Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: evaluationfunction tuning

Author: Albert Bertilsson

Date: 12:33:01 09/07/03

Go up one level in this thread


If your program doesn't beat TSCP I don't think it's because you lack tuning of
your evaluation. I remember when Sharper couldn't beat TSCP, then after
improving search adding hash-tables etc it started to beat TSCP more often. On a
very basic level I think it is important to get rid of bugs, if you add lots of
evaluation it's possible that you've added bugs in the evaluation. Try betting
good results with little evaluation first to see if the tactics are correct. The
downside of having little eval is that many engines will beat the crap out of
your engine (I've seen it happen to Sharper alot) but if you have few buggs you
can always add evaluation later and see improvement.

I think many engines have buggs so keep testing against TSCP if you beat it with
little evaluation your engine probably has few buggs. Also run testsuites like
WAC to find buggs.

If you really want to focus on evaluation lock for the testsuites called quiet,
the test evaluation and run fast too!

/Regards Albert



This page took 0.06 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.