Player Modeling, Search Algorithms and Strategies in Multi-player Games

Abstract

For a long period of time, two person zero-sum games have been in the focus of researchers of various communities. The efforts were mainly driven by the fascination of special competitions such as Deep Blue vs. Kasparov, and of the beauty of parlor games such as Checkers, Backgammon, Othello, and Go.

Multi-player games, however, have been investigated considerably less, and although literature of game theory fills books about equilibrium strategies in such games, practical experiences are rare. Recently, Korf, Sturtevant and a few others started highly interesting research activities. We focused on investigating a four-person chess variant, in order to understand the peculiarities of multi-player games without chance components. In this contribution, we present player models and search algorithms that we tested in the four-player chess world. As a result, we may state that the more successful player models can benefit from more efficient algorithms and speed, because searching more deeply leads to better results. Moreover, we present a meta-strategy, which beats a paranoid α-β player, the best known player in multi-player games.