One way to create a champion level poker agent is to compute a Nash Equilibrium in an abstract
version of the poker game. The resulting strategy is then used to play in the full game. With this
approach, translation is required between the full and abstract games in order to use the abstract
strategy. In limit poker this translation step is defined when the abstraction is chosen. However,
when considering no-limit poker the translation process becomes more complicated. We formally
describe the process of translation and investigate its consequences. We examine how the current
method, hard translation, can result in exploitable agents and introduce a new probabilistic method,
soft translation, that produces more robust players. We also investigate how switching between
strategies with different underlying abstractions affects the performance of an agent.
Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:AEU.10048/508 |
Date | 11 1900 |
Creators | Schnizlein, David |
Contributors | Bowling, Michael, Szafron, Duane, Schaeffer, Jonathan (Computing Science), Kolfal, Bora (Business) |
Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
Language | en_US |
Detected Language | English |
Type | Thesis |
Format | 391606 bytes, application/pdf |
Relation | David Schnizlein, Michael Bowling, and Duane Szafron. Probabilistic state translation in extensive games with large action sets. In Proceedings of the Twenty-first International Joint Conference on Artificial Intelligence (IJCAI-09), pages 278–284, 2009. |
Page generated in 0.0019 seconds