Computer Chess Club Archives




Subject: Re: learning to tune parameters by comp-comp games

Author: Josť Carlos

Date: 06:27:02 12/29/00

Go up one level in this thread

On December 28, 2000 at 15:18:54, Uri Blass wrote:

>How much rating can programs earn by playing against themselves?
>I think that it is possible to improve the rating of programs by playing a lot
>of games between the program and itself when you change one parameter(for
>example increasing the value of pawn by 5%).
>It is possible to play a lot of games and stop only when there is a difference
>of 70 in order to learn if increasing the value of pawn by 5% is a good change
>or a bad change(we need big difference because the difference from small change
>is usually small and we can get often wrong results if we stop only at small
>If you find that increasing the value of pawn by 5% is productive you do the
>change and the program learned to increase the value of pawn.
>After it you continue in doing similiar tests.

  I think you miss an important point here: the eval function is
multidimensional, so changing only one parameter at a time can only make you
reach a local optimum in one dimension, that can actually be far from an
"absolute n-dimensional optimum".
  There're many technichs whose idea is similar to yours, but using
multidimensional optimization, for example, using the gradient.

  Anyway, playing against itself is not enough, because it is well known that
the better way for improving one's strenght is playing people slightly stronger
than you, so you can learn (because they are stronger, but not so strong that
you don't understand why you lose).

  Just my opinion.

  Josť C.

This page took 0.16 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.