Author: J. Wesley Cleveland
Date: 10:39:34 01/16/02
Go up one level in this thread
On January 16, 2002 at 07:41:28, Graham Laight wrote: >It has occurred to me that it is wrong to evaluate a position in terms of >relative pawns (the "de facto" standard - whereby an evaluation of 2 means that >you're approximately the equivalent of 2 pawns ahead). > >This means that many aspects of evaluation have to be squeezed into a dimension >which is not appropriate at all. > >A better way would be to evaluate "winning probability". If a position was a >draw, the value would be 0.50 (or 50%). If the player should win 3 out of 4 >times, the eval should be 75%. If the player must win from here, then the >evaluation should be 100%. > >It seems strange when you think about it that all programmers have chosen to >adopt the traditional "pawn equivalence" standard. > >-g A few more thoughts on this. 1. You need both probability of winning and probability of drawing, e.g. 40% winning and 50% drawing is better than 50% winning and 0% drawing. 2. You need some way of distinguishing between won positions as to which are closer to checkmate, or your program could just shuffle pieces in a won position. 3. This would make it clear that value of material varies with the position, e.g. if you are a rook ahead, it would be stupid to make a move that wins a pawn but might get you mated, while if you are a pawn behind, it may be your best chance. 4. It makes it clear that evaluations are inexact estimates, and allows statistical and Monte Carlo techniques to be applied to them. 5. Probability of winning or drawing is the only relevant information needed to evaluate chess positions. Material values are a heuristic developed to help humans, especially inexperienced humans evaluate chess positions for probability of winning or drawing. Computers don't need this second step.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.