Adaptive learning for poker
Luigi Barone, Lyndon While · UWA Profiles and Research Repository (University of Western Australia) · 2000
Evolutionary algorithms are more than function optimisers — they adapt and learn in dynamic environments. In this paper, we use this implicit learning characteristic of evolutionary algorithms to create a computer poker player capable of adapting to different opponent strategies. We identify several important poker principles and use these as the basis for a hypercube of evolving populations of poker playing candidates. We simulate the most commonly employed strategies of human players and report experiments that demonstrate the emergent adaptive behaviour of our evolving poker player. In particular, we show that our evolving poker player develops different techniques to counteract the different playing styles employed by its opponents in order to maximise personal winnings. We compare the strategies evolved by our poker player with a competent static player to highlight the importance of adaptation and demonstrate the improved performance of our approach. Finally we discuss a real-world implementation of our model that recently competed in the annual rec.gambling.poker Hold'em poker elimination tournament. Our poker player had some success, winning a few hands, finishing in the top 22% of players.