Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Testing engines at slow rate of play?

Author: Dann Corbit

Date: 12:44:49 01/01/02

Go up one level in this thread


On January 01, 2002 at 15:11:54, Ted Sutton wrote:

>Why does everyone test engines at a very fast time control?

Here are 1000 games played at G/90 on PII 300MHz or equivalent:
ftp://cap.connx.com/pub/battle%20of%20the%20crowns/

Here are 1600 games played at G/90 on 900 MHz Athlon or equivalent:
ftp://cap.connx.com/pub/The%20Gauntlet/contest/
(It will be 16,000 when complete).

Here are 80,000 games played at 40 moves in 2 hours:
http://home.swipnet.se/~w-36794/ssdf/nr000.htm

This contest is at 40/40 (which is fairly slow paced):
http://home.hccnet.nl/leo.dijksman/index.html

There are many contests here at various time controls:
http://www.geocities.com/chessfun_1999/

This one is borderline blitz at G/30:
http://www.geocities.com/lyapko/lgwbl.htm

Here are 2,500 games played at 40/40 time control:
http://www.amateurschach.de/schach/cbase/cce_main.htm

>    It gives us irrelevant information, though it is easy to do the tests.

It gives blitz information.  While disgusting to me, it appears that most of the
planet prefers this disgusting and vagely chesslike form of entertainment over
the real thing.

>      What about testing engine vs. engine where each move is 12 hours, to see
>which one is better in correspondence mode?

Lots of correspondence chess games are actually played that way.  The problem
with the correspondence chess approach is that it takes so long to create enough
data to get reliable results with it.  On the other hand, I have participated in
just such a match:
http://www.gambitsoft.com/kup.htm

(Here is the previous contest):
http://www.gambitsoft.com/turnier/kupold.htm



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.