Public balance record
Bunfight's current mixed-Crew record comes from a deterministic simulator and targeted rules-engine checks. The repaired Always Be Chewing expansion now has its own current rules regression too. This page separates those records from older automated runs and from the human table testing that still has to happen.
Always Be Chewing after repair
The current 98-card expansion source was replayed card by card after its rules repair. All 20 declared actions and all 29 declared mechanics executed, and four fixed-seed deck-building lanes accounted for every valid and invalid attempt.
This proves rules and data coverage. It does not count as a played match, a balance result, a Draft pod, a fun score, or a human session.
Seven mixed Kitchens
The proof used representative unrestricted-Crew Kitchens derived from the Core Set's card-type distribution. It tested Crew mixing; it did not invent a second set of deck-building rules.
Execution coverage
Ordinary games executed effects from 107 unique cards. Separate targeted checks then ran all 420 cards and all 12 structured action families through the rules engine. That distinction matters: 420 cards were not all played in ordinary games.
Method and provenance
The simulator used seven legal representative Kitchens under the current unrestricted Crew-mixing rule. It played every pair 100 times and alternated which Kitchen went first.
Evidence record SHA-256: fc0075cffe7c2a1325e3478b922d40c7ed1d4176d55b1260e12b45ea0d8a5fa6
TCG source commit: 924d7c973f797d01de779d0849ab0c104a8c80f6
Automated evidence archive
These records answer different questions and sometimes reuse the same cards and formats. Do not add them together into one game total.
Current rules evidence
Seven representative mixed-Crew Kitchens played every pair 100 times. All seven overall win rates and the first-player rate stayed inside the project’s 40–58% automated sanity gate.
This checks one representative Kitchen for each Crew. It does not test every possible 40-card Kitchen or prove that a matchup is fun.
Historical automated smoke test
The seven starter Kitchens played every pair 100 times under an earlier automated lane. The first-player rate was 51.0%, with no ties; starter win rates ranged from 42.3% to 54.8%.
Historical deterministic autoplay only. It is not current mixed-Crew evidence, a teaching test, or a record of people learning the game.
Historical automated format test
Six unnamed programmed strategies each ran in Starter, Constructed, Draft, Sealed, and Pack Rush. Each format has its own 1,000-game record.
This lane predates the final unrestricted-Crew Kitchen architecture. Programmed strategies are not people or agent-led playtesters, and this record does not prove drafting skill, table behavior, or enjoyment.
Current rules and package checks
The rules harness validated 420 cards, sampled 500 Play Boosters with no collation failures, built 24 Draft decks and 24 Sealed decks with legal starting Heroes, and ran 60 Pack Rush games.
These checks validate data, collation, deck construction, and rule execution. They do not replace physical card review or a live Draft pod.
Superseded evidence
This earlier constructed matrix recorded a 48.0% first-player rate and zero ties under the former off-Crew restriction.
It is not current Kitchen evidence. Grill Villains finished at 34.2% and Shake Squad at 60.2%, and the off-Crew rule was later removed.
Historical pre-repair expansion evidence
Nine archived Codex-simulated expansion scenarios from May 9, 2026—three constructed, three expansion-heavy, and three limited-format—document historical design testing before the expansion repair. A simplified curve-pressure proxy also attempted 1,200 runs: 1,017 valid and 183 invalid random builds.
These are pre-repair historical design records, not human playtests or current balance approval. The separate current regression checks repaired rules/data contracts, not gameplay.
Current deterministic contract regression
The repaired 98-card Always Be Chewing source executes every declared card, action, and mechanic contract. Four fixed-seed construction lanes account for every valid and invalid build attempt.
This is deterministic rules and data regression, not gameplay, balance, fun, human playtesting, retail collation, or commercial approval.
Printable human playtest sheet
Use your browser's Print command. One completed sheet counts as one session only after people have played and the notes have been kept with the testing record.
Play the current package
The public Print and Play package and the worksheet above are the next useful proof. Record the first confusing moment, funniest moment, slow turn, card-readability problem, and any rule wording that makes two players stop and argue.
