论文信息 - Strategy Effectiveness of Game-Theoretical Solution Concepts in Extensive-Form General-Sum Games

Strategy Effectiveness of Game-Theoretical Solution Concepts in Extensive-Form General-Sum Games

Game theory describes the conditions for the strategies of rational agents to form an equilibrium. However, game theory can fail from the prescriptive viewpoint and can serve only as a heuristic recommendation for agents. There exists a plethora of game theoretic solution concepts, however, their effectiveness has never been compared; hence, there is no guideline for selecting correct algorithm for a given domain. Therefore, we compare the effectiveness of solution-concept strategies and strategies computed by Counterfactual regret minimization (CFR) and Monte-Carlo tree search in practice. Our results show that (1) CFR strategies are typically the best, and (2) the effectiveness of the refinements of NE depends on the utility structure of the game.

Branislav Bosanský | Nicola Gatti | Jiri Cermak

[1] R. McKelvey,et al. Quantal Response Equilibria for Normal Form Games , 1995 .

[2] Peter Bro Miltersen,et al. Computing a quasi-perfect equilibrium of a two-player game , 2010 .

[3] Mark H. M. Winands,et al. Nested Monte-Carlo Tree Search for Online Planning in Large MDPs , 2012, ECAI.

[4] Jan Willemson,et al. Improved Monte-Carlo Search , 2006 .

[5] E. Damme. A relation between perfect equilibria in extensive form games and proper equilibria in normal form games , 1984 .

[6] Michael H. Bowling,et al. A New Algorithm for Generating Equilibria in Massive Zero-Sum Games , 2007, AAAI.

[7] D. Koller,et al. Efficient Computation of Equilibria for Extensive Two-Person Games , 1996 .

[8] Branislav Bosanský,et al. Practical Performance of Refinements of Nash Equilibria in Extensive-Form Zero-Sum Games , 2014, ECAI.

[9] Pierre Hansen,et al. A New Sequence Form Approach for the Enumeration and Refinement of All Extreme Nash Equilibria for Extensive Form Games , 2009, IGTR.