A least-certainty heuristic for selective search

Abstract

We present a new algorithm for selective search by iterative expansion of leaf nodes. The algorithm reasons with leaf evaluations in a way that leads to high confidence in the choice of move at the root. Performance of the algorithm is measured under a variety of conditions, as compared to minimax with α/β pruning, and to best-first minimax.