Best Play for Imperfect Players and Game Tree Search; Part Ii - Experiments
نویسندگان
چکیده
In part 1 we proposed the BPIP algorithm for game playing. BPIP maintains a probabilistic model of its uncertainty and uses it to grow its search tree in the most relevant directions, and to value the tree accounting in a detailed way for the relevance of each leaf. Here we describe the rst implementation of this algorithm and report on experiments comparing it with the standard (alpha-beta, or AB) approach, and the \naive probability update" (NPU) approach, in several popular games. BPIP is seen to evaluate a xed tree more accurately than either AB or NPU in a large variety of games. BPIP is seen to beat strong alpha-beta programs in Othello and Warri even when the alpha-beta programs are given substantially greater computational resources. We have invented several important BPIP-gameplayer engineering tricks in the course of this research, speciically, the \multispike trick," and new methods of evaluation function design. More such tricks undoubtably remain to be discovered.
منابع مشابه
Evolving Adaptive Play for the Game of Spoof
For game playing in general it is important for players to be adaptive, this is particularly true for games where no optimal fixed strategy is known to exist. Adaptive artificial opponents capable of learning and opponent modelling are highly desirable in computer games. Typically, a great deal of a game’s ability to maintain the interest of human players is provided by multiplayer functionalit...
متن کاملIntrinsic vs. Extrinsic Motivation in an Interactive Engineering Game
In this paper, we study intrinsic vs. extrinsic motivation in players playing an electrical engineering gaming environment. We used UNTANGLED, a highly interactive game to conduct this study. This game is developed to solve complex mapping problem from electrical engineering using human intuitions. Our goal is to find whether there are differences in the ways anonymous players solved electrical...
متن کاملStrategic Planning for Imperfect - Information Games Stephen
Although game-tree search works well in perfect-information games, there are problems in trying to use it for imperfect-information games such as bridge. The lack of knowledge about the oppo-nents' possible moves gives the game tree a very large branching factor, making the tree so immense that game-tree searching is infeasible. In this paper, we describe our approach for overcoming this proble...
متن کاملProb-Maxn: Playing N-Player Games with Opponent Models
Much of the work on opponent modeling for game tree search has been unsuccessful. In two-player, zero-sum games, the gains from opponent modeling are often outweighed by the cost of modeling. Opponent modeling solutions simply cannot search as deep as the highly optimized minimax search with alpha-beta pruning. Recent work has begun to look at the need for opponent modeling in n-player or gener...
متن کاملBest Play for Imperfect Players and Game Tree Search; Part I - Theory
1. The point of game tree search is to insulate oneself from errors in the evaluation function. The standard approach is to grow a full width tree as deep as time allows, and then value the tree as if the leaf evaluations were exact. This has been eeective in many games because of the computational eeciency of the Alpha-beta algorithm. But as Bayesians, we want to know the best way to use the i...
متن کامل