Abstract
Risk dominance strategy is a complementary part of game theory decision strategy besides payoff dominance. It is widely used in decision of economic behavior and other game conditions with risk characters. In research of imperfect information games, the rationality of risk dominance strategy has been proved while it is also wildly adopted by human players. In this paper, a decision model guided by risk dominance is introduced. The novel model provides poker agents of rational strategies which is relative but not equals simply decision of "bluffing or no". Neural networks and specified probability tables are applied to improve its performance. In our experiments, agent with the novel model shows an improved performance when playing against our former version of poker agent named HITSZ_CS_13 which participated Annual Computer Poker Competition of 2013. The macroscopical results and analysis of details all cofirms the effectiveness of works provided in this paper.
| Original language | English |
|---|---|
| Pages (from-to) | 555-566 |
| Number of pages | 12 |
| Journal | Journal of Information Hiding and Multimedia Signal Processing |
| Volume | 5 |
| Issue number | 3 |
| State | Published - Jul 2014 |
| Externally published | Yes |
Keywords
- Imperfect information
- Poker
- Risk dominance
Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver