License the Baccarat Engine
A punto banco engine whose correctness is proven by exhaustive enumeration, not asserted. Rust core, WASM boundary, runs in a browser or on a server.
Why this one
Most casino-game engines ask you to trust their paytables. This one ships the proof, and you can re-run it yourself in about three seconds.
- Every reachable coup, enumerated. All 1,659,001 of them, from a fresh 8-deck shoe, each with its exact hypergeometric weight — no Monte Carlo, no error bars.
- Self-checking. Leaf probabilities sum to 0.999999999978, and the outcome frequencies reproduce the canonical 8-deck figures to six decimals: Player 44.624661%, Banker 45.859742%, Tie 9.515597%.
- Eleven side bets, all within 0.0033pp of published analysis — pairs, Dragon 7, Panda 8, both Dragon Bonus sides, and the full Tiger family.
- Cross-checked against regulator sources (Nevada GCB Rules of Play, NJAC 13:69F). That check is what caught a retired 35:1 Tiger Tie paytable and corrected it to 45:1.
- Reproducible hands. The shoe RNG is pinned by name, so a seeded shoe is bit-identical across dependency updates — an auditor can replay any hand.
- 149 engine tests plus a seeded statistical suite.
What's included
- The rules engine: full third-card tableau, naturals, commission and commission-free (EZ) settlement, and eleven side bets.
- All five pit scoreboards — Bead Plate, Big Road, Big Eye Boy, Small Road, Cockroach Pig.
- Optional presentation layer: the card squeeze, and a teaching dealer that explains each drawing decision by rule.
- A multiplayer table server, if you want live seats rather than a single-player module.
The code is MIT — here's what you'd actually be paying for
The engine is MIT licensed, so you can ship it commercially today, for free, with no permission needed. What teams pay for is the work around it:
- Rule variants and integration. Super 6, no-commission toggles, alternate paytables, a different settlement model, or wiring the engine into your platform. Fixed-scope, fixed-price.
- A compliance audit report. The enumeration results and the regulator cross-check, written up for a compliance or certification reader, with the reproduction steps so your auditor can verify it independently. Useful precisely because "we ran some tests" doesn't survive that conversation.
- A support window. A named contact for correctness questions, and fixes prioritised against your release schedule.
- Custom builds. Additional side bets, other games built to the same standard of proof, or a branded reskin of the whole table.
MIT means you never have to ask. It also means nobody else is going to answer your auditor's questions.
Get in touch
Open an issue or start a discussion on the GitHub repository with what you're building and which tier fits, and you'll get a reply.