Technical research into poker AI and game theory.
Poker AI · Evaluation · Qualification
Research note 04
Qualification Reliability in Agentic Poker Tournaments
A study of how Swiss rounds, match length, adaptive pairing and duplicate play affect top-six qualification in 60-agent poker tournaments.
Poker AI · Evaluation · Variance
Research note 03
Variance and Sample Efficiency Across Poker Formats
A benchmark of how many hands it takes to tell two poker agents apart across seven formats, and how much of that cost variance reduction can remove.
Poker AI · Evaluation · Benchmark Design
Research note 02
Ranking, Exploitability and Benchmark Design in Poker AI
An exploration of how tournament performance, exploitability and benchmark design can produce different notions of what makes a poker agent strong.
Poker AI · Game Theory · MCCFR
Research note 01
A Delayed-Linear Weighted-Averaging MCCFR Study at 8BB Effective Stacks
A validated 4-max all-in-or-fold strategy produced using delayed-linear weighted-averaging MCCFR, including convergence, exploitability and held-out validation.
FullHouse Research produces original technical research at the intersection of poker AI and game theory, alongside the FullHouse Hackathon.