Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Measuring program improvement

Author: Jon Dart

Date: 19:35:22 11/17/01

Go up one level in this thread


On November 16, 2001 at 19:14:21, Christophe Theron wrote:

>
>You should just stop relying on ICC ratings.

Christophe's advice is good. I am somewhat influenced by watching Arasan play
online and sometimes am convinced that some recent change has made it worse. But
ratings are a highly imperfect measure, subject to short-term fluctuations, and
also easily distorted by the nature of the opponents you're matching, and other
factors.

Test suites are also imperfect but have the advantage of giving you objective,
repeatable numbers. So I rely on those more.

Another thing you can try is running your own tournament offline against other
programs. If you have a spare machine and can run it 24 hours for a few days,
you could accumulate several hundred games, enough to get an idea of your
program's strength against its opponent.

--Jon



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.