Implementation of a dynamic tree searching algorithm in a chess programme

Abstract

Algorithms searching game trees for games like chess, checkers etc. must have two features: They must find a good solution by inspecting all relevant parts of the tree and they must do it within acceptable time. According to that, most existing algorithms have two principal weaknesses: inconsistency and redundancy. By inconsistency I understand that not all relevant parts of the tree are considered or that parts are considered or pruned for incorrect reasons. Typical inconsistencies are fixed depth limitations, incorrect quiescence analysis and the ordinary methods of forward pruning. Consistency is a sufficient, but fortunately not always necessary condition for correct results.