Kleij, A.A.J. van der (2010) Monte Carlo Tree Search and Opponent Modeling through Player Clustering in no-limit Texas Hold'em Poker. Master's Thesis / Essay, Artificial Intelligence.
|
Text
AI-MAI-2010-A.A.J.van.der.Kleij.pdf - Published Version Download (5MB) | Preview |
Abstract
Texas Hold'em Poker is a challenging card game in which players have to deal with uncertainty, imperfect information, deception and multiple competing agents. Predicting your opponents' moves and guessing their hidden cards are key elements. A computer Poker player thus requires an opponent modeling component that performs these tasks. We propose a clustering algorithm called "K-models clustering" that clusters players based on playing style. The algorithm does not use features that describe the players, but models that predict players' moves instead. The result is a partitioning of players into clusters and models for each of these clusters. Learning to reason about the hidden cards of a player is complicated by the fact that in over 90% of the games, a player's cards are not revealed. We introduce an expectation-maximization based algorithm that iteratively improves beliefs over the hidden cards of players whose cards were not revealed. After convergence, we obtain models that predict players' hidden cards. We built a computer Poker player using these algorithms and Monte Carlo Tree Search: an anytime algorithm that estimates the expected values of moves using simulations. The resulting program can hold its own in 2 player games against both novices and experienced human opponents.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 15 Feb 2018 07:44 |
| Last Modified: | 15 Feb 2018 07:44 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/9370 |
Actions (login required)
 |
View Item |