Identity coverage by legend pair: Chaos/Fury has two finishers and none that pays by turn four

Issue #491 (a slice of #200), 2026-10-27. Constructed, Duel (485). Rules version 2026-07-16. This is the first slice of the identity-first pass decided on #200 on 2026-10-27: count the finishers each legend domain pair can play, take the pair with the fewest, and try to build lines for it that beat the free Hold curve of the unopposed clock.

Chaos/Fury is the target pair, and Body/Calm is the runner-up. No Chaos/Fury finisher in the catalogue pays by T4, and it has two finishers in all, the fewest of the fifteen pairs. This slice walked two new Chaos/Fury lines. One beats the pair's Hold baseline on the clock and enters the catalogue as a candidate; the other pays on the baseline's own turn and is a negative result.

1. What is counted

The pairs. A deck's Domain Identity is its legend's domains, 103.1.b.2: "Your deck’s Domain Identity is dictated by the domains of your Champion Legend." A card fits it by 103.1.b.3: "If a card has a single Domain, then that card is permitted in the Domain Identity that corresponds to the same Domain." and 103.1.b.4: "If a card has more than one Domain, then that card is permitted only in a Domain Identity that contains all of the indicated Domains on that card." Six domains give fifteen pairs.

The entries. Every data/combos.json entry of class INFINITE, BURST, CHAIN or ALT_WIN, 85 at e3101e5. An entry fits a pair when every card in its uses does: each card's domains from data/cards.json, matched on base codes, and a colourless card fits every pair. A mono-domain entry fits the five pairs that contain its domain, so the columns do not sum to 85. Both statuses are counted: 83 of the 85 are verified and 2 are candidate, power-nexus-sentinel-renata-time-warp-chain and power-nexus-sentinel-three-time-warps, both Mind CHAINs at T6.

The clock. npm run adversarial -- --turns prints a turn for every finisher: the turn its card set is first paid for on a perfect curve, plus one if it pays at a Hold, with nothing cast against it. The clock prices an entry together with the entries it folds in for fuel or payoff, and prints the domains of that whole set. A row counts as paying by T4 in a pair when it pays by T4, those printed domains fit the pair, and none of its cards is banned in Constructed (data/legality.json). This matters on one row: bottled-constellation-time-warp uses only Mind cards, but its T4 needs lux-infinite-energy and lux-infinite-power, which are Mind/Order, so its T4 counts in Mind/Order and nowhere else. Six of the 85 entries have a priced set wider than their own cards, all listed in §4.

ALT_WIN sits apart. An ALT_WIN does not score points; The Grand Plaza or Gutter Palace ends the game. The clock still prints a turn for it, and that turn is reported in its own column and not used to pick the target.

The baseline. The clock's Hold baseline is T5 for a pair with Body or Calm in it, because only those two domains print a unit at Energy 1 or less, and T6 for the other six pairs. The derivation is in the unopposed clock play.

The target rule. Fewest INFINITE, BURST and CHAIN entries paying by T4; ties broken on fewest finishers of all four classes, then alphabetically.

2. The table

One current table (#537 collapsed the five that slices 1 to 5 appended here). It is the command of §7 run on this PR's tree: master aec0933 (scripts/ byte-identical to it) plus slice 6's entry in data/combos.json. 91 finishers.

The order, (clocked by T4, all four): chaos/fury (0, 4), body/calm (0, 5), body/chaos (0, 7), body/fury (0, 8), body/order (0, 8), chaos/order (0, 11), calm/order (0, 12), body/mind (0, 15), fury/mind (0, 15), chaos/mind (0, 16), calm/mind (0, 22), calm/chaos (1, 9), fury/order (1, 9), calm/fury (2, 11), mind/order (2, 31). The first six are worked (slices 1 to 6), so the first pair not yet worked is Calm/Order (0, 12), baseline T5. §4 lists the entries whose priced set is wider than their own cards, now five of 91.

Changelog, one line per slice. Each line is the same command at the sha named, before that slice added anything, and what moved since the line above. The earlier tables are in git history (git show 07a93ad:docs/plays/2026-10-27-identity-coverage-by-legend-pair.md).

• 2026-10-27, e3101e5, #491 (slice 1): 85 finishers; target Chaos/Fury (0, 2), runner-up Body/Calm (0, 4).
• 2026-10-27, d8f796b, #496 (slice 2): slice 1's up-from-the-deep-flow-grand-plaza (mono-Chaos ALT_WIN, T4) adds an ALT_WIN, an ALT_WIN paying by T4 and one to "all four" in every Chaos pair; #494's anyBodies charge moves no row across T4; Body/Calm (0, 4) worked.
• 2026-10-27, c200d59, #512 (slice 3): slice 2's frisky-hunter-grand-plaza (mono-Calm ALT_WIN, T4) does the same in every Calm pair; the readying (#509) and [Accelerate] (#518) models move no row across T4; 87 finishers; Body/Chaos (0, 5) worked.
• 2026-10-27, 2fbe030, #524 (slice 4): slice 3's towering-combatant-up-from-the-deep-grand-plaza (Body/Chaos ALT_WIN, T4) gives Body/Chaos (0, 6); 88 finishers; Body/Fury (0, 7) worked, ahead of Body/Order at the same count alphabetically.
• 2026-10-31, 07a93ad, #529 (slice 5): slice 4's towering-combatant-hidden-marai-grand-plaza (Body/Fury ALT_WIN, T4) gives Body/Fury (0, 8); dragonstorm-confront-grand-plaza clocks T7, not T12; 89 finishers; Body/Order (0, 7) worked.
• 2026-10-31, aec0933, #537 (slice 6): slice 5's galio-vanguard-captain-grand-plaza (Body/Order ALT_WIN, T4) gives Body/Order 7 ALT_WINs; #528 (18374ab) prices ready-recruits-grand-plaza by Order alone at T4, so it now pays by T4 in every Order pair (ALT_WIN, pay by T4: body/order 1 to 3 with Galio, calm/order 4 to 5, chaos/order 3 to 4, fury/order 1 to 2) and leaves §4; #532 and #544 move dragonstorm-confront-grand-plaza T7 to T6 to T5; #530 and #540 move no row; no clocked count moved; 90 finishers; Chaos/Order (0, 10) worked. Then slice 6's own entry, yasuo-ride-the-wind-hold-curve-turn-five (mono-Chaos CHAIN, T5), adds one CHAIN and one to "all four" in every Chaos pair and moves no clocked count, because it pays on T5: the table above, 91 finishers.

Why each slice took its pair

Slice 1 (#491, at e3101e5): Chaos/Fury. 0 clocked finishers by T4 and 2 finishers in all: tryndamere-brambleback-conquer (mono-Fury BURST, T7 against T6) and spiderling-swarm-grand-plaza (mono-Chaos ALT_WIN, T6). Eleven pairs had 0 clocked finishers by T4; the count of all four classes is the tiebreak, and 2 was the lowest of the fifteen. The runner-up was Body/Calm, 0 by T4 and 4 in all, tied with Body/Chaos and first alphabetically: ahri-trinity-svellsongur-hold (BURST, T6), nasus-svellsongur-conquer-burst (Calm BURST, T6), flurry-of-feathers-grand-plaza-win (Calm ALT_WIN, T3; T4 after #494 / PR #497) and dragonstorm-confront-grand-plaza, a Body ALT_WIN counted in Body/Calm only by its own cards.

Slice 2 (#496, at d8f796b): Body/Calm. By the target rule alone Chaos/Fury was still first, but it was the pair slice 1 worked, and #200's decision takes one pair at a time and counts consecutive pairs for its stop rule, so a slice moves to the next pair in the order that has not been worked. That was Body/Calm, at (0, 4); Body/Chaos had gone to 5 with slice 1's Chaos ALT_WIN.

Slice 3 (#512, at c200d59): Body/Chaos, at (0, 5), the first pair not yet worked once the slice-1 and slice-2 candidates were counted. Its five: two CHAINs (trinity-force-yasuo-battle-mistress-hold-and-moves T7, draven-yasuo-battle-mistress-contested-chain T9) and three ALT_WINs (up-from-the-deep-flow-grand-plaza T4, spiderling-swarm-grand-plaza T6, dragonstorm-confront-grand-plaza priced with its Fury fuel at T12).

Slice 4 (#524, at 2fbe030): Body/Fury, at (0, 7), tied with Body/Order and first alphabetically. Its seven: gemdragon-henge-vi-blind-fury (INFINITE, T12), three BURSTs (brambleback-trinity-skyfall-conquer T6, tryndamere-brambleback-conquer T7, dragonstorm-brambleback-trinity-conquer T12), two CHAINs (trinity-skyfall-arena-second-battlefield-chain T5, trinity-tryndamere-both-phases-chain T7) and dragonstorm-confront-grand-plaza.

Slice 5 (#529, at 07a93ad): Body/Order, at (0, 7). Its seven: one BURST (ivern-arena-trinity-body-order-hold T7) and six ALT_WINs (grand-plaza-recruit-vanguard T5, karthus-machine-evangel-renata-plaza T6, plaza-armory-miss-fortune T6, vanguard-captain-manufactor-plaza T4, ready-recruits-grand-plaza priced Mind/Order at T4, dragonstorm-confront-grand-plaza priced Body/Fury at T7). Only vanguard-captain-manufactor-plaza paid by T4 inside the pair.

Slice 6 (#537, at aec0933): Chaos/Order, at (0, 10), the first pair not yet worked: the five worked before it are the first five in the order. Its ten: one INFINITE, pursuer-herald-recruits, which clocks T4 and is banned in Constructed; two CHAINs, ivern-ride-the-wind-double-conquer T6 and ivern-arena-draven-chaos-order-chain T7; and seven ALT_WINs, four of them paying by T4 (ready-recruits-grand-plaza, zed-clone-eye-recruits, vanguard-captain-manufactor-plaza and up-from-the-deep-flow-grand-plaza). No legal INFINITE, BURST or CHAIN of the pair paid by T4, which is why the issue asked for a BURST or CHAIN shape (§5).

3. The two Chaos/Fury lines

The Plaza line puts three Energy-2 bodies on The Grand Plaza by the end of T2, and one Up from the Deep cast twice ([Flow]) adds four Tentacles on T3: seven units for twelve Energy and no Power, and the T4 Hold wins. On the clock that is two turns ahead of the Hold curve. In the walk, the opponent's ready bodies attack the Plaza on their T3 and kill five of the seven, because four of them are 1-Might Tentacles. The cheapest answer costs nothing.

The Tryndamere line is the pair's only BURST. Its 1 + T x (1 + K) needs two Tryndamere, Barbarian and three Red Brambleback in one attack. Seal of Rage brings that attack from T7 to T6, and T5 is out of reach: a Tryndamere costs E7, so going first the two can only be bought on T4 and T5, and that leaves no turn with 4 Energy for the third Brambleback. T6 is the Hold curve's own turn, so the line is a negative result. Each play doc carries its own Breaks to, Needs and Not verified.

So Chaos/Fury now has one finisher paying by T4 on the clock's terms, and it is an ALT_WIN. The clocked count for the pair stays at 0: no Chaos/Fury BURST pays before T6, and no CHAIN was searched. [2026-10-31, #537: slice 6's yasuo-ride-the-wind-hold-curve-turn-five is a mono-Chaos CHAIN at T5, so it fits Chaos/Fury too and beats its T6 baseline on the clock; it pays on T5, so the clocked column, which counts pay by T4, stays at 0.]

4. The six entries whose priced set is wider than their own cards

The clock folds fuel and payoff entries into a row, and for six finishers that widens the domains. They count in a pair by their own cards in the "all four" column and by the priced set in the "pay by T4" columns:

Read by own cards alone, bottled-constellation-time-warp would count as paying by T4 in all five Mind pairs. It cannot: its T4 is bought with the output of two Mind/Order loops. [2026-10-31, #529: at 07a93ad dragonstorm-confront-grand-plaza clocks T7, not T12; the other five turns are unchanged, and the six are still the six of 89 (§2).] [2026-10-31, #537: at aec0933 they are five of 90. #528 prices ready-recruits-grand-plaza by its own Order cards, so it is no longer wider; dragonstorm-confront-grand-plaza clocks T5 (#532, #544); the other four turns are unchanged.]

5. Stop rule

The #200 decision switches to the adversarial pass after three consecutive pairs with no line that beats its baseline.

The next pair in the order is Body/Calm (baseline T5). [2026-10-27, #496: Body/Calm was slice 2; the next pair not yet worked is Body/Chaos (baseline T5), §2.] [2026-10-27, #512: Body/Chaos was slice 3; the next pair not yet worked is Body/Fury (baseline T5), §2.] [2026-10-27, #524: Body/Fury was slice 4; the next pair not yet worked is Body/Order (baseline T5), §2.] [2026-10-31, #529: Body/Order was slice 5; the next pair not yet worked is Chaos/Order (baseline T6), §2.] [2026-10-31, #537: Chaos/Order was slice 6; the next pair not yet worked is Calm/Order (baseline T5), §2.]

Slice 2, Body/Calm (#496). One line was walked: two Frisky Hunters on the Plaza. Two Energy-1 bodies take The Grand Plaza on T2, a Frisky Hunter and its Bird join them on T2 and another on T3 with a Mutated Mouser: seven units for twelve Energy and no Power, and the T4 Hold wins. That is T4 on the clock against the Body/Calm Hold baseline of T5, so it enters as a candidate. In the walk, against the same deck and draws as slice 1's Plaza play, the Mouser's [Tank] raises the price of killing one of the seven from 1 Might to 3, and the opponent's two T2 bodies attack with 6 anyway: 3 to 3 after their T4. It is castable by T3 in 0.56% of games. No other line was walked: the other Body/Calm cards that put two or more units on the board by T3 are Flurry of Feathers, already catalogued in flurry-of-feathers-grand-plaza-win, and Desert's Call, which makes units at the Hunter's rate as a spell and rests on [Repeat] (the play's Leads). The pair's two catalogued BURSTs pay at T6 on the clock, and no new BURST or CHAIN was searched, so its clocked count stays at 0.

Slice 3, Body/Chaos (#512). One line was walked: a Towering Combatant on the Plaza. Two Determined Sentries take The Grand Plaza on T2 and Towering Combatant, a [Tank] that is 5 Might as a defender, is played onto it. On T3 Up from the Deep, a third Sentry and Pit Rookie, whose buff makes the Combatant 6, make seven: twelve Energy, no Power, and the T4 Hold wins. That is T4 on the clock against the Body/Chaos Hold baseline of T5, so it enters as a candidate, and its kill price is 6 where slice 1's was 1 and slice 2's 3. In the walk, against the same deck and draws as the two earlier Plaza plays, their T3 attack with exactly 6 kills the Combatant and nothing else, so the T4 Hold sees six. The [Flow] casting and a second Combatant rebuild to nine behind a new 6-Might Tank, their 6 can again kill only it, and the T5 Hold wins the game: the first identity-first Plaza walk to win, on the baseline's own turn. It is castable with its Tank by T3 in 0.36% of games. No other line was walked: the pair's other cheap multi-unit card is Up from the Deep itself, already the slice-1 entry, and its two catalogued CHAINs pay at T7 and T9 on the clock. No new BURST or CHAIN was searched, so its clocked count stays at 0.

Slice 4, Body/Fury (#524). One line was walked: a hidden Marai on the Plaza. Body/Fury has no card at Energy 4 or less that makes two units for one card, so going first the seven cards seen by T3 have to be the seven units, and under a legend that adds no Energy seven with Towering Combatant among them cost 13 Energy against 12. Mischievous Marai (Fury) is hidden at the Plaza on T2 for one Power from a rune already exhausted for Energy, and played from there for no Energy: two Sentries on T1, the Combatant and the hide on T2, a third Sentry, Pit Rookie and Pouty Poro on T3. Seven units, 11 Energy and 1 Power, kill price 6. That is T4 on the clock against the Body/Fury Hold baseline of T5, so it enters as a candidate. In the walk, against the same deck and draws as the three earlier Plaza plays, their two 3-Might attackers kill only the Combatant; Marai, played inside that showdown, deals 2 and kills neither, and the T4 Hold sees six. The hand is empty after T3, so there is no rebuild, and it does not win by T5. It is castable with its Tank on the board by the end of T3 in 0.0011% of games (0.0077% counting a T3 hide that needs the opponent's window). No other line was walked: the pair's catalogued CHAINs and BURSTs pay at T5 or later on the clock. The same shell under the other Fury/Body legend, Butcher of the Sands, pays the thirteenth Energy with no Marai and is about fifteen times as castable (0.0160%); it is a Lead of the play, not walked. No new BURST or CHAIN was searched, so its clocked count stays at 0.

Slice 5, Body/Order (#529). One line was walked: Galio and a Vanguard Captain on the Plaza. Order has both things the Body slices lacked: a [Tank] that is 6 Might for three Energy, Galio, Indefatigable, and a card that is three units, Vanguard Captain, once another card has been played that turn. Two Determined Sentries take The Grand Plaza on T2, and Galio and a third Sentry are played onto it; on T3 Pit Rookie buffs Galio to 7 and turns the Captain's [Legion] on. Eight units for 11 Energy and 2 Order Power, both Powers runes that had already paid their Energy, and two kills cost 8 Might. That is T4 on the clock against the Body/Order Hold baseline of T5, so it enters as a candidate. In the walk, against the same deck and draws as the four earlier Plaza plays, their two 3-Might attackers bring 6, all of it must go to Galio, he survives, both of theirs die, and the T4 Hold wins the game: the first identity-first Plaza walk to win on T4. Eight with Galio by the end of T3 happen in 0.0116% of games and seven with Galio in 0.4405%; the two Order Powers fail on 5.63% of six-and-six rune splits. No other line was walked: the pair's other cheap multi-unit cards, Faithful Manufactor and Royal Guard, make one token each, and the Manufactor shape is already catalogued. No new BURST or CHAIN was searched, so its clocked count stays at 0.

Slice 6, Chaos/Order (#537). One line was walked: Yasuo fires on turn four off the Hold curve. This slice had to search a BURST or CHAIN shape, and none in the pair reaches 8 by T4: the T4 ceiling is the Hold curve's 5, plus one Yasuo point, plus Power Nexus's 1, which is 7. Shapes checked: Yasuo, Windrider with Ride the Wind (a point per copy per turn); Shen, Leader of the Kinkou Order (E6 P2, T5 at best); Power Nexus (four rainbow Power at a Hold, one game in three); Draven, Audacious (needs a combat won); the two catalogued Ivern CHAINs (T6, T7); Reckoner's Arena and Karthus, Eternal. The CHAIN that pays is the Yasuo one, on T5. One Yasuo played to base on T3 shuttles base to A, A to base with Ride the Wind, which readies him, and base to A again on T4. He starts T5 at A, so T5 runs the mirror, A to base, Ride the Wind to A, and A to base. The point is a 194.1.c gain, so 470 does not cap it. On top of the Hold curve that is 6 on T4 and 8 on T5. The Chaos/Fury shuttle play had the same cards at T6 because it fired three Yasuos together. That is T5 on the clock against the Chaos/Order Hold baseline of T6, so it enters as a candidate CHAIN. In the walk, against the same deck and draws as the Galio walk, their free T3 attack takes the second battlefield back while Yasuo has used the T3 mana, and it wins on T6, against the free curve's T8 on the same board. Like the slice-4 entry, the walk loses to the clock. The cheapest break is Hidden Blade, in their own identity: hidden for one Power, played for no Energy. On curve it happens in 6.76% of games with nine Energy-2 bodies. Its entry is mono-Chaos and adds a CHAIN to every Chaos pair, but it pays on T5, so the pair's clocked count stays at 0.

6. Not verified

The table counts entries, not decks. An entry fits a pair if its cards do; whether a real list in that pair would run it is not asked. The clock is the optimistic lower bound its own header states: perfect draws, nothing cast against the line, no interaction. It also does not charge an entry's anyBodies, and §3 shows one row where that moves the turn. The other pay-by-T4 rows were not re-priced by hand. SETTLED 2026-10-27: #494 / PR #497 made the clock charge anyBodies. ALT_WIN turns are reported, not ranked.

7. The command

Run from the repository root of a checkout of e3101e5, with coverage.mjs saved anywhere. On a later checkout the table moves: this slice's own entry, up-from-the-deep-flow-grand-plaza, is a mono-Chaos ALT_WIN and adds one to the ALT_WIN and "all four" columns of every Chaos pair. [2026-10-31, #537: §2 now holds one table, this command on aec0933 plus slice 6's entry, and its changelog names the sha and the movers of each earlier measurement.]

npm run -s adversarial -- --turns | node coverage.mjs

// #491: finishers by legend domain pair. stdin = `npm run adversarial -- --turns`.
import { readFileSync } from "node:fs";
const read = (f) => JSON.parse(readFileSync(f, "utf8"));
const dom = new Map(read("data/cards.json").cards.map((c) => [c.base.toUpperCase(), c.domains || []]));
const banned = new Set(read("data/legality.json").entries
  .filter((r) => r.format === "constructed" && r.status === "banned").flatMap((r) => r.bases.map((b) => b.toUpperCase())));
const clock = new Map();
for (const l of readFileSync(0, "utf8").split("\n")) {
  const m = l.match(/^\s+T\s*(\d+) vs T\d baseline\s+(?:SLOWER)?\s+\w+\s+(\S+)\s+(\S+)/);
  if (m) clock.set(m[3], { pays: +m[1], closure: m[2] === "colourless" ? [] : m[2].split("/") });
}
const F = ["INFINITE", "BURST", "CHAIN", "ALT_WIN"];
const fin = read("data/combos.json").combos.filter((e) => F.includes(e.class)).map((e) => ({
  id: e.id, cls: e.class, ...clock.get(e.id),
  own: [...new Set(e.uses.flatMap((u) => dom.get(u.card.toUpperCase())))],
  banned: e.uses.some((u) => banned.has(u.card.toUpperCase())) }));
const D = ["body", "calm", "chaos", "fury", "mind", "order"];
const fits = (ds, p) => ds.every((d) => p.includes(d));
const rows = [];
for (let i = 0; i < 6; i++) for (let j = i + 1; j < 6; j++) {
  const p = [D[i], D[j]];
  const own = fin.filter((f) => fits(f.own, p));
  const n = (c) => own.filter((f) => f.cls === c).length;
  const t4 = (cls) => fin.filter((f) => cls(f.cls) && fits(f.closure, p) && f.pays <= 4 && !f.banned);
  rows.push({ p: p.join("/"), base: p.includes("body") || p.includes("calm") ? 5 : 6,
    inf: n("INFINITE"), burst: n("BURST"), chain: n("CHAIN"), alt: n("ALT_WIN"), all: own.length,
    t4: t4((c) => c !== "ALT_WIN"), altT4: t4((c) => c === "ALT_WIN"), banned: own.filter((f) => f.banned) });
}
console.log("| pair | Hold baseline | INFINITE | BURST | CHAIN | clocked, pay by T4 | ALT_WIN | ALT_WIN, pay by T4 | all four |");
console.log("|---|---|---|---|---|---|---|---|---|");
for (const r of rows) console.log(`| ${r.p} | T${r.base} | ${r.inf} | ${r.burst} | ${r.chain} | ${r.t4.length}${r.t4.length ? ` (${r.t4.map((f) => `\`${f.id}\` T${f.pays}`).join(", ")})` : ""} | ${r.alt} | ${r.altT4.length} | ${r.all}${r.banned.length ? ` (${r.banned.map((f) => `\`${f.id}\` banned`).join(", ")})` : ""} |`);
const order = [...rows].sort((a, b) => a.t4.length - b.t4.length || a.all - b.all || a.p.localeCompare(b.p));
console.log(`\norder: ${order.map((r) => `${r.p} (${r.t4.length}, ${r.all})`).join(", ")}`);
const drift = fin.filter((f) => !fits(f.closure, f.own) || !fits(f.own, f.closure));
console.log(`own cards vs priced closure differ on ${drift.length} of ${fin.length}: ${drift.map((f) => `\`${f.id}\` ${f.own.join("/")} -> ${f.closure.join("/")} T${f.pays}`).join(", ")}`);