Distributed Computing Collusion in Online Poker Pays Off Bachelor's Thesis Benjamin Zehnder
[email protected] Distributed Computing Group Computer Engineering and Networks Laboratory ETH Z¨urich Supervisors: Tobias Langner, Dr. Thomas Locher Prof. Dr. Roger Wattenhofer June 17, 2012 Abstract In this thesis we describe the implementation of a colluding poker robot which plays online against human players. Our robot uses the graphical user interface of the poker platform to get the information and to interact with the poker platform. We tried to get an advantage through collusion without relying on complicated opponent simulations or a lot of domain knowledge. We let play up to four robots together against human opponents at tables of six players with play and real money. Our results indicate that our robots are playing comparatively at play money tables but bad at real money tables. Our results further indicate that collusion provides a relevant advantage at play and real money tables, which increases with every additional colluding player. In order to win at real money tables, col- luding robots need a more complicated decision making process or more domain knowledge than ours had. Another result shows that human players play more carefully with real money compared to play money. i Contents Abstracti 1 Introduction1 2 Related research3 3 Materials and methods5 3.1 No limit Texas Hold'em.......................5 3.2 Robot design.............................6 3.3 Interface reading and interaction..................7 3.4 Odds calculation...........................9 3.5 Collusion............................... 10 3.6 Strategies............................... 11 3.6.1 Big stack strategy...................... 13 3.6.2 Hutchison strategy.....................