Using Stereotypes to Improve Early-Match Poker Play
Agent modelling is a critical aspect of many artificial intelligence systems. Many different techniques are used to learn the tendencies of another agent, though most suffer from a slow learning time. The research proposed in this paper examines stereotyping as a method to improve the learning time of poker playing agents. Poker is a difficult domain for opponent modelling due to its hidden information, stochastic elements and complex strategies. However, the literature suggests there are clusters of similar poker strategies, making it an ideal environment to test the effectiveness of stereotyping. This paper presents a method for using stereotyping in a poker bot, and shows that stereotyping improves performance in early-match play in many scenarios.
Keywordsopponent modelling agent modelling stereotypes poker games
Unable to display preview. Download preview PDF.
- 1.Billings, D., Burch, N., Davidson, A., Holte, R., Schaeffer, J., Schauenberg, T.: Approximating game-theoretic optimal strategies for full-scale. In: IJCAI 2003, pp. 661–668 (2003)Google Scholar
- 2.Billings, D., Davidson, A., Schauenberg, T., Burch, N., Bowling, M., Holte, R.: Game-Tree Search with Adaptation in Stochastic Imperfect-Information Games. Computers and Games, 21–34 (2006)Google Scholar
- 3.Johanson, M.B.: Robust Strategies and Counter-Strategies: Building a Champion Level Computer Poker, Masters thesis, University of Alberta (2007)Google Scholar
- 5.Vorhaus, J.: Killer poker online 2: Advanced strategies for crushing the internet game. Kensington Publishing (2006)Google Scholar
- 6.Billings, D., Papp, D., Schaeffer, J., Szafron, D.: Opponent modeling in poker. In: 15th National Conference on Artificial Intelligence (AAAI 1998), pp. 493–498 (1998)Google Scholar
- 7.Billings, D., Kan, M.: A tool for the direct assessment of poker decisions. The International Association of Computer Games Journal (2006)Google Scholar
- 8.Gilpin, A., Sandholm, T.: A competitive texas holdem poker player via automated abstraction and real-time equilibrium computation. In: 21st National Conference on Artificial Intelligence, pp. 1007–1013 (2006)Google Scholar