Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: a standard for evaluation score.

Author: KarinsDad

Date: 12:10:39 05/19/00

Go up one level in this thread


When I read your title, I was actually thinking that your were asking for
something else.

I thought you were asking for a standardized evaluation display where we come up
with a bunch of rules for the evaluation with set values. Then, whenever we
change the PV in our output, we not only calculate the evaluation that the
program normally comes up with, but also the "standardized evaluation".

For example:

pawn +1
knight +3
bishop +3
rook +5
queen +9
mobility square per piece +0.02
controlled normal square +0.1
controlled square +0.15
controlled square next to enemy king +0.2
etc.

This could have the potential to improve chess programs since the programmers
could compare how their programs arrived at a score for a given position versus
how the standard arrived at it's score for a given position.

Sometimes, you will see two programs playing against each other where both
programs think that they are losing. Programmers could take the data from such
positions and figure out which evaluation parameters in their code may be
causing their program to evaluate incorrectly since the standardized evaluation
will either think it is a draw or that one program is winning (this of course
assumes that the standardized evaluation is reasonably correct in most
positions).

I doubt anyone will do this, though, since getting a list of evaluation
parameters which are reasonably correct for a majority of positions is probably
unrealistic. But, it was a fun idea to toy around with.

KarinsDad :)



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.