The Bots

Every opponent at the table is a computer player with its own style. Studying how they play is a practical way to learn to recognize — and beat — the same patterns at real tables.

How they think

• Fair by design: bots never see your hole cards or the deck. They decide using only their own cards, the board, the pot, stack sizes and the actions everyone has made in public.
• Hand strength: each decision estimates how often their hand wins by simulating many possible run-outs against likely opponent hands.
• Personality: one shared decision engine runs a different set of numeric tendencies for each character — how loose, how aggressive, how often they bluff, how big they bet.
• Adapting to you: during a session, bots track what every player does in public (how often they raise, fold to bets, go to showdown) and adjust. Nothing is carried over between sessions.

The roster

Jimmy Rebuy — the maniac

A successful tradesman who fires off bets and raises constantly. Occasionally wins a big stack; often reaches for his extra rebuy. Lesson: let him bluff into your strong hands.

George H. Devol — the riverboat gambler

Self-proclaimed king of the riverboat gamblers. Loves big pots and big swings. Lesson: value bet thinly; he pays off.

Daryl — loose and passive

A clown who deals when he's broke and plays a sneaky semi-pro game when he's focused. Calls a lot, raises less. Lesson: when he finally raises, respect it.

Mike — loose-aggressive

There's always a Mike. He plays plenty of hands and plays them hard. Lesson: widen your calling range with strong hands and pick your spots.

Doc Holliday — the nit

Patient and very selective. When he puts money in, he usually has it. Lesson: steal his blinds and fold to his big bets.

Greg — the professor

Thoughtful and tricky, known for check-raise traps out of position. Lesson: be wary of betting medium hands into his checks.

Keith — tight-aggressive

The big-picture planner. Solid fundamentals and well-timed aggression make him one of the toughest players at the table.

Kal — the disciplinarian

The most balanced player: mixes value and bluffs so he's hard to read. Lesson: play solid, avoid predictable patterns.

Friendships at the table

Jimmy, Daryl and Greg run together and tend to avoid betting into each other — but not always. This behavior isn't scripted by name; it emerges from how each one reads the others' public play.

Difficulty modes

• Easy: personalities play more on raw hand strength with more randomness and bigger leaks.
• Medium: personalities think in ranges and adapt more to observed tendencies.
• Hard (coming soon): currently disabled while we build a stronger foundation.

Hard mode roadmap: Project Khan

Hard mode is being rebuilt around a game-theory approach called Counterfactual Regret Minimization (CFR) — the family of algorithms behind the strongest poker AIs.

• Blueprint strategy: trained offline over millions of simulated hands, grouping similar situations into about 200 post-flop "buckets" so the strategy stays compact.
• Algorithm research: comparing CFR+, Discounted CFR and Monte Carlo sampling variants to pick the fastest, most stable trainer.
• In-game search: on the turn and river, a fast solver compiled to WebAssembly refines the blueprint for the exact situation, running in the background so the table stays smooth.
• Rigorous testing: every change is measured over hundreds of thousands of hands on fresh deals, since short samples in no-limit Hold 'em are mostly luck.
• Still personal: Hard mode opponents keep their personalities on top of the stronger core, and everything still runs offline on your device.