Comparative Analysis of Extensive Form Zero-Sum Games Solving Algorithms: Evaluating Optimal Algorithms in Poker-Like Games

Abstract The solving of extensive form zero-sum games has been accomplished by the implementation of a number of algorithms, each of which has advantages and disadvantages. The selection of an ideal algorithm is one of the most important challenges, and it has been investigated in this paper by applying the categories of famous algorithms that are currently in existence. Initially, the optimal parameters of each algorithm have been chosen, and then evaluation has been carried out by using Poker-like games. The exploitability and the utility average of the algorithms have been compared via the use of two measures. In general, there are two primary stage involved in comparing algorithms. In the first stage, the algorithms are assessed in the order of Kuhn, Leduc, and Royal Poker while playing in the two-player format. The second stage involves evaluating the Kuhn Poker algorithm with three players, four players, and five players, and then analyzing the findings that result from this evaluation. As a consequence of this comparison, the four algorithms that are considered to be the most optimum are DCFR, RCFR, CFR+, and FSP accordingly.