{ "schema": "s2s-bench-v1", "exhibit": "chair-trials", "published_utc": "2026-08-13", "status": "PUBLISHED — live at research.strata2signal.com.", "licence": "CC BY 4.0", "attribution": "strata→signal research, research.strata2signal.com", "hardware": "One 96G VRAM workstation, which also answered live requests for three of our apps throughout these trials — two public, one serving playtests for humans and agents alike. Contention is measured in both directions and published with the figures rather than assumed away.", "dataset": "chair-trials-judge-cases", "what_this_is": "All twenty-one judge cases, whole: the claim as the model saw it, the rulebook span it was asked to check the claim against, and the expected verdict. Twelve seeded defects across five classes and nine true claims, over eleven games.", "artifact": "judge-c1.json", "sha256": "fede4e32154f92ac81ae68709fbde2cf1f91f3d1c76e7c83941023d986919702", "frozen_at": "2026-08-12T09:05-06:00", "n": 21, "n_kill": 12, "n_preservation": 9, "verify_receipt": "canary catch 6/6, 0 defects in frozen 21 (verification-report.json)", "verification": "Adversarially verified before freezing: three independent lenses per case, six deliberately planted defective canaries, and a four-case holdout to a differently prompted lens. The fan caught six of six canaries and found zero defects in the frozen twenty-one. A verification that catches none of its own canaries is a verification that proves nothing, which is why the canary count publishes with the set.", "source": "Foster's Complete Hoyle, 1914 — public domain in the US. Project Gutenberg #53881, PG boilerplate stripped: the header carries PG's trademark terms, the 1914 text does not. Text sha256 7927555a…", "limits": [ "Synthetic entailment over public-domain text is a weaker instrument than a human-verified field key. It is audit-ready precisely because every case is here.", "The defects are ours. A defect class we did not think to seed is a class this instrument cannot see." ], "cases": [ { "id": "c1-p-bez-01", "game": "Bézique", "category": "preservation", "claim_text": "Bézique is played with a combined deck of 64 cards.", "quote": "_=CARDS.=_ Bézique is played with two packs of thirty-two cards each,\nall below the Seven being deleted, and the two packs being then\nshuffled together and used as one. It is better to have both packs\nof the same colour and pattern, but it is not absolutely necessary.\nThe cards rank, A 10 K Q J 9 8 7; the Ace being the highest, both in\ncutting and in play.", "quote_line_ranges": [ [ 20453, 20459 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; BÉZIQUE, OR SIXTY-FOUR-CARD BINOCLE — CARDS; source lines 20453-20458", "notes": "Strict entailment by arithmetic only: two packs of thirty-two cards each, shuffled together and used as one, is 2 x 32 = 64. No fact outside the quote is needed. The quote's other content (rank order, colour of the backs) is irrelevant distractor material.", "spare": false }, { "id": "c1-p-six-01", "game": "Sixty-six", "category": "preservation", "claim_text": "In Sixty-six the Ace of a suit counts 11 and the Ten counts 10, so the two together count 21.", "quote": "A player may reach sixty-six by winning tricks containing certain\ncounting cards; by holding and announcing marriages, which are the King\nand Queen of any suit; and by winning the last trick.\n\nThe various counts for these are as follows:--\n\n For King and Queen of trumps, _=Royal Marriage=_, 40\n For King and Queen of any plain suit, _=Marriage=_, 20\n For the Ace of any suit, 11\n For the Ten of any suit, 10\n For the King of any suit, 4\n For the Queen of any suit, 3\n For the Jack of any suit, 2\n For the last or twelfth trick, 10", "quote_line_ranges": [ [ 22074, 22088 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; SIXTY-SIX — counting table; source lines 22074-22087", "notes": "Strict entailment by arithmetic over two rows of the quoted table: Ace 11 + Ten 10 = 21. The claim is deliberately confined to the counting values the span states; it does not assert HOW those cards are won, because the quote's first sentence covers tricks, marriages and the last trick together and does not attach a mechanism to a single row. Shares its span with c1-k-six-w1 so the judge is tested on the span, not on topic priors.", "spare": false }, { "id": "c1-p-nap-01", "game": "Napoleon", "category": "preservation", "claim_text": "A table of three players is a legal game of Napoleon, though four is the best number.", "quote": "_=PLAYERS.=_ Any number from two to six can play; but four is the best\ngame. If five or six play it is usual for the dealer to give himself no\ncards.", "quote_line_ranges": [ [ 17117, 17120 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; NAPOLEON, OR NAP — PLAYERS; source lines 17117-17119", "notes": "Strict entailment by instantiating a stated range: 'Any number from two to six can play; but four is the best game.' Three lies inside two-to-six. Requires a small inference (range -> member), not recall.", "spare": false }, { "id": "c1-p-piq-01", "game": "Piquet", "category": "preservation", "claim_text": "A Piquet player who ends the game with 48 points is lurched and loses a double game.", "quote": "_=STAKES.=_ Piquet is played for so much a game of 100 points; but if\nthe loser has not reached 50 points he is lurched, and loses a double\ngame.", "quote_line_ranges": [ [ 24840, 24843 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; PIQUET — STAKES; source lines 24840-24842", "notes": "Strict entailment by numeric comparison: the quote sets the lurch threshold at 'has not reached 50 points', and 48 < 50. Pairs deliberately with c1-k-bez-h1, where the same double-game consequence is hedged rather than ruled.", "spare": false }, { "id": "c1-p-dom-01", "game": "Dominoes", "category": "preservation", "claim_text": "In dominoes the five suit holds 56 pips, seven more than the four suit's 49.", "quote": "_=The Suits.=_ All the dominoes with the same number upon either end\nbelong to the same suit; the seven bones with a 4 forming the 4 suit;\nthose with a 6 the 6 suit, and so on. The number of pips in each suit\nmay be easily remembered by observing that the ace suit has the same\nnumber of pips as the pieces in the set, 28; and that each suit above\nthe ace has seven pips more than the number of pieces in the set, while\nthe blank suit has seven pips less, so that each suit progresses in\nregular order, seven pips at a time, as shown in the margin.\n\n Blank 21\n Ace 28\n Deuce 35\n Trey 42\n Four 49\n Five 56\n Six 63", "quote_line_ranges": [ [ 32806, 32822 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; DOMINOES — The Suits, with the pip table; source lines 32806-32821", "notes": "Strict entailment: the quoted margin table gives Five 56 and Four 49, and 56 - 49 = 7, which the prose also states ('each suit progresses in regular order, seven pips at a time'). The claim deliberately does NOT say 'double-six set' — the span never names the set, it only implies 28 pieces, and canary c1-can-03 seeds exactly that over-reach as a defect.", "spare": false }, { "id": "c1-p-chk-01", "game": "Chuck-luck", "category": "preservation", "claim_text": "At Chuck-luck the house pays 60 on a total of 4, less than the 71 to 1 odds against that total.", "quote": "The percentage of the house, even in a square game, may be seen from\nthe following table, which gives the odds against the event, and the\nodds which the house pays:--\n\n The odds against 3 or 18 are 215 to 1; the house pays 180\n ” 4 ” 17 ” 71 ” 1 ” 60\n ” 5 ” 16 ” 35 ” 1 ” 29\n ” 6 ” 15 ” 20½ ” 1 ” 18\n ” 7 ” 14 ” 13¼ ” 1 ” 12\n ” 8 ” 13 ” 9¼ ” 1 ” 8\n ” 9 ” 12 ” 7¼ ” 1 ” 6\n ” 10 ” 11 ” 7 ” 1 ” 6", "quote_line_ranges": [ [ 28535, 28547 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; CHUCK-LUCK — house percentage table; source lines 28535-28546", "notes": "Strict entailment by reading one table row plus the lead-in sentence that names the two columns ('the odds against the event, and the odds which the house pays'), then a numeric comparison 60 < 71. The claim says 'pays 60', not 'pays 60 for one' — the for-one convention is stated in a different paragraph that this span does not cover. Shares its span with c1-k-chk-w1.", "spare": false }, { "id": "c1-p-con-01", "game": "Conquian", "category": "preservation", "claim_text": "A Conquian player wins by laying down eleven cards, one more than the ten he was dealt.", "quote": "_=Dealing.=_ Ten cards are dealt to each player in five rounds of two\nat a time, and the twenty remaining in the stock are laid upon the\ntable face down, between the players, but a little to the left of the\ndealer. The stock may be slightly spread, to facilitate the process of\ndrawing cards from it.\n\n_=Objects of the Game.=_The object of each player is to form triplets,\nfours, and sequences, by combining the cards dealt him with others\ndrawn from the stock. These combinations are laid upon the table, face\nup, and the player wins the game who first succeeds in laying down\neleven cards in this manner.", "quote_line_ranges": [ [ 25934, 25945 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; CONQUIAN — Dealing and Objects of the Game; source lines 25934-25944", "notes": "Strict entailment across the two paragraphs in the span: ten cards dealt to each player, and the game won by the first to lay down eleven cards; 11 - 10 = 1. Neither paragraph alone entails the claim, which is why the span covers both.", "spare": false }, { "id": "c1-p-spf-01", "game": "Spoil Five", "category": "preservation", "claim_text": "In Spoil Five the ace of hearts is always one of the four best trumps and always ranks third among them.", "quote": "In the trump suit the same order of cards is retained, except that four\ncards are always the best trumps. These are the Five, Jack, and ace of\nthe suit itself, and the ace of hearts, the latter being always the\nthird best. This gives us the rank of the cards as follows, when the\nsuit is trump:--", "quote_line_ranges": [ [ 17379, 17384 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; SPOIL FIVE — rank of the trump suit; source lines 17379-17383", "notes": "Strict entailment, near-verbatim on the load-bearing words: 'four cards are always the best trumps... and the ace of hearts, the latter being always the third best'. Claim is deliberately kept to what 'always' covers; it does NOT assert which cards stand first and second, because the quote's list order is not stated to be a ranking.", "spare": false }, { "id": "c1-p-bkg-01", "game": "Backgammon", "category": "preservation", "claim_text": "A game of Backgammon is played with thirty men in all, fifteen to each of the two players.", "quote": "Backgammon is played by two persons, each of whom is provided with\nfifteen men, two dice, and a dice-box. The men on each side are of\ndifferent colours, black and white, and the players are distinguished\nby the colour of the men with which they play. The board is divided\ninto two tables, _=inner=_ and _=outer=_, and at the beginning of the\ngame the men may be set up in either of the positions shown in the\ndiagram.", "quote_line_ranges": [ [ 31617, 31624 ] ], "expected_verdict": "PASS", "kill_class": null, "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; BACKGAMMON, OR TRIC-TRAC — opening paragraph; source lines 31617-31623", "notes": "Strict entailment by arithmetic: two persons, each provided with fifteen men, is 2 x 15 = 30. Shares its span with c1-k-bkg-w2 (spare), whose only defect is the dice count in the same sentence.", "spare": false }, { "id": "c1-k-piq-c1", "game": "Piquet", "category": "kill", "claim_text": "In Piquet a trump suit is fixed before the play begins, and trumps beat all plain suits.", "quote": "_=CARDS.=_ Piquet is played with a pack of thirty-two cards, all below\nthe Seven being deleted. The cards rank: A K Q J 10 9 8 7, the Ace\nbeing the highest both in cutting and in play. There is no trump in\nPiquet, and all suits are equal in value. Two packs are sometimes used,\none by each player in his proper turn to deal.", "quote_line_ranges": [ [ 24818, 24823 ] ], "expected_verdict": "FAIL", "kill_class": "contradiction", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; PIQUET — CARDS; source lines 24818-24822", "notes": "Directly contradicted by the quote: 'There is no trump in Piquet, and all suits are equal in value.' No world knowledge can rescue it — Piquet genuinely has no trump. A judge that answers PASS has read the rank order (A K Q J 10 9 8 7) as a trump statement.", "spare": false }, { "id": "c1-k-bez-c1", "game": "Bézique", "category": "kill", "claim_text": "In Bézique a player must follow suit to the card led so long as cards remain in the stock.", "quote": "_=METHOD OF PLAYING.=_ The pone begins by leading any card he chooses,\nto which his adversary may play any card he pleases. A player is not\nobliged to follow suit, nor to trump; but may renounce or trump at\npleasure until the stock is exhausted, after which the method of play\nundergoes a change. If a player follows suit, the higher card wins the\ntrick, and if identical cards are played to the same trick, such as two\nJacks of clubs, the leader wins. Trumps win plain suits. The winner of\nthe trick takes in the cards, turning them face down; but before he\nleads for the next trick he has the privilege of announcing and scoring\nany one of certain combinations that he may hold in his hand. After,\nor in the absence of any such announcement, and before leading for\nthe next trick, he draws a card from the top of the stock and places\nit in his hand, without showing or naming it. His adversary draws the\nnext card, so that each player restores the number of cards in his\nhand to eight. This method of drawing from the stock is open to many\nobjections, and in France the pone always draws first, no matter who\nwins the trick.", "quote_line_ranges": [ [ 20515, 20532 ] ], "expected_verdict": "FAIL", "kill_class": "contradiction", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; BÉZIQUE — METHOD OF PLAYING; source lines 20515-20531", "notes": "Directly contradicted: 'A player is not obliged to follow suit, nor to trump; but may renounce or trump at pleasure until the stock is exhausted.' The claim inverts the quote's own scope clause, so the sentence that would rescue it is the sentence that kills it.", "spare": false }, { "id": "c1-k-cal-c1", "game": "Calabrasella", "category": "kill", "claim_text": "In Calabrasella a misdeal costs the offender the deal, which passes to the player on his left.", "quote": "_=Dealing.=_ The cards are presented to the pone to be cut, and at\nleast four must be left in each packet. The dealer then distributes\nthem to the players four at a time, until each has twelve, four\nremaining in the stock or talon, which is left upon the table face\ndown. No trump is turned. The deal passes to the left.\n\nThere are no misdeals. If the cards are not properly distributed, or\nfour cards are not left in the talon, the same dealer must deal again,\nwithout penalty.", "quote_line_ranges": [ [ 26118, 26127 ] ], "expected_verdict": "FAIL", "kill_class": "contradiction", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; CALABRASELLA — misdeals; source lines 26118-26126", "notes": "Directly contradicted: 'There are no misdeals... the same dealer must deal again, without penalty.' World knowledge cannot rescue it; loss-of-deal is the common rule in neighbouring games, which is exactly the prior this case probes. The span deliberately also carries 'The deal passes to the left', so the second half of the claim is separately true of the normal rotation — a judge that pattern-matches one supported fragment will pass a claim whose main assertion the span denies.", "spare": false }, { "id": "c1-k-con-c1", "game": "Conquian", "category": "kill", "claim_text": "In Conquian a card drawn from the stock is taken into the player's hand until he can use it.", "quote": "_=Method of Playing.=_ The cards dealt, each player sorts his hand into\nsequences and triplets, and determines what cards he wants to complete\nhis runs, so that he may be on the lookout for them. The pone then\ndraws the top card from the stock and turns it face up on the pack. If\nthis card can be used in combination with any of those in his hand, he\ndraws it over to his side of the table, and takes from his hand the\ncards completing the combination of three cards, leaving them all face\nup. Even if he has cards enough in his hand to increase the combination\nto four or more cards, he should not show them. The cards drawn from\nthe stock must never be taken into the hand.", "quote_line_ranges": [ [ 25965, 25975 ] ], "expected_verdict": "FAIL", "kill_class": "contradiction", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; CONQUIAN — Method of Playing; source lines 25965-25974", "notes": "Directly contradicted by the quote's closing sentence: 'The cards drawn from the stock must never be taken into the hand.' The claim describes ordinary rummy practice, so a model answering from genre familiarity rather than from the span will answer PASS.", "spare": false }, { "id": "c1-k-dom-n1", "game": "Dominoes", "category": "kill", "claim_text": "In dominoes, if both players draw dominoes of the same weight and neither is a double, they shuffle and draw again.", "quote": "_=Shuffling and Cutting.=_ Dominoes are provided with a small brass\npin in the centre of the face, which enables one to spin them round,\npush them about on the table, and so to shuffle them thoroughly. There\nare three methods of determining who shall have the first play, or\n_=set=_, as it is called: _=1.=_ The player having the higher double;\nor, failing any double in either hand, the _=heavier=_ domino, that\nis, one with a greater number of pips on its face than any held by his\nadversary. _=2.=_ One player selects any two dominoes, face down, and\npushes them toward his adversary, who chooses one. Both are then turned\nup, and whichever gets the lighter domino has the first set. _=3.=_\nEach player draws a domino, face down, and the one getting the lower\ndouble sets first. If neither draws a double the lighter domino sets.", "quote_line_ranges": [ [ 32823, 32835 ] ], "expected_verdict": "FAIL", "kill_class": "non-entailment", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; DOMINOES — Shuffling and Cutting; source lines 32823-32834", "notes": "The quote enumerates three complete methods of settling the first set and is silent on ties under the third method. Verified unsupported beyond the span too: the DOMINOES chapter (source lines 32793-33044) states no tie-breaking rule for the draw, so world knowledge of this text cannot rescue the claim. UNCERTAIN is an acceptable non-PASS verdict here.", "spare": false }, { "id": "c1-k-eca-n1", "game": "Écarté", "category": "kill", "claim_text": "In Écarté a player who reaches five points before his adversary has scored at all wins a double game.", "quote": "_=MARKERS.=_ In France, the game is always marked with the ordinary\nround chips or counters, never with a marker. As five points is the\ngame, four of these counters are necessary for each player.", "quote_line_ranges": [ [ 16487, 16490 ] ], "expected_verdict": "FAIL", "kill_class": "non-entailment", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; ÉCARTÉ — MARKERS; source lines 16487-16489", "notes": "The quote fixes the game at five points and says nothing about doubling. Verified unsupported beyond the span: the ÉCARTÉ chapter (source lines 16471-17091) contains no double-game or lurch rule at all — its only bonus is the vole, worth two points. Contrast c1-p-eca-02 (spare), which keeps to what this same span does entail.", "spare": false }, { "id": "c1-k-spf-n1", "game": "Spoil Five", "category": "kill", "claim_text": "In Spoil Five the player who takes the first deal also leads to the first trick.", "quote": "_=CUTTING.=_ This is unknown at Spoil Five. The players take their\nseats at random, and one of them deals a card face up to each in\nsuccession. The first Jack takes the first deal. Some note should be\nmade of the player who gets the first deal, as the rules require that\nwhen the game is brought to an end the last deal shall be made by the\nplayer on the right of the first dealer.", "quote_line_ranges": [ [ 17401, 17407 ] ], "expected_verdict": "FAIL", "kill_class": "non-entailment", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; SPOIL FIVE — CUTTING; source lines 17401-17406", "notes": "The quote settles only who deals first and who deals last; it is silent on the opening lead. Verified false beyond the span as well, so world knowledge cannot rescue it: 'The eldest hand begins by leading any card he pleases' (source line 17454) — the eldest hand sits on the dealer's left, not in the dealer's chair.", "spare": false }, { "id": "c1-k-piq-s1", "game": "Piquet", "category": "kill", "claim_text": "In Piquet the twelve cards are always dealt two at a time.", "quote": "_=DEALING.=_ The cards shuffled, they are presented to the pone to be\ncut, and at least two cards must be left in each packet. Twelve cards\nare dealt to each player, either two or three at a time; but whichever\nmethod is first selected must be maintained throughout the game. In\nEngland the cards are always dealt by twos. No trump is turned. The\nremaining eight cards are placed face downward on the table, the five\ntop cards being laid crosswise on the three at the bottom. These eight\ncards are called the _=talon=_ or stock. Each player deals in turn.", "quote_line_ranges": [ [ 24844, 24852 ] ], "expected_verdict": "FAIL", "kill_class": "scope-shift", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; PIQUET — DEALING; source lines 24844-24851", "notes": "Scope-shift, not contradiction: the quote does contain 'always dealt by twos', but bounded — 'In England the cards are always dealt by twos.' The general rule in the same paragraph is 'either two or three at a time'. The claim lifts a national practice to a universal rule while keeping the quote's own word 'always', which is what makes it hard.", "spare": false }, { "id": "c1-k-con-s1", "game": "Conquian", "category": "kill", "claim_text": "The forty-card Conquian pack is made by discarding the King, Queen and Jack of each suit.", "quote": "_=Cards.=_ Conquian is played with the Spanish pack, forty cards, the\n8 9 10 of each suit being deleted. In America, it is much more common\nto play with a pack of forty cards from which the three court cards, K\nQ J, have been discarded, leaving each suit an unbroken sequence from\nthe Ace to the Ten. Some persons play with the full pack, but it spoils\nthe game; as it is then possible to win on a sequence of a single suit.\nThere are no trumps, and the cards have no value as to rank, a sequence\nof 6 7 J being no better than one of 2 3 4. The Ace is not in sequence\nwith the King.", "quote_line_ranges": [ [ 25907, 25916 ] ], "expected_verdict": "FAIL", "kill_class": "scope-shift", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; CONQUIAN — Cards; source lines 25907-25915", "notes": "Scope-shift: the quote's own forty-card pack is the Spanish pack with the 8 9 10 deleted; the K Q J pack is introduced as an American practice ('In America, it is much more common'), not as what the pack is. Both packs are forty cards, so the number in the claim is correct and only the scope is wrong — deliberately kept clear of the wrong-number class.", "spare": false }, { "id": "c1-k-bez-h1", "game": "Bézique", "category": "kill", "claim_text": "In Bézique a loser who has not reached 500 points loses a double game.", "quote": "_=STAKES.=_ Bézique is played for so much a game, 1,000 points up; or\nfor so much a point, the score of the loser being deducted from that of\nthe winner. When a partie of five games is agreed upon, it is usual to\nhave an extra stake upon the odd game, and when three games have been\nwon by the same player, the partie is at an end. It is usual to count\nit a double game if the loser has not reached 500 points.", "quote_line_ranges": [ [ 20474, 20480 ] ], "expected_verdict": "FAIL", "kill_class": "hedge-dropped", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; BÉZIQUE — STAKES; source lines 20474-20479", "notes": "Hedge-dropped: the quote reads 'It is usual to count it a double game if the loser has not reached 500 points.' Threshold, consequence and game are all reproduced correctly; only the modality is strengthened from custom to rule. The paired preservation case c1-p-piq-01 shows the same consequence stated flatly as a rule in Piquet, so a judge cannot pass this by treating 'double game' as inherently hedged.", "spare": false }, { "id": "c1-k-six-w1", "game": "Sixty-six", "category": "kill", "claim_text": "In Sixty-six a royal marriage counts 40 and a marriage in a plain suit counts 30.", "quote": "A player may reach sixty-six by winning tricks containing certain\ncounting cards; by holding and announcing marriages, which are the King\nand Queen of any suit; and by winning the last trick.\n\nThe various counts for these are as follows:--\n\n For King and Queen of trumps, _=Royal Marriage=_, 40\n For King and Queen of any plain suit, _=Marriage=_, 20\n For the Ace of any suit, 11\n For the Ten of any suit, 10\n For the King of any suit, 4\n For the Queen of any suit, 3\n For the Jack of any suit, 2\n For the last or twelfth trick, 10", "quote_line_ranges": [ [ 22074, 22088 ] ], "expected_verdict": "FAIL", "kill_class": "wrong-number", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; SIXTY-SIX — counting table; source lines 22074-22087", "notes": "Wrong-number: the quoted table gives Royal Marriage 40 (correct in the claim) and plain-suit Marriage 20, not 30. One number right and one wrong inside a single sentence, so a judge that stops at the first checkable figure passes a false claim.", "spare": false }, { "id": "c1-k-chk-w1", "game": "Chuck-luck", "category": "kill", "claim_text": "At Chuck-luck the odds against a total of 9 are 9¼ to 1 and the house pays 6.", "quote": "The percentage of the house, even in a square game, may be seen from\nthe following table, which gives the odds against the event, and the\nodds which the house pays:--\n\n The odds against 3 or 18 are 215 to 1; the house pays 180\n ” 4 ” 17 ” 71 ” 1 ” 60\n ” 5 ” 16 ” 35 ” 1 ” 29\n ” 6 ” 15 ” 20½ ” 1 ” 18\n ” 7 ” 14 ” 13¼ ” 1 ” 12\n ” 8 ” 13 ” 9¼ ” 1 ” 8\n ” 9 ” 12 ” 7¼ ” 1 ” 6\n ” 10 ” 11 ” 7 ” 1 ” 6", "quote_line_ranges": [ [ 28535, 28547 ] ], "expected_verdict": "FAIL", "kill_class": "wrong-number", "provenance": "Foster's Complete Hoyle (1914), Project Gutenberg #53881, sha256 7927555a…; CHUCK-LUCK — house percentage table; source lines 28535-28546", "notes": "Wrong-number: the quoted table gives 9 as 7¼ to 1 paying 6; 9¼ to 1 is the row for 8. The payout half of the claim is correct, and the wrong figure is a real value from an adjacent row, so the error survives a careless 'is this number in the table?' check.", "spare": false } ], "contamination_caveat": "Published 2026-08-13. Models with a later training cutoff may have seen these sets. We author fresh sets each cycle; this one is not a standard, it is our kit, yours to reuse.", "not_a_standard": "Our own trials, our own hardware, for our own chairs. Not first independent numbers, and not a benchmark." }