On self-play computation of equilibrium in poker

TL;DR

This paper compares genetic algorithms and counterfactual regret minimization in computing near-equilibrium strategies for simplified poker games, analyzing their performance against analytical Nash equilibria and deviations.

Contribution

It provides a comparative analysis of two algorithms for equilibrium computation in simplified poker, highlighting their effectiveness and limitations.

Findings

Counterfactual regret minimization outperforms genetic algorithms in accuracy.

Both algorithms approximate Nash equilibria in simplified poker.

Performance varies when facing opponents deviating from equilibrium.

Abstract

We compare performance of the genetic algorithm and the counterfactual regret minimization algorithm in computing the near-equilibrium strategies in the simplified poker games. We focus on the von Neumann poker and the simplified version of the Texas Hold'Em poker, and test outputs of the considered algorithms against analytical expressions defining the Nash equilibrium strategies. We comment on the performance of the studied algorithms against opponents deviating from equilibrium.