Computer Chess Club Archives




Subject: Re: evaluationfunction tuning

Author: Tom Likens

Date: 15:46:44 09/07/03

Go up one level in this thread

On September 07, 2003 at 11:35:48, JW de Kort wrote:

>Hi friends,
>I have recently spend a lot of time improving my evaluationfunction. I have
>added a lot of features and studied a number of sources (Crafy, Gerbil, TSCP,
>Phalanx, Beowulf etc.). Still my engine is not able to beat TCSP 181 in all
>games. Now i want to fine tune the evaluation features. Can anybody give me some
>advice how to ga ahead with this process? Is there a EPD set available of good
>testing positions for this purpose?
>Thnak in advance.
>Jan Willem

One feature you might want to implement is a "flip/flop" command.
Basically, these commands allow you to mirror the position around
the x and/or y-axes.  This is useful because you can use it as one
way to check that all your fancy new evaluation changes work the
same way for both black and white as well as side-to-side.  It's
a quick bug check and pretty useful.

You can even go one step further and implement a command to run an
evaluation-only test on a set of positions.  I have a command that will
run an entire test suite of positions through the evaluator only
(no search is performed).  The original position is checked as well
as the three other possibilities (flopped around the y-axis, then
flipped around the x-axis and finally, again flopped around the
y-axis).  If any of these show large evaluation anomalies, then
it's time to break out the debugger.

Will this help you tune the evaluation, probably not, but it will
let you catch bugs.  One thing to note about this is that if you
have evaluation terms dependent on the castling rights they will
legitimately change after you start swapping things around.


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.