Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is SSDF list level is now badly inflated?!

Author: Gunnar Andersson

Date: 02:23:35 09/28/99

Go up one level in this thread


On September 27, 1999 at 13:59:58, Robert Hyatt wrote:

>Picking good book moves is an issue.  The commercial guys seem to bank on
>having a human tune their book.  This is impractical, for lots of reasons.
>Letting the program tune its own book is also very difficult.  There really
>isn't a good solution for this problem yet, either human solution or computer
>solution...

Isn't this because the evaluation functions haven't reached the level where the
values they return have a good global interpretation? I.e. comparable between
different game stages and different types of games.

I've been involved in computer Othello for a while now. It's considerably easier
to tune a good evaluation function automatically and all the top programs (which
are much stronger than the human world champion) tune their book automatically.
In the future, evaluation functions in chess programs might reach this level too
and then having humans tune the book won't be necessary (or a good idea at all).

/ Gunnar



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.