Dead Hand Poker teaches table reading. Each hand is dealt at a six-seat table against computer opponents with distinct styles, and you rate each opponent, judge bluffs and choose your action before you may look at your own hand. Five curriculum levels add range narrowing, pot odds, EV and regret scoring, and blocker reports.
The poker engine is plain TypeScript with no UI or I/O, so the browser app and a separate REST API run the same code. The API makes the hidden information real: every response passes through one redaction function that withholds the deck, the opponents' cards and your own hole cards until the reveal. Opponent ranges are modeled with card removal in mind, and equity is computed against the modeled range rather than the real cards.
Status A prototype: the frontend runs the engine in the browser and does not yet call the server, and server sessions are in memory with no auth.
Zach’s part Built the engine, the trainer and the API.