game-row .rec{font-size:var(--fs-xs);color:var(--text-muted)}
.game-row .sc{margin-left:auto;font:700 var(--fs-xl)/1 var(--font-num);font-variant-numeric:tabular-nums}
.game-row.losing .sc{color:var(--text-secondary)}
.game-note{margin-top:var(--sp-2);font-size:var(--fs-xs);color:var(--text-muted);border-top:1px solid var(--border-subtle);padding-top:var(--sp-2)}
.game-note b{color:var(--text-secondary);font-weight:600}

/* ============================================================ FOOTER */
.footer{margin:var(--sp-5) 0;padding-top:var(--sp-3);border-top:1px solid var(--border-subtle);
  display:flex;gap:var(--sp-4);color:var(--text-muted);font-size:var(--fs-xs);line-height:1.6}
.footer .right{margin-left:auto;text-align:right}

/* ================================================================ THE KEY
   It was one run-on paragraph carrying four different kinds of thing: the
   state of the view, two independent colour axes, the card layout, and the
   keyboard map. You do not READ a key, you find your way to one row of it -
   so the state stays a sentence and everything else became swatch-and-meaning
   in named groups.

   The grouping is load-bearing. The paragraph had to SAY that "movement over
   time is a separate axis", because cyan/orange and green/red are two palettes
   answering two different questions; as groups that is structural and needs no
   sentence. */
.footer{flex-direction:column;align-items:stretch}
.bkey-state{margin:0 0 var(--sp-3);max-width:92ch}
.bkey{display:grid;gap:var(--sp-3) var(--sp-4);
  grid-template-columns:repeat(auto-fit,minmax(210px,1fr))}
.bkey-grp{min-width:0}
.bkey-t{color:var(--text-secondary);font:700 10px/1 var(--font-label);
  letter-spacing:.14em;text-transform:uppercase;margin-bottom:5px}
.bkey-n{color:var(--text-muted);margin-bottom:7px}
/* dt/dd as a two-track grid so every meaning starts on the same x no matter
   how wide its swatch is - a key whose second column wanders is a key you have
   to read instead of scan. */
.bkey-l{display:grid;grid-template-columns:auto 1fr;gap:4px var(--sp-2);
  margin:0;align-items:baseline}
.bkey-l dt{justify-self:start;white-space:nowrap}
.bkey-l dd{margin:0;color:var(--text-muted);min-width:0}
.bkey-f{margin-top:7px;color:var(--text-dim);font-style:italic}
/* One column on a phone: two 210px tracks in a 359px wrapper is a wrap, not a
   grid, and the groups stop reading as groups. */
@media (max-width:620px){
  .bkey{grid-template-columns:1fr;gap:var(--sp-3)}
  .bkey-state{margin-bottom:var(--sp-3)}
}

/* ============================================================ CHEAT SHEET
   Screen-hidden, print-only. Printed in INK ON WHITE, not the terminal
   palette: a draft room is a lit room with a cheap printer, and a dark
   surface there is unreadable and a waste of toner. The validated screen
   palette is a screen decision, not a paper one. */
/* THE CHEAT SHEET, defined ONCE.
   These rules used to live entirely inside @media print, which meant the only
   way to see the sheet was to open a print dialog - clicking a button called
   "Cheat sheet" fired window.print() with no chance to look at it first.
   Defined at top level now, so the on-screen preview and the printed page are
   rendered by the same CSS and cannot drift: what you read is what prints. */
.sheet{display:none;font:10px/1.35 -apple-system,BlinkMacSystemFont,"Segoe UI",Helvetica,Arial,sans-serif;color:#000}
/* On screen it is a paper preview laid over the app. */
.sheet.show{display:block;position:fixed;inset:0;z-index:40;overflow:auto;
  background:rgba(3,6,9,.75);padding:56px 20px 40px}
.sheet.show > .s-page{max-width:820px;margin:0 auto;background:#fff;
  padding:26px 30px;border-radius:6px;box-shadow:0 24px 60px rgba(0,0,0,.6)}
.sheet-bar{position:fixed;top:0;left:0;right:0;height:48px;z-index:41;
  display:flex;align-items:center;justify-content:flex-end;gap:var(--sp-2);
  padding:0 var(--sp-4);background:var(--bg-header);
  border-bottom:1px solid var(--border-subtle)}
.sheet-bar .caps{margin-right:auto;color:var(--text-secondary)}

.s-hd{border-bottom:2px solid #000;padding-bottom:5px;margin-bottom:9px}
.s-hd h1{font:800 italic 17px/1 sans-serif;letter-spacing:.02em}
.s-meta{font-size:9px;color:#444;margin-top:3px}
.sheet h2{font:700 11px/1 sans-serif;text-transform:uppercase;letter-spacing:.08em;
  margin:11px 0 4px;border-bottom:1px solid #999;padding-bottom:2px}
.sheet h2 small{font-weight:400;text-transform:none;letter-spacing:0;color:#666;margin-left:6px}
.s-tbl{width:100%;border-collapse:collapse}
.s-tbl td{padding:1.5px 4px;border-bottom:1px solid #e4e4e4;vertical-align:top}
.s-adp{width:26px;text-align:right;font-weight:700;font-variant-numeric:tabular-nums}
.s-mkt{width:26px;text-align:right;color:#666;font-variant-numeric:tabular-nums}
/* The gap is the actionable number, so it survives a black-and-white
   printer: the SIGN carries the meaning and weight carries the emphasis. */
.s-gap{width:26px;text-align:right;color:#666;font-variant-numeric:tabular-nums}
.s-gap.wait{font-weight:700;color:#000}
.s-gap.reach{font-weight:700;color:#000;font-style:italic}
.s-key{font-size:8.5px;color:#444;margin:2px 0 4px}
.s-pos{width:52px;font-weight:700}
.s-t{color:#666;font-weight:400;margin-left:3px}
.s-nm{font-weight:600}
.s-tm{color:#666;font-weight:400;margin-left:4px}
.s-bye{color:#666;font-weight:400;margin-left:5px}
/* injuries print as text in a box - no color survives a mono printer */
.s-inj{margin-left:5px;border:1px solid #000;padding:0 3px;font-size:8px;font-weight:700}
.s-note{color:#333;font-style:italic}
.s-mv{margin-left:5px;font-size:8px;color:#666}
.s-avoid{font-size:9px;color:#333}
.s-empty{color:#666;font-style:italic}
.s-band{break-inside:avoid;page-break-inside:avoid;margin-bottom:7px}
.s-bh{font:700 9px/1 sans-serif;letter-spacing:.06em;background:#eee;padding:3px 4px}
.s-bh span{float:right;font-weight:400;color:#555}
.s-ft{margin-top:10px;padding-top:5px;border-top:1px solid #999;font-size:8px;color:#555}

@media print{
  @page{size:letter portrait;margin:11mm}
  html,body{background:#fff!important;color:#000!important}
  .topbar,.view,.pcard,.pcard-scrim,.footer{display:none!important}
  .sheet{display:block;font:10px/1.35 -apple-system,BlinkMacSystemFont,"Segoe UI",Helvetica,Arial,sans-serif;color:#000}
  /* The preview chrome is screen-only; the page underneath is the artefact. */
  .sheet-bar{display:none!important}
  .sheet.show{position:static;inset:auto;background:none;padding:0;overflow:visible}
  .sheet.show > .s-page{max-width:none;margin:0;background:none;padding:0;
    border-radius:0;box-shadow:none}
}

/* ============================================================ RESPONSIVE */
/* The draft room needs ~1000px for two real columns; below that, stack. */
@media (max-width:1000px){
  .room{grid-template-columns:1fr}
  .stat-strip{grid-template-columns:1fr 1fr}
}
@media (max-width:640px){
  .wrap{padding:0 var(--sp-2)}
  /* ONE ROW, NOT TWO. 2026-08-11 (6th): the bar was wrapping on a phone - the
     44px toggle plus the two format selects plus the countdown come to a few
     pixels more than 402 - so it stood 70pt tall before a word of content, and
     the board scrolled underneath it. He reported it as a padding issue and it
     is: the height was a WRAP, not a padding, which is why tightening the
     padding under it never moved the number.

     `nowrap` on the bar and a shrinkable meta group is what fixes it at the
     cause. The vertical padding comes down with it, and --topbar-h is measured
     by a ResizeObserver, so everything that pins beneath the bar - the filter
     chips and now the search - follows the new height without a second edit. */
  .topbar-inner{flex-wrap:nowrap;gap:var(--sp-2);padding:4px 0;min-height:0}
  .topbar-meta{flex-wrap:nowrap;gap:var(--sp-2);min-width:0;
    overflow-x:auto;scrollbar-width:none}
  .topbar-meta::-webkit-scrollbar{display:none}
  .topbar-meta>*{flex:none}
  .sn-toggle{width:38px;height:38px;flex:none;margin-right:var(--sp-1)}
  /* MEASURED, and it is 11px. The meta group came to 363 in a 352 box - the
     two format selects at 211 and the countdown at 145 - so the countdown was
     clipped mid-word by a rounding error's worth of space. Eleven pixels of
     select padding is a cheaper thing to spend than the last four characters
     of the one number in the bar that changes daily. */
  .brand{border-right:none;padding-right:0;height:auto}
  /* .hide-m is the meta strip's priority order made explicit: the ADP stamp
     and the league format are constants of this build (the format is restated
     in every ADP column's tooltip and on the cheat sheet), so they yield to
     the countdown - which changes daily and turns red inside the last week -
     and to the room selector, the only control in the strip. Without this the
     strip still fits, it just wraps to a second line and costs the phone
     another 16px of sticky header. */
  .brand-sub,.hide-m{display:none}
  .search{margin-left:0;order:9;width:100%}
  .fsearch{width:100%}
  .clock-time{font-size:24px}
  .scores{grid-template-columns:1fr}
  /* THE PHONE GETS FIVE COLUMNS AND TWO LINES, sized against a MEASURED 325px
     of content rather than against the 720px this was designed at. Eight
     columns needed 474px in a 357px row and the whole page scrolled sideways.

     What goes: the market price and the room's delta against it (context, not
     the decision at hand) and the survival bar (the word or the number says it
     in a fifth of the width). What stays: team and bye, folded UNDER the name
     rather than dropped, because the bye is the one thing here you cannot
     reconstruct from anywhere else on the screen.

     THE HEADER USED TO GO WITH THEM, and that was the defect: eight labels
     over five columns would name the wrong things, so all eight were dropped -
     including the one over the number that is the LOUDEST OBJECT IN THE ROW at
     exactly this width. The nearest naming of it was the panel sentence, some
     160px up and outside the scroller the rows live in, which is to say
     nowhere. The header stays and shows the three labels that still have a
     column under them; the rest hide with the cells they named. */
  :root{--ba-cols:26px 26px minmax(96px,1fr) 56px 52px}
  .ba-head{gap:2px 8px;padding:7px var(--sp-4)}
  .ba-head > span{display:none}
  .ba-head > span:nth-child(1){display:block;grid-column:1;text-align:right}
  .ba-head > span:nth-child(3){display:block;grid-column:3}
  /* Spans the value track AND the draft button's, whose own label is blank:
     "GONE BY 12.11" does not fit 56px and the 52px beside it is empty. */
  .ba-head > span:nth-child(7){display:block;grid-column:4/6;text-align:right;
    letter-spacing:.08em}
  .ba{gap:2px 8px;padding:8px var(--sp-4)}
  .ba .rk,.ba .pos,.ba .surv,.ba .draftbtn{grid-row:1/3;align-self:center}
  .ba .rk{grid-column:1}
  .ba .pos{grid-column:2}
  .ba .nm{grid-column:3;grid-row:1;font-size:var(--fs-lg)}
  .ba .tb{grid-column:3;grid-row:2}
  .ba .surv{grid-column:4}
  .ba .draftbtn{grid-column:5}
  .ba .mkt,.ba .room,.surv .meter{display:none}
  .surv .pct{width:auto;text-align:right}
  /* THE PHRASE WRAPS RATHER THAN CLIPS, and it costs the row nothing. This
     column is 56px here and `gone by 2.12` is 60px on one line, so it takes
     two - and the cell already spans both of the row's grid rows (name over
     team-and-bye), which is 34px of vertical box for 23px of type. Measured
     in Chrome at 375: the row does not grow. Clipping it instead would have
     produced `gone by 2.1`, which is the one failure mode this whole column
     was rebuilt to remove - a horizon that cannot be trusted. */
  .surv .pct.sure{white-space:normal;flex:0 1 auto;min-width:0}
  .ba-legend{flex-direction:column;gap:var(--sp-1)}
  .ba-legend .k{margin-left:0;text-align:left}
}

/* ==================================================== STATE EXPORT/IMPORT */
.statebox{background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3) var(--sp-4);margin-top:var(--sp-4)}
.sb-hd{display:flex;justify-content:space-between;align-items:baseline;gap:var(--sp-3);
  margin-bottom:var(--sp-3);flex-wrap:wrap}
.sb-actions{display:flex;align-items:center;gap:var(--sp-2);flex-wrap:wrap}
.sb-note{margin-top:var(--sp-3);font-size:var(--fs-xs);color:var(--text-muted);line-height:1.6}

/* ============================================================ DRAFT GRID
   Ported from design/edge-draft-board.html (v2 handoff).

   THE CELL TREATMENT IS THE POINT OF THIS HANDOFF. v1 filled each cell with a
   saturated position colour; across 150 cells that shouts and the eye has
   nowhere to rest. This mixes the position hue into the dark cell surface in
   oklab (38% at the top falling to 20%), adds a specular sweep and an inner
   bottom shadow, and puts the position on a small chip. Same information, read
   at a glance, calm enough to sit in front of for three hours.

   Position colour is never alone: every cell writes its position, club and
   bye in words on the identity line, with the pick coordinate opposite. */
/* The board and the panel that explains it, one positioned box - so the key
   can hang OVER the grid without a measurement and without moving anything. */
.dg-board{position:relative}

/* ===================================================== THE BOARD CAPTION
   The half of the key that may not sit behind a tap.

   Two facts about this grid have to be readable in the first second: that
   almost every cell on it is a FORECAST, and what the signed marks measure.
   Both used to live only inside #dgKey, and the count only in an 11px grey
   line at the far corner of the top bar - so a board of thirty-five named
   players in position colours read as a draft that had already happened,
   which is the exact opposite of what it says. A truth stated in the
   quietest ink on the screen, in a different object from the thing it is
   about, is not stated.

   It costs the board ~22px, once, and it is the cheapest 22px on this
   screen: the alternative is a reader who mis-reads every cell. */
.dg-cap{display:flex;flex-wrap:wrap;align-items:center;
  gap:4px var(--sp-3);padding:0 0 var(--sp-2)}
.dgc-state{flex:1 1 220px;min-width:0;
  font:600 var(--fs-xs)/1.4 var(--font-ui);color:var(--text-secondary)}
.dg-hint{display:none}
.dgc-legend{flex:0 1 auto;display:flex;align-items:center;flex-wrap:wrap;
  gap:3px 4px}
/* --text-secondary, not --text-muted, and 9px is a floor rather than a
   default: this legend sits on --bg-app rather than on a white card, where
   the muted step measured 4.15:1 at 8px. It is the only naming these marks
   get without opening a panel, so it has to be read, not glimpsed. */
.dgc-lab,.dgc-w{font:400 9px/1 var(--font-num);letter-spacing:.1em;
  text-transform:uppercase;color:var(--text-secondary);white-space:nowrap}
.dgc-lab{margin-right:2px}
.dgc-w{margin-right:var(--sp-2)}
.dgc-w:last-child{margin-right:0}
/* A mark and its name wrap as one item or not at all: at 390 the loose list
   stranded "reach" on its own line under the badge it names. */
.dgc-pair{display:inline-flex;align-items:center;gap:4px;
  margin-right:var(--sp-1)}
.dgc-pair:last-child{margin-right:0}
/* The word wears the treatment it names - the legend teaches the pencil by
   showing it (round 9). */
.dgc-proj-w{font-style:italic}
/* The swatches are the board's own marks, wearing the board's own classes -
   the same law #dgKey and BOARD_KEY obey. Only the flow margin is stated. */
/* line-height 1 and no vertical padding, for the same reason the identity
   line's copy of this badge carries them: a legend row should be as tall as
   the row, not as tall as one item's leading. The badge's own 1.3 made each
   wrapped legend line 13.7px, and in the full-screen mode the caption is
   subtracted from the board's height - so a badge's leading was costing the
   board seven pixels of a fixed budget. */
.dgc-legend .dg-sig{margin-left:0;line-height:1;padding-top:0;padding-bottom:0}
/* THE CELL SWATCHES, at a quarter of #dgKey's 34x20 - this caption is one line
   of a running draft's screen and the key's box would set the caption's height
   on its own. 18x11 still shows both things the swatch has to show: the fill,
   and whether the edge is broken. `container-type:normal` because .dg-cell
   declares an inline-size container and a 16px one makes every container query
   in the cell fire against a swatch. */
.dgc-sw{position:relative;flex:0 0 18px;height:11px;display:block}
.dgc-sw .dg-cell{position:absolute;inset:0;min-height:0;padding:0;
  cursor:default;container-type:normal}

.dg-wrap{overflow:auto;padding-bottom:var(--sp-2);max-height:62vh}
.draft-grid{--dg-cols:12;position:relative;display:grid;
  grid-template-columns:72px repeat(var(--dg-cols),minmax(104px,1fr));
  gap:3px;min-width:calc(72px + var(--dg-cols) * 107px);
  /* LEVERAGE H1 (2026-08-24): FIGURE AND GROUND. The table takes the page's
     own ground so a forecast - which paints nothing - recedes INTO the
     table, while a pick's filled card stands ON it. The round-9 panel's
     residual charge was that forecasts resemble picks; both were white
     objects on a white panel, and the state had to be re-read per cell from
     border, slant and a 7px word. The reference never has this problem: its
     table is dark and only real picks get a card. An existing validated
     ground, not a new colour; the tint lives on the grid, never on the
     ghost, so the cell-level "no paint on a projection" law is untouched.
     The 3px gaps read as the table showing through between cards. */
  background:var(--bg-app)}
/* ONE outline around the whole YOU column, positioned by renderGrid from the
   header cell's measured box. */
/* z-index 1, deliberately BELOW the sticky round gutter's 2: the rail is a
   later sibling, so at equal z it paints its border on top of the gutter while
   the column it outlines scrolls underneath - a floating cyan edge over the
   round numbers, on film from his mock at 4.05. */
.dg-you-rail{position:absolute;top:0;bottom:0;border:1px solid var(--accent-border);
  border-radius:var(--r-sm);pointer-events:none;z-index:1}

/* header row: team names across the top, YOU called out */
.dg-team{position:sticky;top:0;z-index:3;padding:7px 6px;text-align:center;
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-sm);font:700 var(--fs-xs)/1 var(--font-ui);
  letter-spacing:var(--track-caps);color:var(--text-muted);white-space:nowrap;
  overflow:hidden;text-overflow:ellipsis}
.dg-team.you{color:var(--accent)}
.dg-corner{position:sticky;top:0;left:0;z-index:4;background:var(--bg-app)}

/* ------------------------------------------------------------ THE KEY
   The legend lives in the corner the board already had and could not use: the
   square over the round gutter, which is the one box on this screen that
   carried nothing. It is a button, at the header row's full height (44px, the
   thumb floor, which the team headers beside it already meet), and it costs
   the board not one pixel of height. */
/* IT LABELS THE COLUMN IT HEADS, AND CARRIES THE DOOR BESIDE THE LABEL.
   This button used to read "KEY" in the team headers' own tracked caps, in
   the team header row, directly over the column of round numbers - so it read
   as that column's heading, and the board appeared to carry a column called
   KEY with nothing in it. Three blind readers reported exactly that. The word
   is now RD, which is what the gutter holds, and the door is a marked control
   under it rather than a caption over an empty rail. */
/* THE THUMB FLOOR IS TWO NUMBERS, and this control only ever stated one.
   The rule above claims the header row's 44px height and says nothing about
   width, so the button took whatever the round gutter's track gave it: 72px on
   the desktop, and 42px on the phone, where it measured 42 x 44 in the round-6
   capture - the one genuinely sub-44 target on that screen, and it is the door
   to the board's own legend, which is the control a confused reader reaches
   for. `min-width` rather than a width: the desktop track is already 72 and a
   floor that has to come out exact is a floor that fails again.
   The 2px it takes past the phone's 42px track land in the grid's own 2px gap,
   so the button ends exactly where the first team column begins and not one
   pixel of any other control is covered. The TRACK is untouched on purpose -
   widening that moves every column on the board. */
.dg-keybtn{display:flex;flex-direction:column;align-items:center;
  justify-content:center;gap:2px;min-width:44px;
  border:1px solid var(--border-subtle);border-radius:var(--r-sm);
  cursor:pointer;color:var(--text-muted);padding:0}
.dgkb-rd{font:700 9px/1 var(--font-num);letter-spacing:.14em}
.dgkb-q{display:flex;align-items:center;justify-content:center;
  width:15px;height:15px;border-radius:50%;
  border:1px solid var(--accent-border);background:var(--accent-soft);
  color:var(--accent);font:700 10px/1 var(--font-ui)}
.dg-keybtn:hover,.dg-keybtn.on{color:var(--accent);border-color:var(--accent-border)}
.dg-keybtn.on .dgkb-q{background:var(--accent);color:var(--on-accent);
  border-color:var(--accent)}

/* THE PANEL OVERLAYS, and that is the whole design constraint: a legend that
   pushes the player list down is a legend that costs you a pick. It hangs off
   the board's top-left corner, over the cells, and the next tap dismisses it. */
/* The caption sits above the grid now, and its height is MEASURED and handed
   to the stylesheet by positionYouRail (the same seam --dg-track and
   --dg-gutter already come through), because the sentence wraps to two lines
   at phone widths and a number written here would be right at one width. */
.dg-key{position:absolute;z-index:30;top:calc(var(--dg-cap-h,0px) + 44px);left:0;
  width:min(340px,calc(100% - 12px));
  background:var(--bg-surface-raised);border:1px solid var(--border-strong);
  border-radius:var(--r-md);box-shadow:0 10px 30px rgba(0,0,0,.35);
  padding:var(--sp-3);cursor:pointer}
/* THE REVEALED FACT. Deliberately the same surface as .dg-key above rather than
   a second visual language for "a small overlay attached to the thing you just
   tapped" - one idiom, learned once. z-index 46 sits one above .dg-key's 45 in
   the mode, and under .intro's 60.
   NO PERSISTENT MARKER IS ADDED TO THE ANNOTATED ELEMENTS, and that is measured
   rather than stylistic: Chrome counts ::before content in innerText, so a
   one-character dot on 142 elements would put ~142 chars into a resting census
   pinned at 7817 with a 2% (157 char) budget. The affordance would have eaten
   nine tenths of the allowance before anything else spent a byte. */
.fact-pop{position:fixed;z-index:46;max-width:min(280px,calc(100vw - 24px));
  background:var(--bg-surface-raised);border:1px solid var(--border-strong);
  border-radius:var(--r-md);box-shadow:0 10px 30px rgba(0,0,0,.35);
  padding:var(--sp-2) var(--sp-3);
  font:400 var(--fs-xs)/1.45 var(--font-ui);color:var(--text-secondary)}
.dgk-list{list-style:none;margin:0;padding:0;display:grid;gap:7px}
.dgk-list li{display:flex;align-items:flex-start;gap:9px;
  font:400 var(--fs-xs)/1.45 var(--font-ui);color:var(--text-secondary)}
.dgk-list b{color:var(--text-primary);font-weight:700}
/* THE SWATCHES ARE CELLS. Not a palette copied out of the board's rules - the
   board's own classes, sized down, exactly as the Board tab's legend is built.
   That is what makes a legend maintainable: change a cell treatment and the
   key changes with it, at both saturations, in both skins, or it does not
   change at all. Only geometry is stated here. */
.dgk-sw{position:relative;flex:0 0 34px;height:20px;
  display:flex;align-items:center;justify-content:center}
.dgk-sw .dg-cell{position:absolute;inset:0;min-height:0;padding:0;
  cursor:default;container-type:normal}
/* The YOU rail is drawn by JS from a measured header cell on the board; here
   it only has to be the same border, so it is given the box instead. */
.dgk-sw .dg-you-rail{left:0;right:0}
/* THE VOCABULARY KEY. Deliberately quiet at rest: one small labelled control,
   not a banner. The panel it opens carries no size of its own until it opens,
   so the resting board is unchanged. */
.vk-btn{display:inline-flex;align-items:center;gap:6px;margin:var(--sp-2) 0 0;
  min-height:44px;padding:0 var(--sp-3);background:none;
  border:1px solid var(--line);border-radius:var(--r-sm);cursor:pointer;
  font:600 var(--fs-xs)/1 var(--font-ui);letter-spacing:var(--track-caps);
  color:var(--text-muted)}
.vk-btn:hover,.vk-btn.on{color:var(--text-primary);border-color:var(--text-muted)}
.vk-panel{margin:var(--sp-2) 0 0;padding:var(--sp-3);border:1px solid var(--line);
  border-radius:var(--r-sm);background:var(--bg-raised)}
.vk-hd{margin:0 0 var(--sp-3);font:700 var(--fs-xs)/1.35 var(--font-ui);
  color:var(--text-muted);letter-spacing:var(--track-caps)}
.vk-list{list-style:none;margin:0;padding:0;display:grid;gap:7px}
.vk-list li{font:400 var(--fs-xs)/1.45 var(--font-ui);color:var(--text-secondary)}
.vk-list b{color:var(--text-primary);font-weight:700}
.vk-foot{margin:var(--sp-3) 0 0;font:400 var(--fs-xs)/1.4 var(--font-ui);
  color:var(--text-muted)}
.dgk-hd{margin:0 0 var(--sp-3);font:700 var(--fs-xs)/1.35 var(--font-ui);
  color:var(--text-muted);letter-spacing:var(--track-caps);text-transform:none}
.dgk-foot{margin:var(--sp-3) 0 0;font:400 var(--fs-xs)/1.4 var(--font-ui);
  color:var(--text-muted)}

/* round labels down the left edge */
.dg-rd{position:sticky;left:0;z-index:2;display:flex;align-items:center;
  justify-content:flex-end;padding-right:var(--sp-2);background:var(--bg-app);
  font:700 var(--fs-xs)/1 var(--font-ui);letter-spacing:var(--track-caps);
  color:var(--text-muted);white-space:nowrap}

/* the cell */
.dg-cell{position:relative;background:var(--bg-cell);
  border:1px solid var(--border-subtle);border-radius:var(--r-sm);
  /* 6px of side padding, not 8. A 12-team board holds its cells at the 104px
     track floor whatever the window does, and of that the identity line and
     the pick coordinate have to share what the padding and the 4px position
     edge leave: at 8px that was 83px, and six of 179 cells ellipsed the bye
     off the end. The four pixels bought here are the cheapest of the three
     places to find them (the other two are the name's face and the line's). */
  padding:7px 6px 6px;min-height:48px;overflow:hidden;
  /* The meta row's contents are sized against the CELL, not the viewport - a
     12-team room compresses these to 72px at any window size, and "DST WAS
     15.10" wants 29px more than that. */
  container-type:inline-size}
/* A 4px LEFT EDGE in the position hue: the cell's identity moves to its edge,
   so the fill can stay a whisper and the chip stops being the only carrier. */
.dg-cell.filled{border-color:rgba(140,180,200,.10);cursor:pointer;border-left-width:4px}
.dg-cell.filled.pos-QB{border-left-color:var(--pos-qb)}
.dg-cell.filled.pos-RB{border-left-color:var(--pos-rb)}
.dg-cell.filled.pos-WR{border-left-color:var(--pos-wr)}
.dg-cell.filled.pos-TE{border-left-color:var(--pos-te)}
.dg-cell.filled.pos-K{border-left-color:var(--pos-k)}
.dg-cell.filled.pos-DST{border-left-color:var(--pos-dst)}
/* Kept is a STATE: a dashed edge on three sides and a marker. The club code
   stays where it was, because that is a different fact. */
.dg-cell.keeper{border-left-width:4px;border-style:dashed dashed dashed solid}
.dg-cell.keeper.pos-QB{border-left-color:var(--pos-qb)}
.dg-cell.keeper.pos-RB{border-left-color:var(--pos-rb)}
.dg-cell.keeper.pos-WR{border-left-color:var(--pos-wr)}
.dg-cell.keeper.pos-TE{border-left-color:var(--pos-te)}
.dg-cell.keeper.pos-K{border-left-color:var(--pos-k)}
.dg-cell.keeper.pos-DST{border-left-color:var(--pos-dst)}
/* KEPT takes the pick coordinate's end of the identity line: a keeper's round
   is in the gutter, and "kept" is the thing about that cell you cannot read
   anywhere else on the board. */
/* SIZED OFF THE LINE IT RIDES ON, not off a global step. At `--fs-xs` this
   marker was 10px on an 8.5px identity line - a label larger than its own row -
   and the 1.4px tracking on top took it to 29.6px against the pick
   coordinate's 23.6px in the same slot. That 6px is exactly what pushed
   `WR · DEN (10)KEPT` over: 5px at 768 and 1400, 3px at 1990, measured. The
   phone's own override below is UNCHANGED and still earned - at 7.5px this
   inherits into a 78px cell and clips by 4px, which is what that rule was
   measured down to answer. */
.dg-kept{margin-left:auto;flex:none;font-family:var(--font-num);
  font-size:inherit;font-weight:400;line-height:1;
  letter-spacing:.08em;color:var(--text-muted)}
.dg-cell.filled:hover{filter:brightness(1.08)}
/* a filled cell is a FLAT surface with a whisper of the position hue - the
   identity line and the left edge do the identification. The old 38-40% gradient plus a
   specular sweep made 150 cells compete for the eye; the user called it, and
   he was right. */
.dg-cell.filled.pos-QB{background:color-mix(in oklab,var(--pos-qb) 12%,var(--bg-cell))}
.dg-cell.filled.pos-RB{background:color-mix(in oklab,var(--pos-rb) 12%,var(--bg-cell))}
.dg-cell.filled.pos-WR{background:color-mix(in oklab,var(--pos-wr) 12%,var(--bg-cell))}
.dg-cell.filled.pos-TE{background:color-mix(in oklab,var(--pos-te) 12%,var(--bg-cell))}
.dg-cell.filled.pos-K{background:color-mix(in oklab,var(--pos-k) 9%,var(--bg-cell))}
.dg-cell.filled.pos-DST{background:color-mix(in oklab,var(--pos-dst) 9%,var(--bg-cell))}
/* your own column reads as a channel: raised surface + ONE exterior outline
   drawn by .dg-you-rail, never a border per card (user's call, and the rail
   reads as a single object where 15 bordered cells read as 15 alarms) */
.dg-cell.mine{background:var(--bg-cell-hover)}
.dg-cell.keeper{background:var(--bg-surface-raised);
  box-shadow:inset 0 0 0 1px var(--border-strong)}
/* drafted in the followed room but not on this board: the cell is consumed,
   and says so, rather than reading as an open pick */
.dg-cell.filled.off .dg-nm{color:var(--text-secondary)}
/* A FOLLOWED PICK IS A PICK, so it keeps the solid edge. This was dashed - it
   was the cell's way of saying "we do not have a row for this player" - and
   dashed is now the mark of a slot nobody has taken, which is the one thing
   this cell is NOT: the pick happened, in the real room, and drawing it like a
   forecast is the exact error the channel exists to prevent. What it does not
   know is stated where it belongs, in words, on the identity line: `off
   board`. Secondary ink on the name keeps it a half-step behind a cell we can
   name in full. */
.dg-cell.filled.off{border-style:solid}

.dg-top{display:flex;align-items:baseline;position:relative;z-index:1}
.dg-nm{flex:1;font:600 var(--fs-md)/1.2 var(--font-ui);color:var(--text-primary);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.dg-pk{flex:none;margin-left:6px;font:600 var(--fs-xs)/1 var(--font-ui);
  font-variant-numeric:tabular-nums;color:var(--text-muted)}

/* --------------------------------------------------- THE IDENTITY LINE
   `RB · BUF (7)` at one end, the pick coordinate at the other, ABOVE the
   name. It replaced a meta row that carried a position chip, a club code and
   the pick number, and it is a straight trade: the chip said in a coloured
   box what the fill and the left edge already said twice, and the bye - the
   one fact here that no other surface of the draft room states beside a name -
   was on none of them.

   Set as a WATERMARK, not as content: it inherits the cell's own ink at low
   opacity, the way the reference frame darkens its cell colour to write the
   same line, so a board of 150 of these reads as texture and the NAME is
   still the only thing you see from a metre away. */
/* THE UI FACE, NOT THE MONO, and it is a fit decision rather than a taste
   one: `DST · WAS (14)` is fourteen characters, and at a monospace advance
   they cost 63px of a 72px cell before the pick coordinate has said anything.
   The proportional face carries the same string in two thirds the width, which
   is the difference between printing the bye and ellipsing it. The coordinate
   opposite keeps the mono, because a column of pick numbers is a column of
   numbers to compare. */
/* 8.5px, and it is a MEASUREMENT rather than a step off the type scale: the
   line has to hold `WR · LAR (11)` beside `12.11` inside an 83px content box
   (a 104px cell less its padding and its position edge), and at the 10px
   `--fs-xs` 152 of 179 cells ellipsed. It is a watermark, and the reference
   frame sets its own at about a tenth of the cell's width; this is a twelfth.
   The NAME under it is the object that has to read from a metre away. */
/* --text-secondary AT FULL OPACITY, AND IT IS THE STEP THIS BOARD HAS ALREADY
   TAKEN TWICE for the same reason (2026-08-22). The caption's legend took it
   at 9px ("the only naming these marks get without opening a panel, so it has
   to be read, not glimpsed"), and the ghost cell took it at 8.5px three
   hundred lines down, measured, because --text-dim under this rule's own .9
   came out at 1.57:1. What was left on --text-muted at .9 was the MADE cell -
   so the board was drawing its forecasts' identity line more legibly than its
   facts'. Round 6's accessibility lens named grid text as the smallest type
   on screen and also the fastest-read information; this is the half of that
   charge that survives measurement, and the fix was already written one
   selector away. The type scale does the receding here; the ink only has to
   be readable. */
.dg-id{display:flex;align-items:baseline;gap:4px;position:relative;z-index:1;
  margin-bottom:2px;font:700 8.5px/1 var(--font-ui);
  letter-spacing:.02em;color:var(--text-secondary)}
.dg-id .dg-pk{font-size:8.5px}
.dg-idt{flex:0 1 auto;min-width:0;white-space:nowrap;overflow:hidden;
  text-overflow:ellipsis}
/* The pick coordinate keeps the far end, and it is the first thing to go when
   the cell is too narrow to hold both: the round gutter states the round, so
   of the two items on this line it is the only one stated elsewhere. */
.dg-id .dg-pk{margin-left:auto;font-weight:400}
/* THE WORD IN THE COORDINATE'S SLOT (round 9). A forecast with no mark ends
   its identity line in PROJ where a pick ends in its number. Same box, same
   /1 leading, same muted ink as the number it replaces, so it cannot set a
   row height the coordinate could not - but it is a WORD, in the mono caps
   voice every state marker on this app speaks. 16.1px at 375 against the
   23.0px of `15.10`: the slot only gained room. The round-9 panel is why it
   exists: with no in-cell word, five of six critics re-failed the forecast
   channel; with the word as an under-layer, round 8 failed it as texture.
   In the line, in words, is the synthesis. */
.dg-projw{flex:none;margin-left:auto;font:600 8.5px/1 var(--font-num);
  letter-spacing:.08em;color:var(--text-muted)}
/* 76px is the CONTENT box, not the cell: container-type sizes against the
   padding box, so a 104px cell queries as 88 and a 72px one as 56. Measured
   both ways rather than guessed - at 96 this hid the pick on every cell. */
@container (max-width:76px){ .dg-id .dg-pk{display:none} }
/* ------------------------------------------------------ THE QUIET BOARD (4a)

   AN UNMADE PICK IS NOT A CARD. It keeps its number and gives up everything
   else - box, border, fill, and the direction arrow that repeated what the
   round gutter and the numbering already say. 96 of 150 cells stop competing,
   which is the entire point of the build: the six that HAVE something to say
   are now the only things on the board that look like anything. */
.dg-cell.dg-open{background:none;border-color:transparent;
  display:flex;align-items:flex-start;justify-content:flex-end}
.dg-cell.dg-open .dg-pk{margin-left:0;color:var(--text-ghost);font-weight:400}
.dg-cell.dg-open.mine{background:none}

/* ---------------------------------------------------------- THE GHOST BOARD

   An unmade pick that the market has an opinion about, which is nearly all of
   them. It is drawn at a fraction of a made pick's weight so 4a's rule holds -
   the six cells that HAPPENED still own the board - but it is no longer a bare
   number, because the market's projection is a fact this app has and Sleeper's
   board does not print. Dashed edge is HALF the legend and colour is the other
   half: dashed and colourless means nobody has picked here yet. */
/* THE 4px POSITION STRIPE BELONGS TO A SLOT A PLAYER OWNS, so a ghost gives it
   up along with the hue. It was a solid 4px edge - the one solid edge on an
   otherwise dashed cell - carrying a colour that is now reserved, so what it
   would carry after the change is emphasis with nothing behind it. A
   projection is 1px and dashed on all four sides, which is the whole shape of
   the claim it makes. */
.dg-cell.ghost{border:1px dashed var(--border-subtle);
  cursor:pointer;background-color:transparent}
/* ROUND 9: THE STATE MOVED INTO THE GLYPHS, and the under-layers are deleted
   rather than restyled. The hatch and the rotated watermark scored 2 pass /
   4 fail in round 8 with both passes from the cheapest judges, and four
   critics named the board's texture as its biggest flaw; the ruling on the
   record is "simplify the cell rather than stack a fourth signal on it".
   What rounds 7 and 8 actually measured is that every projection signal sat
   AROUND the name while the name itself printed exactly like a pick's - the
   ballots' own words were "forecast names look like completed picks". So the
   name now carries the state: penciled - italic, book weight, secondary ink -
   against a pick's upright 600 in primary ink. The slant is synthesized (the
   Plex subset ships no italic face) and that is the point, not a compromise:
   the claim is slant, which survives a 10px two-line clamp where a
   background pattern measurably did not. The dashed edge stays as the
   cell-level half; the legend's own "projected" word wears the same pencil,
   so the treatment is named by demonstration where it is read. */
.dg-cell.ghost .dg-nm{font-weight:400;font-style:italic;color:var(--text-secondary)}
/* MEASURED, 2026-08-15: this was --text-dim, and --text-dim under the .9 the
   base rule then carried is 1.57:1 at 8.5px on a ghost's transparent fill. The
   watermark idiom this borrows from works because the reference DARKENS a
   saturated cell; a ghost has no fill to darken, so on paper the same gesture
   is just light grey on white - and this line carries the bye week and the
   pick coordinate, two facts stated beside a name nowhere else in the room.
   The RULE ITSELF IS GONE (2026-08-22): the base rule at the top of the board
   now states exactly this, for every cell rather than for the forecast alone,
   so keeping a copy here would be the same declaration written twice - and
   the day they disagree is the day the ghost and the pick claim different
   things about one line. The measurement above is why the base says it. */
.dg-cell.ghost:hover{background-color:var(--bg-cell-hover);border-color:var(--border-strong)}
.dg-cell.ghost:hover .dg-nm{color:var(--text-primary)}
/* THE WATERMARK RULE IS GONE WITH ITS SPAN (round 9). The word survives in
   two places that never scroll away: every ghost's own title, and the PROJ
   tag inside the live cell, where being wrong costs a pick. */
/* ================== THE FACT CHANNEL, AND IT IS COLOUR ==================
 *
 * A PROJECTION HAS NO POSITION HUE. Not a whisper of one, not 42% of one.
 *
 * This rule used to run the other way - the ghost took the position colour at
 * 42% of a made cell's edge, on the argument that "a run of one colour down a
 * column is still a colour question at scan speed, quiet enough that a
 * projection never reads as a pick". The second half of that sentence was a
 * hope, and it was measured false: four of six blind critics read this board
 * as a draft that had already gone six rounds, with 179 of 180 cells still
 * projections and the header beside them reading `0 picked`. Thirty named
 * players in position colours IS a draft board that has happened. The only
 * thing standing against it was a 1px dashed border nothing on screen named.
 *
 * So colour gets ONE job and nothing else may borrow it: a cell carries the
 * position hue if and only if a real player owns that slot - picked, kept, or
 * drafted off-board. Everything the market merely EXPECTS is drawn in the
 * board's neutrals. It is not a fourth encoding; it is the fill encoding,
 * given a single meaning, and the caption's two swatches name it in the one
 * place a reader is already looking.
 *
 * The channel is worth more as the draft runs than before it starts, which is
 * the test the old scheme failed: colour spreads across the grid exactly as
 * picks land, so "how far has this room got" is answerable from across the
 * table, and "the forecast and the history are now indistinguishable" - the
 * sharpest thing said about the old board - stops being possible to say.
 *
 * The position is not lost with the hue. Every ghost states it in words on its
 * identity line (`RB · BUF (7)`), which the old pastel could not do and which
 * is why that line was added.
 */
/* Written out rather than deleted, and the suite reads it: these six
   selectors are the exact place a position colour would come back onto a
   forecast, so they stay on the page holding the neutral. A rule that states
   the law is cheaper to keep true than an absence that has to be remembered. */
.dg-cell.ghost.pos-QB,.dg-cell.ghost.pos-RB,.dg-cell.ghost.pos-WR,
.dg-cell.ghost.pos-TE,.dg-cell.ghost.pos-K,.dg-cell.ghost.pos-DST{
  border-color:var(--border-subtle)}

/* THE MARK. The one thing on this board that no draft board draws: the cells
   where our composite and the market disagree past the bar. A BUY is lifted,
   because a projection you should be planning around is not background - it is
   the reason to look at the board at all.
   IT IS LIFTED IN THE ACCENT, NEVER TOWARD A PICK. Accent is the misprice
   channel - the same ink the +16 badge in this very cell is set in, legended
   two swatches along - and it is not a position hue, so it says "our number
   disagrees here" without saying "somebody drafted here". The EDGE is what the
   old rule got wrong and it is only touched in colour: it stays dashed, at 1px,
   because solid is now the other half of the fact channel and a forecast may
   never wear it however loudly it deserves to be read. */
.dg-cell.ghost.sig-val{background-color:var(--accent-soft);border-color:var(--accent-border)}
.dg-cell.ghost.sig-val .dg-nm{color:var(--text-primary);font-weight:600}
/* THE FADE BORDER IS CUT (2026-08-15). It was a THIRD border treatment - a red
   edge alongside the dashed projection and the solid pick - spent on the one
   verdict this board never acts on: a player you are not going to take, priced
   above our number, in a cell that is not yours. The badge already says it, in
   the same red, with the magnitude attached. A legend has to be able to name
   every encoding on the board in six lines; an encoding that cannot earn a
   line is the one to remove, not the one to explain. */
.dg-sig{flex:none;margin-left:5px;padding:1px 4px;border-radius:3px;
  font:800 9px/1.3 var(--font-num);font-variant-numeric:tabular-nums;
  letter-spacing:.02em}
.dg-sig.val{color:var(--positive-ink);background:var(--accent-soft)}
.dg-sig.fade{color:var(--negative-ink);background:var(--negative-soft)}
/* NO CONTAINER QUERY ON THIS ONE, deliberately, and it is the opposite call to
   the pick number six rules up. That number is dropped at narrow widths
   because the round gutter already states it; this is three characters that
   are stated NOWHERE else on the board, and they are the reason this board is
   worth more than a list of who is next.
   WHAT YIELDS THE SPACE IS THE COORDINATE, NOT THE NAME (2026-08-15). The mark
   used to sit on the name's row and take 24.7px of a 104px cell, which left the
   name 57.3px and clipped it on eighteen desktop cells - every one of them a
   cell carrying a mark, because the mark is what took the room. It sits on the
   identity line now, in the slot the pick coordinate held, and the coordinate
   is the item that can go: the round gutter states the round, the column header
   states the seat, and the cell's title still carries the pair in full. */
/* line-height 1, and it is a ROW HEIGHT rule rather than a type rule. The
   badge's own `1.3` made the identity band 10.4px in a cell carrying a mark
   against 7.5px in one that does not, and marks start in round 3 - so the
   first measurement after this move read 41.5px for rounds 1 and 2 and 44.4px
   for rounds 3 through 15. Fixing round 1 by making thirteen other rounds
   taller is not fixing it. On this line the badge takes the line's leading,
   not its own. */
/* AND THE SIZE COMES FROM THE LINE, which is the half of this fix that was
   missing. Zeroing the padding and the leading got the badge to 9px on an
   8.5px line, and a grid row is as tall as its tallest cell, so the thirteen
   rounds that carry marks stood 0.5px taller than the two that do not. The
   640px rule below already does exactly this - 7.5px on the phone's 7.5px
   line - so this is the same rule finished at the desktop scale. Third marker
   in this cell to learn it after `.dg-kept`: a marker takes the size of the
   line it rides on, never its own. */
.dg-id .dg-sig{margin-left:auto;flex:none;line-height:1;
  font-size:inherit;padding-top:0;padding-bottom:0}

/* THE ONLY LIT CELL ON THE BOARD, so it is findable from across the room.

   INVERTED, not accented (2026-08-15). It was a solid accent plate, which on
   the lit skin is the same blue as your own column's channel, the YOU header
   and every primary button on the page - the one cell that has to be
   unmistakable was wearing the app's most-used colour. It now takes `--turf`,
   the dark band this design already owns, so the three objects that say "this
   is the moment" - the state rail, the top-strip chip and this cell - are one
   gesture drawn three times. The accent survives as the ring and the drop,
   which is what an accent is for. */
.dg-cell.now{border-color:var(--turf,#0f1a22);background:var(--turf,#0f1a22);
  color:#fff;box-shadow:0 0 0 2px var(--accent-border),0 0 26px var(--accent-glow)}
.dg-cell.now .dg-nm{color:#fff;font-weight:700}
.dg-cell.now .dg-pk{color:#fff;opacity:.55}
/* ---- THE LIT CELL IS TWO BANDS, LIKE EVERY OTHER CELL ON THE BOARD ----
 *
 * A GRID ROW IS AS TALL AS ITS TALLEST CELL, and this cell is always in round
 * 1 before a draft starts. It carried three stacked bands - name, then a clock
 * on 8px of top margin, then the market's read on another 5 - against the two
 * that every other cell uses, and the arithmetic came out in Chrome as 65.5px
 * for round 1 against 41.5px for rounds 2 to 15 at 375, 64.4 against 48.0 at
 * 1280. The round whose picks decide a season was half again the least dense
 * row on the board, and nothing about that was a decision anybody made - it
 * was three bands paying for two.
 *
 * The clock and the market's read share the second band now: they answer the
 * two halves of one question (how long, and who the room takes) and neither is
 * a headline on its own. The margins that stacked them are gone, and the clock
 * spends them on SIZE - 15px on the phone against the 13px it had, which is the
 * largest face on the board and the correct outcome for the one object here
 * that is counting down.
 */
.dg-nowb{display:flex;align-items:center;gap:3px;min-width:0}
.dg-clock{flex:none;
  font:800 var(--fs-lg)/1 var(--font-num);font-variant-numeric:tabular-nums;color:#fff}
.dg-clock.low{color:#ff8a94}
/* WHO THE MARKET HAS HERE, inside the cell that is on the clock. The one
   question a live pick raises that a draft board never answers: not "whose
   turn" - the cell already says that - but "what does the room think happens
   next". Same projection the ghosts below are drawn from. */
/* TWO LINES, LABEL OVER NAME, and that is what lets it sit beside the clock
   instead of under it. `MKT · J. Gibbs` on one line is 75.6px measured at 9px,
   and the room left beside a 31px clock in an 87px cell is 52px. Stacked, the
   widest line is the name at ~38px and the pair is 16px tall - shorter than
   the 18px the margin-and-line arrangement cost, and it fits. */
/* THE INK SAYS FORECAST TOO, and the two steps are measured on this plate
   rather than chosen. The lit cell's ground is --turf (#0f1a22); against it,
   white at .55 is 6.0:1, at .46 is 4.6:1 and at .42 is 4.0:1. So the NAME
   drops a step - it is the thing that was reading as a locked-in pick, and
   4.6:1 keeps it over the floor at 6.5px - while the LABEL, which is the word
   that says the name is a guess, comes UP out of a whisper that was under the
   floor at 6px. The annotation is now the louder half of the pair, which is
   the whole point: the data that is not real yet may not look like the data
   that is. */
.dg-mkt{flex:1 1 auto;min-width:0;text-align:right;white-space:nowrap;
  overflow:hidden;text-overflow:ellipsis;
  font:700 8px/1 var(--font-num);letter-spacing:.06em;color:rgba(255,255,255,.46)}
.dgm-l{display:block;font-size:6.5px;letter-spacing:.12em;
  color:rgba(255,255,255,.58);margin-bottom:1px}
/* The suggestion's name is penciled like every other forecast name on the
   board (round 9): one treatment, learned once, wherever a name is a guess. */
.dgm-n{font-style:italic}
/* THE BOARD'S DASHED EDGE, REPEATED ON THE PLATE, AND IT IS THIS BOX'S OWN
   BORDER. Everywhere else on the board a broken line means "nobody has picked
   here"; the lit cell is the one place that fact was carried by a label alone.
   IT HAD TO BE THE BOX, and that is measured rather than stylistic. The first
   two attempts hung the hairline off the NAME - a border, then a dashed
   underline - and both were invisible at 375 and 390 for the same reason: the
   name's inline box runs from 207.8 to 217.8 while this element's line boxes
   end at 216.3, and `overflow:hidden` cuts everything under it. An element's
   own border is painted whatever its overflow does.
   It costs one pixel of BORDER BOX (12.5 to 13.5) and it cannot set the band:
   the clock beside it is 15px, so the slack was already there. Verified at
   375, 390 and 1440 - every round still one height. */
.dg-mkt{border-bottom:1px dashed rgba(255,255,255,.34);padding-bottom:1px}
/* THE LIVE CELL IS REAL; THE NAME INSIDE THIS BOX IS NOT. A full dashed frame,
   the PROJ word and the penciled name keep the suggested player from
   inheriting the authority of the ticking plate. The diagonal cut it used to
   share with the ghost cells went when theirs did (round 9): one state, one
   set of marks, everywhere. Outline is inset so it costs no width and cannot
   make the row taller. */
.dg-mkt.is-projected{outline:1px dashed rgba(255,255,255,.42);outline-offset:-1px}
/* 48vh was 390px of a 812px phone, all of it above the two panels a draft is
   actually run from - Best Available and the pick path started below the
   fold, so every pick meant scrolling past the board to reach the controls.
   The grid keeps its own scroller and keepLivePickVisible() already tracks
   the live cell inside it, so a shorter window loses no information: it still
   shows four rounds, and it lifts everything below it by ~150px. */
@media (max-width:640px){ .dg-wrap{max-height:30vh} }

/* ================================================ BEST-AVAILABLE POS TABS */
.ba-tabs{display:flex;gap:4px;padding:var(--sp-2) var(--sp-4) 0;flex-wrap:wrap}
/* The position chips are MONO KEYS: label and count on one line, the way his
   button system draws a counted key ("QB 1"), instead of the two-line box they
   were. The keycap surface comes from the shared primitive above. */
.batab .bt-l{font:inherit;letter-spacing:inherit}
.batab .bt-n{font:inherit;font-variant-numeric:tabular-nums;color:var(--text-muted)}
.batab.active .bt-n{color:var(--on-accent-soft)}
/* HIS 6a/7a INVERTS THIS. A filled slot used to light up; now it goes quiet
   and the UNFILLED one wears the amber, because you do not need to be told
   about what is already done. The count is still the signal - colour agrees
   with a fraction that is already on screen, and never carries it alone. */
.batab.done .bt-n{color:var(--text-secondary)}
.batab.need{border-color:var(--risk-border)}
.batab.need .bt-l,.batab.need .bt-n{color:var(--risk)}
.batab.active.need{border-color:var(--accent-border)}
/* THE ACTIVE KEYCAP IS A LIGHT FACE, so its ink is the dark one every other
   active chip uses. This said `var(--accent)` - accent lettering on the accent
   keycap - so selecting a position you still need made its own label
   disappear: measured 95,212,240 on the light face, which is not a contrast
   ratio so much as a rumour. Found by looking at the phone's circles, where
   the chip is 52px and the failure is unmissable; it was on the desktop pill
   the whole time. The amber belongs to the UNSELECTED need state, and being
   selected already has a louder signal than colour. */
.batab.active.need .bt-l{color:var(--on-accent)}
.batab.active.need .bt-n{color:var(--on-accent-soft)}
.batab.tog{justify-content:center}
.ba.gone{opacity:.45}
.ba.gone .nm{text-decoration:line-through}
.draftbtn.off{display:inline-block;text-align:center;color:var(--text-muted);
  border-style:dashed;cursor:default;line-height:24px}

/* ==================================================================== SOS
   Direction VERIFIED against FantasyPros' own markup (4/5 = is-good): more is
   better, 5 = easiest schedule. The number is always printed beside the pips,
   so the meter is decoration and never the signal. */
.sos-hint{display:block;font:400 8px/1 var(--font-num);color:var(--text-muted);
  letter-spacing:0;text-transform:none;margin-top:2px}
.sos-m{display:inline-flex;gap:1.5px;margin-right:5px;vertical-align:1px}
.sos-m i{width:3px;height:9px;border-radius:1px;background:var(--border-strong)}
.sos-m i.on{background:var(--text-secondary)}

/* ==================================================== OVERNIGHT MOVERS STRIP
   Direction is a glyph plus a number; the colour agrees with the existing
   room-delta rule (later = cheaper = favorable) and never carries it alone. */
.moved{display:flex;align-items:center;gap:var(--sp-2);flex-wrap:wrap;
  background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-2) var(--sp-3);margin-bottom:var(--sp-3)}
.moved>.caps{white-space:nowrap}
.mv-none{font-size:var(--fs-sm);color:var(--text-muted)}
.mv-chip{display:inline-flex;align-items:center;gap:5px;min-height:36px;
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-sm);padding:4px 8px;cursor:pointer}
.mv-chip:hover{border-color:var(--border-strong);background:var(--bg-row-hover)}
.mc-n{font:600 var(--fs-sm)/1 var(--font-ui);color:var(--text-primary);white-space:nowrap}
.mc-d{font:700 var(--fs-xs)/1 var(--font-num);font-variant-numeric:tabular-nums}
.mv-src{margin-left:auto;font-size:var(--fs-xs);color:var(--text-muted);white-space:nowrap}
/* multi-day drift; bolder outline when several rooms agree */
/* `.conv` = the same move is running across several rooms. A box, not a
   second number: conviction is a STRENGTH on the move, not another fact. */
.mv.conv{padding:2px 4px;border-radius:3px;box-shadow:inset 0 0 0 1px currentColor}

/* ============================================== DRAFT COUNTDOWN + DATE */
.draftclock{color:var(--text-secondary)}
/* inside a week, this is the reserved caution slot - the word DAYS carries it */
.draftclock.soon{color:var(--risk)}

/* ------------------------------------------------- THE TOP-STRIP PICK BAR
   Same element as the countdown, wearing the room's state while the room is
   the tab on screen. It is the one object visible at every scroll position on
   every width, which is the only place a pick clock can live and still be a
   pick clock: the on-the-clock cell can be four rounds up a scrolled board.
   Three parts, in reading order - which pick, whose, how long. */
.dr-tick{display:inline-flex;align-items:center;gap:8px;padding:4px 4px 4px 9px;
  border:1px solid var(--border-subtle);border-radius:var(--r-md);
  background:var(--bg-surface-raised);white-space:nowrap}
.drt-pk{font:800 var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-primary);letter-spacing:.01em}
.drt-st{font:700 9px/1 var(--font-num);letter-spacing:.13em;color:var(--text-muted)}
/* THE COUNTDOWN IS THE PLATE, not another grey number beside two others. It is
   the only thing in this strip that moves, and it has to read from arm's
   length across a phone held sideways at a draft table. */
.drt-c{display:inline-block;min-width:52px;padding:4px 7px;border-radius:var(--r-sm);
  background:var(--turf,#0f1a22);color:#fff;text-align:center;
  font:800 var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums}
.drt-c.low{background:var(--negative-ink);color:#fff}
/* YOUR TURN inverts the whole chip, exactly as it inverts the whole rail and
   the whole cell - one gesture for one moment, drawn three times in three
   places rather than three different highlights. */
.dr-tick.you{background:var(--turf,#0f1a22);border-color:var(--turf,#0f1a22)}
.dr-tick.you .drt-pk{color:#fff}
.dr-tick.you .drt-st{color:var(--accent)}
.dr-tick.you .drt-c{background:rgba(255,255,255,.14);color:#fff}
.dr-tick.you .drt-c.low{background:var(--negative-ink)}
@media (max-width:520px){
  /* The strip also holds two selects at this width. The pick and the clock are
     the two facts that cannot be dropped; the state word is stated by the rail
     one object down and by the inverted cell on the board. */
  .drt-st{display:none}
  .dr-tick{gap:6px;padding-left:7px}
  .drt-c{min-width:46px;padding:4px 5px}
}
#draftDate{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);color:var(--text-primary);
  font:600 var(--fs-sm)/1 var(--font-ui);padding:9px 8px;min-height:36px}
#draftDate:focus{outline:none;border-color:var(--accent)}

/* ====================================================== BOARD DRAFT BANNER */
.draftbanner{display:flex;align-items:center;gap:var(--sp-3);
  background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-2) var(--sp-3);margin:var(--sp-3) 0;
  font-size:var(--fs-sm)}
.draftbanner .tag{font:700 var(--fs-xs)/1 var(--font-num);letter-spacing:var(--track-caps);
  color:var(--text-muted);text-transform:uppercase;white-space:nowrap}
.draftbanner .body{color:var(--text-secondary)}
.draftbanner .body b{color:var(--text-primary);font-weight:600}
.draftbanner .fchip:first-of-type{margin-left:auto}

/* ============================================== ROS SPARKLINE (player card) */
.pc-spark{width:100%;height:46px;display:block;margin-top:var(--sp-3)}
.pc-spark .sp{fill:var(--text-muted)}
/* playoff weeks are the ones that decide it; the tooltip says so too */
.pc-spark .sp.po{fill:var(--accent)}
.pc-spark .sp-none{fill:var(--border-strong)}
/* PAR column */
.par{color:var(--text-secondary);font-weight:600}

/* the chip inside a grid cell is the same ghost chip the board uses */

/* ============================================================== MY BOARD */
.mb-row{display:flex;align-items:flex-start;gap:var(--sp-3);flex-wrap:wrap}
.mb-rank{flex:1;min-width:130px;background:var(--bg-surface-raised);
  border:1px solid var(--border-subtle);border-radius:var(--r-md);padding:var(--sp-3)}
.mb-rank .k{font-size:var(--fs-xs);letter-spacing:var(--track-caps);
  text-transform:uppercase;color:var(--text-muted);font-weight:600}
.mb-rank .v{font:700 var(--fs-xl)/1.1 var(--font-num);font-variant-numeric:tabular-nums;margin-top:3px}
.mb-rank .n{font-size:var(--fs-xs);color:var(--text-muted);margin-top:2px}
.mb-d{font-size:var(--fs-sm)}
.mb-btns{display:flex;flex-direction:column;gap:var(--sp-2);min-width:120px}
.mb-pin{background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);color:var(--text-primary);font:600 var(--fs-sm)/1 var(--font-ui);
  padding:9px 10px;min-height:36px;width:100%}
.mb-pin:focus{outline:none;border-color:var(--accent)}
/* a moved player is marked on the board itself */
.mk-mv{color:var(--accent);font-weight:800}

/* ============================================================ PREP DRAWER */
.prep{background:var(--bg-surface);border:1px solid var(--border-strong);
  border-radius:var(--r-md);padding:var(--sp-3) var(--sp-4);margin-bottom:var(--sp-3)}
.prep-hd{display:flex;align-items:center;gap:var(--sp-3);margin-bottom:var(--sp-2)}
.prep-hd .caps:first-child{color:var(--accent)}
.prep-hd .fchip:nth-last-child(2){margin-left:auto}
.prep-empty{font-size:var(--fs-sm);color:var(--text-muted);line-height:1.7}
.prep-sec{margin-top:var(--sp-3);padding-top:var(--sp-2);border-top:1px solid var(--border-subtle)}
.prep-sec>.caps{display:block;margin-bottom:6px}
.prep-row{display:flex;flex-wrap:wrap;gap:6px}
.prep-chip{display:inline-flex;align-items:center;gap:5px;min-height:36px;
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  border-radius:var(--r-sm);padding:4px 8px;cursor:pointer}
.prep-chip:hover{border-color:var(--border-strong);background:var(--bg-row-hover)}
.prep-chip .pc-n{font:600 var(--fs-sm)/1 var(--font-ui);color:var(--text-primary);white-space:nowrap}
.prep-chip .pc-d{font:700 var(--fs-xs)/1 var(--font-num)}
.prep-note{font-size:var(--fs-sm);color:var(--text-secondary);line-height:1.6;padding:3px 0}
.prep-note b{color:var(--text-primary);margin-right:6px}

/* ---------------------------------------------------------------- WHY panel
 * COLOR IS A BUDGET here, exactly as the third palette established. The loud
 * end of the wheel is spent on DISAGREEMENT, because disagreement is where
 * this product says edge lives: a DISLOCATED verdict gets accent (buy) or
 * warm (fade), CROWDED gets the risk yellow, and CONVICTION - the verdict
 * that says "the price is fine" - is deliberately QUIET. A card that shouts
 * agreement teaches the eye to ignore the shout.                            */
.why-verdict{display:flex;flex-direction:column;gap:2px;padding:var(--sp-3);
  border:1px solid var(--border-subtle);border-left-width:3px;
  border-radius:var(--radius-sm);margin-bottom:var(--sp-3);
  background:var(--bg-surface-raised)}
.why-vlabel{font:700 var(--fs-xs)/1 var(--font-num);letter-spacing:.08em;
  text-transform:uppercase;color:var(--text-secondary)}
.why-vdetail{font-size:var(--fs-sm);color:var(--text-secondary);line-height:1.5}
.why-verdict.dislocated{border-left-color:var(--accent);background:var(--accent-soft)}
.why-verdict.dislocated .why-vlabel{color:var(--accent-hover)}
.why-verdict.crowded{border-left-color:var(--risk);background:var(--risk-soft)}
.why-verdict.crowded .why-vlabel{color:var(--warning)}
.why-verdict.conviction{border-left-color:var(--border-strong)}
.why-verdict.none{border-left-color:var(--border-subtle);background:transparent}
.why-verdict.none .why-vlabel{color:var(--text-muted)}

.why-row{display:grid;grid-template-columns:92px 1fr;gap:var(--sp-3);
  padding:var(--sp-2) 0;border-top:1px solid var(--border-subtle)}
.why-row:first-of-type{border-top:0}
.why-lab{font:600 var(--fs-xs)/1.5 var(--font-num);letter-spacing:.06em;
  text-transform:uppercase;color:var(--text-muted)}
.why-list{margin:0;padding:0;list-style:none;display:flex;flex-direction:column;
  gap:var(--sp-2)}
.why-list li{font-size:var(--fs-sm);color:var(--text-primary);line-height:1.5}
.why-note{font-size:var(--fs-xs);color:var(--text-muted);line-height:1.5;
  margin-top:2px}

/* The provenance chip. Every receipt carries one - constitution 34 - and the
 * four confidence classes are visually RANKED, so "verified" and "estimate"
 * can never be mistaken for each other at a glance. */
.why-chip{display:inline-block;font:600 10px/1.4 var(--font-num);
  letter-spacing:.04em;text-transform:uppercase;padding:1px 5px;
  border-radius:3px;border:1px solid var(--border-subtle);
  color:var(--text-muted);white-space:nowrap;vertical-align:1px}
.why-chip.verified{color:var(--accent-hover);border-color:var(--accent-border)}
.why-chip.reported{color:var(--text-secondary);border-color:var(--border-strong)}
.why-chip.prior{color:var(--warning);border-color:rgba(250,178,25,.4)}
.why-chip.estimate{color:var(--text-muted);border-style:dashed}

/* The range band, in WHOLE points - constitution 35. Deliberately drawn as an
 * interval with two endpoints and no midpoint marker: a dot in the middle
 * would read as a prediction, which is the precise claim this band refuses to
 * make. */
.why-band{display:flex;align-items:center;gap:var(--sp-2);
  margin:var(--sp-2) 0 0 calc(92px + var(--sp-3))}
.why-band-lo,.why-band-hi{font:700 var(--fs-md)/1 var(--font-num);
  color:var(--text-primary)}
.why-band-bar{position:relative;flex:0 0 64px;height:3px;
  background:var(--border-subtle);border-radius:2px}
.why-band-bar i{position:absolute;top:0;bottom:0;background:var(--accent);
  border-radius:2px;opacity:.55}
.why-band-cap{font-size:var(--fs-xs);color:var(--text-muted)}

@media (max-width:640px){
  .why-row{grid-template-columns:1fr;gap:var(--sp-1)}
  .why-band{margin-left:0}
}

/* ---------------------------------------------- the panel, template pieces */
/* The star: a real control wired to the queue, not the template's decoration. */
.pc-star{position:absolute;top:var(--sp-3);left:var(--sp-3);background:none;border:none;
  color:var(--text-muted);font-size:17px;line-height:1;padding:5px 7px;border-radius:var(--r-sm)}
.pc-star:hover{color:var(--accent);background:var(--bg-row-hover)}
.pc-star.on{color:var(--accent)}
.pc-id{padding-left:26px}
.pc-team .pcd-logo{width:15px;height:15px;vertical-align:-3px;margin-right:5px}

/* Chips carry MEANING in their colour, as the design does: the tier band and
   the read tag are judgements, bye and jersey are just facts. */
.pc-sub .f.tier{border-color:color-mix(in srgb,var(--warning) 45%,transparent);color:var(--warning)}
.pc-sub .f.rank{border-color:var(--border-strong);color:var(--text-primary)}

/* Verdict: a bold two-tone headline over the sentence, per the template. */
.pc-vhead{font:800 var(--fs-xl)/1.2 var(--font-num);letter-spacing:.04em;margin-bottom:5px}
.pc-vhead b.val{color:var(--positive-ink)}
.pc-vhead b.fade{color:var(--negative-ink)}
.pc-vhead b.none,.pc-vhead b{color:var(--text-primary)}
.pc-vhead b.val,.pc-vhead b.fade{font-weight:800}
.pc-vhead span{color:var(--text-muted);font-weight:600;font-size:var(--fs-lg)}
.pc-vhead b.weak{opacity:.8}

/* MARKET CONSENSUS: where he actually goes, with your pick marked. */
.pc-mc{margin-top:var(--sp-3)}
.mc-hd{display:flex;align-items:baseline;gap:var(--sp-3)}
.mc-num{font:700 26px/1 var(--font-num);font-variant-numeric:tabular-nums;color:var(--text-primary)}
.mc-cap{font-size:var(--fs-xs);color:var(--text-muted);line-height:1.4}
.mc-chart{position:relative;margin-top:var(--sp-4)}
.mc-bars{display:flex;align-items:flex-end;gap:2px;height:56px}
.mc-b{flex:1;display:flex;align-items:flex-end}
.mc-b i{display:block;width:100%;border-radius:2px 2px 0 0;background:var(--border-strong)}
.mc-b.you i{background:var(--accent);box-shadow:0 0 10px color-mix(in srgb,var(--accent) 45%,transparent)}
.mc-you{position:absolute;top:-16px;transform:translateX(-50%);white-space:nowrap;
  font:800 9px/1 var(--font-num);letter-spacing:.06em;color:var(--on-accent);
  background:var(--accent);border-radius:3px;padding:2px 5px}
.mc-ax{display:flex;justify-content:space-between;margin-top:5px;
  font:600 var(--fs-xs)/1 var(--font-num);color:var(--text-muted)}

/* ---------------------------------------------- momentum + injury tabs */
.pagehead{font:700 var(--fs-xl)/1.2 var(--font-ui);margin:var(--sp-4) 0 var(--sp-2)}
.method{border:1px solid var(--border-subtle);border-radius:var(--r-md);
  background:var(--bg-surface);padding:var(--sp-3) var(--sp-4);
  font-size:var(--fs-sm);color:var(--text-secondary);line-height:1.6;
  margin-bottom:var(--sp-3)}
.method .caps{display:block;margin-bottom:4px}
.method code{color:var(--text-primary);font-size:var(--fs-xs)}
#moveTable td,#injTable td{text-align:right}
#moveTable td.tl,#injTable td.tl{text-align:left}

/* prep: the checks that say whether a queue holds together */
.prep-warn{list-style:none;margin:var(--sp-2) 0 0}
.prep-warn li{font-size:var(--fs-sm);color:var(--text-secondary);line-height:1.5;
  padding:5px 0 5px 14px;position:relative}
.prep-warn li::before{content:"";position:absolute;left:0;top:11px;width:5px;height:5px;
  border-radius:50%;background:var(--risk)}
.prep-warn b{color:var(--text-primary)}
.prep-shape{margin-left:var(--sp-3);display:inline-flex;gap:5px;align-items:center;
  font-weight:600;color:var(--text-secondary)}
.prep-chip.gone{opacity:.45;text-decoration:line-through}

/* A projection the vendors disagree about is not the same claim as one they
   agree on, and the card must not present them identically. */
.pc-kv.contested{border-color:var(--risk-border);background:var(--risk-soft)}
.pc-kv.contested .k,.pc-kv.contested .v{color:var(--risk)}

/* ══════════════════ COMMAND-CENTRE PANELS ══════════════════
   The KPI row and rank boards the Ice Editorial template opens both rooms
   with. Unfilled cards on the page ground, hairline rules, serif figures -
   the same three moves as the masthead, so the rooms and the board read as
   one publication rather than as two apps. */
/* EXPLICIT COLUMN COUNTS, not auto-fit. A spanning child and `auto-fit`
   disagree: auto-fit sizes the track list from the items it can see one at a
   time, so the 2-column tile pushed the count past what fits and left a
   trailing 0px track - measured "280px 280px 280px 280px 0px" at 1400. That is
   the same `1fr`-shrinks-below-its-contents trap the card foot hit today, and
   the same lesson: when a grid contains a span, the author picks the count. */
.kpi-strip{display:grid;gap:var(--sp-3);margin:var(--sp-4) 0 var(--sp-5);
  grid-template-columns:repeat(6,minmax(0,1fr))}
@media (max-width:1500px){.kpi-strip{grid-template-columns:repeat(4,minmax(0,1fr))}}
@media (max-width:1020px){.kpi-strip{grid-template-columns:repeat(2,minmax(0,1fr))}}
.kpi{position:relative;overflow:hidden;padding:16px 18px;
  border:1px solid var(--border-subtle);border-radius:var(--r-lg)}
.kpi.wide{grid-column:span 2}
.kpi .caps{display:block;margin-bottom:var(--sp-2)}
.kpi-name{font:500 19px/1.2 var(--font-display);color:var(--text-primary);
  margin-bottom:2px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.kpi-val{font:400 40px/1.05 var(--font-display);font-variant-numeric:tabular-nums;
  color:var(--text-primary)}
.kpi-val.up{color:var(--positive)} .kpi-val.dn{color:var(--negative)}
.kpi-val.risk{color:var(--risk)}
.kpi-note{margin-top:6px;font-size:var(--fs-sm);color:var(--text-muted);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.kpi .spark,.kpi svg.spark{position:absolute;right:0;bottom:0;opacity:.5}
.donut{position:absolute;right:16px;top:50%;transform:translateY(-50%);
  width:74px;height:74px}
.donut .dt{fill:none;stroke:var(--border-subtle);stroke-width:4}
.donut .dseg{fill:none;stroke-width:4;stroke-linecap:butt}
.donut .dseg.dn{stroke:var(--negative)} .donut .dseg.risk{stroke:var(--risk)}
.donut .dseg.mut{stroke:var(--accent)} .donut .dseg.mut2{stroke:var(--neutral)}

/* Section headings are the masthead's serif one step down, so the page has a
   type hierarchy instead of four sizes of the same uppercase label. */
.sect{margin:var(--sp-5) 0 var(--sp-3);font:400 24px/1.15 var(--font-display);
  color:var(--text-primary);letter-spacing:-.004em}
/* repeat(5), not auto-fit: these boards hold exactly five cards, and auto-fit
   produced "272px x5 then 0px 0px" at 1990 - two empty tracks, the same trap
   the KPI strip hit. When the CONTENT fixes the count, the author states it. */
.rb-grid{display:grid;gap:var(--sp-3);
  grid-template-columns:repeat(5,minmax(0,1fr))}
@media (max-width:1100px){.rb-grid{grid-template-columns:repeat(3,minmax(0,1fr))}}
@media (max-width:700px){.rb-grid{grid-template-columns:repeat(2,minmax(0,1fr))}}
.rc{position:relative;overflow:hidden;padding:14px 16px 16px;cursor:pointer;
  border:1px solid var(--border-subtle);border-radius:var(--r-lg)}
.rc:hover{border-color:var(--accent-border);background:var(--bg-row-hover)}
.rc-top{display:flex;align-items:center;justify-content:space-between;gap:8px}
.rc-n{font:400 15px/1 var(--font-display);color:var(--text-muted)}
.rc-face{height:74px;margin:6px 0 2px;display:flex;align-items:flex-end}
.rc-face img{height:74px;width:auto;object-fit:contain}
.rc-name{font:500 17px/1.2 var(--font-display);color:var(--text-primary);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.rc-sub{font-family:var(--font-num);font-size:var(--fs-xs);letter-spacing:.1em;
  text-transform:uppercase;color:var(--text-muted);margin-top:2px}
.rc-val{margin-top:8px;font:400 27px/1 var(--font-display);
  font-variant-numeric:tabular-nums}
.rc-val.up{color:var(--positive)} .rc-val.dn{color:var(--negative)}
.rc-val.risk{color:var(--risk)} .rc-val.mut{color:var(--text-secondary)}
.rc-note{margin-top:7px;font-size:var(--fs-sm);color:var(--text-muted);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.rc-rank{margin-top:6px}
.rc .spark{position:absolute;right:0;bottom:0;opacity:.45;pointer-events:none}
@media (max-width:620px){
  .kpi-strip{grid-template-columns:minmax(0,1fr)}
  .kpi.wide{grid-column:auto}
}

/* ══════════════════ PRIMETIME PLAYER CARD ══════════════════
   The team's own broadcast graphic. Every colour in the head derives from one
   hex in TEAM_COLORS via --tc / --tc-d / --tc-l / --tc-p, so a Lions card is
   Honolulu blue and a Chiefs card is red without a second rule being written.
   The gold is NOT team-derived: it is the fixed primetime accent, the one
   thing constant across all 32, which is what makes the set read as a series.

   The oblique is a SKEW, not a synthesised italic. Archivo Black ships no
   italic, so `font-style:italic` asks the browser to slant it for us and every
   engine does that slightly differently; a transform is the same instruction
   everywhere and leaves the glyphs' own spacing alone. */
:root{--pt-gold:#ffd23f}
.pt-hd{position:relative;isolation:isolate}
.pt-hero{position:relative;min-height:208px;overflow:hidden;
  background:linear-gradient(105deg,var(--tc-d) 0%,var(--tc) 58%,var(--tc-d) 100%)}
/* The diagonal jersey stripe, at 4% - present as texture, gone as pattern. */
.pt-hero::before{content:"";position:absolute;inset:0;pointer-events:none;
  background:repeating-linear-gradient(115deg,rgba(255,255,255,.05) 0 2px,
    transparent 2px 15px)}
/* The copy starts BELOW the back pill rather than behind it. The pill is
   absolute, so nothing reserves this space unless it is stated. */
.pt-hero-in{position:relative;z-index:2;padding:64px 28px 16px;max-width:64%}
.pt-name{margin:0;font:400 clamp(34px,4.6vw,58px)/.92 "Archivo Black",
  var(--font-ui);text-transform:uppercase;color:#fff;letter-spacing:-.01em;
  transform:skewX(-9deg);transform-origin:left center;
  text-shadow:0 3px 14px rgba(0,0,0,.4)}
.pt-chips{display:flex;flex-wrap:wrap;gap:7px;margin-top:15px;align-items:center}
.pt-chip{font:600 13px/1 var(--font-label);letter-spacing:.11em;
  text-transform:uppercase;padding:5px 10px;transform:skewX(-9deg);
  background:#fff;color:var(--tc-d)}
/* The BOX is skewed and the TEXT is put back upright inside it - a
   parallelogram chip whose label is legible, which is what the template
   draws. The verdict TAG is the deliberate exception: "AT PRICE" is slanted
   there, and it is the one word on the card that should read as motion. */
.pt-chip>i{display:inline-block;font-style:normal;transform:skewX(9deg)}
.pt-chip.rank{background:var(--pt-gold);color:#111}
.pt-chip.dark{background:rgba(0,0,0,.62);color:#fff}
.pt-chip.inj{background:#2a1215;color:#ff8a80}
.pt-bio{margin-top:13px;font:600 13px/1.4 var(--font-label);
  letter-spacing:.13em;text-transform:uppercase;color:var(--tc-p)}
/* The headshot rides the right edge and dissolves into the band rather than
   sitting in a box - the template's single most identifying move. */
.pt-face{position:absolute;right:0;top:0;bottom:0;width:44%;z-index:1;
  display:flex;align-items:flex-end;justify-content:flex-end}
.pt-face img{height:100%;width:100%;object-fit:cover;object-position:top center;
  -webkit-mask-image:linear-gradient(100deg,transparent 0%,#000 42%);
  mask-image:linear-gradient(100deg,transparent 0%,#000 42%)}

/* The stat strip IS `.pc-band`, restyled. Five cells, not the template's
   four, because the fifth is VALUE - the one number that says the price is
   wrong, and the reason to open the card at all. auto-fit is avoided here on
   today's lesson: an explicit count, since the cell count is known. */
/* COLUMNS FROM THE CELL COUNT, not a number I typed. The band ships four
   cells or five depending on whether a verdict outlier exists - Gibbs is AT
   PRICE, so he has no VALUE cell - and a hard `repeat(5)` left him with a
   dead fifth column. `grid-auto-flow:column` sizes to whatever arrives, which
   is the honest answer when the content decides the count. */
.pt-hd .pc-band{display:grid;gap:0;margin:0;padding:0;border:none;
  grid-auto-flow:column;grid-auto-columns:minmax(0,1fr);
  background:#111417;border-top:3px solid var(--pt-gold)}
.pt-hd .pc-band .b{padding:13px 18px;border-right:1px solid #23282e;min-width:0;
  display:block;background:none}
.pt-hd .pc-band .b:last-child{border-right:none}
.pt-hd .pc-band .bk{display:block;font:600 11px/1 var(--font-label);
  letter-spacing:.2em;text-transform:uppercase;color:var(--pt-gold)}
.pt-hd .pc-band .bv{display:block;margin-top:5px;
  font:400 clamp(20px,2.1vw,29px)/1 "Archivo Black",var(--font-ui);color:#fff;
  transform:skewX(-9deg);transform-origin:left center;
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.pt-hd .pc-band .bv.d-up{color:var(--positive)}
.pt-hd .pc-band .bv.d-dn{color:var(--negative)}
.pt-hd .pc-band .bn{display:block;margin-top:5px;
  font:500 12px/1 var(--font-label);letter-spacing:.1em;
  text-transform:uppercase;color:#7d8790;
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
/* The bio line, in the hero's own light rather than the app's grey. */
.pt-hero .pc-bio{margin-top:13px;font:600 13px/1.4 var(--font-label);
  letter-spacing:.13em;text-transform:uppercase;color:var(--tc-p)}

/* The verdict banner: a skewed tag butted against a skewed body, which is the
   template's strongest single element and the reason the card reads as a
   call rather than as a readout. */
.pt-verdict{display:flex;align-items:stretch;gap:0;padding:16px 22px 18px;
  background:#0a0c0e}
.pt-vtag{flex:none;display:grid;place-items:center;padding:10px 20px;
  transform:skewX(-9deg);background:var(--pt-gold);color:#111;
  font:400 21px/1 "Archivo Black",var(--font-ui);text-transform:uppercase}
.pt-vtag.val{background:var(--positive);color:#04211b}
.pt-vtag.fade{background:var(--negative);color:#2a0f0a}
.pt-vbody{flex:1;min-width:0;display:flex;align-items:center;padding:10px 18px;
  transform:skewX(-9deg);background:#fff;color:#14171a;
  font:500 14px/1.42 var(--font-label);letter-spacing:.01em}
.pt-vbody>i{display:block;font-style:normal;transform:skewX(9deg)}
.pt-vbody b{font-weight:700}
@media (max-width:620px){
  /* THE PORTRAIT GETS ITS OWN COLUMN, which is what the desktop hero has
     always done and what the phone one had given up.

     Two wrong answers came first and both are worth recording. The original
     dropped the photo to opacity .35 and let the copy run over it: the text
     sat on a busy field AND the player was a ghost. The second put the photo
     back to full strength under a gradient scrim - "the player is even more
     obscured now", correctly, because a 62% wash landed straight across the
     face while the copy still overlapped it. Both were trying to make two
     things share one space.

     They do not have to. At 402px the name sets at 34px and measures ~250px,
     so the copy fits in two thirds and the portrait takes the last third
     clean. No scrim, no opacity, no overlap - the photo is anchored right so
     the SUBJECT is in the clear column rather than the middle of the frame,
     and its existing left-edge mask does the blending where the two columns
     meet. */
  /* THE CHIPS WERE THE HEIGHT, and the column was the reason. Four chips -
     RB1, RB · DETROIT LIONS, TIER 1 · BYE 6, QUESTIONABLE - each want about
     150px and none of them got it, so every one took a row of its own and
     left a wide empty gutter beside it. Four rows of chip plus two of name
     plus two of bio is what made the box tall, and the portrait then sat at
     the bottom of a space sized by text it has nothing to do with.

     Widening this column is what SHRINKS the hero: at 74% the chips pair up
     two to a row, which removes two rows outright. The portrait then took the
     width it actually needed rather than the width the old layout gave it, and
     his +20% on top of that lands it at 43% - still under the 350px source, so
     it renders at or below native size and stays sharp. The two columns
     overlap by a few percent where the mask feathers, which is free. */
  .pt-hero-in{max-width:74%;padding:52px 14px 14px}
  .pt-chips{margin-top:10px;gap:6px}
  /* THE LAST ROW, bought for two points of type. TIER 1 · BYE 6 and
     QUESTIONABLE came to ~320px against a 297px column and so took a row
     each; at 11px with tighter cheeks they total ~260 and share one. A chip is
     a label at a glance, not something read at length, so the size it needs is
     the size that fits two to a line. */
  .pt-chip{font-size:11px;padding:4px 8px;letter-spacing:.08em}
  .pt-hero .pc-bio{margin-top:9px;font-size:12px}
  /* THE ELEMENT SPANS THE WHOLE HERO, and that is what finally kills the seam.

     Three attempts fought the symptom: opacity .35 (a ghost player), a scrim
     over the top (worse - a wash straight across the face), then a wider box
     with a softer gradient. All three left a vertical line in the same place,
     because the line was never the gradient. `.pt-face` was a 62%-wide
     absolutely positioned BOX, so its left edge is a hard boundary in the
     middle of the picture - the photo simply stops existing there - and no
     mask can blend an edge the element itself creates.

     Full width removes the boundary. The subject stays right via
     object-position and the mask does the only job left: a pure-horizontal
     ramp with the whole hero to travel in, clear under the copy and solid
     under the portrait. 90deg rather than 100 because a tilted gradient's
     axis is computed from the box diagonal, so on a tall box the horizontal
     part of the transition is far shorter than the percentages suggest -
     which is what made every earlier fade read as a step. */
  /* HIS CALL, and it is the one that stops fighting the picture: "why don't
     you render it in the very right part of the box at a smaller size?"

     Everything before this scaled a 350px source UP to fill a 314px-tall hero
     with object-fit:cover, which crops it to a band and blows it past its own
     resolution - so the compromises were all about hiding a picture that was
     too big and too soft. Contained at 44% it renders near its native size,
     sharp, whole, and entirely clear of the copy. No crop to hide, so the mask
     shrinks to a short left-edge feather instead of a fade across half the
     card. Anchored bottom-right, which is where a cut-out portrait sits. */
  /* THE PORTRAIT FILLS ITS COLUMN, TOP TO BOTTOM, and the box comes in to
     meet it. Contained-at-44% left a tall wedge of empty blue above the
     player's head: the picture sat on the floor of a box sized by the COPY
     beside it, so the two never agreed on a height. Now the image takes the
     full height of the hero and keeps its own aspect, and the hero stops
     reserving space nothing was using. */
  .pt-face{opacity:1;left:auto;right:0;width:43%;top:0;bottom:0;
    align-items:stretch;justify-content:flex-end;padding:0}
  .pt-face img{width:100%;height:100%;
    object-fit:contain;object-position:bottom right;
    -webkit-mask-image:linear-gradient(90deg,transparent 0%,#000 22%);
    mask-image:linear-gradient(90deg,transparent 0%,#000 22%)}
  /* TWO LINES IS THE RIGHT ANSWER for this name and not a defect: "JAHMYR"
     alone measures 201px of a 249px column in Archivo Black, so one line would
     need ~19px - a caption, not a hero. The size is set for the column and the
     wrap is allowed. */
  .pt-name{font-size:clamp(26px,7.4vw,58px)}
  .pt-hd .pc-band{grid-auto-flow:row;grid-template-columns:repeat(2,minmax(0,1fr))}
  .pt-hd .pc-band .b:nth-child(2n){border-right:none}
  .pt-verdict{flex-direction:column;gap:10px}
  .pt-vtag,.pt-vbody{transform:none}
  .pt-vbody>i{transform:none}
}
@media (prefers-reduced-motion:reduce){.pt-hero::before{display:none}}

/* THE ROOKIE TAG, his call: "we just denote they're a rookie, paint that in
   the top corner of the card."
   Top LEFT, because the close and star already own the top right. Skewed and
   hard-edged so it reads as part of the hero's plate language rather than a
   sticker; no gold, because gold is the verdict tag's alone and this is a fact
   about the player, not a call on him. It answers the season plate's empty
   state before the reader gets there - a card with no game log should say why
   at the top, not 2,000px down. */
.pt-rookie{position:absolute;top:14px;left:16px;z-index:3;
  display:inline-block;padding:5px 13px 5px 11px;
  transform:skewX(-9deg);background:rgba(6,9,11,.72);
  border:1px solid rgba(255,255,255,.22);
  /* --font-cond named the condensed face this app carried until the Lions
     skin deleted the Barlow files; it has been declared nowhere since, so
     this has always resolved to --font-ui. Written as what it is, because a
     fallback nothing can override is not a fallback. */
  font:700 10px/1 var(--font-ui);letter-spacing:.22em;
  text-transform:uppercase;color:var(--tc-p,#fff)}
@media (max-width:620px){.pt-rookie{top:10px;left:12px;padding:4px 10px}}

/* ══════════════════ PLAYER CARD BODY ═══════════════════════════════════════
 * From his 2026-08-07 handoff, option 2a, appended after the .pt-* hero block
 * exactly as the README specifies. Unmodified: the design is his and it states
 * its own contract (quantity is luminance never hue; team colour on plates and
 * numerals but never on running text; gold stays on the verdict alone).
 *
 * The four --tc/--tc-d/--tc-l/--tc-p custom properties are already set on
 * .pt-hd from teamRamp(), and they cascade in here - so nothing below asks for
 * a team colour and nothing below calls teamRamp().
 * ═══════════════════════════════════════════════════════════════════════════ */
/* ============================================================================
 * PLAYER CARD BODY - everything under .pt-hd, added 2026-08-08.
 *
 * Scope: `.pcard .pb`. The hero (.pt-hero, .pt-band, .pt-verdict) is untouched.
 * This block is the rest of the card: the conviction tiles, the evidence /
 * room / season plates, the floor rail, and the 2025 game log.
 *
 * Three rules govern everything here, and breaking any of them is what made
 * the first pass look generated rather than designed:
 *
 *  1. LABELS ARE GREY, VALUES ARE WHITE, TEAM COLOUR IS STRUCTURE.
 *     --text-secondary carries every caps label. Archivo Black italic carries
 *     every number. var(--tc) appears only on section plates, entry numerals
 *     and the Drafted button - never on running text. A label set in the team
 *     hue at 11px tracked .22em is unreadable, which is exactly what the first
 *     conviction header did.
 *
 *  2. QUANTITY IS ENCODED AS LUMINANCE, NEVER AS HUE.
 *     The floor rail and the game-log heat both ramp dark-slate to bone using
 *     the app's own neutrals. A team-tinted ramp reads fine on ATL and dies on
 *     LAC or GB. See .pb-rail and .pb-log .heat.
 *
 *  3. HALF-WIDTH PLATES COME IN PAIRS.
 *     .pb-grid is a 2-up auto-fit. An odd number of half-width children always
 *     orphans one and leaves a visible void; every .wide child must sit on a
 *     row boundary. See the ordering note on .pb-grid.
 * ========================================================================= */

/* The body field. The hero's diagonal ruling continues underneath at a much
   lower alpha - it is what ties the two halves of the card together, and
   removing it is what made this section read as a different product. */
.pcard .pb{background-color:#0d0b0d;
  background-image:repeating-linear-gradient(115deg,rgba(255,255,255,.014) 0 1px,transparent 1px 8px);
  border-top:1px solid color-mix(in srgb,var(--tc-l) 30%,transparent);
  padding:var(--sp-5) var(--sp-5) var(--sp-1)}

/* ---- plates ------------------------------------------------------------ */
/* Hard 2px corners, not the 8-11px of the rest of the app: the hero is built
   from hard-edged skewed plates and these have to belong to it. */
.pcard .pb-plate{border:1px solid var(--border-subtle);border-top:2px solid var(--border-strong);
  border-radius:2px;background-color:#100e10;
  background-image:repeating-linear-gradient(115deg,rgba(255,255,255,.016) 0 1px,transparent 1px 8px);
  overflow:hidden}
.pcard .pb-plate.open{border-color:color-mix(in srgb,var(--tc-l) 30%,transparent);border-top-color:var(--tc)}

/* GRID PARITY. Order children so half-width plates pair up and every .wide
   sits at a row boundary, e.g.  evidence | room  /  season | finish  /
   [log .wide] / [news .wide]. Five half-width children is a bug, not a layout. */
.pcard .pb-grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(320px,1fr));
  gap:var(--sp-3);align-items:start}
.pcard .pb-plate.wide{grid-column:1/-1}
/* EACH COLUMN IS ITS OWN STACK. A plain grid flows row by row, so a fifth
   plate lands UNDER the first column and never in the gap beside it - which is
   how Evidence ran ten findings deep on the left while The room ended early
   and left a dead column on the right. Two .pb-col children keep the pair
   counting as two half-width items, so the parity rule is untouched. */
.pcard .pb-col{display:flex;flex-direction:column;gap:var(--sp-3);min-width:0}

/* THE STALENESS NUDGE. Raised only when the worker has actually seen a newer
   data layer, so it is never decoration: if it is on screen, the board really
   is out of date. Sized to the topbar's own numeric type and given the risk
   hue rather than the accent, because it reports a problem rather than
   offering a feature. */
.freshen{border:1px solid var(--risk);background:rgba(250,178,25,.10);
  color:var(--risk);cursor:pointer;padding:4px 10px;border-radius:2px;
  font:700 10px/1 var(--font-label);letter-spacing:.16em;text-transform:uppercase}
.freshen:hover{background:rgba(250,178,25,.18)}
/* THE FRESHNESS SLO STRIP. .freshen answers "newer exists than this tab";
   this answers "nothing newer exists because the harvest is down". A full-
   width strip rather than a topbar chip because the stamp legitimately
   yields its topbar space on a phone (.hide-m's documented priority order)
   and a dead pipeline may not: warning tokens, which both grounds define,
   so the strip reads on the lit views and the graphite board alike. */
#sloBanner{background:var(--warning-soft);color:var(--warning);
  border-bottom:1px solid var(--warning);padding:7px var(--sp-4);
  font:600 11.5px/1.45 var(--font-label);letter-spacing:.04em;
  text-align:center}
/* The league-scoring control, sized to sit in the topbar's numeric run so it
   reads as part of the provenance strip rather than a form. */
.fmt{display:inline-flex;align-items:center;gap:7px}
.fmt-teams{color:var(--text-muted)}
/* THEY HAVE TO LOOK LIKE CONTROLS. These were a transparent box behind a
   --border-subtle hairline, which was legible on the old #0b0e10 header and
   stopped being legible the moment the dark surfaces moved onto the card
   grid's #0b0d14: "the toggles between 12 teams .5 ppr is invisible."

   A hairline at .12 alpha is a rule for separating things that are already
   distinct - it was never enough to say "this is pressable", and the ground
   change is what exposed that it had been carrying that job. They take the
   heavier edge and a fill one step off the header, so they read as objects on
   either ground; the caret moves up a step in the ink ramp for the same
   reason. Both are tokens, so the light topbar answers with its own values. */
.fmt select{appearance:none;-webkit-appearance:none;cursor:pointer;
  background:var(--bg-surface);border:1px solid var(--border-strong);border-radius:var(--r-sm);
  padding:3px 20px 3px 8px;color:var(--text-secondary);
  font:700 10px/1 var(--font-label);letter-spacing:.16em;text-transform:uppercase;
  background-image:linear-gradient(45deg,transparent 50%,var(--text-secondary) 50%),
    linear-gradient(135deg,var(--text-secondary) 50%,transparent 50%);
  background-position:calc(100% - 11px) 52%,calc(100% - 7px) 52%;
  background-size:4px 4px,4px 4px;background-repeat:no-repeat;
  /* background-image overrides the shorthand's colour, so the fill is restated
     as its own longhand or the box goes transparent again. */
  background-color:var(--bg-surface)}
.fmt select:hover{color:var(--text-primary);border-color:var(--border-strong)}
.fmt select:focus-visible{outline:2px solid var(--accent);outline-offset:2px}

/* ══════════════════ BOARD CONTROL BAR - his "Classic (6a)" ════════════════
   Eleven controls on one line wrapped differently at every width, and the
   count that explains the filtering sat 400 rows below the chips that caused
   it. Two rows now: filter the pool and say what survived, then choose how the
   survivors are drawn and ordered. Row rules are shared; only the grouping
   changed, so every id and handler is untouched. */
.ctlbar{display:flex;flex-direction:column;border:1px solid var(--border-subtle);
  border-radius:2px;background:var(--bg-panel);margin-bottom:var(--sp-3)}
.ctlrow{display:flex;align-items:center;gap:var(--sp-2);flex-wrap:wrap;
  padding:var(--sp-3) var(--sp-4)}
.ctlrow+.ctlrow{border-top:1px solid var(--border-subtle)}
/* Pushed right, and the FIRST thing in each row that is allowed to grow -
   so the readout and the exports hold the right edge at every width. */
.ctlcount{margin-left:auto;color:var(--text-muted);font-size:var(--fs-sm);
  letter-spacing:.08em;white-space:nowrap}
.ctlend{margin-left:auto;display:flex;align-items:center;gap:var(--sp-2)}
@media (max-width:760px){
  .ctlrow{padding:var(--sp-2) var(--sp-3)}
  /* On a phone the readout and the exports stop being right-aligned: a
     margin-left:auto inside a wrapped row strands them on a line of their own. */
  .ctlcount,.ctlend{margin-left:0}
}

/* ====================================================================
   THE PHONE BOARD.

   Measured before any of this was written: on a 375x812 screen the first
   player card sat 1,339px down. One and two thirds of a screen of
   furniture before a single row of the thing the app is for - and the
   filters that decide WHAT the board shows sat at the top of that pile,
   so changing position twenty rows in meant scrolling back past all of
   it and then scrolling down again to see what changed.

   Nothing here is deleted, and nothing is reordered. The context you
   read ONCE gets compact; the controls you use WHILE reading follow you
   down the page. Everything still exists at every width - a phone gets
   the same board, not a lesser one.
   ==================================================================== */
@media (max-width:620px){
  /* Read once, on a tab you already chose to open - and true of EVERY tab,
     not just the board. Stays in the DOM because the Start-here cover derives
     its destination descriptions from it, and textContent survives
     display:none. */
  .mh-dek{display:none}

  /* A FINGER IS NOT A CURSOR, and this is the one token that says so for every
     table at once - the board, momentum and injuries all measure their rows
     from it. 36px is a fine desktop row and a fiddly phone target: 295 player
     buttons on the momentum table and 102 on injuries were all under 44. */
  :root{--row-h:44px}
  /* The name is inline text inside a data row, so it grows its HIT BOX with
     padding and pulls the box back with a matching negative margin - the
     technique already used to reach 36px, retuned to fill the taller row.
     16px of line + 28px of padding = 44, and the text does not move. */
  .pname{padding:14px 0;margin:-14px 0}

  /* THE WIDE TABLES ARE STILL WIDE, and that is written down rather than
     half-fixed. Momentum and Injuries are 1,180px of columns inside a 359px
     wrapper; they scroll, and scrolling out to the numbers takes the player
     they belong to off the left edge.

     A pinned identity column was built and BACKED OUT, because it needs a
     table refactor this rule cannot fake. Three things were learned and are
     worth not re-discovering:
       - a scroll SHADOW cannot work here: every row paints an opaque
         background over the container, so the container's gradient never shows;
       - --bg-row is `transparent`, so a pinned cell shows every scrolled number
         straight through the name unless it carries a surface of its own;
       - a table under AUTO layout ignores max-width on a td and sizes columns
         from the 1,180px total, so the pinned name came out 322px inside a
         357px wrapper - it hid the very numbers it was meant to sit beside.
     Doing it properly means table-layout:fixed and an explicit width per
     column, which is a real change to a nine-column table and wants its own
     pass. See HANDOFF. */

  /* THE FILTERS FOLLOW YOU. One line that scrolls sideways, pinned under
     the topbar - which is sticky and 53px, so this sits at 53. This is
     the whole reason the pass was worth doing: the position chips are
     the control you touch most and they were the furthest away.

     `display:contents` on the BAR is what makes the pin actually reach the
     board. A sticky element only sticks within its PARENT's box, and the
     bar is 288px tall - so the first attempt pinned the row to a strip it
     was already inside and it scrolled away with everything else at
     scrollY 4000. Dissolving the bar makes the rows children of .wrap,
     which spans the whole board, so the range is the board. The bar's own
     border and margin move onto the rows below. */
  #posFilters{display:contents}
  #posFilters .ctlrow:first-child{
    /* MEASURED, NOT GUESSED. This was a literal 53px - the topbar's height on
       the day it was written - and it broke the first time the topbar changed:
       adding the league-size control wrapped it to two lines at 103px and the
       filter row went and sat underneath it, invisible. A sticky offset is a
       fact about another element, so it is read from that element (see the
       ResizeObserver that writes --topbar-h) instead of remembered here. */
    position:sticky;top:var(--topbar-h,53px);z-index:9;
    flex-wrap:nowrap;overflow-x:auto;overscroll-behavior-x:contain;
    scrollbar-width:none;background:var(--bg-panel);
    /* The DST chip was sliced to "DS" at 390 with the scrollbar hidden - a
       scroller that hides its bar has to say some other way that it scrolls
       (round 9). Same fade as the draft board's edge: one idiom. */
    -webkit-mask-image:linear-gradient(to right,#000 calc(100% - 16px),transparent);
    mask-image:linear-gradient(to right,#000 calc(100% - 16px),transparent)}
  #posFilters .ctlrow:first-child::-webkit-scrollbar{display:none}
  #posFilters .ctlrow:first-child>*{flex:none}
  #posFilters .ctlrow:first-child::after{content:"";flex:none;width:8px}
  /* THE COUNT SCROLLS WITH THE ROW, and that is a reversal.
     It was pinned to the right edge so "did that filter do anything" stayed
     answerable at the moment a chip was pressed. Two problems, both only
     visible with real state: it is opaque and permanently covers whatever chip
     is beneath it, and the string GROWS when a filter is on - "397 shown"
     became "28 of 397 shown" and buried the very chip that had just been
     tapped. Pinning a variable-width label over a scroller is a trade that
     gets worse exactly when it is needed. It rides at the end of the row now,
     one flick away, covering nothing. */
  #posFilters .ctlrow:first-child .ctlcount{margin-left:var(--sp-2)}

  /* THE SEARCH TRAVELS WITH THEM. His call, 2026-08-11 (6th): "the search bar
     should travel with the hero and the buttons." The chips pinned and the
     search did not, so the one control you reach for while scrolling a
     397-row board was four hundred rows above you.

     It pins UNDER the chip row, so its offset is two facts about two other
     elements - and the chip row is exactly the thing that wraps at the widths
     where this matters. Both are measured and published by ResizeObservers in
     app.js; the fallbacks are only for the first paint before either has run.
     Writing either number down here is the literal that this file already
     records paying for once, when a 53px topbar became 103px and the filter
     row went to sit underneath it. */
  /* IT BLEEDS TO THE EDGES, and it has to. Once the bar is dissolved the
     search is a child of .wrap, which carries `padding:0 var(--sp-2)` at this
     width - so a background painted on the search itself stops short of the
     screen on both sides, and the board's rows travel up through those two
     gaps as it pins. Reported as "that search bar is bleeding on the margin".
     Negative inline margin cancels the wrap's padding and the same amount is
     given back as padding, so the STRIP spans the viewport while the input
     inside it stays on the page's own column - which is what the chips row
     above it already does via its own scroller. */
  /* LONGHANDS, because the shorthand did not survive. `margin:0 calc(...) X`
     computed -8px on the left and 0 on the right, so the strip bled one side
     and left an 8px gutter on the other for the board to travel up through -
     which is the artifact, not the fix for it. Stated as two properties that
     cannot be half-applied. */
  #posFilters .search{position:sticky;z-index:8;
    top:calc(var(--topbar-h,53px) + var(--filters-h,52px));
    margin-left:calc(-1 * var(--sp-2));margin-right:calc(-1 * var(--sp-2));
    margin-top:0;margin-bottom:var(--sp-2);
    /* AND width:auto, which is the whole reason the negative margins looked
       like they were half-applied. `.search` carries `width:100%` from the
       control bar, and a percentage width is resolved against the CONTAINING
       BLOCK - so the strip kept a 386px width and the negative margins only
       slid it left, bleeding one edge and leaving the other 8px gutter open.
       Both margins were correct in the computed style the whole time. */
    width:auto;
    padding:6px var(--sp-2);background:var(--bg-panel);
    border-bottom:1px solid var(--border-subtle)}

  /* The dissolved bar's own dress, reapplied to the rows that survived it. */
  #posFilters .ctlrow{border:1px solid var(--border-subtle);
    background:var(--bg-panel)}
  #posFilters .ctlrow:first-child{border-bottom:none}
  #posFilters .ctlrow:last-child{margin-bottom:var(--sp-3)}

  /* 44px, because a finger is not a cursor. These were 32.
     Set on the BUTTON SYSTEM's own group selector rather than per screen, so a
     control added anywhere is thumb-sized on a phone by construction - the same
     reason the system exists at all. Only the height moves; the gradient face,
     the lip and the ink are untouched, which is what the key guard checks. */
  .key,.fchip,.segbar .seg,.batab,button.primary{height:44px;padding:0 14px}
  /* The natives, which size themselves from font and padding - plus the four
     that pin their own min-height at a desktop size and therefore need the
     same specificity to move: the two sort controls, the lens and baseline
     selects, and the draft date. */
  select,input[type="text"],input[type="search"],input[type="date"],
  input[type="number"]{min-height:44px}
  .fsort select,.fdir,#baselineSel,#lensSel,#draftDate{min-height:44px}
  .fdir{min-width:44px}
  .pc-actions .fchip{min-height:44px}
  /* The three that carry their own geometry: the hero's step arrows, the
     movement chips in the overnight strip, and the draft grid's team headers. */
  .hero-step{min-width:44px;min-height:44px}
  .mv-chip{min-height:44px}
  .dg-team{min-height:44px}
  /* THE ONE THAT MATTERS MOST, and the one a hand check missed: `.draftbtn` is
     the button you press to draft a player out of Best Available, and it is
     bare 10px text - a 10px target for the action taken under a draft clock.
     It only renders while a draft is running, which is why walking the tabs by
     hand never saw it and the suite did. Same padding-plus-negative-margin hit
     box as the player name: 10px of line + 34px of padding = 44, text unmoved. */
  .draftbtn{position:relative;z-index:0;padding:17px 10px;margin:-17px 0;
    display:inline-block;color:var(--accent);font-weight:700}
  /* AND IT LOOKS LIKE WHAT IT IS (round 9; louder in round 11). The 44px hit
     box fixed the thumb and left the eye: two panels running called the
     visible 10px text an under-floor control, and after round 9's hairline
     ring the round-10 generalist still read the pill as DISABLED. Enabled
     now looks enabled: full-strength accent ring, soft accent fill, 700
     text. The ring is a pseudo-element inside the same hit box - the
     clickable area is unchanged - and it sits under the label in the
     button's own stacking context so the fill can never wash the word.
     Off-state stays bare text: a disabled ring invites a dead control. */
  .draftbtn::before{content:"";position:absolute;inset:9px 0;z-index:-1;
    border:1.5px solid var(--accent);background:var(--accent-soft);
    border-radius:14px}
  .draftbtn.off{color:var(--text-dim)}
  .draftbtn.off::before{content:none}

  /* The movers are a glance, not a grid. Ten chips wrapped into 274px of
     a phone; one scrollable line says the same thing in 60. */
  #movedStrip{flex-wrap:nowrap;overflow-x:auto;overscroll-behavior-x:contain;
    scrollbar-width:none}
  #movedStrip::-webkit-scrollbar{display:none}
  #movedStrip>*{flex:none}

  /* Both questions kept, in less room. The hero band is a headline and a
     line of evidence; at 148px each it was spending a fifth of the screen
     on two of them. */
  .hz-ban{min-height:0}
  .hz-ban .hz-in{padding:12px 14px}
}
.freshen:focus-visible{outline:2px solid var(--risk);outline-offset:2px}
/* The stamp is what claimed the old date, so it is what goes quiet. */
#stamp.stale{color:var(--text-muted);text-decoration:line-through}
@media (max-width:620px){.freshen{padding:3px 7px;letter-spacing:.1em}}

/* The skewed label plate. -9deg and the counter-skew on the inner <i> are the
   hero's, exactly; the negative margin lets it break the plate's left edge. */
.pcard .pb-head{width:100%;display:flex;align-items:stretch;padding:0;
  background:none;border:none;cursor:pointer;text-align:left}
.pcard .pb-tab{display:grid;place-items:center;padding:9px 18px 9px 14px;
  background:var(--tc);transform:skewX(-9deg);margin-left:-4px}
.pcard .pb-tab.q{background:rgba(140,180,200,.12)}
.pcard .pb-tab i{display:block;font-style:normal;transform:skewX(9deg);
  font:700 12px/1 var(--font-label);letter-spacing:.2em;text-transform:uppercase;color:#fff}
.pcard .pb-tab.q i{color:var(--text-primary)}
.pcard .pb-head .meta{flex:1;display:flex;align-items:center;justify-content:flex-end;
  gap:var(--sp-2);padding:0 15px}
.pcard .pb-head .mark{font:400 14px/1 var(--font-num);color:var(--text-secondary);
  width:10px;text-align:center}
.pcard .pb-body{border-top:1px solid var(--border-subtle)}

/* ---- the three type roles ---------------------------------------------- */
.pcard .pb-k{font:700 10px/1 var(--font-label);letter-spacing:.2em;
  text-transform:uppercase;color:var(--text-secondary)}
.pcard .pb-k.sm{font-size:9px;letter-spacing:.16em}
.pcard .pb-v{font:400 18px/1 "Archivo Black",sans-serif;font-style:italic;color:var(--text-primary)}
.pcard .pb-v.lg{font-size:22px}
.pcard .pb-u{font:700 9px/1 var(--font-label);letter-spacing:.16em;
  text-transform:uppercase;color:var(--text-muted)}
.pcard .pb-n{font:400 13px/1.45 var(--font-ui);color:var(--text-secondary)}
.pcard .pb-n.hi{color:var(--text-primary)}

/* Annotations sit behind a hairline so they read as subordinate rather than as
   leftover grey body copy. */
.pcard .pb-note{padding-left:11px;border-left:1px solid var(--border-strong);
  font:400 12px/1.55 var(--font-ui);color:var(--text-muted)}

/* Provenance badges. Deliberately quiet - mono, hairline, no colour. They
   qualify a claim; they are not the claim. */
.pcard .pb-badge{font:500 9px/1 var(--font-num);letter-spacing:.12em;color:var(--text-muted);
  border:1px solid var(--border-strong);border-radius:2px;padding:3px 5px}
.pcard .pb-badge.bare{border:none;padding:0}

/* ---- tiles: the one container pattern ---------------------------------- */
/* A 1px gutter of the grid's own background showing between lifted cells. Used
   for the conviction pillars, every stat strip and the vegas/usage rows, so the
   card has ONE way of drawing a group of numbers rather than four. */
.pcard .pb-tiles{display:grid;gap:1px;background:rgba(140,180,200,.10);
  border:1px solid rgba(140,180,200,.10)}
.pcard .pb-tiles.c2{grid-template-columns:repeat(2,minmax(0,1fr))}
.pcard .pb-tiles.c3{grid-template-columns:repeat(3,minmax(0,1fr))}
.pcard .pb-tiles.c4{grid-template-columns:repeat(4,minmax(0,1fr))}
.pcard .pb-tiles.c5{grid-template-columns:repeat(5,minmax(0,1fr))}
.pcard .pb-tiles.fit{grid-template-columns:repeat(auto-fit,minmax(230px,1fr))}
.pcard .pb-tile{background:#0d0b0d;padding:9px 11px}
/* Conviction pillars get the top-lit variant so six of them read as six tiles
   rather than as quadrants of one field. NOT alternated light/dark: the grid is
   auto-fit, so an odd/even pattern checkerboards at 3 columns and stripes at 2. */
.pcard .pb-tile.lift{background:linear-gradient(180deg,#171418,#100e11);padding:13px 16px}

/* ---- numbered entries (evidence, news) --------------------------------- */
.pcard .pb-entry{display:grid;grid-template-columns:34px 1fr;padding:14px 16px 14px 0;
  border-bottom:1px solid rgba(140,180,200,.08);text-decoration:none}
.pcard .pb-entry:last-child{border-bottom:none}
.pcard .pb-num{text-align:center;padding-top:2px;color:var(--tc);
  font:400 13px/1 "Archivo Black",sans-serif;font-style:italic}
.pcard .pb-entry .hd{display:flex;align-items:baseline;justify-content:space-between;gap:var(--sp-2)}

/* ---- floor rail -------------------------------------------------------- */
/* One skewed segment per point between floor and ceiling, ramping in LUMINANCE
   only. Endpoints are the app's neutrals, so the gauge is legible on all 32
   cards; team identity is the 9% wash behind it and nothing else. The white
   notch is ringed in the surface colour so it survives wherever it lands. */
.pcard .pb-rail{flex:1;position:relative;padding:5px 3px;
  background:color-mix(in srgb,var(--tc) 9%,transparent)}
.pcard .pb-rail .segs{display:flex;gap:3px}
.pcard .pb-rail .seg{flex:1;height:22px;transform:skewX(-9deg)}
.pcard .pb-rail .notch{position:absolute;top:-4px;bottom:-4px;width:2px;background:#fff;
  transform:skewX(-9deg);box-shadow:0 0 0 2px rgba(13,11,13,.85)}
.pcard .pb-railwrap{display:flex;align-items:center;gap:var(--sp-4)}
.pcard .pb-railwrap .end{font:400 26px/1 "Archivo Black",sans-serif;font-style:italic;
  color:var(--text-primary)}
.pcard .pb-railkeys{display:flex;justify-content:space-between;margin-top:10px}

/* ---- game log ---------------------------------------------------------- */
/* Tracks sum to --pb-logw. Do NOT set a min-width larger than that total: the
   surplus becomes dead space to the right of the last column and buys a
   horizontal scrollbar that reveals nothing. */
.pcard .pb-log{--pb-logw:652px;overflow-x:auto}
.pcard .pb-log .inner{min-width:var(--pb-logw)}
/* The template is set INLINE per position now (see PB_LOG_SETS): a receiver
   has no rushing band and a quarterback has a passing one, so the track list
   cannot be a constant here. The RB set is the fallback and is the same 14
   tracks this rule always carried. */
.pcard .pb-log .r{display:grid;
  grid-template-columns:var(--pb-cols,30px 50px 44px 52px 120px 52px 38px 46px 40px 30px 38px 36px 46px 30px);
  align-items:center;min-width:var(--pb-logw);border-bottom:1px solid rgba(140,180,200,.07)}
.pcard .pb-log .grp{font:700 9px/1 var(--font-label);letter-spacing:.2em;text-transform:uppercase;
  color:var(--text-secondary);text-align:center;padding-bottom:5px;margin:0 4px;
  border-bottom:1px solid var(--border-strong)}
.pcard .pb-log .grp.lead{border-bottom-color:rgba(140,180,200,.3)}
.pcard .pb-log .h{font:700 9px/1 var(--font-label);letter-spacing:.14em;text-transform:uppercase;
  color:var(--text-secondary);padding:7px 0;border-bottom:1px solid rgba(140,180,200,.16)}
.pcard .pb-log .c{padding:8px 8px 8px 0;text-align:right;
  font:500 11px/1 var(--font-num);color:var(--text-primary)}
.pcard .pb-log .c.dim{color:var(--text-secondary)}
.pcard .pb-log .c.wk{text-align:left;padding-left:0;color:var(--text-secondary);font-weight:600}
.pcard .pb-log .c.opp{text-align:left;font:700 11px/1 var(--font-label);letter-spacing:.1em}
/* Scoring weight, shaded by brightness only - see rule 2. Alpha is set inline
   from the value; the class only fixes the type. */
.pcard .pb-log .c.heat{padding:8px;font:400 14px/1 "Archivo Black",sans-serif;font-style:italic}
/* The column that earns the table. Actual minus projection, diverging off a
   centre line, so boom and bust weeks are visible without reading 17 numbers. */
.pcard .pb-log .delta{padding:8px 10px 8px 4px;display:flex;align-items:center;gap:9px}
.pcard .pb-log .delta .track{position:relative;flex:1;height:14px}
.pcard .pb-log .delta .track::before{content:"";position:absolute;left:50%;top:0;bottom:0;
  width:1px;background:rgba(140,180,200,.18)}
.pcard .pb-log .delta i{position:absolute;top:4px;height:6px;transform:skewX(-9deg)}
.pcard .pb-log .delta i.up{left:50%;background:rgba(61,220,151,.55)}
.pcard .pb-log .delta i.dn{right:50%;background:rgba(255,92,114,.55)}
.pcard .pb-log .delta b{font:600 10px/1 var(--font-num);width:30px;text-align:right;font-weight:600}
.pcard .pb-log .delta b.up{color:var(--positive-ink)}
.pcard .pb-log .delta b.dn{color:var(--negative-ink)}
.pcard .pb-log .snapbar{margin-top:4px;height:2px;background:rgba(140,180,200,.14)}
.pcard .pb-log .snapbar i{display:block;height:100%;background:var(--text-secondary)}
/* Bye is a rule, not an empty row - an empty row reads as missing data. */
.pcard .pb-log .bye{display:flex;align-items:center;gap:var(--sp-3);padding:9px 0;
  min-width:var(--pb-logw);border-bottom:1px solid rgba(140,180,200,.07)}
.pcard .pb-log .bye .ln{flex:1;height:1px;
  background:repeating-linear-gradient(90deg,rgba(140,180,200,.22) 0 4px,transparent 4px 9px)}
/* Totals are NEUTRAL. A team-tinted band means this row is maroon on ATL and
   navy on BAL; the summary of a season should look the same on every card. */
.pcard .pb-log .tot{border-top:2px solid rgba(140,180,200,.34);
  background:linear-gradient(180deg,#191c1f,#111417)}
.pcard .pb-log .tot .c{padding-top:10px;padding-bottom:10px}

/* ---- schedule grid ----------------------------------------------------- */
.pcard .pb-wk{padding:7px 2px;border-radius:5px;cursor:pointer;text-align:center;
  font-family:var(--font-num);background:var(--bg-surface-raised);
  border:1px solid var(--border-subtle);color:var(--text-secondary);
  transition:background .12s,border-color .12s}
.pcard .pb-wk .n{display:block;font:600 11px/1 var(--font-num)}
.pcard .pb-wk .o{display:block;margin-top:4px;font:500 11px/1 var(--font-num)}
.pcard .pb-wk.po{background:color-mix(in srgb,var(--tc) 14%,transparent);
  border-color:color-mix(in srgb,var(--tc-l) 45%,transparent);color:var(--tc-p)}
.pcard .pb-wk.bye{background:transparent;border-style:dashed;color:var(--text-dim)}
.pcard .pb-wk.on{background:var(--tc);border-color:var(--tc-l);color:#fff}

/* Year tabs inside the season plate. */
.pcard .pb-yr{padding:7px 14px;border:none;cursor:pointer;transform:skewX(-9deg);
  background:rgba(140,180,200,.09);color:var(--text-secondary);
  font:700 11px/1 var(--font-label);letter-spacing:.2em;text-transform:uppercase}
.pcard .pb-yr i{display:block;font-style:normal;transform:skewX(9deg)}
.pcard .pb-yr.on{background:var(--tc);color:#fff}

/* ---- action bar -------------------------------------------------------- */
/* Three states doing three different jobs. Queue is a shortlist MARK, so it is
   bright neutral and carries no hue - a bookmark is not a judgement. Avoid takes
   the negative ink. Drafted is the only committing action and therefore the only
   one that takes the team fill, at 1.35 width, so a drafted card is unmistakable
   from across the room. 44px floor is the touch target. */
.pcard .pb-acts{position:sticky;bottom:0;z-index:3;display:flex;gap:var(--sp-2);
  padding:var(--sp-3) var(--sp-5);background:var(--bg-app);
  border-top:1px solid color-mix(in srgb,var(--tc-l) 30%,transparent)}
.pcard .pb-act{flex:1;min-height:44px;display:flex;align-items:center;justify-content:center;
  gap:9px;padding:0 14px;border-radius:2px;cursor:pointer;transition:all .12s;
  font:700 13px/1 var(--font-label);letter-spacing:.2em;text-transform:uppercase;
  background:linear-gradient(180deg,#171418,#100e11);
  border:1px solid rgba(140,180,200,.16);color:var(--text-secondary)}
.pcard .pb-act:hover{filter:brightness(1.35)}
/* A control that cannot act. Same plate as its live sibling so the bar keeps
   its shape, but visibly inert and not a button - the dead-control rule. */
.pcard .pb-act.off{cursor:default;opacity:.55;filter:none;
  background:linear-gradient(180deg,#141216,#0e0c0f);border-style:dashed}
.pcard .pb-act.off:hover{filter:none}
.pcard .pb-act.on{background:linear-gradient(180deg,#1d2126,#141719);
  border-color:rgba(238,246,250,.5);color:var(--text-primary)}
.pcard .pb-act.avoid.on{background:linear-gradient(180deg,#231216,#180d10);
  border-color:color-mix(in srgb,var(--negative) 50%,transparent);color:var(--negative-ink)}
.pcard .pb-act.primary{flex:1.35;background:transparent;
  border:1px dashed rgba(140,180,200,.24);color:var(--text-muted)}
.pcard .pb-act.primary.on{border:1px solid var(--tc-l);color:#fff;
  background:linear-gradient(180deg,var(--tc),var(--tc-d));
  box-shadow:inset 0 1px 0 rgba(255,255,255,.18)}
.pcard .pb-act.primary:hover{filter:brightness(1.15)}

/* ---- narrow ------------------------------------------------------------ */
@media (max-width:560px){
  .pcard .pb{padding:var(--sp-4) var(--sp-3) var(--sp-1)}
  .pcard .pb-tiles.c4,.pcard .pb-tiles.c5{grid-template-columns:repeat(2,minmax(0,1fr))}
  .pcard .pb-tiles.c3{grid-template-columns:1fr}
  .pcard .pb-acts{padding:var(--sp-3)}
  .pcard .pb-act{letter-spacing:.12em;font-size:12px}
}

/* ══════════════════ EDGE CARDS — the board in the Primetime language ══════
   Same grammar as the player card: a hard clip on the lower-right corner, a
   skewed rank rail in the position's colour, Archivo Black obliques, and the
   numbers in a mono-ish row along the foot.

   THE RAIL IS THE CLUB, THE CHIP IS THE POSITION. The rail matches the
   Primetime player card - open Gibbs and the panel is Honolulu blue, so his
   card on the board is too - while the position chip keeps the hue
   constitution 18 measured. Both were briefly keyed to one variable, which
   made the chip a redundant second statement of the club and took position
   out of the board's colour language. */
/* container-type so the NAME can size against the CARD rather than the
   viewport. The card's width is set by a grid whose column count changes four
   times between 375px and 1440px, so a viewport-keyed font size is guessing at
   one remove; cqw is the actual relationship and it stops being re-tuned every
   time the grid does. */
.pcd{container-type:inline-size;
  /* flex-direction:ROW, stated. The card's previous rule set `column`, and a
     new `display:flex` does not undo a direction - so the rail, the body and
     the headshot stacked vertically at full width and the whole layout read
     as a completely different design. Nothing looked broken; it looked wrong,
     which is harder to trace. */
  position:relative;display:flex;flex-direction:row;align-items:stretch;
  gap:0;padding:0;overflow:hidden;
  background:#0b0d14;border:none;border-radius:0;
  clip-path:polygon(0 0, 100% 0, 100% 88%, 94% 100%, 0 100%);
  transition:transform .15s ease}
.pcd:hover{transform:translateY(-3px)}
.pcd-rail{flex:none;width:46px;align-self:stretch;display:flex;
  align-items:flex-start;justify-content:center;padding:11px 0;
  background:linear-gradient(180deg,var(--pc),color-mix(in srgb,var(--pc) 62%,#000));
  clip-path:polygon(0 0, 100% 0, 72% 100%, 0 100%)}
.pcd.feat .pcd-rail{width:56px}
/* A THREE-DIGIT RANK IN A 46px RAIL. "165" at 26px Archivo Black is wider
   than the stripe, and the skew pushes it further left - the leading digit was
   sliced off from pick 100 down. The rail is where the board is scanned, so
   the numeral yields rather than the stripe. */
.pcd-rank.d3{font-size:19px;letter-spacing:-.02em}
.pcd.feat .pcd-rank.d3{font-size:23px}
.pcd-rank{font:400 26px/1 "Archivo Black",var(--font-ui);color:#0b0d14;
  transform:skewX(-10deg);font-variant-numeric:tabular-nums}
.pcd.feat .pcd-rank{font-size:30px}
.pcd-body{flex:1;min-width:0;display:flex;flex-direction:column;gap:6px;
  padding:13px 8px 11px 10px}
/* ONE LINE. The serif runs wider than the condensed face it replaced, so the
   kicker wrapped to two rows on every card below the featured three and made
   the whole grid taller. Tracking down, no wrap - this is a locator, not a
   sentence, and it is the one line on the card that may ellipsis freely. */
.pcd-kick{font:600 10px/1.2 var(--font-label);font-style:italic;
  letter-spacing:.06em;text-transform:uppercase;color:#5c6478;
  display:flex;align-items:center;gap:5px;flex-wrap:nowrap;
  white-space:nowrap;overflow:hidden;min-width:0}
.pcd-name{display:block;margin:0;min-width:0;
  font:400 clamp(10px,2.35cqw,18px)/.95 "Archivo Black",var(--font-ui);
  text-transform:uppercase;
  color:var(--text-primary);transform:skewX(-10deg);transform-origin:left center}
.pcd.feat .pcd-name{font-size:clamp(12px,3.6cqw,23px)}
.pcd-name .nm{display:block;overflow:hidden;text-overflow:ellipsis;
  white-space:nowrap;min-width:0}
.pcd-chips{display:flex;gap:6px;align-items:center;flex-wrap:wrap;
  font:600 10px/1 var(--font-label);font-style:italic}
.pcd-pr,.pcd-tier{padding:3px 9px;transform:skew(-10deg);display:inline-block;
  border-radius:0;letter-spacing:.06em}
/* TWO VARIABLES, TWO QUESTIONS. `--pc` is the club and paints the left rail,
   so a card matches the panel it opens. `--pp` is the position and paints the
   chip, so RB6 still reads green at a glance. Keying both to one colour made
   the chip a second statement of the team and dropped position from the
   board's colour language entirely. */
.pcd-pr{background:var(--pp);color:#0b0d14;border:none}
.pcd-tier{background:none;border:1.5px solid #2a2f3d;color:var(--text-secondary)}
.pcd-pr>i,.pcd-tier>i{display:inline-block;font-style:normal;transform:skew(10deg)}
/* The foot keeps its five cells and its deterministic height; only the dress
   changed. `auto-flow:column` for the same reason the Primetime band uses it -
   the featured card carries SPREAD and the rest do not. */
/* THREE UNEQUAL CELLS. The foot is always PAR / EDGE / READ, and they do not
   hold the same amount: "AT PRICE" is the longest string on the row and EDGE
   is usually a dash. Equal thirds gave READ 65px for a 62px word and it
   rendered "AT ..." on every featured card. Explicit tracks, since the cell
   count is now fixed at three. */
/* CONTENT-SIZED, NOT FRACTIONAL. PAR is a short number and EDGE is usually a
   dash; READ carries "AT PRICE", the longest string on the row. Fixed
   fractions gave EDGE a share it did not need while READ starved - and the
   right ratio changed every time the section rail opened or closed, which is
   how this got tuned three times for three different widths. The first two
   tracks take exactly what they hold and READ takes the rest. */
/* READ TAKES ITS OWN ROW ON A REGULAR CARD. Every attempt to fit PAR, EDGE
   and READ on one line failed somewhere - five cells, then equal thirds, then
   content-sized tracks - because "AT PRICE" needs about a third of a 180px
   body on its own and the section rail can remove 212px from every card at
   any moment. Two rows always fit, and a card that always reads the same is
   worth more than one that fits on the widths I happened to test. */
.pcd-foot{margin-top:auto;display:grid;
  grid-template-columns:minmax(0,auto) minmax(0,1fr);gap:6px 10px;padding:9px 0 0;
  border-top:1px solid #1c2030;background:none;font-style:italic}
/* overflow:hidden, not just min-width:0 - a grid track can be narrower than
   its content and the content will happily paint over the next cell, which is
   exactly what EDGE and SIGNAL did at 1400. */
.pcd-foot .c{min-width:0;display:block;overflow:hidden}
/* PAR is the one foot figure the template colours, and it is the one the
   board is actually read on. */
.pcd-foot .c:first-child .v{color:var(--accent)}
/* `.wide` is a leftover from the two-row foot. Both of its old rules are
   gone: one spanned the full row, the other pinned the featured card's SIGNAL
   cell with justify-self:end - which slid it left, over READ, at every width.
   auto-flow:column decides placement now, and nothing overrides it. */
.pcd-foot .c.wide{grid-column:auto;justify-self:stretch}
.pcd-foot .c:nth-child(3){grid-column:1/-1}
.pcd-foot .k{display:block;font:600 9px/1 var(--font-label);
  letter-spacing:.06em;white-space:nowrap;overflow:hidden;text-overflow:clip;text-transform:uppercase;color:#5c6478}
.pcd-foot .v{display:block;margin-top:4px;min-height:0;
  font:400 clamp(11px,3.1cqw,17px)/1 "Archivo Black",var(--font-ui);color:var(--text-primary);
  transform:skewX(-10deg);transform-origin:left center;
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
/* The featured foot is three cells in a ~197px body. "+148.3" at 17px
   Archivo Black is 62px against a 65px track - it rendered "+14...". The cap
   comes down rather than the cell count, which is already at three. */
.pcd.feat .pcd-foot .v{font-size:clamp(11px,1.95cqw,14px)}
/* THE FEATURED CARD USES THE SAME TWO ROWS. It kept a single row on the
   reasoning that it is wider - true only with the section rail CLOSED. Open
   the rail and 212px comes off every column, and the featured foot was the
   first thing to break ("+14..."). One layout that always fits beats two that
   each fit in one rail state. */
.pcd-foot .v.tag{font-size:12px}
/* The featured card had a LARGER tag than the regular ones, which is why
   "AT PRICE" fitted at 290px and clipped at 380px - the one place the extra
   emphasis cost the word itself. Featured now matches. */
.pcd.feat .pcd-foot .v.tag{font-size:12px}
/* The EDGE chip carries the longest token on the row after READ - "ESPN -104"
   is nine characters in a box with a border and padding. */
.pcd-foot .v.sm,.pcd-foot .v.sig{font-size:9px;transform:none}
.pcd-foot .pcd-sig{padding:2px 5px;letter-spacing:.02em}
/* THE HEADSHOT TAKES THE SLACK ON ITS LEFT. The body was sized when the foot
   was fighting for every pixel; with READ on its own row the names and numbers
   finish well short of the photo, and that space is worth more to a face than
   to padding.

   AND THE DIAGONAL IS THE REST OF IT. The gap he could see between the longest
   name and the headshot on every card was never layout - the photo box already
   reaches under it. It was the clip path removing the left 14% of the box AT
   THE TOP, which is exactly the row the name sits on. Shrinking the cut to 3%
   hands that width back to the face and costs the layout nothing, which is why
   it works where widening the box did not: the box is capped by the longest
   NAME, and this is not the box. */
.pcd-photo{flex:none;width:74px;align-self:stretch;
  clip-path:polygon(3% 0, 100% 0, 100% 100%, 0 100%)}
.pcd.feat .pcd-photo{width:108px}
/* `cover` on a narrow box crops the sides of the face; these are portrait
   crops with the head at the top, so the frame is pinned to the very top of
   the image: whatever falls outside is torso, never scalp. 12% was still
   taking the top of the head on the taller portraits. */
.pcd-photo .pcd-face{width:100%;height:100%;object-fit:cover;
  object-position:50% 0;opacity:.95}
.pcd-photo .pcd-face.logo{object-fit:contain;padding:14px;opacity:.55}
.pcd.nophoto .pcd-photo{display:none}
.pcd.gone{opacity:.42}
.pcd.gone .pcd-name{text-decoration:line-through}
/* A NARROW CARD WRAPS ITS FOOT rather than crushing it. Four cells across a
   135px body is 33px each, which renders "+148.3" as "+..." and "AT PRICE" as
   "AT ..." - the numbers the board exists to show, spent on fitting them in
   one line. A container query, so the rule keys off the CARD and not the
   viewport: the same card is wide in the featured row and narrow below it. */
@container (max-width:330px){
  .pcd-foot{grid-auto-flow:row;grid-template-columns:repeat(2,minmax(0,1fr));
    gap:7px 10px}
}
@container (max-width:250px){
  .pcd-photo{display:none}
}
@media (max-width:620px){.pcd-photo{display:none}}
/* The featured photo is 150px of a card that can be 340px wide on a tablet. */
@media (max-width:900px){.pcd.feat .pcd-photo{width:96px}}
@media (max-width:760px){.pcd-photo,.pcd.feat .pcd-photo{display:none}}

/* ══════════════════ HERO — chips + flag banners ══════════════════
   The EDGE Cards masthead. A row of skewed pills carries the counts that used
   to occupy two full tiles, and the width they freed goes to the two things
   you act on: the top value and the top fade, each led by a coloured flag. */
.stat-strip{display:grid;gap:var(--sp-3);grid-template-columns:repeat(2,minmax(0,1fr))}
/* THE CHIPS SPAN THE TOP. They were pills huddled at the left at 11px, which
   is a footnote's size for the four numbers that describe the whole board -
   reported as "too small, hurt my eyes". Four equal columns across the full
   width, at a size meant to be read from a seat rather than a scroll. */
.hz-chips{grid-column:1/-1;display:grid;gap:10px;
  grid-template-columns:repeat(4,minmax(0,1fr))}
.hz-chip{display:grid;place-items:center;
  font:600 16px/1 var(--font-label);font-style:italic;
  letter-spacing:.1em;text-transform:uppercase;padding:13px 14px;
  transform:skew(-10deg);background:var(--chip-ground);color:var(--card-ink-soft);
  white-space:nowrap;overflow:hidden}
.hz-chip>i{display:inline-block;font-style:normal;transform:skew(10deg)}
.hz-chip.on{background:var(--chip-on-face);color:var(--chip-on-ink)}
.hz-chip.warn{background:var(--chip-warn-face);color:var(--chip-warn-ink)}
.hz-chip.ghost{background:none;border:1px solid var(--chip-ghost-edge);
  color:var(--card-ink-dim)}

.hz-ban{container-type:inline-size;position:relative;display:flex;align-items:stretch;gap:0;padding:0;
  overflow:hidden;border:none;border-radius:0;background:#0b0d14;
  clip-path:polygon(0 0, 100% 0, 100% 86%, 96% 100%, 0 100%)}
/* The flag is the template's single strongest hero element: a skewed slab of
   the pole's colour with the direction drawn in it, so which way a card leans
   is legible before a word of it is read. */
.hz-flag{flex:none;width:54px;display:grid;place-items:center;font-size:15px;
  clip-path:polygon(0 0, 100% 0, 72% 100%, 0 100%)}
.hz-ban.val .hz-flag{background:linear-gradient(180deg,var(--positive),
  color-mix(in srgb,var(--positive) 58%,#000));color:#04211b}
.hz-ban.fade .hz-flag{background:linear-gradient(180deg,var(--negative),
  color-mix(in srgb,var(--negative) 58%,#000));color:#2a0f0a}
.hz-in{flex:1;min-width:0;padding:14px 18px 15px 14px;display:flex;
  flex-direction:column;gap:8px}
.hz-top{display:flex;align-items:center;justify-content:space-between;gap:10px}
.hz-ban .caps{margin:0}
.hz-ban .caps b{color:var(--text-secondary);font-weight:700}
/* `.stat.hero .hero-nav` above is (0,3,0) and pins this absolute; this rule
   was written at (0,2,0) and has therefore never applied, so the arrows have
   been floating over the banner since the day it was added. A long name runs
   straight under them - "fonts are bleeding over the toggle boxes arrows" -
   because an absolutely positioned box reserves no space for the text beside
   it. Matched on the same three classes so the intent finally outranks what
   it was written to replace. */
.stat.hero.hz-ban .hero-nav{position:static;display:flex;gap:6px}
/* THE NAME WINS. Adding the direction word to this row pushed "TreVeyon
   Henderson" into an ellipsis - the one string on a hero banner that must
   never be abbreviated, since it is the answer the banner exists to give.
   The row wraps, so the delta and its word drop to a second line before the
   name gives up a character. */
.hz-line{display:flex;align-items:baseline;gap:8px 10px;min-width:0;flex-wrap:wrap}
.hz-ban .stat-val.name{flex:1 1 100%}
@container (min-width:520px){.hz-ban .stat-val.name{flex:0 1 auto}}
.hz-ban .stat-val.name{font:400 clamp(19px,2.5cqw,30px)/1 "Archivo Black",var(--font-ui);
  text-transform:uppercase;color:var(--text-primary);transform:skewX(-9deg);
  transform-origin:left center;white-space:nowrap;overflow:hidden;
  text-overflow:ellipsis;min-width:0;flex:0 1 auto}
.hz-d{flex:none;font:400 clamp(17px,2.1cqw,26px)/1 "Archivo Black",var(--font-ui);
  transform:skewX(-9deg);font-variant-numeric:tabular-nums}
.hz-d.d-up{color:var(--positive)} .hz-d.d-dn{color:var(--negative)}
/* The direction in words, sized to sit under the figure rather than compete
   with it - the number is the magnitude, this is what it means. */
.hz-dw{flex:none;font:600 10px/1.1 var(--font-label);
  font-style:italic;letter-spacing:.14em;text-transform:uppercase;opacity:.9}
.hz-dw.d-up{color:var(--positive)} .hz-dw.d-dn{color:var(--negative)}
.hz-ban .hero-worth{flex:none;margin:0}
.hz-ban .stat-note{margin-top:0;width:auto;white-space:nowrap;overflow:hidden;
  text-overflow:ellipsis;font-size:var(--fs-sm)}
.hz-ban .spark{position:absolute;right:0;bottom:0;opacity:.4;pointer-events:none}
@media (max-width:900px){
  .stat-strip{grid-template-columns:minmax(0,1fr)}
  .hz-chips{grid-template-columns:repeat(2,minmax(0,1fr))}
  .hz-chip{font-size:13px;padding:11px 10px}
}

/* The room lens, announced at the head of the board legend. Accent-coloured
   because it is a STATE the reader is in, not a warning. */
.lens-on{color:var(--accent);font-weight:600}

/* ============================== PHONE TABLES ==============================
   The momentum leaderboard and the injury tracker are 1180px of columns; at
   phone width the scroller showed ONLY the name column and every number sat
   off-screen with no hint it existed (HANDOFF: the pinned-column attempt was
   backed out - transparent row backgrounds and auto table layout). This is
   the third approach: SAME MARKUP, rows restyled into stacked card-rows, so
   there is no second implementation to drift. Appended at end of file on
   purpose - source order is the tiebreak that made .hide-m inert. */
@media (max-width:640px){
  /* Both tables leave table layout entirely. */
  #moveTable, #injTable,
  #moveTable thead, #injTable thead,
  #moveTable tbody, #injTable tbody { display:block; width:100%; min-width:0; }
  #moveTable tr, #injTable tr {
    display:flex; flex-wrap:wrap; align-items:center;
    row-gap:2px; column-gap:8px; padding:10px 2px;
    /* `--line` was never defined anywhere, so the fallback WAS the rule - and
       it is graphite's blue-grey, chosen before either table could sit on a
       lit plate. --border-subtle carries that exact value on graphite (an
       identity swap there) and the paper hairline on the plate. */
    border-bottom:1px solid var(--border-subtle);
  }
  #moveTable td, #injTable td, #moveTable th { border:0; padding:0; }

  /* Momentum: the header row stays - it IS the sort control - as chips. */
  #moveTable thead tr { padding:6px 2px; row-gap:6px; }
  #moveTable th.sortable {
    /* Same swap; .2 becomes --border-strong's .22 on graphite, which is a
       difference no eye resolves and one fewer literal to drift. */
    font-size:11px; padding:6px 10px; border:1px solid var(--border-strong);
    border-radius:6px; min-height:32px; display:inline-flex; align-items:center;
  }
  #moveTable th.tape-hd { display:none; }

  /* Momentum rows: badge | name+team | big delta, then labeled smalls,
     then the night tape on its own full-width line. */
  #moveTable td:nth-child(2) { order:-2; }            /* pos badge */
  /* min-width forces the wrap: with basis 0 every cell just shrank onto one
     line (the name cell measured 14px and the delta painted over it). */
  #moveTable td.player { order:-1; flex:1 1 auto; min-width:56%; }
  #moveTable td.overall { order:0; margin-left:auto; font-size:15px; }
  #moveTable td:nth-child(3), #moveTable td:nth-child(4),
  #moveTable td:nth-child(5), #moveTable td:nth-child(8),
  #moveTable td:nth-child(9) { order:1; flex:0 0 auto; white-space:nowrap;
    font-size:12px; }
  /* The desktop zebra painted per-cell; as flex items those cells become
     floating grey boxes. The row keeps the tint, the cells lose it. */
  #moveTable tbody td { background:none; height:auto; }
  #moveTable tr.mv-row { background:rgba(140,180,200,.02); }
  #moveTable td:nth-child(3)::before { content:"bye "; }
  #moveTable td:nth-child(4)::before { content:"was "; }
  #moveTable td:nth-child(5)::before { content:"now "; }
  #moveTable td:nth-child(8)::before { content:"o/n "; }
  #moveTable td:nth-child(9)::before { content:"days "; }
  #moveTable td:nth-child(3)::before, #moveTable td:nth-child(4)::before,
  #moveTable td:nth-child(5)::before, #moveTable td:nth-child(8)::before,
  #moveTable td:nth-child(9)::before {
    color:var(--text-dim, #6b7f8c); font-size:10px; letter-spacing:.08em;
    text-transform:uppercase;
  }
  /* display:block overrides the earlier mobile rule that hid the tape
     outright - the tape is exactly the data a phone was missing. */
  #moveTable td.tapecell { display:block; order:2; flex-basis:100%;
    overflow-x:auto; }

  /* Injuries: no sortable headers, so the thead goes and labels come from
     the cells. Line 1: badge, name, team. Line 2: status + body part.
     Line 3: labeled numbers. The pedge colour cell is redundant with the
     position badge and becomes noise as a flex item. */
  #injTable thead { display:none; }
  #injTable td.pedge { display:none; }
  #injTable td:nth-child(3) { order:-2; }             /* pos badge */
  #injTable td:nth-child(2) { order:-1; flex:1 1 auto; min-width:48%; } /* name */
  #injTable td:nth-child(4) { order:0; font-size:12px; color:var(--text-dim, #6b7f8c); }
  #injTable td:nth-child(7), #injTable td:nth-child(8) { order:1;
    flex:0 1 auto; font-size:13px; }
  #injTable tbody td { background:none; height:auto; }
  #injTable td:nth-child(8)::before { content:"· "; color:var(--text-dim, #6b7f8c); }
  #injTable td:nth-child(5), #injTable td:nth-child(6),
  #injTable td:nth-child(9), #injTable td:nth-child(10) { order:2; font-size:12px; }
  #injTable td:nth-child(5)::before { content:"mkt "; }
  #injTable td:nth-child(6)::before { content:"cmp "; }
  #injTable td:nth-child(9)::before { content:"as of "; }
  #injTable td:nth-child(5)::before, #injTable td:nth-child(6)::before,
  #injTable td:nth-child(9)::before {
    color:var(--text-dim, #6b7f8c); font-size:10px; letter-spacing:.08em;
    text-transform:uppercase;
  }
  /* Band rows and empty rows span the whole card. */
  #injTable tr.inj-band td, #injTable td.empty, #moveTable td.empty {
    flex-basis:100%;
  }
}

/* ============================ PHONE CARD DENSITY ==========================
   At 375 the card board showed about four players between the topbar and the
   fold: 148px per card, and the foot spent a whole extra row on READ. That
   two-row foot is a DESKTOP constraint - it exists because four cards share a
   row there and "AT PRICE" needs a third of a 180px body. A phone is one
   column, so the card is 359px wide and the body has room the desktop case
   never has. Same markup, one row, and the headshot comes back off the bench
   with the width that buys. */
@media (max-width:640px){
  /* The three numbers fit one line at this width; READ stops taking its own. */
  .pcd-foot{grid-template-columns:auto auto minmax(0,1fr);gap:0 10px;padding:7px 0 0}
  .pcd-foot .c:nth-child(3){grid-column:auto}
  .pcd-body{padding:8px 10px;gap:4px}
  .pcd-chips{gap:5px}
  /* THE HEADSHOT RETURNS. It was hidden at three breakpoints (760/620/and the
     bare rule) back when the body was fighting the foot for every pixel; with
     the foot on one row that fight is over, and a face is the fastest way to
     find a player under a clock. Narrower than desktop's 74px, and it costs
     no HEIGHT at all - it stretches the row the card already has. */
  .pcd-photo,.pcd.feat .pcd-photo{display:block;width:52px}
  /* Unchanged and load-bearing: a player whose headshot 404s collapses to a
     clean text card rather than reserving 52px for nothing. */
  .pcd.nophoto .pcd-photo{display:none}
}

/* ---- the two table-polish items left by the mobile pass ---------------- */
@media (max-width:640px){
  /* MOMENTUM: the delta absorbs the rest of line one, so the labeled smalls
     wrap together instead of BYE riding up beside the headline number. */
  #moveTable td.overall{flex:1 0 auto;margin-left:auto;text-align:right}
  /* INJURIES: age and its freshness meter end the numbers line rather than
     floating mid-row, and the body part loses its leading separator - it was
     written as a mid-line divider and reads as an orphan whenever the status
     above it fills line one, which is most of them. */
  #injTable td:nth-child(10){margin-left:auto;text-align:right}
  #injTable td:nth-child(8)::before{content:none}
  /* The "AS OF" label goes, the DATE stays - constitution 15 requires the
     date beside a designation, not the caps label announcing it, and those
     ~45px are what pushed age and its meter onto a ragged third line on the
     longer rows. Age reads as the qualifier it is once it sits on the same
     line as the date it is derived from. */
  #injTable td:nth-child(9)::before{content:none}
}

/* ---- the Best Available header stops competing for one line ------------ */
@media (max-width:640px){
  /* `.panel-hd` is flex with space-between, so the caps label and the
     "survival on [room] ADP" sentence fought over 375px - and the sentence
     contains a select that the 44px thumb floor made tall, so the result was
     a cramped four-part row rather than a sentence. The label keeps line one;
     the sentence takes its own, where it reads as the sentence it was
     written as. Only headers that HAVE a survline change - the queue and
     roster headers are a label and a count and still sit on one line. */
  .panel-hd{flex-wrap:wrap;row-gap:6px;align-items:center}
  .survline{flex:1 1 100%}
  .survline select{flex:1 1 auto;min-width:0}
}

/* ============================================================ REPORT CARD
   His 2026-08-10 mock, ported. A PAPER DOCUMENT ON THE DARK BOARD: the card
   is a thing the app ISSUES, and the conceit only works if it looks issued.
   Prefixed .rpt- rather than .rc-, which the rail card already owns - the
   panel used to wear `.rc` itself and inherited a pointer cursor from it.
   Literals rather than tokens throughout: this is a light surface inside a
   dark app, so --text-primary and friends are exactly wrong here. */
.rptpanel{max-width:920px}
.rpt-sheet{background:#f2efe6;color:#14181a;border-radius:2px;
  padding:44px 48px 34px;box-shadow:0 1px 3px rgba(0,0,0,.4)}
.rpt-lbl{font-family:var(--font-num);font-size:8px;letter-spacing:.2em;
  color:#5b6165;text-transform:uppercase}
.rpt-lbl-sp{margin-top:22px}

/* Masthead. The rule is 3px DOUBLE, which is what draws the two hairlines. */
.rpt-mast{display:grid;grid-template-columns:1fr auto 1fr;gap:20px;
  align-items:start;padding-bottom:14px;border-bottom:3px double rgba(20,24,26,.5)}
.rpt-word{font:500 27px/1 var(--font-display);letter-spacing:.16em;color:#14181a}
.rpt-kick{font-family:var(--font-num);font-size:8px;letter-spacing:.24em;
  color:#5b6165;margin-top:6px}
.rpt-titles{text-align:center}
.rpt-title{font:italic 400 23px/1 var(--font-display);color:#14181a}
.rpt-term{font-family:var(--font-num);font-size:9px;letter-spacing:.2em;
  color:#5b6165;margin-top:6px}
.rpt-room{font-family:var(--font-num);font-size:9px;letter-spacing:.08em;
  line-height:1.9;color:#5b6165;text-align:right}

/* Manager row. The FINAL MARK box is the one boxed thing on the sheet. */
.rpt-meta{display:grid;grid-template-columns:1fr 1fr 1fr 150px;gap:20px;
  align-items:end;margin-top:22px}
.rpt-field{border-bottom:1px solid rgba(20,24,26,.18);padding-bottom:6px}
.rpt-fv{font:400 19px/1.25 var(--font-display);color:#14181a;margin-top:5px}
.rpt-field:nth-child(2) .rpt-fv,.rpt-field:nth-child(3) .rpt-fv{
  font:400 12px/1.5 var(--font-num);letter-spacing:.02em}
.rpt-final{border:2px solid #14181a;border-radius:3px;padding:6px 0 4px;
  text-align:center}
.rpt-finalv{font:500 44px/1 var(--font-display);color:#14181a}
.rpt-cap{font-family:var(--font-num);font-size:9px;color:#5b6165;
  vertical-align:super;margin-left:1px}

/* The six subjects. THE FLEXIBLE COLUMNS CARRY A REAL FLOOR: a 1fr whose
   container is narrower than the fixed tracks collapses to zero and ellipsis
   renders every subject as "S." - the mistake that cost three passes on 5a. */
.rpt-subjects{list-style:none;margin-top:26px}
.rpt-shead,.rpt-srow{display:grid;
  grid-template-columns:minmax(120px,1fr) 58px 92px 92px minmax(180px,1.5fr);
  gap:0 14px}
.rpt-shead{font-family:var(--font-num);font-size:8px;letter-spacing:.18em;
  color:#5b6165;padding-bottom:8px;border-bottom:1px solid rgba(20,24,26,.35)}
.rpt-srow{align-items:baseline;padding:11px 0;
  border-bottom:1px solid rgba(20,24,26,.1)}
.rpt-srow:last-child{border-bottom:none}
.rpt-subj{font:400 16px/1.3 var(--font-display);color:#14181a}
.rpt-mk{font:600 19px/1 var(--font-display);color:#14181a}
.rpt-mk.rpt-good{color:#0f6f83}
.rpt-mk.rpt-warn{color:#8a5a12}
.rpt-measure{font-family:var(--font-num);font-size:12px;color:#14181a;
  font-variant-numeric:tabular-nums}
.rpt-avg{font-family:var(--font-num);font-size:12px;color:#5b6165;
  font-variant-numeric:tabular-nums}
.rpt-remark{font-family:var(--font-ui);font-size:12px;line-height:18px;color:#3a4145}

/* Remarks + attendance, and the class ranking aside. */
.rpt-split{display:grid;grid-template-columns:minmax(0,1fr) 300px;gap:32px;
  align-items:start;margin-top:26px}
.rpt-prose{font:italic 400 15px/1.65 var(--font-display);color:#14181a;
  margin-top:8px}
.rpt-chart{display:block;width:100%;height:auto;margin-top:10px;overflow:visible}
.rpt-bar{fill:rgba(20,24,26,.55)}
.rpt-bar-low{fill:#8a5a12}
.rpt-med{stroke:#0f6f83;stroke-width:1;stroke-dasharray:3 3}
.rpt-tick{font-family:var(--font-num);font-size:7px;fill:#5b6165;text-anchor:middle}
.rpt-chartnote{font-family:var(--font-num);font-size:8px;letter-spacing:.12em;
  color:#5b6165;margin-top:8px;text-transform:uppercase}
.rpt-issued{font-family:var(--font-num);font-size:8px;letter-spacing:.14em;
  color:#5b6165;margin-top:22px;padding-top:10px;
  border-top:1px solid rgba(20,24,26,.18);display:flex;justify-content:space-between}
.rpt-sig{letter-spacing:.2em;color:#14181a}

.rpt-rank{list-style:none;margin-top:8px}
.rpt-rrow{display:grid;grid-template-columns:16px 30px minmax(0,1fr) 52px;gap:8px;
  align-items:center;padding:5px 0;border-bottom:1px solid rgba(20,24,26,.1)}
.rpt-rrow:last-child{border-bottom:none}
.rpt-rn{font-family:var(--font-num);font-size:9px;color:#8b9196}
.rpt-rrow .rpt-mk{font-size:14px}
.rpt-rteam{font:400 14px/1.2 var(--font-display);color:#14181a;
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.rpt-mine .rpt-rteam{color:#0f6f83}
.rpt-rpar{font-family:var(--font-num);font-size:11px;color:#14181a;
  text-align:right;font-variant-numeric:tabular-nums}
.rpt-scale{display:flex;gap:10px;flex-wrap:wrap;margin-top:10px;
  font-family:var(--font-num);font-size:8px;letter-spacing:.14em;color:#5b6165}
.rpt-foot{font-family:var(--font-num);font-size:8px;letter-spacing:.14em;
  color:#5b6165;margin-top:4px}
.rpt-basis{font-family:var(--font-ui);font-size:11px;line-height:1.6;
  color:#5b6165;margin-top:24px;padding-top:12px;
  border-top:1px solid rgba(20,24,26,.18)}

.rpt-empty{color:var(--text-muted);font-size:var(--fs-sm)}

/* THE PHONE, which is his second frame and a reduction rather than a
   redesign: the same paper, the subject table down to three columns, and the
   two panels that cannot survive 345px (the ranking aside and the week chart)
   dropped rather than crushed. */
@media (max-width:640px){
  .rpt-sheet{padding:20px 16px 18px}
  .rpt-mast{grid-template-columns:1fr auto;gap:10px;row-gap:12px}
  .rpt-word{font-size:21px}
  .rpt-title{font-size:19px;text-align:right}
  .rpt-titles{text-align:right}
  .rpt-room{grid-column:1 / -1;text-align:left;font-size:8px;line-height:1.7}
  /* MANAGER and CLASS STANDING STACK, with the mark box beside them.
     Left to flow, the standing landed in a 88px track and broke "1 OF 10 ·
     373.6 PAR" over two lines while the mark box dropped to full width. The
     roster shape is the one field that goes: it is the longest string on the
     sheet and the least useful under a thumb. */
  .rpt-meta{grid-template-columns:minmax(0,1fr) 96px;gap:10px 12px}
  .rpt-field:nth-child(1){grid-column:1;grid-row:1}
  .rpt-field:nth-child(2){display:none}
  .rpt-field:nth-child(3){grid-column:1;grid-row:2}
  .rpt-final{grid-column:2;grid-row:1 / span 2;align-self:stretch;
    display:flex;flex-direction:column;justify-content:center;gap:2px}
  .rpt-finalv{font-size:34px}
  .rpt-shead,.rpt-srow{grid-template-columns:minmax(0,1fr) 34px 66px;gap:0 8px}
  .rpt-shead span:nth-child(4),.rpt-shead span:nth-child(5),
  .rpt-avg,.rpt-remark{display:none}
  .rpt-shead{letter-spacing:.16em}
  .rpt-subj{font-size:15px}
  .rpt-mk{font-size:17px}
  .rpt-split{grid-template-columns:minmax(0,1fr);gap:18px}
  .rpt-side,.rpt-chart,.rpt-chartnote,.rpt-lbl-sp{display:none}
  .rpt-issued{flex-direction:column;gap:4px}
}

/* ================================================ THE PACED PICK REVEAL
   Opponents land one at a time with a visible beat, the way Sleeper does it.
   PRESENTATION ONLY: S.picks is already final when this runs (see
   advanceAuto), so nothing here can put the board behind the truth. `--pd` is
   the pick's place in the batch, set per cell by renderDraft. */
@keyframes dgLand{
  from{opacity:0;transform:translateY(-5px) scale(.985)}
  60%{opacity:1}
  to{opacity:1;transform:none}
}
.dg-land{animation:dgLand .3s cubic-bezier(.2,.7,.3,1) both;
  animation-delay:var(--pd,0s)}
/* A cell that has not been revealed yet must not be readable through the
   animation's first frame - `both` holds the from-state, which is opacity 0. */
@media (prefers-reduced-motion:reduce){
  .dg-land{animation:none}
}
/* YOUR TURN. The cell already says "You" and carries the board's only glow;
   what was missing is the MOMENT - after a run of opponent picks the board
   simply was your turn, with nothing marking the handover. One pulse, once. */
@keyframes dgTurn{
  0%{box-shadow:0 0 0 0 var(--accent)}
  70%{box-shadow:0 0 0 7px rgba(0,0,0,0)}
  100%{box-shadow:0 0 0 0 rgba(0,0,0,0)}
}
.dg-cell.now.mine{animation:dgTurn 1.1s ease-out 2}
@media (prefers-reduced-motion:reduce){.dg-cell.now.mine{animation:none}}

/* ================================== THE PHONE DRAFT ROOM (Sleeper parity)
   His instruction, with two frames from a screen recording: "replicate".
   What Sleeper's mobile draft actually IS, and the reason you never pan far:
   THE BOARD IS THE BACKGROUND AND THE ACTION LIVES IN A SHEET. You look at
   the board; you act in the sheet. So this is a LAYOUT change, phone-only and
   only while a draft is running - desktop never sees a rule in this block.

   Same markup throughout, on the same principle as the two wide tables: the
   phone gets a different arrangement of the things already rendered, never a
   second render path that can drift from the first. */
@media (max-width:640px){
  /* ---- the board, at Sleeper's density -------------------------------- */
  /* NOT gated on body.drafting, and that is deliberate: the cell has to FIT
     at this width whether or not a draft is running, and the 414px legibility
     contract builds its probe cells with no draft in progress. Gating the
     trim meant the contract measured desktop content in phone-width boxes and
     failed three ways. Only the things that are genuinely draft-only - the
     board's 32vh window and the sheet - stay behind the class.

     80px COLUMNS, measured off his reference frame rather than estimated.
     design/sleeper-mobile-midboard.png is 444px wide for a 375pt phone, so it
     is a 1.184x capture: the 95px column pitch in that image is 80 CSS px on
     the device. Taking the image number would have drawn the board ~18% too
     wide, which is why the notes say to match a mock's PROPORTION, never its
     pixel values. */
  .draft-grid{
    grid-template-columns:42px repeat(var(--dg-cols),minmax(78px,1fr));
    min-width:calc(42px + var(--dg-cols) * 80px);gap:2px}
  /* THE CUT EDGE BECOMES A FADE. Two scored rounds called the right edge
     "sliced" and "horizontally clipped": a column cut mid-glyph reads as a
     rendering fault, not as a viewport onto a wider board. A 12px mask says
     "continues" in the vocabulary every phone list uses. A mask, not an
     overlay, because the panel behind these cells is not one paint: filled
     cells carry a 62% mix an overlay gradient would have to match per
     position. The board always overflows at this width (12 teams x 80px
     against 390), so the fade never lies about more being there; at full
     scroll it softens the last column's own padding, which is the price of
     having no second clock watching scrollLeft. */
  .dg-wrap{-webkit-mask-image:linear-gradient(to right,#000 calc(100% - 12px),transparent);
    mask-image:linear-gradient(to right,#000 calc(100% - 12px),transparent)}
  .dg-rd{font-size:10px;letter-spacing:.04em}
  .dg-rdw{display:none}
  /* The seat you own says YOU and nothing else at this pitch: the full
     "YOU · Team 1" measured wider than the 80px column and ellipsed to
     "YOU · Tea…" - the one header on the board reading as a fault. One
     render path; the phone hides the suffix the desktop keeps. */
  .dg-team .dgt-rest{display:none}
  /* 2px of side padding, not 4 and no longer 3. The name contract is a RATIO
     against the same names unabbreviated, and at 4px it cleared by 0.087
     against a 0.10 bound - 13% of headroom on a guard that has already failed
     CI once because the runner rasterises this font a shade wider than this
     laptop. Every pixel bought here is spent twice now: the name got the
     badge's old room back on the row below, and the identity line has to hold
     `WR · DEN (10)` (49.1px measured) beside a mark (18.7px) in the same box.
     70px of content did not do it; 72px does, with 2.16px over. */
  .dg-cell{min-height:38px;padding:3px 2px 4px}
  /* THE NAME WRAPS AND SHRINKS, which is how Sleeper affords an 80px column.
     Edge truncated to one line and at this width that turned
     "J. Smith-Njigba" into an ellipsis carrying no information. Two clamped
     lines at 10px: a long name costs height, never meaning. */
  /* AN ELLIPSIS, AND A FIXED TWO-LINE BOX. Both are defects the clamp
     introduced and neither is the clamp's fault.
     `text-overflow:clip` hard-cut surnames mid-word in fully visible interior
     cells - "P. Washingt", "D. Montgom" - which reads as a rendering fault
     rather than as truncation. Every one of those cells carries a +/- badge:
     the badge takes ~24px of a 72px content box, so the name has 48px and
     "Washington" wants 55. The name yields the space (that is the badge's own
     ruling, three rules down) but it has to yield it VISIBLY.
     And the box is two lines whether or not the name needs two, because a
     board where "Kyr. Williams" wraps and "J. Chase" does not has a different
     baseline in every cell - the reference's cells are all two lines for
     exactly this reason (first name over last), and its rows read as rows. */
  .dg-cell .dg-nm{font-size:10px;line-height:1.2;min-height:24px;
    white-space:normal;overflow:hidden;text-overflow:ellipsis;
    display:-webkit-box;-webkit-line-clamp:2;-webkit-box-orient:vertical}
  /* THE IDENTITY LINE AT THE REFERENCE'S OWN PROPORTION. Sleeper writes
     `RB - BUF (7)` and `1.9` across a 70pt cell at about 7pt; this is a 78pt
     cell, so 7.5px is that line at this width, measured from the frame rather
     than guessed. It is what makes the club and the bye affordable here at
     all - the old meta row spent 9px and a 22px chip to say less.

     THE TEAM CODE USED TO BE DROPPED AT THIS WIDTH, on the argument that it is
     on the player's own card. That argument was wrong in the one place it was
     applied: the card is two taps and a scroll away with a clock running, and
     dropping the club left the phone cell leaning entirely on an unlabelled
     pastel for position. The chip and the separate club span are what went
     instead; this line says both in less room than either. */
  /* 7.5px and a 2px gap, MEASURED against the longest line this board can
     produce (`DST · WAS (14)` beside `15.10`) rather than chosen: at 8px and a
     3px gap, 31 of 179 cells ellipsed the bye off the end, which is the one
     fact this line was added to carry. */
  .dg-cell .dg-id{gap:2px;margin-bottom:1px;font-size:7.5px;letter-spacing:0}
  .dg-cell .dg-id .dg-pk{font-size:7px}
  /* LEVERAGE H1: the word at the identity line's OWN size and ink, not a
     step below it. Both round-9 projection fails said "despite tiny PROJ
     labels" - the state word was the smallest, dimmest thing in the cell.
     7.5px secondary is what the line beside it already wears; worst case
     `DST · WAS (14)` 52.5px + the word ~17px + 2px gap = 71.5 of 72, and a
     DST ghost can never carry a mark, so the pair still fits by
     construction. */
  .dg-cell .dg-id .dg-projw{font-size:7.5px;color:var(--text-secondary)}
  /* ROUND 9: THE COORDINATE COMES OFF MADE CELLS AT THIS WIDTH. The
     container query above already ruled that the coordinate is the first
     thing a narrow cell gives up, because the round gutter and the column
     header state it; the phone finishes that ruling. The round-8 first-timer
     named "pick decimals" among what makes this board intimidating, and on a
     made cell the number is the one fact stated three other places. A
     forecast's slot now holds the word PROJ instead of a number (see
     cellIdRight), and an OPEN cell keeps its number - a bare coordinate is
     what an empty slot IS, the reference's own idiom. Every cell's title
     still carries the pair in full. The legibility probe appends its
     fixtures to this same grid, so its made-cell fixture loses the
     coordinate here too; that is the shipped board it is supposed to
     measure, and its clipping bound only gets easier with the number gone. */
  #draftGrid .dg-cell.filled .dg-id .dg-pk{display:none}
  /* THE MARK ON THE IDENTITY LINE, MEASURED DOWN TO FIT IT. On the name's row
     the badge was 21.8px of a 70px cell and it was the name that paid; on this
     line it is beside `WR · DEN (10)` at 49.1px and 2px of flex gap, and 8px
     with 2px of padding brings it to 18.7px - 69.8 of the 72px this cell now
     has. The worst identity line the board can produce (`DST · WAS (14)`,
     52.5px) can never carry a mark, because gridSignal refuses K and DST, so
     the two worst cases are disjoint by construction rather than by luck. */
  /* 7.5px, WHICH IS THE IDENTITY LINE'S OWN SIZE, and that is what keeps every
     round the same height: the band is as tall as its tallest item, so a badge
     one pixel over sets a different row height for the thirteen rounds that
     carry marks. Measured after: 41.5px, rounds 1 through 15, marked or not. */
  .dg-cell .dg-id .dg-sig{font-size:7.5px;padding:0 2px;letter-spacing:0}
  /* KEPT is a marker, not a reading, and at the desktop's tracked mono it took
     14px more than the longest keeper identity line (`WR · DEN (10)`) had
     left. Measured down until that worst case cleared with nothing over. */
  .dg-cell .dg-id .dg-kept{font-size:6px;letter-spacing:.06em}

  /* THE POSITION COLOUR FILLS THE PHONE CELL.
     This is NOT a reversal of the desktop decision two thousand lines up -
     that one is about 150 cells at once, where a saturated grid shouts and the
     eye has nowhere to rest. A phone shows about twenty, at 80px, with a name
     clamped to 10px over two lines, and there the mix that keeps a desktop
     board calm reads as no colour at all. His frame answers "how many backs
     went in round 2" from across the room; a 4px stripe cannot.
     The mix is 62%, not 100%: at full saturation the name needs a different
     ink per position, and a cell that changes its text colour six ways is six
     contrast bounds to keep rather than one.

     A KEPT PLAYER IS A FILLED SLOT and fills with the rest of them. On a board
     where colour means "taken at this position", leaving keepers dark would
     have drawn the four cells his league starts with as the only holes in
     round 9. They keep the dashed edge that says they were never picked. */
  .dg-cell.filled.pos-QB,.dg-cell.keeper.pos-QB{background:color-mix(in oklab,var(--pos-qb) 62%,var(--bg-cell))}
  .dg-cell.filled.pos-RB,.dg-cell.keeper.pos-RB{background:color-mix(in oklab,var(--pos-rb) 62%,var(--bg-cell))}
  .dg-cell.filled.pos-WR,.dg-cell.keeper.pos-WR{background:color-mix(in oklab,var(--pos-wr) 62%,var(--bg-cell))}
  .dg-cell.filled.pos-TE,.dg-cell.keeper.pos-TE{background:color-mix(in oklab,var(--pos-te) 62%,var(--bg-cell))}
  .dg-cell.filled.pos-K,.dg-cell.keeper.pos-K{background:color-mix(in oklab,var(--pos-k) 50%,var(--bg-cell))}
  .dg-cell.filled.pos-DST,.dg-cell.keeper.pos-DST{background:color-mix(in oklab,var(--pos-dst) 50%,var(--bg-cell))}
  /* The stripe was the whole signal; against a filled cell it is a seam. */
  .dg-cell.filled{border-left-width:1px}
  /* THE PICK NUMBER STAYS ON THE PHONE. A container query two thousand lines
     up drops it under 76px, which is every cell on this board - it was written
     when this row also carried a position chip and a separate club span and
     three things could not share 72px. The chip is gone (the fill and the word
     both say the position now), so the line holds the identity and the
     coordinate, which is how you read a snake. Without this the line measured
     zero and the phone cell stated only a name. */
  .dg-cell.filled .dg-id .dg-pk,
  .dg-cell.keeper .dg-id .dg-pk{display:block}
  .dg-cell.filled .dg-nm,.dg-cell.keeper .dg-nm{color:var(--text-primary)}
  /* The watermark, on a saturated fill: the cell's own dark ink at low
     opacity, which is what the reference does and what keeps one line of
     secondary text from needing a contrast bound per position hue. */
  .dg-cell.filled .dg-id,.dg-cell.keeper .dg-id{
    color:var(--text-primary);opacity:.62}

  /* THE GHOST SPEAKS THE PHONE BOARD'S LANGUAGE - identity line in words with
     the club and the bye, name clamped to two lines, coordinate or mark at the
     far end - AND IT SPEAKS IT WITHOUT COLOUR. A projection that used a
     different grammar from a pick would be a second board sharing one grid; a
     projection that used the same COLOUR as a pick is what this board actually
     shipped, and it is why thirty forecasts read as a draft already gone.
     This was `20%` of the position hue against the made cell's `62%`, i.e. the
     separation was a saturation step, and a saturation step is exactly the kind
     of difference a reader cannot judge without both samples side by side. The
     phone is where it was worst: at 78px the desktop's 4px stripe is a seam, so
     the wash was carrying the whole distinction alone.
     Colourless now, at the board's own cell ground. The dashed edge stays and
     is the only thing left saying "nobody has picked here", which is what the
     caption's two swatches are there to name. */
  .dg-cell.ghost.pos-QB,.dg-cell.ghost.pos-RB,.dg-cell.ghost.pos-WR,
  .dg-cell.ghost.pos-TE,.dg-cell.ghost.pos-K,.dg-cell.ghost.pos-DST{
    background-color:var(--bg-cell)}
  .dg-cell.ghost{border-width:1px;border-style:dashed;
    border-color:color-mix(in srgb,var(--border-strong) 70%,transparent)}
  .dg-cell.ghost .dg-id .dg-pk{display:block}
  /* THE WATERMARK NEEDS THE CELL'S OWN INK ON A PALE FILL TOO. A ghost is a
     20% wash, so the muted grey the desktop rule leaves on it was 7.5px of
     light grey on near-white - under the floor at the exact size where the
     bye week and the pick coordinate live, and those two facts are stated
     nowhere else beside a name in this room. The filled cell already solves
     this the reference's way, by darkening the cell's own ink; the pale cell
     needs more of it, not less. */
  .dg-cell.ghost .dg-id{color:var(--text-primary);opacity:.72}
  .dg-cell.ghost .dg-nm{color:var(--text-secondary)}
  /* A BUY IS DRAWN LOUD, AND IT IS STILL DRAWN AS A FORECAST.
     This rule read "they stop being projections visually: full saturation,
     solid edge, name in primary ink", and that is the single sharpest source
     of the defect four of six critics named: twenty-eight cells on an untouched
     board, deliberately given a made pick's exact treatment, under a header
     saying `0 picked`. The intent was right - a projection your own number
     disagrees with is the reason to look at this board - and it was spent in
     the one currency that was already reserved.
     The lift moves entirely into the accent, which is the misprice channel and
     not a position hue: accent ring, accent edge COLOUR, primary ink on the
     name. The edge stays dashed and 1px. Loud, and unmistakably not a pick. */
  /* The accent wash is restated here because the phone's own `.pos-*` neutral
     rule three blocks up matches at the same specificity and later in the
     file, so the desktop declaration loses the cascade at this width - the
     first measurement at 375 came back with the buy cells painted plain white
     while 1280 had them in accent-soft. Same lift, both widths. */
  .dg-cell.ghost.sig-val{background-color:var(--accent-soft);border-color:var(--accent);
    box-shadow:inset 0 0 0 1px var(--accent)}
  .dg-cell.ghost.sig-val .dg-nm{color:var(--text-primary);font-weight:700}
  .dg-sig{padding:1px 3px;font-size:8.5px}
  /* THE LIT CELL AT PHONE DENSITY. The name band's 24px floor is for a
     two-line clamped player name; this cell's first band holds the word "You"
     and a coordinate, so it takes its own line and hands the 12px back.
     What the 12px buys is the CLOCK: 15px against the 13px it had, in a cell
     that is now 41.5px instead of 65.5px. The market's read is 6.5px stacked
     under a 6px label at the right of the same band - 33px of the 72px the
     cell has, beside a 36px clock. Measured, both. */
  /* ONE LINE, AND IT IS THE RULE RATHER THAN THE COPY THAT GUARANTEES IT.
     The band above clamps every cell's name to TWO lines with a 24px floor,
     which is right for a wrapped player name; this cell holds a label, so it
     zeroed the floor and took one line - while the label was "You". The
     moment the clock sat on somebody else the label was "On the clock", it
     wrapped inside its 45px box, and the cell went 41.5px to 49.0px, taking
     round 1 with it. `display:block` puts the base rule's nowrap and ellipsis
     back in charge of this one band, so the cell's height stops depending on
     how long a word somebody types into it. */
  .dg-cell.now .dg-nm{min-height:0;line-height:1.2;display:block;
    white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
  .dg-cell.now .dg-top{margin-bottom:1px}
  .dg-clock{font-size:15px}
  .dg-mkt{font-size:6.5px;letter-spacing:0}
  .dgm-l{font-size:6px;letter-spacing:.1em;margin-bottom:0}
  /* THE FORECAST'S NAME GETS THE BOARD'S OWN TREATMENT, which it was the one
     name on the screen not getting. A ghost cell wraps its name to two lines
     and ellipses what is left over, and 0 of 181 of them clip; this box kept
     nowrap in 33px and cut 133 of the 179 names the market can put here,
     including the one the cell exists to show. Two lines cost NOTHING here -
     measured, the cell still stands at 41.5px against 41.5 for the other
     fourteen rounds, because the pair still fits inside the height the 15px
     clock beside it already sets. 166 of 179 now fit whole.
     The alternatives were measured and are worse: giving the name the whole
     band on its own row fits all 179 but takes the cell to 54.5px, and
     reversing round 7's 15px clock to 13px fits only 95. */
  .dgm-n{display:-webkit-box;-webkit-line-clamp:2;-webkit-box-orient:vertical;
    white-space:normal;overflow:hidden;text-overflow:ellipsis;line-height:1}

  /* ==================================================================== */
  /* PRE-DRAFT ON A PHONE IS A ROOM, NOT A FORM.
   *
   * The mode below already owns the RUNNING draft. Everything before pick 1
   * was still the old page: a masthead, a three-line dek, five fields and nine
   * buttons, with the board itself starting at 1,200px. Measured on a 390x844
   * screen, the first screenful of the draft room contained no draft.
   *
   * The fix is the same argument the mode makes, applied one state earlier.
   * The room's state (#drRail) and what the market is getting wrong right now
   * (#drEdge) are the screen; the controls that SHAPE a draft are a drawer
   * behind one chip, opening the row that has always been there, in place,
   * rather than a second settings surface to keep in step with the first.
   *
   * The chip and the drawer live at THIS width only, because only here is
   * there a choice to make. Desktop keeps every control on screen and gains
   * only the ordering. */
  #panel-draft .masthead .mh-dek,
  #panel-draft .masthead .mh-kicker{display:none}
  #panel-draft .masthead{margin-bottom:var(--sp-3)}
  #panel-draft .masthead .mh-title{font-size:30px}
  #btnDrSetup{display:inline-flex}
  #panel-draft .draft-setup{display:none}
  body.dr-setup-open #panel-draft .draft-setup{display:flex}
  .drr-act{padding:8px var(--sp-3)}
  .drr .clock-card{padding:12px var(--sp-3);gap:12px}
  .drr .clock-time{font-size:24px}
  .drr .clock-team{font-size:13px;line-height:1.25}
  .drr .clock-card .caps{font-size:8.5px;letter-spacing:.1em}
  .drr .clock-meta{min-width:0;flex:1 1 auto}
  .drr .clock-until{flex:0 0 auto;white-space:nowrap}
  .drr .clock-until .num{font-size:15px}
  .clock-count{font-size:22px}

  /* THREE ROWS, and the other three are rendered but not drawn. A misprice
     board that pushes the actual draft board off the screen has reproduced the
     defect it was built to fix; three rows is 210px and the grid still opens
     above the fold. The rest are one scroll away, in the same list, rather
     than off the right edge of a horizontal scroller nobody finds. */
  .dre:nth-child(n+4){display:none}
  .dre-dek{display:none}
  .dre-hd{padding:10px var(--sp-3) 8px;border-bottom:1px solid var(--border-subtle);
    gap:6px;font-size:9px}
  .dre-t{letter-spacing:.08em}
  .dre-n{font-size:8px}
  .dre{padding:9px var(--sp-3) 10px}
  .dre-surv{min-width:0;font-size:10px}
  .dre-scale{min-width:46px}
  .dre-read{min-width:48px;font-size:8px}
  /* `.dre .dre-act`, MATCHING THE BASE RULE'S SPECIFICITY, and that is the
     whole fix. Written as a bare `.dre-act` this block asked for 9px text in
     9px of padding and the phone computed 10px in 12px: the base rule carries
     an extra `.dre` deliberately (to out-weigh `.draftbtn`, declared later as
     a plain text link), so at (0,1,0) against (0,2,0) this override lost on
     specificity no matter how far down the file it sat. It was dead the day it
     was written.
     44px is the thumb minimum this app holds every target to, and the pill was
     34px - the panel's whole promise is that a row is ONE TAP from drafting or
     queueing, so it is the last control that may be hard to hit. Costs 10px a
     row, 30px over the three drawn here, measured. */
  .dre .dre-act{padding:0 9px;font-size:9px;min-height:44px}
  .dre-gap{font-size:14px;min-width:30px}
  /* THE MODE TAKES THE CARD AND LEAVES THE WRAPPER. `.clock-card` is the
     middle of the mode's fixed bar, and it is a CHILD of the rail now - so the
     rail may not be display:none, or the mode loses its clock along with the
     chrome around it. `display:contents` removes the box and keeps the child,
     which is exactly the distinction being made. */
  body.dr-full #panel-draft .drr{display:contents}
  body.dr-full #panel-draft .drr-act{display:none}
  /* YOUR TURN INVERTS THE WHOLE BAR IN THE MODE, not the clock card's own box.
     The card is fixed between two measured insets (52 and 100), so painting
     turf on IT drew a black rectangle floating between the X and the pick
     count, with hard edges in the middle of a light bar - a highlight that
     read as a rendering fault. The bar is the object that is on the clock, so
     the bar is what changes colour.

     `#drRail.you ~ .dr-bar` is a plain sibling selector and it only works
     because the rail precedes the bar in the markup. That is not a
     coincidence to rely on silently: the rail is the room's STATE and the bar
     is the mode's chrome, and state is written before the thing that reads
     it. */
  body.dr-full #drRail.you .clock-card{background:none;border:0}
  body.dr-full #drRail.you ~ .dr-bar{
    background:var(--turf,#0f1a22);border-bottom-color:var(--turf,#0f1a22)}
  body.dr-full #drRail.you ~ .dr-bar .dr-x{color:rgba(255,255,255,.72)}
  /* The count's tint on the turfed bar went with the count (2026-08-22, the
     rule ~90 lines down). It could not have fired since: the only state that
     still shows this box in the mode is FIN, and the rail carries `fin`
     exactly when it does not carry `you`. A rule that cannot match is a
     statement about the design that is not true. */
  /* THE COUNTDOWN TAKES THE "YOUR PICK IN" SLOT, and only when there IS a
     countdown. The bar has 238px between its two measured insets; with a
     running clock in it, "your pick in / now" was clipped mid-word against
     the pick count. The label was kept in this bar because "14 picks" is not
     self-describing next to "17 / 180 picks" - a live timer is, and the board
     directly below already draws your own column. With no pick clock set,
     nothing here fires and the bar is exactly what it was. */
  body.dr-full .clock-count ~ .clock-until{display:none}
  body.dr-full .clock-count{margin-left:var(--sp-3);font-size:19px}
  /* The misprice board follows the same law as the queue and the roster: the
     mode does not delete a surface, it moves it to the sheet. `:not(.dr-sheet)`
     is the sentence the other three panels are already written with. */
  body.dr-full #panel-draft .dr-edge:not(.dr-sheet){display:none}
  /* As a sheet face, the list is the scroller and everything else is fixed
     furniture - the same arrangement #availList has, for the same reason. */
  body.dr-full .dr-edge.dr-sheet{padding:0}
  body.dr-full .dr-edge.dr-sheet > *{min-height:0}
  body.dr-full .dr-edge.dr-sheet .dre-hd{flex:0 0 auto}
  body.dr-full .dr-edge.dr-sheet .dre-list{
    flex:1 1 auto;overflow:auto;-webkit-overflow-scrolling:touch}
  /* Every row shows in the sheet: the three-row cut above is a rule about the
     PAGE, where the board has to leave room for the grid. The sheet is the
     board's own screen and the whole point of opening it is to see the rest. */
  body.dr-full .dr-edge.dr-sheet .dre:nth-child(n+4){display:block}

  /* ==================================================================== */
  /* FULL SCREEN: A DRAFT ON A PHONE IS A MODE, NOT A SECTION.
   *
   * His verdict on the first pass was "the two aren't comparable", and he was
   * right: the board and the sheet were both Sleeper's, and they sat under a
   * masthead, a five-field setup row and seven buttons, in a page that
   * scrolled. Sleeper's whole viewport is close / clock / board / sheet.
   *
   * So the mode OWNS THE VIEWPORT. Every piece is fixed and measured off one
   * variable, --sheet-h, so the board can never disagree with the sheet about
   * where they meet. The page behind it does not scroll; everything that
   * scrolls does it inside its own box.
   *
   * The bit is published by renderDraft and the X clears it. Nothing here is
   * gated on the class alone - it is gated on the class INSIDE this query, so
   * a desktop that somehow carries `dr-full` renders exactly as it always has.
   */
  body.dr-full{
    /* 46, and the two pixels over the thumb floor are the strip's own top
       border plus a pixel of margin. This was drawn at 36; the suite named it
       four times (`BUTTON.drtab:35`), and the first fix - 44 exactly - still
       measured 43, because the border is inside the height. A floor met by
       arithmetic that has to come out exact is a floor that will fail again,
       so the button carries its own `min-height` as well. */
    /* 42vh, DOWN FROM 45vh (2026-08-15). The mode exists so the board is the
       biggest thing on the screen, and at 45vh it was not: 405px of sheet
       against 352px of board, and the board came out at 39.1% against the 40%
       floor this mode holds itself to. The shortfall was EIGHT pixels, and
       44vh would have cleared it by one - the bound-that-barely-holds this
       suite has already lost a CI day to. 42vh puts the board at 379px, 42.1%,
       and costs the player list about one row. Measured, not chosen. */
    --sheet-h:42vh;--dr-bar:48px;--dr-tabs:46px;--dr-top:48px;
    /* THE FORECAST COLUMN BUYS 22px FROM THE TWO RANK GUTTERS AND THE NAME.
       `gone by 2.12` is 56px at 10px in Chrome and the phone's shared track is
       56px, which is a clip waiting for one long round label. The gutters hold
       a two- or three-digit rank and a two-letter position chip and were drawn
       at 26 for a 44px thumb target they do not own - the row is the target
       here, not the rank. Measured at 375: 24+24+84+78+52 plus four 8px gaps
       and 32px of padding is 342 of 375, and the name still gets 133px, which
       is 15 characters of the 17px face. */
    --ba-cols:24px 24px minmax(84px,1fr) 78px 52px;
    overflow:hidden}
  /* Open is a state of THE SHEET, and the sheet is whichever panel the tab
     names - so it lives on the body, not on #baPanel. Held on the panel, it
     snapped shut every time you switched face. */
  body.dr-full.dr-open{--sheet-h:78vh}
  /* dvh where it exists: 100vh on iOS Safari is the height with the browser
     chrome GONE, so the sheet's last row sits under the toolbar until you
     scroll something that cannot scroll. */
  @supports (height:100dvh){
    body.dr-full{--sheet-h:42dvh}
    body.dr-full.dr-open{--sheet-h:78dvh}
  }

  /* ---- the bar --------------------------------------------------------- */
  body.dr-full .dr-bar{
    display:flex;position:fixed;top:0;left:0;right:0;height:var(--dr-bar);
    z-index:41;align-items:center;justify-content:space-between;
    padding:0 var(--sp-2);background:var(--bg-app);
    border-bottom:1px solid var(--border-subtle)}
  .dr-x{width:44px;height:44px;padding:0;border:0;background:none;
    color:var(--text-secondary);font:300 26px/1 var(--font-ui);cursor:pointer}
  .dr-x:active{color:var(--text-primary)}
  /* TWO LINES, RIGHT ALIGNED. The string it carries stopped being "0 / 180
     picks" and became "0 picked · 152 projected", which is the sentence that
     stops the board being misread - and it does not fit the 92px this slot
     measures. It wraps inside its own box rather than widening it: the box is
     what the clock card's right inset is measured against. */
  .dr-count{font:var(--fs-xs)/1.25 var(--font-num);color:var(--text-muted);
    padding-right:var(--sp-2);font-variant-numeric:tabular-nums;
    max-width:92px;text-align:right;white-space:normal}
  /* ...AND IN THE MODE IT IS NOT ON THE BAR AT ALL (2026-08-22).
   *
   * The bar is 48 of the 78 pixels above the board, and it was spending its
   * right-hand 92px on a two-line string parked against the pick timer. Four
   * of six critics in round 6 read the strip over this board as jargon to be
   * decoded, sitting exactly where the eye should lock onto the turn - and the
   * turn callout is the one thing that round records Edge winning outright.
   * So the bar gives up the count and becomes the turn alone: the X, the pick
   * coordinate with the seat, and the countdown.
   *
   * NOTHING IS LOST AND NOTHING IS DELETED. `setupMeta()` still has one writer
   * and three sinks; this hides one of them at one width, and the caption one
   * object down now prints the same fact as a sentence (see `.dgc-state`
   * below). The element stays because the sink stays.
   *
   * FIN IS THE EXCEPTION, because at FIN this box is not the count: renderDraft
   * replaces its contents with the report-card door, which is a control and the
   * payoff of the whole draft. The rail's own state says which face is up -
   * `#drRail.fin ~ .dr-bar` is the sibling idiom `.you` already uses here, and
   * it works for the same reason: the rail is the room's STATE and the bar is
   * the mode's chrome, so state is written before the thing that reads it. */
  body.dr-full .dr-count{display:none}
  body.dr-full #drRail.fin ~ .dr-bar .dr-count{display:block}
  /* The bar's right side at FIN: the count's slot becomes the door to the
     grade. Drawn like the sheet's draft pill - the other control this mode
     asks a thumb to hit. */
  .dr-report{min-height:34px;padding:0 14px;border:1px solid var(--accent-border);
    border-radius:17px;background:var(--accent-soft);color:var(--accent);
    font:600 11px/1 var(--font-ui);letter-spacing:.08em;text-transform:uppercase;
    cursor:pointer}
  /* THE CLOCK CARD IS THE BAR'S MIDDLE, in place - not a second reading of
     the pick number drawn into new markup. The insets keep it clear of the X
     and the count, and the z-index puts it over the bar's background rather
     than under it. The pick TIMER is not here on purpose: Edge ticks it inside
     the on-the-clock cell, which is where his reference frame draws it too. */
  /* The insets are MEASURED, not guessed: 52 clears the 44px X plus the bar's
     padding, and the right inset clears whatever the bar's right side holds.
     At the 70 this was drawn with, the clock's box lay over the count by 30px
     and over the X by 6 - invisible, because the ink inside it is centred, and
     a box that overlaps a button still takes its taps.
     SYMMETRIC AT 52 NOW (2026-08-22), because the count left the bar and the
     turn is the only thing on it. Nothing on the right side to clear, and a
     card centred between two equal insets is centred in the bar. The FIN face
     is the one that still holds something there - the report-card door, an
     84x34 box measured inside its 92px slot - so it keeps the 112 the count
     was measured for: 92px of box, 8px of its own padding and 12px of air.
     Same rail-state idiom as the count rule above, and it has to be here or
     the card's fixed box would sit over the door and take its taps. */
  body.dr-full .clock-card{
    position:fixed;top:0;left:52px;right:52px;height:var(--dr-bar);z-index:42;
    margin:0;padding:0;background:none;border:0;border-radius:0;
    justify-content:center;gap:var(--sp-3);overflow:hidden}
  body.dr-full #drRail.fin .clock-card{right:112px}
  /* THE TURN GETS THE ROOM THE COUNT GAVE UP. 16px was what a 226px card with
     a two-line string beside it could hold; the card is 286px at 390 now and
     the pick coordinate is the object the whole bar exists to state. 26px is
     the largest face this 48px bar takes with the tabular numerals' own box
     clear of both edges - measured at 375 and 390, not chosen - and the seat
     beside it follows one step, 12px to 14px. The countdown is untouched at
     19px: the pair IS the turn, and round 6 records both halves of it
     passing. */
  body.dr-full .clock-time{font-size:26px}
  body.dr-full .clock-card .caps{display:none}
  body.dr-full .clock-team{font:600 14px/1.2 var(--font-ui)}
  body.dr-full .clock-until{margin-left:0}
  body.dr-full .clock-until .num{font-size:12px}
  /* THIS ONE LABEL STAYS. Everything else on the bar is self-describing, but
     "14 picks" is not - the count at the far right of the same bar reads
     "17 / 180 picks", and two different pick counts an inch apart with only
     one of them named is a bar that has to be decoded. */
  body.dr-full .clock-until .caps{
    display:block;font:400 7px/1 var(--font-num);letter-spacing:.1em;
    color:var(--text-dim)}

  /* ---- what the mode hides, and the one way back ----------------------- */
  /* The page's own chrome, and the draft tab's pre-draft surfaces. They are
     not lost: the X brings the whole page back with every one of them, which
     is why there is no settings sheet in here to keep in step. */
  body.dr-full .topbar,
  body.dr-full #panel-draft .masthead,
  body.dr-full #panel-draft .draft-setup,
  body.dr-full #panel-draft .keepers,
  body.dr-full #panel-draft .rptpanel,
  body.dr-full .sidenav{display:none}
  /* The aside's panels are not hidden - they BECOME the sheet when their tab
     is chosen. The one that is not the sheet is the one that is not on
     screen. */
  body.dr-full #panel-draft aside > .panel:not(.dr-sheet),
  body.dr-full #panel-draft .room-main > .panel:not(.dr-sheet),
  body.dr-full #panel-draft .pickpath:not(.dr-sheet){display:none}
  body.dr-full #panel-draft .wrap{padding:0}
  /* `> .wrap >`, and it is not decoration. `.room` is TWO different things in
     this stylesheet: the draft room's two-column layout, and the column in a
     Best Available row that prices the sorted source against the market. A
     descendant selector here matched both, and the row's price column - which
     the phone deliberately drops - came back as a stray em dash under every
     rank. The class-as-a-WORD lesson, one turn further: check what else
     already answers to the name before you write a rule for it. */
  body.dr-full #panel-draft > .wrap > .room{display:block;gap:0}
  body.dr-full #panel-draft .room-main{padding-bottom:0}
  /* MANUAL ENTRY SURVIVES THE MODE, AND ONLY THEN. With opponents on manual
     the typeahead is the ONLY way any pick gets logged, so hiding it would
     strand him exactly where his first real draft already went wrong. With
     opponents on auto it is a second door onto the sheet's DRAFT button, and
     the mode's whole argument is that 52px of screen is worth more than a
     second door. `dr-manual` is published beside the other two bits. */
  body.dr-full:not(.dr-manual) #taWrap{display:none}
  body.dr-full #taWrap{
    position:fixed;top:var(--dr-bar);left:0;right:0;z-index:39;height:52px;
    margin:0;padding:var(--sp-2) var(--sp-3);background:var(--bg-app);
    border-bottom:1px solid var(--border-subtle)}
  /* ...and when it is there, the board starts under it. TWO variables, not
     one: `--dr-bar` is where the typeahead hangs from and `--dr-top` is where
     the board begins, and folding them together put the input 52px below
     itself. */
  body.dr-full.dr-manual{--dr-top:100px}
  /* The way back IN, once the X has been used. Draft-only and phone-only,
     which is the whole life of the mode. */
  body.drafting:not(.dr-full) #btnFullDraft{display:inline-flex}

  /* ---- the board gets the space that frees ----------------------------- */
  /* 32vh was chosen against a page that still had 400px of controls above it.
     With those gone the board is everything between the bar and the sheet -
     about 55vh on his phone, against Sleeper's ~60%. `max()` because the sheet
     opens to 78vh and a negative height is not a height. */
  /* THE CAPTION IS FIXED FURNITURE IN THE MODE, because the board is. The
     wrap is taken out of flow here, which would leave the caption floating on
     a collapsed .dg-board behind the bar; it hangs off the same --dr-top the
     board does, and the board starts under its MEASURED height. */
  body.dr-full .dg-cap{
    position:fixed;top:var(--dr-top);left:0;right:0;z-index:38;
    gap:1px var(--sp-2);padding:3px var(--sp-3) 4px;
    background:var(--bg-app);border-bottom:1px solid var(--border-subtle)}
  /* AND AT LOW CLOCK THE CAPTION SAYS THE CLOCK. Every token here is lifted
     off `.dgc-state` rather than merely resembling it, because this element
     STANDS IN for that sentence in that slot: same flex basis, same face,
     same colour, so the strip does not move when the two trade places. The
     swap is a swap and not an addition for a measured reason - .dg-cap is
     `flex-wrap:wrap`, so a second string wraps to a second row and takes the
     strip 36px -> 51px, pushing the board down. Self-restoring: `low` clears
     the instant startClock() runs, which is only ever from logPick(). */
  body.dr-full #dgHint.low{
    display:inline-block;flex:1 1 220px;min-width:0;
    font:600 var(--fs-xs)/1.4 var(--font-ui);color:var(--text-secondary)}
  body.dr-full #dgHint.low ~ #dgCapState{display:none}
  /* IN THE MODE THE SENTENCE IS THE CAPTION, AND THE CAPS STRIP GOES.
   *
   * This rule ran the other way until 2026-08-22, and the trade it recorded
   * was honest at the time: the bar two objects up stated the count, so the
   * sentence was a duplicate and the swatches said the encoding better than
   * the clause did. R1 above removed the first half of that argument - the
   * count is not on the bar any more - and round 6 removed the second: four
   * of six critics read `PICKED / PROJECTED / PICKS VS OUR BOARD / +16 VALUE /
   * -12 REACH`, two wrapped lines of tracked caps, as unexplained jargon
   * sitting exactly where the eye should enter the board. A legend is only a
   * shortcut if the reader already has the vocabulary; this reader does not,
   * and the plain-English answer was already written one element to the left
   * and switched off.
   *
   * So the swap: the sentence shows, the strip hides, every fact survives, and
   * the vocabulary changes from abbreviations to English. `boardStateLine()`
   * names the count, the fact channel and - when the rendered board actually
   * carries one - what the +/- marks measure.
   *
   * NOTHING IS DELETED. The swatches stay in the markup (they are the desktop
   * caption, and they are real cells wearing the board's own classes), and in
   * the mode the same two swatches are still one tap away behind the board's
   * own RD/? door, which carries the full key.
   *
   * The page outside the mode keeps BOTH: there the board sits in a scroller
   * rather than in a fixed height budget, so the caption is not spending the
   * board's pixels and the two halves can both be on screen. */
  body.dr-full .dgc-legend{display:none}
  body.dr-full .dg-wrap{
    position:fixed;top:calc(var(--dr-top) + var(--dg-cap-h,0px));left:0;right:0;
    height:max(0px,calc(100vh - var(--dr-top) - var(--dg-cap-h,0px)
      - var(--dr-tabs) - var(--sheet-h)));
    /* No side padding: content scrolls THROUGH a scrollport's padding, so a
       2px pad was a 2px lane at the left edge where cells slid out from under
       the sticky gutter. */
    max-height:none;margin:0;padding:0;overflow:auto;
    transition:height .22s ease}
  @supports (height:100dvh){
    body.dr-full .dg-wrap{
      height:max(0px,calc(100dvh - var(--dr-top) - var(--dg-cap-h,0px)
        - var(--dr-tabs) - var(--sheet-h)))}
  }
  /* The key follows the board. In the mode the board is FIXED, so its wrapper
     is no longer the box the panel can hang off - it hangs off the same two
     variables the board itself is measured from, which is the only way the two
     cannot drift apart. Still an overlay, still zero height taken from the
     list below. */
  body.dr-full .dg-key{position:fixed;
    top:calc(var(--dr-top) + 46px + var(--dg-cap-h,0px));
    left:6px;right:6px;width:auto;z-index:45}

  /* ---- the sheet, and its faces ---------------------------------------- */
  /* The tab strip sits on the sheet's top edge and is measured from the same
     variable, so the two cannot drift apart. */
  body.dr-full .dr-tabs{
    display:flex;position:fixed;left:0;right:0;bottom:var(--sheet-h);
    z-index:41;height:var(--dr-tabs);align-items:stretch;
    background:var(--bg-app);border-top:1px solid var(--border-subtle);
    transition:bottom .22s ease}
  .drtab{flex:1 1 0;min-height:44px;background:none;border:0;
    border-bottom:2px solid transparent;
    color:var(--text-muted);cursor:pointer;
    font:600 10px/1 var(--font-ui);letter-spacing:.12em;text-transform:uppercase}
  .drtab.active{color:var(--accent);border-bottom-color:var(--accent)}

  /* Whichever panel the tab named. `position:fixed` rather than sticky: the
     sheet must not scroll away from the thumb when the board behind it moves. */
  body.dr-full .dr-sheet{
    position:fixed;left:0;right:0;bottom:0;z-index:40;
    height:var(--sheet-h);display:flex;flex-direction:column;
    margin:0;border-radius:14px 14px 0 0;
    background:var(--bg-surface);border:1px solid var(--border-subtle);
    border-top-color:var(--accent-border);
    box-shadow:0 -10px 30px rgba(0,0,0,.55);
    overflow:hidden;transition:height .22s ease}
  /* THE HEADER IS THE HANDLE, and it says so with a grab bar. A sheet whose
     only affordance is a 4px edge is a sheet nobody opens. Written against
     whatever the panel calls its header, because the pick path is not a
     .panel and draws a .pp-hd instead. */
  body.dr-full .dr-sheet > .panel-hd,
  body.dr-full .dr-sheet > .pp-hd{
    position:relative;padding-top:14px;cursor:pointer;flex:0 0 auto;
    touch-action:manipulation}
  body.dr-full .dr-sheet > .panel-hd::before,
  body.dr-full .dr-sheet > .pp-hd::before{
    content:"";position:absolute;top:5px;left:50%;transform:translateX(-50%);
    width:34px;height:4px;border-radius:2px;background:var(--text-dim)}
  /* THE SHEET'S HEADER IS 92px OF A 42vh SHEET UNTIL IT IS TOLD OTHERWISE -
     a title the tab strip has already said, over a sentence wrapped to three
     lines by a full-width select. The mode keeps the CONTROL (which room's ADP
     the survival column follows is a real decision, and the off-format badge
     hangs off it) and drops the naming. */
  body.dr-full .ba-title{display:none}
  body.dr-full #baPanelHd{
    flex-wrap:nowrap;align-items:center;gap:var(--sp-2);
    padding:12px var(--sp-3) 4px;font-size:10px}
  body.dr-full #baPanelHd .survline{
    display:flex;align-items:center;gap:6px;min-width:0;flex:0 0 auto}
  /* The pick this column is measured to is now the COLUMN HEADER, one row
     above the numbers. Repeating it here would be the same fact twice on one
     screen while the search field it is crowding does not fit. */
  body.dr-full .sl-to{display:none}
  body.dr-full .ba-search{height:32px;flex:1 1 auto;max-width:none}
  body.dr-full #adpSource{height:26px;max-width:118px;padding:0 20px 0 7px;
    font-size:11px}
  body.dr-full #survNote{margin-left:auto;text-align:right;font-size:9px}
  /* THE SHEET'S FIXED FURNITURE IS A TAX ON THE LIST, and the list is the
     reason the sheet exists. Measured at 45vh on a 812pt phone (the sheet is
     42vh since 2026-08-15, so these are the roomier numbers, not the current
     ones): a 67px header
     and a 72px chip row left 224px - three and a half rows - and the advisory,
     when it appears, took it to 174 and under three. On his phone, with
     Safari's toolbar over the bottom of it, that is barely two. Every number
     below is the same furniture drawn tighter: 44 and 58 instead of 67 and 72,
     which buys a whole extra player without giving up the 44px thumb floor
     (the circles are 48). */
  /* One line, and it is advice - it may not cost a row of the thing it is
     advising about. */
  body.dr-full #baAdvisory{
    padding:5px var(--sp-3);font-size:10px;line-height:1.3;gap:6px;
    max-height:32px;overflow:hidden}
  /* The last row has to clear the home indicator when Safari's toolbar
     collapses, and the inner scroll must not chain to the page behind it -
     a 200px scroller inside a fixed sheet is exactly where iOS hands the
     gesture to the wrong element. */
  body.dr-full #availList{
    padding-bottom:env(safe-area-inset-bottom,0px);
    overscroll-behavior:contain}

  /* Everything after the header scrolls inside the sheet, whichever face it
     is wearing. The chips and the column labels are the exception: they are
     the sheet's own header and stay put while the list under them moves,
     because "RB 2/2" is the one readout you want fixed while shopping. */
  body.dr-full .dr-sheet > *{min-height:0}
  body.dr-full #baTabs,body.dr-full #baHead,body.dr-full #baAdvisory{flex:0 0 auto}
  body.dr-full #availList,body.dr-full #queuePanel,body.dr-full #myRoster{
    flex:1 1 auto;overflow-y:auto;-webkit-overflow-scrolling:touch}
  /* The pick path is one flowing document rather than a header over a scroll
     box, so it scrolls whole and its header sticks. */
  body.dr-full .dr-sheet.pickpath{display:block;overflow-y:auto;padding:0;
    -webkit-overflow-scrolling:touch}
  body.dr-full .dr-sheet.pickpath > .pp-hd{
    position:sticky;top:0;z-index:2;background:var(--bg-surface)}

  /* THE PICK PATH IN A 375px SHEET.
   *
   * The ledger is seven tracks summing to 1,120px, and inside the sheet that
   * meant one column visible and six behind a sideways scroll - the same
   * failure the Momentum and Injuries tables were rewritten to escape, and
   * with the same fix: SAME MARKUP, rows restyled to wrap, each cell stating
   * its own name from `data-l` rather than from a header row that cannot fit.
   * The header goes, because a header over columns that no longer exist names
   * the wrong things.
   *
   * The mode may not drop this panel: the pick path is the answer this whole
   * app computes, and on his phone it was 600px off the right edge. */
  body.dr-full #pickPath .pp-ledger{overflow-x:visible}
  body.dr-full #pickPath .pp-lhead{display:none}
  body.dr-full #pickPath .pp-lrows{min-width:0;gap:6px}
  body.dr-full #pickPath .pp-lrow{
    display:flex;flex-wrap:wrap;align-items:center;gap:5px 10px;
    min-width:0;padding:9px 11px}
  body.dr-full #pickPath .pp-rank{width:26px;height:26px;font-size:12px}
  body.dr-full #pickPath .pp-cand{flex:1 1 120px}
  body.dr-full #pickPath .pp-cand b{font-size:15px;white-space:normal}
  body.dr-full #pickPath .pp-num{font-size:11px}
  body.dr-full #pickPath .pp-thinwrap,
  body.dr-full #pickPath .pp-rest{flex:1 1 100%;white-space:normal}
  /* The label is the column name the header used to carry, in the cell that
     used to sit under it. */
  body.dr-full #pickPath [data-l]::before{
    content:attr(data-l) " ";
    font:400 8px/1 var(--font-num);letter-spacing:.14em;text-transform:uppercase;
    color:var(--text-dim)}
  body.dr-full #pickPath .pp-num{display:flex;align-items:baseline;gap:5px}
  body.dr-full #pickPath .pp-thinwrap{display:flex;align-items:center;gap:6px}
  body.dr-full #pickPath .pp-lfoot,
  body.dr-full #pickPath .pp-method{display:none}
  body.dr-full #pickPath .pp-hd{gap:var(--sp-2)}
  body.dr-full #pickPath .pp-hd .caps{font-size:9px}
  body.dr-full #pickPath .pp-verdict5{margin:12px 0 14px;padding:0 11px}
  body.dr-full #pickPath .pp-fig{font-size:26px}
  body.dr-full #pickPath .pp-vclaim{font-size:16px}
  body.dr-full #pickPath .pp-grid{padding:0 11px 11px}
  /* The keyboard legend is desktop advice on a device with no j/k. */
  body.dr-full #baLegend{display:none}

  /* THE NEED CHIPS BECOME CIRCLES, his frame's treatment and a better target
     than the 28px pill: 52px clears the thumb floor by eight. The row scrolls
     rather than wrapping - nine chips over two lines is a third of the sheet
     spent on the filter. */
  body.dr-full #baTabs{
    flex-wrap:nowrap;overflow-x:auto;-webkit-overflow-scrolling:touch;
    gap:6px;padding:5px var(--sp-3);scrollbar-width:none}
  body.dr-full #baTabs::-webkit-scrollbar{display:none}
  /* 48, down from 52, and the four pixels are a row of the list - see the
     furniture note above. Still four over the thumb floor. */
  body.dr-full .batab{
    flex:0 0 auto;flex-direction:column;justify-content:center;gap:2px;
    width:48px;height:48px;padding:0;border-radius:50%}
  body.dr-full .batab .bt-l{font-size:10px}
  body.dr-full .batab .bt-n{font-size:9px}
  /* "Drafted" is a switch, not a position: it keeps the pill. */
  body.dr-full .batab.tog{width:auto;height:48px;padding:0 12px;border-radius:24px}

  /* A DRAFT BUTTON IS A THUMB TARGET. The phone pass already put a 44px floor
     on tap targets; this restates it for the one control the sheet exists to
     deliver, because a sheet you cannot draft from is just a list. */
  body.drafting #availList .draftbtn{min-height:44px;min-width:56px}
  /* ...AND IT HAS TO LOOK LIKE ONE. `.draftbtn` is a bare accent-on-hover
     label, which is a fine affordance on a machine with a pointer and no
     affordance at all on a phone: his frame draws a filled DRAFT pill because
     that is the control the sheet exists to deliver. A 44px hit box nobody
     can see is still a control nobody presses. */
  /* ...AND IT IS AN OUTLINE, NOT A FILLED PILL (2026-08-15). It shipped as a
     filled accent pill, which made the loudest object on every row the button
     that is the same on every row. The number that differs - the chance he is
     still there at your next pick - was two thirds its weight in the column
     beside it. That is the emphasis exactly inverted: the survival percentage
     is the reason to press the button, and a row full of identical bright
     pills is a row you scan for nothing. A hairline outline at thumb size is
     still unmistakably a control (the reference's own DRAFT pill is quieter
     than this) and it stops competing with the reading. */
  body.dr-full #availList .draftbtn{
    display:flex;align-items:center;justify-content:center;
    border:1px solid var(--border-strong);border-radius:22px;
    color:var(--text-secondary);background:none;
    font-size:10px;letter-spacing:.08em;padding:0;margin:0;text-align:center}
  body.dr-full #availList .draftbtn:active{
    border-color:var(--accent);color:var(--accent);background:var(--accent-soft)}
  body.dr-full #availList .draftbtn.off{
    border-color:var(--border-subtle);background:none;color:var(--text-dim)}

  /* THE SURVIVAL NUMBER IS THE SECOND-LOUDEST THING IN THE ROW, after the
     name. Two blind critics named it as the one number this screen has that
     the reference does not: an ADP is a price, this is a probability about
     YOUR next pick. It was set one step under the row's smallest heading. */
  /* Right-aligned over the ADP line under it, so the two numbers in this
     column share an edge instead of hanging off opposite ones. */
  body.dr-full .ba .surv{justify-content:flex-end}
  /* 16px against the name's 15, and the ceiling is deliberate: the row's
     identity is still the name, and this is the loudest thing AFTER it. */
  body.dr-full .ba .surv .pct{font-size:16px;font-weight:800;
    line-height:1;color:var(--accent)}
  /* AND A SETTLED ONE IS QUIET. The loud ink belongs to the rows where a
     decision still exists: at the top of an untouched board every name is
     gone before your next turn, and four identical words at 16px/800 are the
     same dead emphasis the DRAFT pill was demoted for. `gone by 2.12` and
     `safe to 2.12` are statements, not scores, so they read at the row's
     smallest weight and the percentage - the only value that varies - keeps
     the accent.
     --text-muted rather than the --text-dim the desktop rule leaves: at 12px
     that measured 2.1:1 on this ground and the phrase still has to be read.
     10px AND ONE LINE, which is what the widened column below is for. This
     was 12px/700 carrying the bare word `gone`, three columns from an enabled
     DRAFT button, on players who were on the board - the round-5 defect. The
     horizon is what fixes that, and a horizon that wrapped or clipped here
     would not be one. */
  body.dr-full .ba .surv .pct.sure{font-size:10px;font-weight:600;
    line-height:1.15;color:var(--text-muted);white-space:nowrap}

  /* THE PRICE IS BACK ON THE ROW, under the survival number rather than beside
     it. His frame carries ADP and PTS; the phone pass had dropped both to fit
     five columns. This restores the one you cannot reconstruct from anywhere
     else on the screen and leaves PTS out: the row already states two numbers,
     and survival is the one this app computes and Sleeper does not. */
  body.dr-full .ba .surv{grid-row:1;align-self:end}
  body.dr-full .ba .mkt{display:block;grid-column:4;grid-row:2;
    font-size:10px;text-align:right;color:var(--text-muted);align-self:start}
  body.dr-full .ba .mkt::before{content:"ADP ";letter-spacing:.08em}

  /* ...AND SO IS THE MISPRICE, which is the number this whole app exists to
   * compute and the one the phone was drafting without. Sleeper's named gap is
   * that it carries no value signal at all; Edge computes one, draws it on the
   * board where round 6 says it competes with the turn callout, and hid it in
   * the row where the decision is actually made (`.ba .room` is display:none
   * at this width, four hundred lines up).
   *
   * IT LANDS ON THE DEMOTED SECOND LINE and never in the accent: the survival
   * percentage at 16px/800 is the row's second-loudest object by decision, and
   * the DRAFT pill was demoted to a hairline outline for competing with it.
   * 10px in the poles the column already owns, on the strip that already holds
   * the club, the bye and the ADP - six facts a row, ESPN's shape, because the
   * sometimes-wanted numbers belong under the identity rather than beside it.
   *
   * SHARES THE CLUB'S CELL, at the far end of it. `.tb` is `DET · bye 6` - a
   * three-letter club and a two-digit bye, 12 characters at its longest, ~60px
   * of a 133px track at 375 - and the widest thing this can print is
   * `+21 value` at ~48px, so the two cannot meet. Measured at 375 and 390.
   *
   * AND AN UNPRICED PLAYER PRINTS NOTHING. The desktop cell holds an em dash to
   * keep its column; on a strip with no column an em dash is ink that says
   * nothing, and `.up`/`.dn` is exactly the class pair that means "there is a
   * reading here". */
  body.dr-full .ba .room{display:block;grid-column:3;grid-row:2;
    justify-self:end;align-self:start;
    font:600 10px/1 var(--font-num);text-align:right}
  body.dr-full .ba .room:not(.up):not(.dn){display:none}
  body.dr-full .ba .room-w{display:inline;margin-left:4px;
    font:400 10px/1 var(--font-ui);letter-spacing:.02em}
}

/* The mode's own furniture does not exist off the phone - the mode does not
   either. Stated once, globally, so no rule inside a media query has to
   remember to turn them off. */
.dr-bar,.dr-tabs,#btnFullDraft{display:none}

/* ============================= PHONE POLISH ==============================
   2026-08-10, from a hypercritical pass over every tab at 375px against his
   mock-draft recording. Each rule names the defect it removes. Desktop is
   untouched: everything here is scoped to a width or to the draft mode. */
@media (max-width:640px){
  /* A keyboard legend on a touch screen documents controls the reader does
     not have. */
  .bkey-keys{display:none}
  /* The table view scrolled sideways was a wall of numbers no row could be
     attributed to - the name is the row's identity, so it stays put while
     the sources scroll. Solid background, or the columns it covers read
     through it. */
  .board-scroll th.tl{position:sticky;left:0;z-index:3;background:var(--bg-surface)}
  .board-scroll td.player{position:sticky;left:0;z-index:2;background:var(--bg-surface)}
  /* The card's one exit was 29px of muted slate on a navy hero - invisible in
     every frame of his recording that shows the card. A keycap at thumb size,
     on its own surface. */
  /* THE PHONE'S OWN COPY OF THIS RULE IS WHAT HE WAS LOOKING AT. It pinned a
     44px circle to the top RIGHT - directly over the portrait, which is the
     one part of the hero with the most going on behind it - and it comes later
     in the file than the desktop rule, so the left-hand pill above never
     applied on the device at all. Same side, same shape, same words as the
     desktop control now; only the thumb target and the notch inset are the
     phone's business. */
  .pc-close{top:calc(var(--sp-2) + env(safe-area-inset-top,0px));
    left:var(--sp-2);right:auto;
    min-width:44px;min-height:44px;padding:0 15px 0 11px}
}

/* iOS Safari repaints sticky positions lazily inside a momentum scroll - the
   round gutter and header row tear off the edge of the board mid-fling. Their
   own compositor layer is the documented fix, and fifteen small layers are
   cheap. Stated for every width: the same WebKit runs on an iPad. */
.dg-rd,.dg-team,.dg-corner{transform:translateZ(0)}

/* ══════════════════ THE COMPARE KEY ══════════════════ 2026-08-11 (6th)
   His call: "add a view button in here where we can compare vs rooms. One of
   the biggest points of the app... maybe even make a lens button that is big
   and attempting to draw attention."

   The diagnosis is right and worth writing down, because it is a class of
   mistake rather than one control. Reading the board through another room is
   the only thing this app does that a rankings site cannot - the composite can
   say everyone disagrees with us, but only the lens can say where Yahoo
   disagrees with ESPN. That capability was carried by two of the smallest
   controls on the page, `vs` and `through`, in 9px mono, wedged between a sort
   dropdown and two export buttons. THE SIZE OF A CONTROL IS A CLAIM ABOUT HOW
   IMPORTANT IT IS, and this one was claiming "footnote".

   So the pair becomes one key that speaks the whole sentence. It is a `.key`
   and not a bespoke button: it inherits the face, the lip, the hard edge and
   the press from the system, so it presses like everything else and the
   suite's key check covers it without being told about it. Only its SHAPE is
   local - two lines and full width, because a sentence does not fit in 32px. */
.cmpkey{flex:1 1 100%;min-width:0;height:auto;min-height:52px;
  display:flex;flex-direction:column;align-items:flex-start;justify-content:center;
  gap:3px;padding:9px 34px 9px 14px;position:relative;text-align:left}
.cmpkey-cap{font-family:var(--font-num);font-size:9px;letter-spacing:.22em;
  text-transform:uppercase;color:var(--text-muted);font-weight:500}
/* The sentence itself, and it may never wrap to three lines - the key's job is
   to be readable at a glance from across a table. */
.cmpkey-line{display:block;max-width:100%;font:400 15px/1.3 var(--font-ui);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.cmpkey-line b{font-weight:600}
/* The half of the sentence that is OFF reads quiet, so "vs EDGE through ESPN"
   and "vs EDGE, no lens" are visibly different states rather than two strings
   of equal weight. */
.cmpkey-line i{font-style:normal;color:var(--text-muted)}
.cmpkey-caret{position:absolute;right:14px;top:50%;width:0;height:0;
  margin-top:-2px;border:5px solid transparent;border-top-color:currentColor;
  opacity:.6}
.cmpkey[aria-expanded="true"] .cmpkey-caret{transform:rotate(180deg);
  margin-top:-7px}

/* THE SHEET. Two groups, each a radiogroup, each built from its select. */
.cmpsheet{display:flex;flex-direction:column;gap:var(--sp-4);
  padding:var(--sp-4);margin-bottom:var(--sp-3);
  background:var(--bg-panel);border:1px solid var(--border-subtle);
  border-radius:2px}
.cmpgrp{display:flex;flex-direction:column;gap:9px}
.cmprows{display:flex;flex-wrap:wrap;gap:8px}
/* THE SELECTS ARE THE STATE and the rows are their mirror, so the selects are
   taken off screen rather than out of the document. Not display:none - a
   hidden select can still be the thing a script reads and writes, which is
   exactly what it is here. */
.cmphid{position:absolute;width:1px;height:1px;overflow:hidden;
  clip-path:inset(50%);white-space:nowrap}

/* ── HIS THREE BOARD NOTES, 2026-08-11 (6th) ──────────────────────────────
   Reported off the simulator, on the lit board in TABLE view.

   1. "PAINT THE HERO THE SAME AS THE GRID ON THE TABLE VIEW." The hero
      banners are graphite objects, and that is right in CARD view, where they
      sit at the head of 397 graphite cards and read as the first two of them.
      In TABLE view there are no cards - the grid is white rows on paper - so
      the same two slabs are the only dark things on the page and read as
      orphaned rather than deliberate. The object did not change; the company
      it keeps did. So on a lit board in table layout the hero takes the
      table's own surface, and the graphite islands stand down.
   2. "REMOVE VERTICAL PADDING so that empty space isn't taking away from the
      card/table view." On a 402pt screen the head of this page spends most of
      a viewport on chrome before a single row of data.
   3. "PUT THE BLUE BUTTON HIGHLIGHTS ON THE BORDER OF THE BOXES so they
      stand out." Done in the SKIN'S TOKENS, not in a new rule: the lit key
      keeps a white face and takes a blue edge and blue ink. That keeps the
      key system one system - every control still reads --key-on-face,
      --key-on-edge and --key-on-ink, and only the values move - which is the
      same argument the card islands make one screen up. A filled blue slab
      also fights the Honolulu-blue chip row directly above it; an outline
      does not, and it lets the box itself be the thing that stands out,
      which is what he asked for. */
/* STATED AS COLOURS, NOT AS ISLAND TOKENS. Redeclaring --card-ink here is
   the fork the island exists to prevent - it would put a second set of values
   for a card's material inside the skin, which is exactly the shape that lost
   this app 37 rules earlier today. The hero in list view is not a card wearing
   different tokens; it is a different object, so it says so directly. */
.lions.lay-list .hz-ban{
  background:var(--bg-surface);
  color:var(--text-primary);
  box-shadow:0 1px 3px rgba(15,26,34,.10)}
.lions.lay-list .hz-ban .stat-val.name{color:var(--text-primary)}
.lions.lay-list .hz-ban .caps b,
.lions.lay-list .hz-ban .stat-note{color:var(--text-secondary)}

/* 2. THE HEAD OF THE BOARD GIVES BACK ITS SPACE, phone only - at desktop
      width the page is not fighting for a viewport. */
/* `#panel-board.lions`, NOT `.lions #panel-board`. The first version of this
   block used the descendant form and matched NOTHING - the panel IS the
   element carrying `lions`, so asking for a `#panel-board` INSIDE a `.lions`
   asks for a second one that does not exist. Every rule was well-formed, the
   stylesheet parsed, nothing errored, and the padding simply never moved. He
   caught it by looking: "you didn't execute the suggestions i made regarding
   the padding." A selector that matches nothing is the quietest failure in
   this file - which is why the phone-density claim is now measured in the
   suite rather than trusted. */
@media (max-width:640px){
  #panel-board.lions .masthead{margin-bottom:var(--sp-3)}
  #panel-board.lions .mh-dek{display:none}
  #panel-board.lions .stat-strip{gap:var(--sp-2)}
  #panel-board.lions .hz-chips{gap:6px}
  #panel-board.lions .hz-ban .hz-in{padding:10px 14px 11px 10px}
  #panel-board.lions .fmtnote{padding:9px 12px;margin:var(--sp-2) 0}
  #panel-board.lions .ctlrow{padding:var(--sp-2) var(--sp-3)}
  /* A VERTICAL DIVIDER IS A ROW'S IDEA, and this row is not a row on a phone -
     it wraps to four lines. Each `.fdiv` then survives as a stray 1px tick at
     the head of whatever wrapped after it, indenting that group by its own
     width plus two margins: "draft order/tier view sticks out like a sore
     thumb", and it was the only group with a divider in front of it. The rule
     it separates two things ACROSS is gone once they are on separate lines, so
     the divider goes with it. */
  #panel-board.lions .ctlrow .fdiv{display:none}
  /* One gap and one left edge for every group, so the wrapped rows read as a
     stack rather than as a row that happened to break. */
  #panel-board.lions .ctlrow{gap:var(--sp-2) var(--sp-2);align-items:stretch}
  #panel-board.lions .ctlrow>.segbar,
  #panel-board.lions .ctlrow>.ctlend{margin-left:0}
  #panel-board.lions .cmpsheet{padding:var(--sp-3);gap:var(--sp-3)}
  #panel-board.lions .mh-title{font-size:clamp(26px,7vw,34px)}
  #panel-board.lions .wrap{padding-top:0}

  /* A SECOND PASS, measured rather than eyeballed. The head above the first
     row came to 1008px on a 874pt screen - more than a full viewport of
     chrome before a single player. Where it actually went:
       stat strip 374  (two 140px hero banners + a 78px chip row)
       control rows 130, moved strip 74, fmtNote 72, masthead 88
     So the strip is the cost, and the hero banners are the strip. They are
     sized for a desktop column and stacked here, which is what makes them
     tall; nothing about what they SAY needs 140px. */
  #panel-board.lions .masthead{margin-top:var(--sp-2)}
  #panel-board.lions .stat-strip{margin:var(--sp-2) 0}
  /* 44px ON THE WHOLE ROW, not just the one that is a button. Making the room
     chip a control brought it under the thumb-target rule and the suite caught
     it at 32px - correctly, since it is now pressable. Sizing only that one
     would leave a 44px chip beside three 32px ones in a four-up grid, so the
     row goes up together: they are one object, they align, and the chips this
     app was told were "too small, hurt my eyes" get the benefit. */
  #panel-board.lions .hz-chip{padding:9px 10px;font-size:12px;min-height:44px}
  #panel-board.lions .hz-chips{gap:5px}
  #panel-board.lions .hz-ban .hz-in{padding:8px 12px 9px 8px;gap:5px}
  #panel-board.lions .hz-ban .stat-val.name{font-size:clamp(17px,5.2vw,22px)}
  #panel-board.lions .hz-ban .stat-note{font-size:var(--fs-sm)}
  #panel-board.lions .hz-flag{width:38px;font-size:12px}
  #panel-board.lions .fmtnote{margin:var(--sp-2) 0 0}
  #panel-board.lions .moved{margin-bottom:var(--sp-2);padding:6px 10px}
}

/* ══════════════════ THE TOPBAR FITS ON ONE PHONE ROW ══════════════════
   2026-08-11 (6th). These four declarations were written first inside the
   phone block up at the top of the mobile section and did absolutely nothing,
   because `.fmt select` and `.draftclock` are DEFINED LATER in this file at
   the same specificity - so the base rule won every time and the countdown
   went on losing its last four characters ("23 days to draft cuts off").

   A media query is not a specificity boost. Restated here, after everything
   they have to beat, which is the only thing that makes them rules rather
   than intentions. Measured: the meta group ran 363 in a 352 box; 16
   characters at .44px of tracking is 7px, and a point of size is another 13,
   so either alone closes an 11px gap and both leave room for the longer
   strings this element renders ("DRAFT TOMORROW", "DRAFTED 12D AGO"). */
@media (max-width:640px){
  .fmt{gap:var(--sp-1)}
  .fmt select{padding-left:6px;padding-right:16px;letter-spacing:.06em}
  /* `.topbar-meta > span` (0,1,1) is what actually sizes this element, so a
     bare `.draftclock` (0,1,0) loses on SPECIFICITY and moving it later in the
     file - which is what the paragraph above did - changed nothing. Measured
     in the browser rather than reasoned about a third time: the winning rule
     was read off the cascade and this one matches its shape. */
  .topbar-meta > .draftclock{font-size:10px;letter-spacing:.02em;white-space:nowrap}
}

/* ══════════════════ THE OVERNIGHT STRIP SCROLLS ITSELF ══════════════════
   2026-08-11 (6th), his call: "make the moved overnight auto scroll." Ten
   chips sat in a row that showed three on a phone, with nothing to say the
   other seven existed - the strip looked like a short list rather than a
   clipped one.

   The renderer writes the chips TWICE and this translates the pair by exactly
   half its width, so at the moment the animation resets the second copy is
   standing where the first began and the seam never appears. Duration scales
   with the content rather than being a flat number, or ten chips and three
   would travel at different speeds.

   IT STOPS WHEN YOU REACH FOR IT. Hover and focus-within pause it, because a
   moving target you are trying to tap is worse than a static one you have to
   scroll; and reduced-motion turns it off entirely and hands the row back to
   the finger, which is what it was before. */
/* THE EDGES FADE. Hard `overflow:hidden` cuts whatever chip is crossing the
   boundary clean in half, and at the left that half-chip sits immediately
   against the "MOVED OVERNIGHT" label, which reads as a rendering fault
   rather than as motion. A mask makes the same clip a fade, so a chip
   arriving and a chip leaving both look intentional. Two properties because
   Safari still wants the prefix, and this runs on his phone. */
.mv-vp{flex:1;min-width:0;overflow:hidden;display:block;
  -webkit-mask-image:linear-gradient(90deg,transparent 0,#000 14px,
    #000 calc(100% - 22px),transparent 100%);
  mask-image:linear-gradient(90deg,transparent 0,#000 14px,
    #000 calc(100% - 22px),transparent 100%)}
.mv-track{display:inline-flex;gap:var(--sp-2);align-items:center;
  will-change:transform;animation:mv-marquee 42s linear infinite}
.mv-dup{display:inline-flex;gap:var(--sp-2);align-items:center;margin-left:var(--sp-2)}
.moved:hover .mv-track,.moved:focus-within .mv-track{animation-play-state:paused}
@keyframes mv-marquee{from{transform:translateX(0)}to{transform:translateX(-50%)}}
@media (prefers-reduced-motion:reduce){
  .mv-track{animation:none}
  .mv-vp{overflow-x:auto;scrollbar-width:none}
  .mv-dup{display:none}
}

/* ============================================================ STRATEGY
 * VISION-STRATEGY. The matrix is the poster object: EDGE card language at
 * table scale - Archivo Black numerals, skewed chips, the winner lit on the
 * accent, quiet grid rules. Everything tokens, both grounds. */
.st-sec{margin-top:calc(var(--sp-5) * 1.5)}
.st-sec:first-child{margin-top:0}
.st-h{font:700 var(--fs-hero)/1.1 var(--font-display);letter-spacing:.01em}
.st-h3{font:700 var(--fs-lg)/1.2 var(--font-display);margin-bottom:var(--sp-2)}
.st-dek{color:var(--text-muted);max-width:70ch;margin:var(--sp-2) 0 var(--sp-3)}
.st-absent{color:var(--text-muted);font-style:italic}
.st-n{color:var(--text-muted);margin-top:var(--sp-2)}
.st-scroll{overflow-x:auto}
.st-matrix{border-collapse:collapse;min-width:420px;width:100%;table-layout:fixed}
.st-matrix th,.st-matrix td{border-bottom:1px solid var(--border-subtle);
  padding:10px 14px;text-align:right;vertical-align:baseline}
.st-matrix thead th{font:500 9px/1 var(--font-num);letter-spacing:.1em;
  text-align:center;padding:0 2px 7px;
  letter-spacing:.08em;text-transform:uppercase;color:var(--text-muted)}
/* The mock's seat gutter is 34px and it is a two-digit mono number, not a
   poster figure - it labels the row, it does not head it. */
.st-seat{width:30px;font:500 var(--fs-sm)/1 var(--font-num);
  font-variant-numeric:tabular-nums;color:var(--text-muted);text-align:left;
  padding:1px 0}
.st-cell{position:relative;padding:1px 2px;border:0}
/* .st-par / .st-beats / .st-refcol / .st-win::after are GONE with the cell
   they styled. The old matrix drew a two-line cell - a skewed Archivo figure
   over an italic "58% of room" - and the ported one is a single mono figure on
   an ink chip. `.st-win` survives with a new meaning, so the old ::after rule
   had to go with the rest or it would lay an accent underline across every
   inked cell; a class that changes what it decorates is exactly where a dead
   rule stops being dead. */
/* ------------------------------------------------- THE EVIDENCE, PORTED
 * His mock's compact matrix: short column codes, the seat's best build INKED
 * and every other cell the same ink at 3-15% alpha ramped on the share of the
 * room it beat. The table reads as a SHAPE first - the inked run across the
 * front and the inked run after the handoff ARE the one law - and as numbers
 * second. Values taken from the mock's own script, not eyeballed. */
.st-evidence{display:grid;grid-template-columns:1.35fr 1fr;gap:44px;
  align-items:start}
/* MIN-WIDTH:0, and it is load-bearing. A grid item defaults to
   `min-width:auto`, which refuses to shrink below its content's min-content
   width - so the matrix's own min-width propagated up and pushed the PAGE
   sideways instead of letting `.st-scroll` scroll the table. On a phone that
   rendered as the whole strategy page bled off the right edge with the
   headings clipped on the left. The overflow belongs to the table, not the
   document. */
.st-evidence>.st-sec{margin-top:0;min-width:0}
.st-evhead{display:flex;justify-content:space-between;align-items:baseline;
  gap:var(--sp-3);flex-wrap:wrap;margin-bottom:2px}
.st-evkey{color:var(--text-muted);letter-spacing:.12em}
.st-chipcell{font:500 var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:center;padding:7px 2px;border-radius:2px;color:#3c4a52}
.st-chipcell.st-win{background:var(--text-primary);color:var(--bg-surface)}
.st-none{color:var(--text-muted)}
/* YOUR ROW, in a table of twelve. Two channels and neither is colour alone: a
   soft fill and a rule down the seat edge, so it survives the graphite ground,
   the paper one and a printout. It marks WHERE to read, and deliberately does
   nothing to the numbers - the evidence is that every seat was measured the
   same way. */
.st-matrix tr.st-mine{background:var(--bg-selected)}
.st-matrix tr.st-mine .st-seat{box-shadow:inset 3px 0 0 var(--accent);
  color:var(--accent)}
.st-slotrow.st-mine .st-slotn{color:var(--accent)}
/* A RULE, NOT A CRESCENT. An inset shadow follows the box's radius, so at the
   Lions skin's 8px it curved away from the edge at both ends and read as a
   blue arc stuck to the row rather than as a mark down its side. Squared, and
   the fill keeps the corner. */
.st-slotrow.st-mine{background:var(--bg-selected);border-radius:2px;
  box-shadow:inset 3px 0 0 var(--accent);padding-left:6px}
.st-legend{margin-top:var(--sp-3);display:grid;gap:6px}
.st-leg{display:flex;gap:var(--sp-2);align-items:baseline;font-size:var(--fs-sm);
  color:var(--text-muted)}
.st-chip{display:inline-block;padding:3px 9px;transform:skewX(-10deg);
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle);
  font:600 var(--fs-xs)/1 var(--font-label);font-style:italic;
  letter-spacing:.06em;color:var(--text-primary);white-space:nowrap}
.st-concl{margin:var(--sp-3) 0 0 1.2em;color:var(--text-muted);
  display:grid;gap:6px;max-width:75ch}
.st-shapes{display:flex;flex-direction:column;gap:var(--sp-5)}
.st-shape{max-width:none}
.st-slotrow{display:flex;align-items:center;gap:var(--sp-2);padding:3px 0}
.st-slotn{width:2ch;text-align:right;font:400 var(--fs-sm)/1 "Archivo Black",var(--font-ui);
  color:var(--text-muted)}
.st-bar{flex:1;height:11px;background:var(--track);border-radius:2px;overflow:hidden}
.st-bar i{display:block;height:100%;background:var(--text-primary);border-radius:2px}
/* THE DREAD SEAT IS AMBER, not red. Red is this app's injury and loss ink; the
   trough seat is not a warning, it is the seat the plan is worth most from -
   which is the sentence the rescue rail above it makes. The mock uses the risk
   amber and it agrees with that reading. */
.st-worst .st-bar i{background:var(--risk)}
.st-worst .st-slotn{color:var(--text-primary);font-weight:600}
.st-barpar{width:7ch;text-align:right;font:500 var(--fs-lg)/1 var(--font-num);
  font-variant-numeric:tabular-nums}
.st-note{color:var(--text-muted);margin-top:var(--sp-2);font-size:var(--fs-sm)}
.st-labrow{display:flex;gap:var(--sp-3);align-items:center;flex-wrap:wrap;
  margin:var(--sp-3) 0}
.st-tog{display:inline-flex;gap:6px;align-items:center;font-size:var(--fs-sm);
  color:var(--text-muted);cursor:pointer}
.st-labout{margin-top:var(--sp-3)}
.st-running{color:var(--text-muted);font-style:italic}
.st-labres{background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3);display:grid;gap:8px}
.st-lr{display:flex;gap:var(--sp-2);align-items:baseline;font-size:var(--fs-md)}
.st-lr-ref{color:var(--text-muted)}
.st-delta{font:600 var(--fs-sm)/1 var(--font-num)}
.st-delta.up{color:var(--positive)}
.st-delta.down{color:var(--negative)}
.st-lr-meta,.st-sample{color:var(--text-muted)}
/* A DEAD LANE NAMES ITSELF. Sized like the number it replaces so the row
   keeps its shape, coloured as a failure so it is never mistaken for one. */
.st-lr-err{color:var(--negative);font-size:var(--fs-sm)}
.st-homelist,.st-cuffs,.st-discs{display:grid;gap:8px;margin-top:var(--sp-3)}
.st-homerow,.st-cuffrow{display:flex;gap:var(--sp-2);align-items:baseline;
  border-bottom:1px solid var(--border-subtle);padding-bottom:6px;
  flex-wrap:wrap}
.st-hname{font-weight:600}
.st-hpos,.st-cteam,.st-cadp{color:var(--text-muted)}
.st-hadp{margin-left:auto;font:600 var(--fs-sm)/1 var(--font-num)}
.st-cstart{display:flex;gap:8px;align-items:baseline;min-width:0}
.st-carrow{color:var(--text-muted)}
.st-frag{display:inline-block;padding:2px 8px;transform:skewX(-10deg);
  font:600 var(--fs-xs)/1 var(--font-label);font-style:italic;
  letter-spacing:.06em;text-transform:uppercase}
.st-frag-fragile{background:rgba(196,59,59,.16);color:var(--negative)}
.st-frag-monitor{background:var(--accent-soft);color:var(--accent)}
.st-disc{background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3);display:grid;gap:6px}
.st-dhead{display:flex;gap:var(--sp-2);align-items:baseline;justify-content:space-between}
.st-dname{font:700 var(--fs-lg)/1.1 var(--font-display)}
.st-dstat{padding:2px 8px;transform:skewX(-10deg);
  background:var(--bg-surface-raised);border:1px solid var(--border-subtle)}
.st-review-open{color:var(--negative)}
.st-served-monitoring{color:var(--accent)}
.st-dhl{font-weight:600}
.st-dfacts{color:var(--text-muted);font-size:var(--fs-sm);max-width:80ch}
.st-dprice{color:var(--text-muted)}
.st-watch{background:var(--bg-surface);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);padding:var(--sp-3);margin-top:var(--sp-3)}
.st-whead{display:flex;gap:var(--sp-2);align-items:baseline}
.st-wtitle{font:700 var(--fs-xl)/1.1 var(--font-display)}
.st-wsub{color:var(--text-muted)}
.st-wthesis{color:var(--text-muted);margin:var(--sp-2) 0;max-width:75ch}
.st-wlive{color:var(--text-primary)}
@media (max-width:640px){
  .st-matrix th,.st-matrix td{padding:8px 9px}
  .st-hadp{margin-left:0;width:100%}
}
/* Strategy board marks: same coin as the injury badge, one skew, tokens. */
.st-badge{display:inline-block;margin-left:6px;padding:2px 6px;
  transform:skewX(-10deg);font:600 var(--fs-xs)/1 var(--font-label);
  font-style:italic;letter-spacing:.06em}
.st-home{background:var(--negative-soft);color:var(--negative-ink)}
.st-hc{background:var(--accent-soft);color:var(--accent)}
.st-fmtnote{color:var(--negative-ink);font-size:var(--fs-sm);
  margin-bottom:var(--sp-2);font-style:italic}
/* The plan in the room: chips read/write the same standing choice the
   Strategy page holds; the note annotates the math, never overrides it. */
.pp-plan{display:flex;gap:6px;flex-wrap:wrap;margin:var(--sp-2) 0}
.pp-planchip{font-size:var(--fs-xs)}
.pp-plannote{color:var(--text-muted);margin:var(--sp-2) 0;line-height:1.5}
.st-delta.flat{color:var(--text-muted)}
.st-wrh{margin-top:var(--sp-4)}

/* ------------------------------------------------------------- THE GATE
 * The strategy page asks which chair you drew before it answers anything, and
 * this is that question at full page scale. Ported from his 2026-08-12 mock;
 * the palette needed no translation at all - #eef1f3, #131c22, #55636b,
 * #6b757c, #d3d9dc, #e7f2f9, #1683c4 and #9a6d10 are the Lions tokens to the
 * literal, which is the fourth mock in a row to agree. Only geometry and the
 * type scale are taken from the file. Two places the design did NOT govern:
 * its #8a949a is --text-muted here, because that step was measured and lifted
 * on 2026-08-11 for failing contrast on this exact ground, and its em dashes
 * are periods, per the standing rule that outranks the mocks' typography. */
/* The gate is a sibling of the page wrap rather than a child, because it IS
   the page while it is up - so it carries the wrap's own measure and gutter
   itself. Without this it renders hard against the rail and the title loses
   its first two letters, which is what a section outside the wrap looks like. */
.st-gate{max-width:1440px;margin:0 auto;padding:var(--sp-5) var(--sp-4)}
.stg{display:flex;flex-direction:column;gap:34px;min-height:640px;
  /* The mock's 96px column rule, at the opacity it draws it. It is the one
     decorative mark on the page and it earns its place by saying "this is a
     board", which is what twelve seats laid side by side are. */
  background-image:repeating-linear-gradient(90deg,
    rgba(19,28,34,.035) 0 1px,transparent 1px 96px)}
.stg-head{display:flex;justify-content:space-between;align-items:flex-start;
  gap:40px;padding-bottom:var(--sp-5);border-bottom:1px solid var(--border-strong);
  flex-wrap:wrap}
.stg-kicker{color:var(--accent);letter-spacing:.26em}
.stg-title{margin:10px 0 0;font:700 clamp(34px,4.4vw,60px)/.94 var(--font-poster);
  letter-spacing:.01em;text-transform:uppercase;color:var(--text-primary)}
.stg-dek{margin:14px 0 0;max-width:58ch;font:300 15px/1.85 var(--font-ui);
  color:var(--text-secondary);text-wrap:pretty}
.stg-frame{display:flex;flex-direction:column;gap:10px;align-items:flex-end}
.stg-chiprow{display:flex;gap:4px;flex-wrap:wrap;justify-content:flex-end}
.stg-chip{background:var(--bg-surface);color:var(--text-secondary);border:0;
  font:400 var(--fs-sm)/1 var(--font-num);padding:9px 12px;border-radius:3px;
  cursor:pointer}
.stg-chip:hover{color:var(--text-primary)}
.stg-chip.on{background:var(--text-primary);color:var(--bg-surface);font-weight:500}
.stg-chip.on.accent{background:var(--accent-soft);color:var(--accent-hover)}
.stg-meta{color:var(--text-muted);letter-spacing:.14em;text-align:right}

.stg-cap{display:flex;justify-content:space-between;gap:var(--sp-4);
  align-items:baseline;margin-bottom:10px;color:var(--text-muted);
  letter-spacing:.18em}
.stg-cap span:last-child{text-align:right;letter-spacing:.1em;
  text-transform:none;max-width:52ch}
/* TWELVE ACROSS WHERE THERE IS ROOM, and however many fit where there is not.
   auto-fit rather than repeat(12,1fr): a phone would otherwise draw twelve
   28px columns, and a seat card that cannot hold its own number is not a
   smaller version of this control, it is a different and useless one. */
.stg-grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(84px,1fr));
  gap:8px}
.stg-card{display:grid;gap:7px;text-align:left;background:var(--bg-surface);
  border:1px solid transparent;border-radius:8px;padding:11px 10px 10px;
  cursor:pointer;transition:transform .12s ease,box-shadow .12s ease}
.stg-card:hover{transform:translateY(-3px);box-shadow:0 4px 14px rgba(15,26,34,.10)}
.stg-card.on{background:var(--bg-selected);border-color:var(--accent);
  box-shadow:0 0 0 4px var(--accent-soft)}
.stg-n{font:700 26px/1 var(--font-poster);color:var(--text-primary)}
/* ONE LINE, ALWAYS. "+12.5 PAR" wrapped to two lines in some cards and not
   others, so a row of twelve seats had cards of two different heights and the
   bars under them no longer formed a readable column - which is the whole job
   of that row. */
.stg-par{font:500 var(--fs-md)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-primary);white-space:nowrap}
.stg-par em{font-style:normal;color:var(--text-muted)}
.stg-tagor{font-style:normal}
.stg-track{display:block;height:5px;background:var(--track);border-radius:2px;
  overflow:hidden}
.stg-track i{display:block;height:100%;background:var(--accent);border-radius:2px}
.stg-tag{font:400 var(--fs-xs)/1.2 var(--font-num);letter-spacing:.06em;
  color:var(--text-muted);overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.stg-card.on .stg-tag{color:var(--accent-hover)}
/* A CALL THE MEASUREMENT CANNOT MAKE. 33 of the 88 banked seats have a winner
   inside what n=1000 resolves, so a clean two-build split across the twelve
   cards would be a picture of a decision that was not measured. The dotted
   rule is the mark this app already uses for "present but not settled", and
   the tag names both builds rather than dropping the second one. */
.stg-tag.tied{border-bottom:1px dotted var(--text-dim);padding-bottom:1px}
.stg-ends{display:flex;justify-content:space-between;gap:var(--sp-4);
  margin-top:10px;color:var(--text-muted);letter-spacing:.14em}
.stg-bare .stg-card{min-height:64px;align-content:center}

/* The bank has not landed yet. An indeterminate rail, because there is no
   progress to report - the honest shape of "waiting" is not a fake percentage. */
.stg-wait{display:flex;flex-direction:column;gap:12px}
.stg-waitcap{color:var(--text-primary);letter-spacing:.18em}
.stg-waitnote{color:var(--text-muted);letter-spacing:.12em;text-transform:none}
.stg-wait .stg-track{height:6px}
.stg-indet{display:block;height:100%;width:34%;background:var(--accent);
  border-radius:2px;animation:stg-slide 1.5s ease-in-out infinite}
@keyframes stg-slide{0%{transform:translateX(-110%)}100%{transform:translateX(320%)}}
@media (prefers-reduced-motion:reduce){
  .stg-indet{animation:none;width:100%;opacity:.5}
  .stg-card{transition:none}
  .stg-card:hover{transform:none}
}

.stg-foot{margin-top:auto;padding-top:var(--sp-5);
  border-top:1px solid var(--border-strong);display:flex;
  justify-content:space-between;align-items:flex-end;gap:40px;flex-wrap:wrap}
.stg-verdict+.stg-foot{border-top:0;padding-top:var(--sp-4);margin-top:0}
.stg-prompt{margin:0;max-width:62ch;font:300 15px/1.85 var(--font-ui);
  color:var(--text-secondary);text-wrap:pretty}
.stg-cue{color:var(--text-muted);letter-spacing:.18em}
.stg-enter{background:var(--accent);color:var(--on-accent);border:0;
  font:500 var(--fs-md)/1 var(--font-num);letter-spacing:.16em;
  text-transform:uppercase;padding:16px 26px;border-radius:4px;
  white-space:nowrap;cursor:pointer}
.stg-enter:hover{background:var(--accent-hover)}

/* ---------------------------------------------------------- THE VERDICT
 * One card, two homes: the gate draws it to earn the reader's agreement and
 * the room draws it because they gave it. See stVerdictCard(). */
.stg-verdict{margin-top:auto;background:var(--bg-selected);
  border:1px solid var(--accent);border-radius:var(--r-lg);
  box-shadow:0 0 0 5px var(--accent-soft);padding:25px 29px 23px}
.st-sec .stg-verdict{margin-top:0}
.stg-vhead{display:flex;justify-content:space-between;align-items:baseline;
  gap:30px;flex-wrap:wrap}
.stg-vwho{color:var(--accent-hover);font-weight:500;letter-spacing:.2em}
.stg-vn{color:var(--text-muted);font-weight:500;letter-spacing:.16em;
  text-align:right}
.stg-vbody{display:grid;grid-template-columns:1.1fr 1fr;gap:48px;
  align-items:end;margin-top:18px}
/* THE STAKE, at poster scale, because it is the number that varies by seat.
   The build name is one of two in every banked room and sat here at 60px
   claiming to be the answer; it is now the line under the answer. */
.stg-vstake{font:700 clamp(40px,5vw,64px)/.92 var(--font-poster);
  letter-spacing:.01em;color:var(--text-primary)}
.stg-vstake em{font-style:normal;font-size:.42em;margin-left:.22em;
  color:var(--text-muted);letter-spacing:.04em}
.stg-vstakecap{display:block;margin-top:6px;color:var(--text-muted);
  letter-spacing:.14em;max-width:44ch}
.stg-vlabel{margin-top:14px;font:700 clamp(22px,2.6vw,34px)/1 var(--font-poster);
  letter-spacing:.02em;text-transform:uppercase;color:var(--text-primary);
  overflow-wrap:anywhere}
.stg-vtied .stg-vor{font:400 .5em/1 var(--font-ui);text-transform:lowercase;
  color:var(--text-muted);letter-spacing:0;margin:0 .18em}
.stg-vdef{margin:12px 0 0;max-width:52ch;font:300 15px/1.85 var(--font-ui);
  color:var(--text-secondary);text-wrap:pretty}
.stg-vdef b{color:var(--text-primary);font-weight:500}
/* The plan's full shape (round 11): what the predicate does after its
   constraint is spent, one step quieter than the banked definition above it
   because it is mechanics, not measurement. */
.stg-vshape{margin:7px 0 0;max-width:52ch;font:300 13.5px/1.7 var(--font-ui);
  color:var(--text-muted);text-wrap:pretty}
.stg-toll{display:flex;flex-direction:column;gap:9px}
.stg-tollhead{display:flex;justify-content:space-between;gap:var(--sp-3);
  color:var(--text-muted);font-weight:500;letter-spacing:.16em}
/* A REAL FLOOR ON THE FLEXIBLE TRACK. A grid whose fixed tracks already fill
   the row leaves `1fr` at exactly zero, and the bar then renders as nothing
   while every number beside it still reads correctly - the failure this repo
   shipped once as a column of candidates called "J.". */
.stg-tollrow{display:grid;grid-template-columns:minmax(58px,88px) minmax(40px,1fr) 60px;
  gap:12px;align-items:center}
.stg-tolln{font:400 var(--fs-sm)/1.2 var(--font-num);color:var(--text-secondary);
  overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
.stg-tollrow .stg-track{height:10px}
/* A BUILD THAT TRAILS BASELINE IS GREY, NOT RED - the mock's own #b6c0c6, and
   it agrees with what red already means here. Red is the injury and loss ink;
   an archetype returning less PAR is not a warning, it is the price of an
   argument, and three red bars in a four-row column read as three alarms. */
/* SCOPED THROUGH THE TRACK ON PURPOSE. `.stg-track i` above is 0-1-1 and this
   was 0-1-0, so the seat cards' accent fill quietly won every toll bar and the
   whole column rendered blue - a sign channel that had stopped signing while
   the numbers beside it still read correctly. Found on the simulator, which is
   the honest witness for a computed style; the class was right and the cascade
   was not. */
.stg-track i.stg-tollbar{background:#b6c0c6}
.stg-track i.stg-tollbar.up{background:var(--accent)}
.stg-tolln{color:var(--text-muted)}
.stg-tolln.up{color:var(--text-primary);font-weight:500}
.stg-tollv{font:500 15px/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:right;color:var(--text-muted)}
.stg-tollv.up{color:var(--accent-hover)}

/* ------------------------------------------------------- THE RESCUE RAIL
 * The argument the page is making, drawn instead of bulleted. */
.st-rescue{display:grid;grid-template-columns:1.2fr 1fr;gap:44px;
  align-items:center;background:var(--bg-surface);border-radius:var(--r-lg);
  padding:30px var(--sp-5)}
.st-rescuedek{margin:10px 0 0;max-width:64ch;color:var(--text-secondary);
  font-size:15px;line-height:1.7;text-wrap:pretty}
.st-rescuer{display:flex;flex-direction:column}
.st-rrow{display:grid;grid-template-columns:minmax(52px,64px) minmax(60px,1fr) auto;
  gap:14px;align-items:center;padding:12px 0;
  border-top:1px solid var(--border-row)}
.st-rseat{color:var(--text-muted);letter-spacing:.14em}
.st-rplay{font-size:var(--fs-lg);line-height:1.4;color:var(--text-primary)}
.st-rgain{font:500 20px/1 var(--font-num);font-variant-numeric:tabular-nums;
  text-align:right;color:var(--accent)}
.st-rrow.st-mine{box-shadow:inset 3px 0 0 var(--accent);padding-left:8px}
.st-rrow.st-mine .st-rseat{color:var(--accent)}

/* ------------------------------------------------------------ OPERATIONS
 * The four things you set around the one thing you run. A band, not five more
 * sections in a column: everything above it is evidence, everything in it is a
 * setting that changes what the evidence is measured under. */
.st-ops{background:var(--bg-surface);border-radius:var(--r-lg);
  padding:var(--sp-5) var(--sp-5) calc(var(--sp-5) * 1.4)}
.lions .st-ops{background:var(--bg-sunken)}
.st-opshead{display:flex;justify-content:space-between;align-items:baseline;
  gap:var(--sp-4);flex-wrap:wrap;margin-bottom:var(--sp-4)}
.st-opscap{color:var(--text-muted);letter-spacing:.12em}
.st-ops .st-sec:first-of-type{margin-top:0}

/* ------------------------------------------------------- THE REGISTER
 * "What this page cannot see", as the mock draws it: a serif head over a
 * two-column numbered list, with the amendments line under a rule. */
.st-dochead{margin:0;font:400 24px/1.15 var(--font-label);letter-spacing:-.004em;
  color:var(--text-primary)}
.st-doc{display:grid;grid-template-columns:1fr 1fr;gap:0 40px}
.st-docrow{display:grid;grid-template-columns:24px minmax(0,1fr);gap:10px;
  padding:11px 0;border-top:1px solid var(--border-strong)}
.st-docn{font:500 var(--fs-sm)/1.7 var(--font-num);color:var(--text-muted)}
.st-doct{margin:0;font:300 14px/1.75 var(--font-ui);color:var(--text-secondary);
  text-wrap:pretty}
.st-docfoot{display:flex;gap:16px;margin-top:var(--sp-5);padding-top:var(--sp-3);
  border-top:1px solid var(--border-strong);color:var(--text-muted);
  letter-spacing:.14em;font-weight:500}
.st-docn2{margin-left:auto;text-align:right;white-space:nowrap}

/* ---------------------------------------------------- THE CONTEXT BAR
 * Which chair the page is speaking from, and the way back to the gate. Never
 * scrolled away from: every number under it is true of ONE seat. */
.st-ctx{display:flex;justify-content:space-between;align-items:center;gap:20px;
  flex-wrap:wrap;background:var(--bg-surface);border-radius:var(--r-md);
  padding:10px var(--sp-4);margin-bottom:var(--sp-4)}
/* --bg-sunken is skin-local by its own declaration - only skin-scoped rules
   may name it - and the mock draws this bar as the recessed strip above the
   masthead, which is exactly what that token is for. */
.lions .st-ctx{background:var(--bg-sunken)}
.st-back{background:none;border:0;padding:0;cursor:pointer;
  font:500 var(--fs-sm)/1 var(--font-num);letter-spacing:.16em;
  text-transform:uppercase;color:var(--accent-hover)}
.st-back:hover{color:var(--text-primary)}
.st-ctxwho{color:var(--text-muted);letter-spacing:.14em;text-align:right}

@media (max-width:1024px){
  /* The evidence and the curve stop sharing a row before the table's own
     min-width forces the page sideways. */
  .st-evidence{grid-template-columns:1fr;gap:var(--sp-5)}
}
@media (max-width:860px){
  /* THE TABLE TAKES THE GUTTER BACK. On a phone the wrap's 16px each side is
     32px the widest object on the page cannot use, and it is spent making a
     five-column table scroll sooner. The scroller bleeds to the screen edges
     and pads its own first and last cell instead, so the measure is unchanged
     for every other section and the table gets the whole viewport. */
  .st-scroll{margin-left:calc(var(--sp-4) * -1);margin-right:calc(var(--sp-4) * -1);
    padding-left:var(--sp-4);padding-right:var(--sp-4)}
  .st-doc{grid-template-columns:1fr}
  .st-docfoot{flex-direction:column;gap:8px}
  .st-docn2{margin-left:0;text-align:left}
  .st-rescue{grid-template-columns:1fr;gap:var(--sp-4);padding:20px 16px}
  .st-ops{padding:var(--sp-4) 12px var(--sp-5)}
  .stg-vbody{grid-template-columns:1fr;gap:var(--sp-5)}
  .stg-head{gap:var(--sp-4)}
  .stg-frame{align-items:flex-start}
  .stg-chiprow{justify-content:flex-start}
  .stg-meta{text-align:left}
  .stg{background-image:none;gap:var(--sp-5);min-height:0}
  .stg-verdict{padding:18px 16px 16px}
  .stg-cap span:last-child{display:none}
  /* A THUMB, NOT A CURSOR. The mock's chips are 29px tall, which is correct
     for the 1336px page it was drawn on and half a target on a phone. Scoped
     by WIDTH like every other touch rule here, so the desktop keeps the
     design's own geometry. */
  .stg-chip{min-height:44px;display:inline-flex;align-items:center;
    padding:0 14px}
  .st-back{min-height:44px;display:inline-flex;align-items:center}
  /* The tie's second build goes, the mark that says there IS a tie stays. */
  .stg-tagor{display:none}
  .stg-par{font-size:var(--fs-sm)}
}
.rpt-rage{width:6ch;text-align:right;color:var(--text-muted);
  font:600 var(--fs-xs)/1 var(--font-num)}
.rpt-rage-flag{color:var(--negative-ink)}

/* ============================================================ OPERATIONS
 * His mock's lab: four labelled rows of chips and a run panel beside them,
 * replacing a row of selects. A select hides its own menu, so the reader could
 * not see that the depth reaches 600 or which settings this board cannot
 * offer, and both of those ARE the lab telling you what it can do. */
.st-labcard{background:var(--bg-surface);border-radius:var(--r-lg);
  padding:24px 26px 22px}
.st-labcols{display:grid;grid-template-columns:1fr 1fr;gap:36px;align-items:start}
.st-labname{font:600 15px/1 var(--font-ui)}
.st-labdek{margin:8px 0 16px;max-width:56ch;color:var(--text-secondary);
  font-size:15px;line-height:1.6}
/* THE LABEL COLUMN IS FIXED AND THE CHIPS FLOW. minmax(0,1fr) with a real
   floor on the chip track: a fixed label plus a wrapping chip row is exactly
   the shape that collapses the flexible track to zero. */
.st-labset{display:grid;grid-template-columns:58px minmax(0,1fr);gap:10px 12px;
  align-items:start}
.st-labl{color:var(--text-muted);letter-spacing:.16em;padding-top:11px}
.st-labchips{display:flex;gap:5px;flex-wrap:wrap}
.st-lchip{background:var(--bg-app);color:var(--text-secondary);border:0;
  font:400 var(--fs-md)/1 var(--font-num);padding:8px 11px;border-radius:3px;
  cursor:pointer}
.st-lchip:hover:not(.na){color:var(--text-primary)}
.st-lchip.on{background:var(--text-primary);color:var(--bg-surface);font-weight:500}
/* THE ROOM ROW LIGHTS ACCENT, the menu rows light ink. One is "this is the
   setting I picked from a list", the other is "this is ON" - the mock draws
   them differently and they are different questions. */
.st-labchips .st-tog.on{background:var(--accent);color:var(--on-accent)}
/* UNAVAILABLE IS STATED, NOT HIDDEN. A setting this board cannot offer stays
   in place saying so; removing it would make the menu look shorter than it is. */
.st-lchip.na{background:var(--track);color:var(--text-muted);cursor:default}
/* --bg-sunken is skin-local by its own declaration; only skin-scoped rules may
   name it. On paper the unavailable chip is the recessed one. */
.lions .st-lchip.na{background:var(--bg-sunken)}
.st-labseats .st-lchip{min-width:34px;text-align:center}
.st-labfoot{display:flex;justify-content:space-between;align-items:center;
  gap:16px;margin-top:18px;padding-top:16px;border-top:1px solid var(--border-row)}
.st-labnote{font:400 var(--fs-sm)/1.5 var(--font-num);color:var(--text-muted)}
.st-labrun{background:var(--accent);color:var(--on-accent);border:0;
  font:500 var(--fs-sm)/1 var(--font-num);letter-spacing:.12em;
  text-transform:uppercase;padding:12px 18px;border-radius:4px;
  white-space:nowrap;cursor:pointer}
.st-labrun:hover:not(:disabled){background:var(--accent-hover)}
.st-labrun:disabled{background:var(--track);color:var(--text-muted);cursor:default}

.st-runpanel{background:var(--bg-app);border-radius:var(--r-lg);
  padding:20px 22px 18px}
.st-runcap{color:var(--text-muted);letter-spacing:.18em}
.st-runidle{margin:12px 0 0;color:var(--text-muted);font-size:var(--fs-lg);
  line-height:1.6}
.st-labres{display:flex;flex-direction:column;gap:14px;margin-top:12px}
.st-rcard{background:var(--bg-surface);border-radius:8px;padding:14px 18px}
/* The build you ran is the lit object; the mock rings it the way the verdict
   card is rung, which is how this page already says "this is the answer". */
.st-rcard.lit{background:var(--bg-selected);border:1px solid var(--accent);
  box-shadow:0 0 0 4px var(--accent-soft);padding:16px 18px 14px}
.st-rchead{display:flex;justify-content:space-between;align-items:baseline;
  gap:14px;flex-wrap:wrap}
.st-rcname{color:var(--text-muted);font-weight:500;letter-spacing:.14em}
.st-rcard.lit .st-rcname{color:var(--accent-hover)}
.st-rcbeats{color:var(--text-muted);letter-spacing:.1em}
.st-rcnum{display:flex;align-items:baseline;gap:8px;margin-top:8px}
.st-rcpar{font:500 34px/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--text-primary)}
.st-rcse{font:400 var(--fs-lg)/1 var(--font-num);color:var(--text-muted)}
/* THE DELTA IS ITS OWN OBJECT. It was a clause hanging off the end of the
   first row, where it read as decoration on that number rather than as the
   answer to the question the run was asked. */
.st-labres .st-delta{display:block;background:var(--risk-soft);border-radius:8px;
  padding:13px 16px;font:500 19px/1 var(--font-num);
  font-variant-numeric:tabular-nums;color:var(--risk)}
.st-labres .st-delta.up{color:var(--accent)}
.st-labres .st-delta.down{color:var(--negative-ink)}
.st-tops{margin-top:6px;padding-top:12px;border-top:1px solid var(--border-row)}
.st-sample{display:grid;grid-template-columns:64px minmax(0,1fr);gap:12px;
  align-items:baseline;padding:8px 0;border-top:1px solid var(--border-row)}
.st-tpar{font:500 var(--fs-xl)/1 var(--font-num);font-variant-numeric:tabular-nums;
  color:var(--accent-hover)}
.st-troster{font:400 var(--fs-md)/1.5 var(--font-num);color:var(--text-secondary)}

@media (max-width:900px){
  .st-labcols{grid-template-columns:1fr;gap:var(--sp-4)}
  .st-labcard{padding:18px 14px 16px}
  .st-labset{grid-template-columns:1fr;gap:4px 0}
  .st-labl{padding-top:10px}
  .st-labfoot{flex-direction:column;align-items:stretch;gap:12px}
  .st-lchip{min-height:44px;display:inline-flex;align-items:center}
  .st-labrun{min-height:44px}
}

/* ------------------------------------------------------- THE HOME ROOM */
.st-homecols{display:grid;grid-template-columns:1.15fr 1fr;gap:36px;
  align-items:start}
.st-homelab{margin:0 0 8px}
/* Thirty-three teams in nine columns: one glance and one tap, where a select
   is one tap, a scroll and a hunt - and it hides which team is set. */
.st-teamgrid{display:grid;grid-template-columns:repeat(9,minmax(0,1fr));gap:4px}
.st-teamgrid .st-lchip{padding:10px 0;text-align:center;justify-content:center}
.st-reachrow{display:flex;align-items:center;gap:12px;margin-top:16px;
  padding-top:14px;border-top:1px solid var(--border-row);flex-wrap:wrap}
.st-reachnote{font:400 var(--fs-sm)/1.5 var(--font-num);color:var(--text-muted);
  margin-left:auto;text-align:right}
.st-homebadge{background:var(--text-primary);color:var(--bg-surface);
  font:500 var(--fs-xs)/1 var(--font-num);letter-spacing:.1em;padding:6px 8px;
  border-radius:3px;white-space:nowrap}
.st-homewhat{display:flex;flex-direction:column;margin-top:12px}
.st-hwrow{display:grid;grid-template-columns:88px minmax(0,1fr);gap:12px;
  padding:11px 0;border-top:1px solid var(--border-strong)}
.st-hwl{color:var(--accent-hover);letter-spacing:.12em;line-height:1.4}
.st-hwt{font-size:14px;line-height:1.55;color:var(--text-secondary)}
.st-hwfoot{margin:14px 0 0;font:400 var(--fs-sm)/1.6 var(--font-num);
  letter-spacing:.06em;color:var(--text-muted)}

/* ------------------------------------------------------------ THE TRIO */
.st-trio{display:grid;grid-template-columns:1fr 1fr;gap:10px;
  margin-top:calc(var(--sp-5) * 1.5)}
.st-trio>.st-sec{margin-top:0;min-width:0}
.st-tile{background:var(--bg-surface);border-radius:var(--r-lg);
  padding:18px 20px;height:100%}
.st-tilename{font:600 14px/1 var(--font-ui)}
.st-tilebody{display:grid;grid-template-columns:minmax(0,1fr) auto;gap:12px;
  align-items:center;margin-top:5px}
.st-tileline{margin:0;font-size:14px;line-height:1.5;color:var(--text-secondary)}
.st-tilectl{background:var(--bg-app);color:var(--text-primary);border:0;
  font:400 var(--fs-sm)/1 var(--font-num);padding:9px 11px;border-radius:3px;
  white-space:nowrap;cursor:pointer}
.st-tilectl:hover{background:var(--bg-selected)}
.st-tilechip{background:var(--accent-soft);color:var(--accent-hover);
  font:500 var(--fs-sm)/1 var(--font-num);padding:9px 11px;border-radius:3px;
  white-space:nowrap}
.st-dock{display:flex;flex-direction:column;gap:6px;margin-top:10px}
.st-dockrow{display:grid;grid-template-columns:minmax(0,1fr) auto;gap:10px;
  align-items:baseline}
.st-dockn{font:400 var(--fs-lg)/1.4 var(--font-ui);color:var(--text-primary)}
.st-docks{color:var(--text-muted)}
.st-dockm{color:var(--text-muted);letter-spacing:.06em;white-space:nowrap}
/* THE LEDGER IS KEPT, NOT DELETED. The mock draws these as summary tiles; the
   eighteen pairs, the receiver ledger and the docket's full facts are real
   research, and a ledger nobody can reach is the same as no ledger. */
.st-tiledet{margin-top:12px;border-top:1px solid var(--border-row);
  padding-top:10px}
.st-tiledet summary{cursor:pointer;color:var(--text-muted);letter-spacing:.14em;
  font-weight:500}
.st-tiledet summary:hover{color:var(--text-primary)}
.st-tiledet[open] summary{margin-bottom:10px}

@media (max-width:900px){
  .st-homecols{grid-template-columns:1fr;gap:var(--sp-4)}
  .st-teamgrid{grid-template-columns:repeat(6,minmax(0,1fr))}
  .st-trio{grid-template-columns:1fr}
  .st-tilebody{grid-template-columns:1fr;align-items:start}
  .st-tilectl,.st-tilechip{justify-self:start;min-height:44px;
    display:inline-flex;align-items:center}
  .st-teamgrid .st-lchip{min-height:44px}
}

/* ------------------------------------------------------- THE PLAY FEED (live)
   The container. Steps 1 and 2 drew a flat one-line row inside it; step 3
   replaces that row with `.ev` below and this shell is all that survives of
   them, because the shell was never the part under test. */
.pfeed{margin-top:var(--sp-4);border:1px solid var(--border-subtle);
  border-radius:var(--r-md);overflow:hidden;background:var(--bg-card)}
.pf-hd{display:flex;align-items:center;justify-content:space-between;gap:var(--sp-3)}
.pf-heading{display:flex;align-items:baseline;gap:8px;min-width:0;flex-wrap:wrap}
.pf-meta{color:var(--text-muted)}
.pf-all{flex:0 0 auto}
.pf-list{list-style:none;margin:0;padding:0;max-height:60vh;overflow-y:auto}
.pf-empty{padding:var(--sp-4);color:var(--text-muted);font-size:var(--fs-sm)}
/* The tier disclosure and the seeded notice. Both are `.alert` variants that
   already ship; the only thing this rule does is stop them inheriting the
   card's own padding twice. */
.pf-tier{padding:0 var(--sp-3) var(--sp-3)}
.pf-tier .alert{margin:0}
.replay-context .body{text-transform:none;letter-spacing:normal}

/* ================================================ THE EVENT ROW (step 3) ===
 * spec-render.md 1.1 through 1.9, and the CSS budget in LIVE-PLAN 7.3 is the
 * constraint that shaped every line of it: `.ev` and its children, `.ev.b0`
 * through `.ev.b3`, `.ev.mine` / `.ev.opp`, `.sig-warn`, `.sig-hold` and
 * `.ev-stake-old`. NO new colour, no new font, no new radius, no new spacing
 * step, no new animation. Every value below is a token this file already
 * ships or a literal that spec-render.md names, and that is not tidiness: the
 * light palette moves --negative-ink from #ff5c72 to #d92d3f, --risk from
 * #e0b46b to #9a6d10 and --font-display all the way to --font-ui, and the row
 * has to follow all three without a second declaration anywhere.
 *
 * THE ROW IS A TWO-COLUMN GRID AND THE LEFT COLUMN IS THE PRODUCT. Real puts
 * its number last and right-aligned, which is right for a chronological feed
 * that the number merely annotates. Here the number IS the sort key, so it
 * belongs where a rank belongs. What both share, and the reason either works,
 * is Real's mechanism 4: the numerals form ONE flush column at ONE x, in
 * every band, so 9.3 beside 1.1 is legible at arm's length without reading a
 * word of either row.
 *
 * BANDS ARE FOUR DISCRETE CLASSES, NEVER A CONTINUOUS RAMP. A ramp makes
 * every row slightly different from its neighbour and nothing reads as a
 * KIND; four bands mean a swing play is recognisable as a class of row from
 * across the desk. The cut points are stake magnitude 0.5 / 2 / 8, which is
 * spec-render's four bands with LIVE-PLAN 3.6's top edge - 6 sat near p75 of
 * the measured pooled stake distribution and would have made a quarter of
 * every feed a "swing" row, and a feed where everything shouts ranks nothing.
 *
 * The literals live HERE as custom properties rather than in the renderer,
 * which is this file's stated habit in the --tape-* comment: tokens, not
 * literals in the renderer.
 */
.ev{display:grid;grid-template-columns:76px minmax(0,1fr);column-gap:var(--sp-4);
  padding:var(--ev-pad-y) var(--sp-4);border-bottom:1px solid var(--border-row);
  background:var(--bg-row);
  --ev-hl:15px;--ev-hlw:400;--ev-stake-fs:15px;--ev-pad-y:9px}
.ev.b1{--ev-hl:17px;--ev-hlw:500;--ev-stake-fs:18px;--ev-pad-y:11px}
.ev.b2{--ev-hl:21px;--ev-hlw:600;--ev-stake-fs:22px;--ev-pad-y:14px}
/* FIVE CHANNELS MOVE TOGETHER between band 0 and band 3, which is what lets a
   big play announce itself without being read: headline 15 to 26px (1.73x),
   headline weight 400 to 600, numeral 15 to 30px (2x), row padding 9 to 18px
   so a swing play owns more of the column, and a 3 percent ground tint. The
   tint is --row-tint and nothing else: no glow, no border, no colour wash.
   The SURFACES comment at the top of this file is explicit that this app's
   cards carry no raised planes and no glow, and three percent is plenty once
   four other channels have already moved. */
.ev.b3{--ev-hl:var(--fs-hero);--ev-hlw:600;--ev-stake-fs:30px;--ev-pad-y:18px;
  background:var(--row-tint)}
/* AFTER THE BANDS, AND THE ORDER IS LOAD-BEARING. `.ev.b3` and `.ev:hover`
   have identical specificity (two classes each; a pseudo-class weighs the
   same as a class), so a hover rule written above the tint loses to it and
   band 3 becomes the one row on the page that does not answer the pointer. */
.ev:hover{background:var(--bg-row-hover)}
/* THE MOCK'S ROW GEOMETRY: a 64px rail, a 16px gutter, and the body's own two
   lines. The grid column narrows from 76px because the tile fills it edge to
   edge and a 76px block of red reads as a panel rather than a marker. */
.gameday .ev{grid-template-columns:64px minmax(0,1fr);column-gap:16px;
  padding-left:0;padding-right:0}
/* HEADLINE FIRST, META SECOND, which is the one ordering change the mock makes
   to the row and it is the right way round: the headline is what the reader is
   scanning for and the meta line is what qualifies it. The row's DOM order is
   unchanged - the meta is still emitted first, because the correction path and
   the suite both match on it - so this is an ORDER property of the flex box
   and not a rewrite of the renderer. */
.gameday .ev-body{display:flex;flex-direction:column}
.gameday .ev-body .ev-hl{order:1}
.gameday .ev-body .ev-meta{order:2;margin-top:6px}
.gameday .ev-body .ev-actor{order:3}
.gameday .ev-body .ev-cp{order:4}
.gameday .ev-body .ev-tags{order:5}
/* The mock sets every headline in the poster face, uppercased, and lets the
   band scale carry the size - which is what --ev-hl already does. */
.gameday .ev-hl{text-transform:uppercase;letter-spacing:.02em}

/* BAND 0 COLLAPSES, and only its headline collapse belongs in CSS. The
   counterparty line, the tag row and the explanation line are simply NOT
   EMITTED on a trace row rather than hidden here, because `display:none` and
   "never drawn" read identically on screen and differently in the DOM - and
   the DOM is what the suite can hold to account. What is left is the ink:
   the headline drops to --text-secondary, which is the same ruling as
   `.d-quiet`'s own comment, "a delta that clears no bar is data, not signal,
   it stays in the margins". */
.ev.b0 .ev-hl{color:var(--text-secondary)}

/* OWNERSHIP IS BINARY AND CATEGORICAL, WHICH IS EXACTLY WHAT AN EDGE BAR IS
   GOOD FOR. Magnitude is already carried by the band scale and by the numeral
   itself; a bar that also tried to carry it would be a third encoding of one
   fact and a second encoding of another. Lifted verbatim from the
   `tr.glow-val` idiom above (inset 3px 0 0), so "this one is yours" is one
   gesture in the draft room and on the live feed. NEITHER class is the third
   state: a play with nobody rostered on it gets no bar, which is correct,
   because there is nothing to mark. */
.ev.mine{box-shadow:inset 3px 0 0 var(--accent)}
.ev.opp{box-shadow:inset 3px 0 0 var(--text-dim)}

/* THE RAIL SPANS EVERY BODY LINE (grid-row 1 / -1) so the numeral column
   cannot wander as the body below it grows a counterparty line or a tag row.
   Right-aligned so the numerals flush against one another regardless of how
   many digits each carries. */
.ev-stake{grid-column:1;grid-row:1 / -1;text-align:right}
.ev-body{grid-column:2;min-width:0}

/* THE NUMERAL'S BOX, AND IT IS THE MOST IMPORTANT DECLARATION IN THIS BLOCK.
   LIVE-PLAN section 4 item 1, called there "the single most important line in
   the plan": no stake numeral on a FORMING row, three dots in a box of the
   band-1 numeral's EXACT width and height, so nothing reflows when the number
   lands. Real prints 0.6 on its in-progress row and 4.6 on the same row 42
   seconds later, and a box that grows when the truth arrives teaches the
   reader that the provisional number was the real one.

   THREE THINGS HOLD THE BOX and each is load-bearing:
     a BLOCK, so its width is the rail's width in every state and the dots,
       the numeral and the word `unscored` cannot disagree about it;
     an EXPLICIT height and line-height derived from the same
       --ev-stake-fs the glyphs are set in, so the box is a function of the
       BAND and never of the content - an auto height would be the font's
       content box and would move the moment the glyphs changed;
     overflow:hidden as the belt, because a numeral that somehow outgrew the
       rail must clip (visibly wrong) rather than shove the body column
       sideways (invisibly wrong).
   1.12 is the ratio, not a line-height token, because this box must scale
   with the band and every line-height token in this file is a constant. */
/* ------------------------------------------------- THE TILE (game day skin)
   The mock's filled plate behind the numeral: red when the play is yours,
   graphite otherwise, with the label word under it.
   IT IS A WRAPPER AND IT DOES NOT OWN THE BOX. `.ev-stake-n` keeps its
   explicit height derived from --ev-stake-fs and stays the element the layout
   probe measures, because the moment a padded tile owns the height, three
   things fail at once and only one of them is visible: the height stops being
   a function of the BAND, tools/live-pair.mjs's UNFIX_CSS unfixes an element
   that no longer decides anything, and the negative control at the end of the
   measurement goes quietly inert - reporting that the box did not move
   because nothing it can reach makes it move. A harness that cannot break the
   thing it is testing is not a control.
   OWNERSHIP IS THE TILE'S COLOUR, which is what frees the label slot below it
   to carry the TIER. LIVE-PLAN 3.7 requires the tier disclosed as a word on
   every row (tier 2 shares essentially none of tier 1's top plays, and the top
   of the feed is where the reader looks); the mock puts YOURS / OPP in that
   slot, which the fill already says. Two encodings of ownership and none of
   the tier would have been the trade, so the colour keeps ownership and the
   word keeps its job. */
/* `align-items` IS LEFT AT stretch AND THAT IS THE LOAD-BEARING WORD.
   Centring the flex children shrink-wraps them to their content, so
   `.ev-stake-n` stopped being a block the width of the rail and became a box
   the width of whatever it happened to contain - and the layout probe caught
   it immediately: the dots box measured [17.62, 28.74] and the number box
   [9.45, 45.08] at both widths. That is the reflow LIVE-PLAN section 4 item 1
   forbids, reintroduced by a line that looks like pure styling. The children
   stretch; the GLYPHS are centred by text-align below. */
.gameday .ev-tile{display:flex;flex-direction:column;
  justify-content:center;gap:2px;padding:8px 4px;border-radius:3px;
  background:var(--gd-tile)}
.gameday .ev.mine .ev-tile{background:var(--gd-red)}
.gameday .ev.mine .ev-stake-n{color:#fff}
.gameday .ev.mine .ev-stake-u{color:rgba(255,255,255,.8)}
/* THE NUMERAL IS CENTRED IN THE TILE, as drawn, and the flush-column rule it
   replaces is preserved rather than dropped: a fixed-width tile centres every
   numeral on ONE x, so the column still reads at arm's length. The suite's
   measurement moves from the ink's right edge to its centre for the same
   reason - same property, measured on the axis this design actually holds. */
.gameday .ev-stake-n{text-align:center}
.gameday .ev-stake-u{text-align:center}
/* The oblique numeral over a colour field, the app's existing Archivo Black
   idiom, and the reason that face ships. Only on the filled tile: obliqued
   type on the bare graphite of a trace row would be decoration. */
.gameday .ev.mine .ev-stake-n{font-family:"Archivo Black",sans-serif;
  transform:skewX(-9deg)}
.ev-stake-n{display:block;font-family:var(--font-num);
  font-variant-numeric:tabular-nums;font-size:var(--ev-stake-fs);
  height:calc(var(--ev-stake-fs) * 1.12);line-height:calc(var(--ev-stake-fs) * 1.12);
  overflow:hidden;white-space:nowrap}
/* THE UNIT LINE IS THE TIER DISCLOSURE, and it is the whole mechanism for the
   tier degradation in one slot with no extra chrome. It is NEVER omitted, so
   a reader can never mistake one tier's row for another's.

   WHAT SHIPS HERE, AND THE TWO PLACES THIS COMMENT USED TO BE WRONG. It said
   "WIN% when both lineups are known, PROJ when only yours is, GAME when no
   roster is connected", which described a plan rather than the page:

     WIN%      tier 1, and it is real as of step 5's caller: both lineups
               known, so the number is percentage points of win probability.
     PTS       tier 2 - NOT `PROJ`. The rail there prints the play's effect on
               my BANKED points and the running total either side of it, and
               those are points already scored, not a projection of anything.
               `PROJ` was the wrong word for that column and was corrected in
               the code long before it was corrected here.
     UNJOINED  no league is joined. NOT `GAME`: tier 3 in stake.js means
               ESPN's game-level win probability, which needs the bulk
               probabilities endpoint this app does not call. Until it does,
               the honest word is the one that says no roster is connected
               rather than one promising a number from a feed that is absent.

   Two more words live in this slot and they are refusals rather than tiers -
   PENDING while the play is still forming and UNSCORED where no kicker or
   D/ST on the board carries a banked espnId - which is what keeps a
   number-shaped hole from ever being read as a zero. */
.ev-stake-u{display:block;color:var(--text-muted)}
/* THE EXPLANATION LINE IS NOT OPTIONAL AND NOT BEHIND A TAP. Real's single
   largest credibility gap, measured across nine surfaces and 45 frames, is
   that its number has no explanation surface anywhere in the product. This is
   the answer: the before and the after, on every row that carries a number,
   always, in the same unit as the numeral above it. */
.ev-stake-x{display:block;font-family:var(--font-num);font-size:var(--fs-xs);
  color:var(--text-muted);white-space:nowrap;overflow:hidden}
/* THE OLD VALUE, STRUCK, for the 20 seconds after a material correction. Step
   4 wires the timer; the rule ships here because it is one of the six
   declarations LIVE-PLAN 7.3 budgets for this row and splitting it across two
   steps would leave the block half-written. */
.ev-stake-old{display:block;font-family:var(--font-num);font-size:var(--fs-xs);
  color:var(--text-dim);text-decoration:line-through}
/* --text-dim rather than --text-muted, so the three dots read as quieter than
   the unit label beneath them. They are the absence of a number, not a value. */
.ev-dots{color:var(--text-dim);letter-spacing:.14em}

/* THE META LINE. Mono, because it is five numbers and a clock and they should
   sit in columns down the feed rather than reflow with the prose above them.
   nowrap and ellipsis because on a 375px phone this is the line that would
   otherwise wrap and push the headline down a row mid-drive. */
.ev-meta{font:500 var(--fs-sm)/1.4 var(--font-num);letter-spacing:.04em;
  color:var(--text-muted);white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
/* ONLY THE HALF OF THE SCORE THAT MOVED comes forward. Straight from Real,
   whose mechanism 2 this is, and it is the cheapest signal in the row: the
   reader learns who scored without reading a word or comparing two numbers. */
.ev-sc-on{color:var(--text-primary);font-weight:600}

/* THE HEADLINE, in --font-display, which is the same face `.ws-item-hl` sets
   the Wire's headlines in - so the feed you watch during the game and the
   feed you read before it speak in one voice. Under the light palette
   --font-display resolves to --font-ui, which is correct and already reasoned
   about where it is declared. Two lines then ellipsis: a third line of
   headline is a headline that failed. */
.ev-hl{font:var(--ev-hlw) var(--ev-hl)/1.2 var(--font-display);
  letter-spacing:-.005em;color:var(--text-primary);text-wrap:balance;
  display:-webkit-box;-webkit-line-clamp:2;-webkit-box-orient:vertical;
  overflow:hidden}
/* THE LOUDEST SIGNAL IN THE ROW, and it is one rule: accent when the play
   actually put points on the board. Real's mechanism 1, and it is readable
   while scrolling too fast to read words. A MISSED field goal stays primary,
   because it scored nothing. */
.ev-hl.ev-scored{color:var(--accent)}

/* THE ACTOR LINE. The name in --text-primary and everything after the first
   middot in --text-secondary and mono, so the sentence reads "who" first and
   "how much" second rather than as one undifferentiated string. */
.ev-actor{font:600 14px/1.4 var(--font-ui);color:var(--text-primary);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
.ev-actor .ev-st{font-weight:400;font-family:var(--font-num);
  color:var(--text-secondary)}
/* THE COUNTERPARTY LINE IS SUBORDINATE IN COLOUR ENTIRELY, INCLUDING THE
   NAME. That is Real's rule at 1.5 and it is the reason the line reads as
   context rather than as a second actor competing with the first. */
.ev-cp{font:400 12.5px/1.4 var(--font-ui);color:var(--text-secondary);
  white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
/* `.sigs` already ships flush RIGHT, which is correct beside a right-aligned
   value column and wrong under a left-aligned headline. This is the whole
   override, and it wraps because a scoring play can carry four chips. */
.ev-tags{justify-content:flex-start;flex-wrap:wrap;margin-top:var(--sp-1)}

/* THE PHONE. 76 + 16 + 16 + 16 leaves 251px of a 375px viewport for a body
   that has to hold a two-line headline and a nowrap meta line, and the meta
   line is the one that loses. The rail narrows to 64px and the gutters to
   --sp-3, which buys 28px back and is enough - measured, not guessed. The
   rail STAYS A FIXED COLUMN, because the property that matters is that the
   numerals share one x, not what that x is. */
@media (max-width:520px){
  .pf-hd{align-items:flex-start}
  .pf-heading{display:block}
  .pf-heading .pf-meta{display:block;margin-top:4px}
  .ev{grid-template-columns:64px minmax(0,1fr);column-gap:var(--sp-3);
    padding-left:var(--sp-3);padding-right:var(--sp-3)}
  .ev.b3{--ev-stake-fs:26px}
  /* THE ACTOR LINE WRAPS ON A PHONE AND THE META LINE DOES NOT, and the
     asymmetry is the point. Measured at 375: `E. Stowers &middot; 6 yd, 2 rec,
     1.6 pts [OPP]` needs 268px in a 251px cell, so nowrap ate the last thing
     on the line - which is the OWNERSHIP CHIP, the one item on the row a
     manager scans for. The meta line keeps its ellipsis because what it loses
     is its LAST slot too, and that slot is the relative age, which is the
     least load-bearing thing in the row and is the only one whose absence
     costs nothing. Truncation is fine when you choose what falls off. */
  .ev-actor,.ev-cp{white-space:normal;text-overflow:clip}
}

/* THE PLAY FEED'S OWN NUMERAL, `.pf-pts` / `.pf-dots` / `.pf-un` / `.pf-own`,
   USED TO SIT HERE AND IS DELETED. Steps 1 and 2 drew a one-line row with a
   fixed 66x15 numeral box; step 3's `.ev-stake-n` above is the numeral now,
   and its box is `calc(var(--ev-stake-fs) * 1.12)` because the band scale
   moves that numeral from 15px to 30px and a constant cannot follow it. The
   no-reflow rule those four rules existed to keep is kept above, by
   measurement rather than by a literal. DECISIONS.md, 2026-08-15. */
