Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How do people do on PET?

Author: Richard Pijl

Date: 12:27:19 07/09/04

Go up one level in this thread


On July 09, 2004 at 14:58:15, Scott Gasch wrote:

>Hey,
>
>I'm wondering what other engines run PET (Pete McKenzie's endgame test) and how
>they do on it?  Monsoon does terribly, for the record.  I don't have an official
>number (because I didn't write the code to score how long a solution is held)
>but it doesn't get more than about 10 of them right and some of those for the
>wrong reason.  I'm thinking a lot about endgame eval lately.
>
>Scott

There are some really difficult positions in PET. The Baron scores between 28-32
solved positions depending on the version (on a Xeon 2.8, 100Mb hash, all 5 men
EGTB's). It happens a lot that performance on PET degrades, but on other endgame
suites the performance is better (e.g. the speelman suite from George Lyapko's
site, run without EGTB's)

But 10 sounds very bad. I've got another engine, basically without anything
(only piece square tables, no hash, no egtb's, no advanced techniques, you name
it, it hasn't got it) and it solves 9 positions (could be for the wrong reason,
I haven't checked).

Richard.



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.