Author: Don Dailey
Date: 21:14:29 01/11/98
Go up one level in this thread
Hi KK, Your method of testing for this is not very precise but is probably about as good as it gets. Since most programs use some kind of trade down bonus you are likely to get a number that is too small. My program values center pawns more than flank pawns and also you tend to create a little positional compensation when you pluck a pawn off the board (like open lines.) It might be useful to do your experiment and consider this a lower bound on the value of a pawn. There is probably also a best pawn to use for this experiment. Do you have any ideas on which one it should be to minimize these side affects? - Don On January 11, 1998 at 22:34:39, Komputer Korner wrote: >There is an easy way to find out. Set up the initial position except >for one side being minus a pawn. Try this for all the different pawns. >Look at the scores. Some programs like M-Chess 7.1 have high evaluation > scores but it isn't clear how much of this is due to overestimation of >the positional score or whether the materiel scores are high. Hiarcs 6 >score is 1.28 units per pawn. Has anybody done any tests? I don't seem >to have time anymore to do tests. > >On January 11, 1998 at 20:56:06, Robert Hyatt wrote: > >>On January 11, 1998 at 18:12:56, Howard Exner wrote: >> >>>Does anyone know how the different programs rate the value of a pawn? >>>I think someone (probably Ed Schroder) said that Rebel's score >>>is about 0.75. >>> >>>Does it make sense to compare eval scores across programs? ie:"this >>>program does better here because of its +4.55 eval versus this >>>other program at +3.62". In reality for program X, 4.55 might be 4 pawns >>>while program Y at 3.62 could be 5 pawns. Is this right? >> >>This is difficult. But you are really asking the wrong question. IE, >>suppose Rebel does this: P=.8, N=3.0, B=3.3, R=5.0 and Q=9.0. And >>suppose >>I did something like this: P=1.0, N=3.5, B=3.8, R=5.5 and Q=10.0... >>what >>would that mean? If you notice, these are actually about equal, as far >>as material goes... >> >>So the value of a pawn is not the real issue... it is the value of a >>unit >>of material... which makes it harder to understand..
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.