Error Minimizing Minimax Avoiding Search Pathology in Game Trees
نویسندگان
چکیده
Game-tree pathology is a phenomenon where deeper minimax search results in worse play. It was was discovered 30 years ago (Nau 1982) and shown to exist in a large class of games. Most games of interest are not pathological so there has been little research into searching pathological trees. In this paper we show that even in non-pathological games, there likely are pathological subtrees. Further, we introduce error minimizing minimax search, an adaptation of minimax that recognizes pathological subtrees in arbitrary games, and cuts off search accordingly (shallower search is more effective than deeper search in pathological subtrees). Finally, we present experimental studies of error minimizing minimax in two different games. In our experiments, error minimizing minimax outperformed minimax, sometimes substantially, and never exhibited pathological characteristics.
منابع مشابه
A study of decision error in selective game tree search
In this paper, we study decision errors caused by the omission of part of the legal candidate moves and the inaccuracy of static evaluation in a selective minimax game tree search. Error upper bounds are presented in Section 2. A simple game tree model, which captures some basic characteristics of the Go game tree, is introduced in Section 3 for a decision error simulation study. Section 4 pres...
متن کاملOn Minimax Game Tree Search Pathology and Node-Value Dependence
Here we are concerned with the mysterious phenomenon of minimax game-tree search pathology, where it appears and how it happens. It is commonly believed that the strategy of searchingdeeper for computer game-playing programs can enhance the accuracy of position evaluation and increase the possibility of detecting the correct move. The strategy has been successfully implemented in practice; but ...
متن کاملBias and pathology in minimax search
This article presents the results of experiments designed to gain insight into the effect of the minimax algorithm on the error of a heuristic evaluation function. Two types of effect of minimax are considered: (a) evaluation accuracy (are the minimax backed-up values more accurate than the heuristic values themselves?), and (b) decision accuracy (are moves played by deeper minimax search bette...
متن کاملWhy Minimax Works: An Alternative Explanation
In game-playing programs relying on the minimax principle, deeper searches generally produce better evaluations. Theoretical analyses, however, suggest that in many cases minimaxing amplifies the noise introduced by the heuristic function used to evaluate the leaves of the game tree, leading to what is known as pathological behavior, where deeper searches produce worse evaluations. In most of t...
متن کاملConfidence Bound Algorithms in Game Trees Effective Monte Carlo Tree Search through Bernstein’s Inequality and the Law of the Iterated Logarithm
Game trees have been utilized as a formal representation of adversarial planning scenarios such as two-player zero-sum games like chess [1, 2]. When using stochastic leaf values based on Bernoulli trials to model noisy game trees, a challenging task is to solve the Monte Carlo Tree Search (MCTS) problem of identifying a best move under uncertainty. Confidence bound algorithms are investigated a...
متن کامل