PokerBot: Hand Strength Reinforcement Learning

Abstract

We sought to explore the problem of teaching a reinforcement learning agent how to play Texas Hold ‘Em (THE), a popular poker game played with a standard 52-card deck. This is an interesting problem because THE, and poker in general, is an incomplete information game in which the best strategy must take into account a significant amount of uncertainty, and for which the input vector of relevant information could be potentially very large. The final product of our research is a simplistic but elegant application of reinforcement learning, with various approaches yielding promising results within the context of THE.