Flipping Poker
AI Integration into Processes
Challenge
Heads-up poker has well-known balance and decision-depth problems at the advanced level. The design challenge was to create a variant that keeps the strategic depth of Hold'em, adds decision density closer to Omaha, and introduces a mechanic that reshuffles read-ranges between streets without adding complexity for complexity's sake.
What we built
We built the game as a rule file on top of the Turn game engine — declarative rules, legal move enumeration per state, win conditions. The three-board flipping mechanic was designed to create asymmetric information resets between betting rounds. Pluggable bot types (random, heuristic, GTO solver) play thousands of self-play games automatically to measure luck factor, balance, and game length after each rule change. An invite-only playable MVP prototype is live.
Result
The variant produces faster, more bluff-heavy games than heads-up Hold'em with deeper per-street decisions than Omaha. Automated metrics confirm the flipping mechanic increases decision variance without wrecking game length. The invite-only MVP prototype is playable and in testing.
Tech used
- Declarative game rule engine (Turn)
- Random, heuristic, and GTO bot types
- Automated luck/balance/duration metrics
- Thousands of self-play simulations per rule change
- Invite-only MVP prototype
delivered within 7 days