King Race

In the last decade many games have been successfully approached by what is commonly known as brute force, i.e., searching as much as possible. This tendency of computer play has been criticized since it differs much from emulating human thoughts which once was one of the primary goals of computer game playing. An alternative way which is much closer to human game playing, is to discover rules automatically for a given game with the aid of Artificial-Intelligence techniques and human thought. Obtaining such rules can be done by building computer game programs. This research line may have a much broader scope (1) to understand better the ideas of a certain game, (2) to investigate how to build automatically learnt policies for a planning problem, and perhaps even (3) to understand a bit better human thinking. To illustrate these three points, we use a small game, that could be called King Race, and we aim at discovering rules to solve it. To discover these rules we use some human thought and a decision tree program. Then we prove mathematically that these automatically obtained rules indeed solve the game. The rules can aid in building computer-chess endgame programs that do not use brute force.