A version of AoBot trained for two days competed in AAAI 2009 and broke even.

currently featuring:

• limit Texas Hold'em game engine, with hand history storage
• fast poker hand evaluation
• parsing and analysis of IRC poker histories
• decision tree based opponent modeling (works better than neural networks), predicting actions and showdown outcomes
• suit isomorphic (canonical) betting sequence encoding, which is a bi-directional map from 0..n to all betting sequences within a round
• game tree reduction by grouping similar situations together and treating them as one (bucketing), percentile and hand strength based abstractions are implemented
• heads-up limit near-equilibrium strategy using counterfactual regret minimization
• counterfactual regret minimization for Kuhn Poker