{ "id": "NM-01", "title": "NOT MINE — A study in passing the blame", "studyDate": "2026-10-13", "timezone": "Asia/Kolkata", "status": "Partial exploratory study; provisional rules; no human participants", "rules": { "rulesVersion": "0.1-test-baseline", "rulesStatus": "provisional-test-baseline", "objectCounts": { "SECRET_PHOTO": 6, "HUSH_MONEY": 6, "SPARE_KEY": 6, "SIGNED_LETTER": 6, "VOICE_RECORDER": 5, "SIGNET_RING": 5, "ALIBI": 2 }, "actionCounts": { "GIFT": 26, "EXCHANGE": 12, "DENY": 6, "REDIRECT": 4 }, "objectsPerPlayer": 5, "handSize": 4, "substituteOnlyWithoutGift": true, "substituteCost": 2, "allowAlibiTransfer": true, "turnCap": 300 }, "sample": { "primaryScripted": 400, "seatRotations": 70, "responseStress": 60, "completedScripted": 530, "invalidatedPilotTurns": 63, "invalidatedPilotDecisions": 111, "correctedPilotTurns": 12, "correctedPilotDecisions": 19, "completedValidAgentGames": 0, "partialAgentFeedback": 4, "formalAgentTablesPlanned": 12, "formalAgentTablesCompleted": 0, "decisionsUsed": 130, "decisionBudget": 1500 }, "definitions": { "group": "Rule hash, control condition, player count, and sample role. Game IDs beginning pilot- are kept in separate pilot groups and excluded from main groups; rotations are separate because they share deals with main games.", "length": "Initiating moveId at termination; cap-hit observations are censored non-wins. Unfinished games excluded from terminal length/rate denominators.", "fairness": "Credit 1/k for k simultaneous winners. Raw winners also reported. Seat 0 is P1; relative seat 0 is the starter. Completed capped games remain non-win assignments.", "bundle": "All offered cards, including ALIBI bundles; mean denominator is fully resolved gifts.", "choices": "Physical legal action count, including interchangeable card IDs. Not a measure of strategic depth; agent-rated non-obvious decisions remain in feedback.", "progress": "Counts and distinct scandalous types after each fully resolved move; setup at move 0. Lead changes compare sets of players tied for minimum scandalous count.", "comeback": "Winner held strictly most / tied most scandalous objects at a prior fully resolved nonterminal move; setup and terminal states excluded.", "kingmaking": "Conservative mechanical opportunity: legal exchange gives own ALIBI for recipient’s final scandalous object. Counts decision windows, not duplicate-ID actions. Does not establish intent or exhaust other forms of help.", "targets": "Original initiating target counts exclude redirects. Consecutive-target streak follows successive initiating moves; same-actor streak follows that actor’s own turns.", "loops": "Hash of complete mechanical state including all hands, deck orders, object zones, active seat, pending and RNG, excluding only monotonically increasing turn/move/reshuffle counters. Latest-spent display and result are retained. No early stopping.", "integrity": "Replays rerun with engine assertions and exact event/hash comparison. Illegal attempts counted from invalid.jsonl; retry/fallback counts require explicit control tags. Unlogged attempts cannot be inferred.", "independence": "Scripted rates describe these reproducible policies, not people or independent LLM players; LLM evidence remains descriptive. No physical timing inferred." }, "primary": [ { "players": 3, "games": 100, "wins": 100, "length": { "n": 100, "min": 4, "median": 19, "p90": 45, "p95": 60, "max": 108, "mean": 23.38 }, "capHit": { "successes": 0, "n": 100, "rate": 0, "wilson95": [ 3.469446951953614e-18, 0.03699349820698568 ] }, "alibiWins": 20, "comebacks": { "winningGames": 100, "winnerPreviouslyStrictMost": 58, "winnerPreviouslyMostIncludingTies": 76 }, "styles": [ { "style": "random-legal", "assignments": 75, "wins": 8 }, { "style": "greedy-shedding", "assignments": 76, "wins": 27 }, { "style": "defensive", "assignments": 76, "wins": 13 }, { "style": "leader-blocking", "assignments": 73, "wins": 52 } ], "starter": { "successes": 29, "n": 100, "rate": 0.29, "wilson95": [ 0.2101483574912227, 0.3853889117557111 ] }, "gifts": { "initiated": 2021, "resolved": 2021, "accepted": { "successes": 1399, "n": 2021, "rate": 0.6922315685304304, "wilson95": null }, "denied": { "successes": 326, "n": 2021, "rate": 0.16130628401781297, "wilson95": null }, "redirected": { "successes": 296, "n": 2021, "rate": 0.14646214745175656, "wilson95": null }, "meanBundleSize": 2.4571994062345373 }, "integrity": { "invalidChoices": 0, "taggedRetries": 0, "fallbacks": 0, "conservationFailures": 0, "replayVerifiedGames": 100, "fullMechanicalLoopGames": 0, "informationLeaks": 0 } }, { "players": 4, "games": 100, "wins": 100, "length": { "n": 100, "min": 5, "median": 20, "p90": 38, "p95": 44, "max": 58, "mean": 22.11 }, "capHit": { "successes": 0, "n": 100, "rate": 0, "wilson95": [ 3.469446951953614e-18, 0.03699349820698568 ] }, "alibiWins": 25, "comebacks": { "winningGames": 100, "winnerPreviouslyStrictMost": 42, "winnerPreviouslyMostIncludingTies": 68 }, "styles": [ { "style": "random-legal", "assignments": 100, "wins": 7 }, { "style": "greedy-shedding", "assignments": 100, "wins": 22 }, { "style": "defensive", "assignments": 100, "wins": 14 }, { "style": "leader-blocking", "assignments": 100, "wins": 57 } ], "starter": { "successes": 28, "n": 100, "rate": 0.28, "wilson95": [ 0.20139685211186212, 0.3748802870992116 ] }, "gifts": { "initiated": 1800, "resolved": 1800, "accepted": { "successes": 1186, "n": 1800, "rate": 0.6588888888888889, "wilson95": null }, "denied": { "successes": 319, "n": 1800, "rate": 0.17722222222222223, "wilson95": null }, "redirected": { "successes": 295, "n": 1800, "rate": 0.1638888888888889, "wilson95": null }, "meanBundleSize": 2.456111111111111 }, "integrity": { "invalidChoices": 0, "taggedRetries": 0, "fallbacks": 0, "conservationFailures": 0, "replayVerifiedGames": 100, "fullMechanicalLoopGames": 0, "informationLeaks": 0 } }, { "players": 5, "games": 100, "wins": 100, "length": { "n": 100, "min": 6, "median": 24, "p90": 44, "p95": 45, "max": 51, "mean": 26.6 }, "capHit": { "successes": 0, "n": 100, "rate": 0, "wilson95": [ 3.469446951953614e-18, 0.03699349820698568 ] }, "alibiWins": 43, "comebacks": { "winningGames": 100, "winnerPreviouslyStrictMost": 44, "winnerPreviouslyMostIncludingTies": 66 }, "styles": [ { "style": "random-legal", "assignments": 125, "wins": 6 }, { "style": "greedy-shedding", "assignments": 125, "wins": 41 }, { "style": "defensive", "assignments": 125, "wins": 14 }, { "style": "leader-blocking", "assignments": 125, "wins": 39 } ], "starter": { "successes": 23, "n": 100, "rate": 0.23, "wilson95": [ 0.15843265880303448, 0.3215438302287378 ] }, "gifts": { "initiated": 2122, "resolved": 2122, "accepted": { "successes": 1325, "n": 2122, "rate": 0.6244109330819981, "wilson95": null }, "denied": { "successes": 411, "n": 2122, "rate": 0.19368520263901978, "wilson95": null }, "redirected": { "successes": 386, "n": 2122, "rate": 0.18190386427898209, "wilson95": null }, "meanBundleSize": 2.502827521206409 }, "integrity": { "invalidChoices": 0, "taggedRetries": 0, "fallbacks": 0, "conservationFailures": 0, "replayVerifiedGames": 100, "fullMechanicalLoopGames": 0, "informationLeaks": 0 } }, { "players": 6, "games": 100, "wins": 100, "length": { "n": 100, "min": 7, "median": 28, "p90": 46, "p95": 50, "max": 61, "mean": 28.98 }, "capHit": { "successes": 0, "n": 100, "rate": 0, "wilson95": [ 3.469446951953614e-18, 0.03699349820698568 ] }, "alibiWins": 53, "comebacks": { "winningGames": 100, "winnerPreviouslyStrictMost": 34, "winnerPreviouslyMostIncludingTies": 58 }, "styles": [ { "style": "random-legal", "assignments": 152, "wins": 4 }, { "style": "greedy-shedding", "assignments": 152, "wins": 24 }, { "style": "defensive", "assignments": 148, "wins": 24 }, { "style": "leader-blocking", "assignments": 148, "wins": 48 } ], "starter": { "successes": 19, "n": 100, "rate": 0.19, "wilson95": [ 0.12514751509768735, 0.27778845379064376 ] }, "gifts": { "initiated": 2274, "resolved": 2274, "accepted": { "successes": 1425, "n": 2274, "rate": 0.6266490765171504, "wilson95": null }, "denied": { "successes": 401, "n": 2274, "rate": 0.17634124890061564, "wilson95": null }, "redirected": { "successes": 448, "n": 2274, "rate": 0.19700967458223395, "wilson95": null }, "meanBundleSize": 2.5338610378188213 }, "integrity": { "invalidChoices": 0, "taggedRetries": 0, "fallbacks": 0, "conservationFailures": 0, "replayVerifiedGames": 100, "fullMechanicalLoopGames": 0, "informationLeaks": 0 } } ], "response": [ { "label": "Ordinary greedy response", "run": "audit-normal-response", "games": 20, "length": { "n": 20, "min": 7, "median": 13, "p90": 20, "p95": 24, "max": 25, "mean": 13.85 }, "capHit": { "successes": 0, "n": 20, "rate": 0, "wilson95": [ 0, 0.16112515805281938 ] } }, { "label": "Always accept", "run": "audit-never-respond", "games": 20, "length": { "n": 20, "min": 7, "median": 11, "p90": 15, "p95": 15, "max": 16, "mean": 11.6 }, "capHit": { "successes": 0, "n": 20, "rate": 0, "wilson95": [ 0, 0.16112515805281938 ] } }, { "label": "Spend defense when available", "run": "audit-always-respond", "games": 20, "length": { "n": 20, "min": 7, "median": 12, "p90": 20, "p95": 21, "max": 30, "mean": 13.45 }, "capHit": { "successes": 0, "n": 20, "rate": 0, "wilson95": [ 0, 0.16112515805281938 ] } } ], "feedback": [ { "player": "P1", "style": "immediate object-shedding", "ratings": { "clarity": 5, "agency": 4, "tension": 4, "repetition": 3 }, "memorableMoves": [ { "moveId": 1, "observation": "My two recorders were redirected from P4 to P3; my count fell from five to three.", "interpretation": "Redirect preserved the sender's progress while changing who suffered the burden." }, { "moveId": 5, "observation": "P2 denied my two-letter gift, leaving my five objects unchanged.", "interpretation": "Target choice and hidden defenses materially affected immediate shedding." }, { "moveId": 12, "observation": "P4 gifted three photos to P2, reducing P4 from six scandalous objects to three.", "interpretation": "Visible matching groups created large potential swings worth tracking outside my own turn." } ], "nonObviousChoices": [ { "moveId": 9, "alternatives": [ "Gift letters to P2", "Gift letters to P4", "Exchange a singleton for P3's ALIBI" ], "reason": "I chose the larger immediate reduction and a recipient lacking letters. Exchanging for ALIBI offered a smaller guaranteed reduction." }, { "moveId": 10, "alternatives": [ "Accept", "Redirect to P3", "Redirect to P4" ], "reason": "Redirecting to P4 kept my count at three and added a new type there; P3 already had a letter bundle." } ], "frictions": [ "Move 2 offered only ACCEPT because my hand lacked defenses; participation did not imply a meaningful response choice.", "Moves 5 and 9 repeated an attempt to shed the same two letters, although the target and resolution differed; no exact repeated turn occurred.", "ALIBI exchanges at moves 3, 4, 6 and 7 repeatedly moved the same ALIBI. This merits observation in a completed game before changing rules." ], "keepUnchanged": [ "Whole matching bundles", "One response window with no redirect chain", "Visible objects and hidden action hands" ], "humanPlaytestQuestions": [ "Do players track opponents' matching groups between turns, or mainly their total counts?", "Does repeated ALIBI movement feel tactical or repetitive?", "Does forced acceptance remain engaging when a player has no defensive card?" ], "partialSession": true, "result": "unfinished" }, { "player": "P2", "style": "defensive resource-saving", "ratings": { "clarity": 5, "agency": 3, "tension": 4, "repetition": 3 }, "memorableMoves": [ { "moveId": 2, "observation": "My two-letter gift was accepted, reducing my scandalous count from four to two.", "interpretation": "Matching bundles provide substantial progress with one action." }, { "moveId": 6, "observation": "I exchanged a recorder for the ALIBI, reducing my count to two; P3 took that ALIBI on move 7.", "interpretation": "Guaranteed reduction was temporary because the ALIBI remained vulnerable to exchange." }, { "moveId": 12, "observation": "After forced acceptance of two recorders on move 11, I had to accept three photos on move 12, reaching seven objects.", "interpretation": "Consecutive incoming gifts can reverse progress without a defensive choice." } ], "nonObviousChoices": [ { "moveId": 5, "alternatives": [ "Accept two letters and retain DENY", "Spend DENY" ], "reason": "I protected my current position. Saving DENY could instead have blocked the later three-photo gift, although subsequent play might differ." }, { "moveId": 10, "alternatives": [ "Gift my lone letter", "Exchange for the ALIBI" ], "reason": "The immediate recapture on move 7 made another ALIBI acquisition less attractive. The gift successfully removed a type despite being redirected." } ], "frictions": [ "ALIBI o35 changed owners through exchanges on moves 3, 4, 6 and 7. This repeated mechanism threatened temporary rather than lasting progress; no identical full turn repeated.", "My ACCEPT responses on moves 11 and 12 were exactly the same action and both forced, although their bundles differed.", "Outside my turns, tracking ALIBI ownership, matching groups and publicly spent responses materially affected decisions." ], "keepUnchanged": [ "Whole matching bundles", "Single response window", "Public objects with private actions" ], "humanPlaytestQuestions": [ "Does repeated ALIBI recapture feel tactical or repetitive?", "Do consecutive forced acceptances sustain attention or reduce perceived agency?", "Do players track distinct object types as well as total counts?" ], "partialSession": true, "result": "unfinished" }, { "player": "P3", "style": "prevent-nearest-opponent-win", "ratings": { "clarity": 5, "agency": 3, "tension": 4, "repetition": 3 }, "memorableMoves": [ { "moveId": 3, "observation": "I traded money for P2's ALIBI, reducing my evidence from seven to six and increasing theirs from two to three.", "interpretation": "An unblockable exchange combined personal progress with delaying the leader." }, { "moveId": 4, "observation": "P4 immediately took my ALIBI in exchange for a letter.", "interpretation": "Holding ALIBI exposed my progress to reversal before my next turn." }, { "moveId": 9, "observation": "P4 redirected two letters to me while I held DENY; my evidence increased from six to eight.", "interpretation": "Off-turn attention mattered, but my defensive card offered no agency against this transfer." }, { "moveId": 11, "observation": "P2 accepted my recorder pair, reducing my evidence to six and increasing theirs to four.", "interpretation": "A matching bundle achieved both catch-up and leader pressure." } ], "nonObviousChoices": [ { "moveId": 3, "alternatives": [ "Gift two photos to P2", "Exchange a singleton for P2's ALIBI" ], "reason": "I chose guaranteed reduction and an additional evidence type for P2 over a larger, blockable transfer." }, { "moveId": 7, "alternatives": [ "Repeat an ALIBI exchange", "Gift a matching pair" ], "reason": "I again prioritized delaying P2, though the previous ALIBI theft demonstrated the vulnerability of that approach." } ], "frictions": [ "Moves 3 and 7 repeated my exchange-for-P2-ALIBI tactic, with different singleton payments; they were strategically repetitive, not identical actions.", "Redirects at moves 1 and 9 loaded me with pairs without a response opportunity. My DENY remained unused.", "The history response was truncated in its middle; earlier assigned observations covered those moves." ], "keepUnchanged": [ "Whole matching-type gifts", "Visible objects with hidden action hands", "Single response window" ], "humanPlaytestQuestions": [ "Do repeated unblockable ALIBI transfers create interesting competition or feel like undone progress?", "Does receiving redirected bundles keep players attentive even when they cannot respond?", "Do players evaluate remaining object types as readily as total evidence?" ], "partialSession": true, "result": "unfinished" }, { "player": "P4", "style": "matching-group consolidation", "ratings": { "clarity": 5, "agency": 4, "tension": 3, "repetition": 3 }, "memorableMoves": [ { "moveId": 4, "observation": "I exchanged my singleton letter for P3's ALIBI. P2 took that ALIBI from me at move 6.", "interpretation": "Guaranteed immediate progress could be reversed before my next turn." }, { "moveId": 8, "observation": "I exchanged money for a third photo.", "interpretation": "Reducing distinct types offered progress even without reducing object count." }, { "moveId": 12, "observation": "P2 accepted my three-photo gift, reducing my scandalous count from six to three.", "interpretation": "The earlier consolidation produced a meaningful bundle transfer." } ], "nonObviousChoices": [ { "moveId": 8, "alternatives": [ "Take ALIBI again", "Gift two photos immediately", "Consolidate three photos" ], "reason": "I chose consolidation after seeing ALIBI repeatedly change owners; the alternatives changed immediate count, exposure, and future gift size." }, { "moveId": 9, "alternatives": [ "Accept letters", "Redirect to P2", "Redirect to P3" ], "reason": "P3 gained a new type, whereas P2 already held letters. Target selection materially changed the burden." } ], "frictions": [ "Moves 3, 4, 6, and 7 repeatedly exchanged a scandalous singleton for the same ALIBI. This was a recurring pattern, not an exact repeated complete turn.", "Outside my turn, ALIBI ownership, spent defenses, and redirect destinations required attention. Move 10 added a letter to me without a response opportunity.", "P2's acceptance at move 11 informed my move 12 target, but acceptance alone did not prove a lack of defense." ], "keepUnchanged": [ "Whole matching-group gifts", "EXCHANGE as a consolidation option", "Single redirect with no response chain" ], "humanPlaytestQuestions": [ "Does repeated ALIBI exchange feel tactical or repetitive over a complete game?", "Do players notice distinct object types as readily as total object count?", "Does forming and transferring a larger matching group justify the setup turn?" ], "partialSession": true, "result": "unfinished" } ], "pilot": [ { "move": 0, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 4, "alibi": true }, { "id": "P3", "count": 5, "alibi": false }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 1, "players": [ { "id": "P1", "count": 3, "alibi": false }, { "id": "P2", "count": 4, "alibi": true }, { "id": "P3", "count": 7, "alibi": false }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 2, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 2, "alibi": true }, { "id": "P3", "count": 7, "alibi": false }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 3, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 3, "alibi": false }, { "id": "P3", "count": 6, "alibi": true }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 4, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 3, "alibi": false }, { "id": "P3", "count": 7, "alibi": false }, { "id": "P4", "count": 4, "alibi": true } ] }, { "move": 5, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 3, "alibi": false }, { "id": "P3", "count": 7, "alibi": false }, { "id": "P4", "count": 4, "alibi": true } ] }, { "move": 6, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 2, "alibi": true }, { "id": "P3", "count": 7, "alibi": false }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 7, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 3, "alibi": false }, { "id": "P3", "count": 6, "alibi": true }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 8, "players": [ { "id": "P1", "count": 5, "alibi": false }, { "id": "P2", "count": 3, "alibi": false }, { "id": "P3", "count": 6, "alibi": true }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 9, "players": [ { "id": "P1", "count": 3, "alibi": false }, { "id": "P2", "count": 3, "alibi": false }, { "id": "P3", "count": 8, "alibi": true }, { "id": "P4", "count": 5, "alibi": false } ] }, { "move": 10, "players": [ { "id": "P1", "count": 3, "alibi": false }, { "id": "P2", "count": 2, "alibi": false }, { "id": "P3", "count": 8, "alibi": true }, { "id": "P4", "count": 6, "alibi": false } ] }, { "move": 11, "players": [ { "id": "P1", "count": 3, "alibi": false }, { "id": "P2", "count": 4, "alibi": false }, { "id": "P3", "count": 6, "alibi": true }, { "id": "P4", "count": 6, "alibi": false } ] }, { "move": 12, "players": [ { "id": "P1", "count": 3, "alibi": false }, { "id": "P2", "count": 7, "alibi": false }, { "id": "P3", "count": 6, "alibi": true }, { "id": "P4", "count": 3, "alibi": false } ] } ], "limits": [ "Scripted policies are not independent LLM players or people.", "Seat rotations share deals with the primary sample; do not pool as independent observations.", "Response conditions share initial states; these change policy, not rules.", "The first agent pilot is excluded because public-action history was incompletely delivered.", "The corrected pilot stopped at the owner's request after 12 turns; no winner was observed.", "Fresh agent assignments and broker audits provide procedural isolation, not a filesystem sandbox.", "P3 disclosed truncated retrospective history output; earlier live views covered those moves.", "Backend model ID was not exposed. No formal rule variants or human tests were completed." ] }