09/01/1991 09:00 AM
Computer Science
Chess has long been used as a testbed for Artificial Intelligence research. There are many good chess playing programs available that perform deep search of the game tree. These programs get their chess-playing power from the depth of the search tree rather than from the sophistication of their position- evaluation heuristics. These programs essentially throw away all the information obtained by evaluating the game tree. This document describes Morph, a chess- playing program based on the completely opposite approach. This program searches only one ply deep and learns from game to game. Morph creates patterns that it uses to evaluate positions that occur in future games.
