Poker as a testbed for AI and reinforcement learning (Part 1)

Poker Handevaluation System

This Github project provides a custom implementation of a Texas Hold ’em handevaluation subsystem as described in chapter 5 of the original Loki Poker Bot research paper. For preflop play, fixed relative strength values are provided based upon simplified monte carlo simulations where all players call to the end of the hand. For postflop play, calculations are provided to determine immediate handstrength and future handpotential via complete scenario enumeration (given the opponent hand probability distributions as input).

From this handevaluation system, betting strategies can be derived and opponent models can be added to provide more realistic and adaptive hand probability distributions. The university of Alberta Computer Poker Research Group has performed extensive research on both fronts and has succesfully used Poker as a testbed for AI and reinforcement learning techniques.

If there is sufficient interest in this project then I might open source and document future updates. Feel free to contact me or leave a reply below.