Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: one general request to CCRL people

Author: Ray Banks

Date: 22:55:40 02/12/06

Go up one level in this thread


On February 12, 2006 at 22:47:03, Kirill Kryukov wrote:
>Both different number of games and different time control are likely the reason of differences. It is known that some engines can benefit more from longer time than the others, resulting in rating differences. Otherwise we would all test in blitz and be happy with that.

Agreed. Although I have no evidence to support this, I'd guess that the speed
increase for an engine would benefit it most at a shorter time control. Then,
when the time control lengthens an engine surely reach a sweet-spot where in the
time available it will give it's best answer 90 % of the time. Then, going even
longer, it surely reaches a point where no matter how long you give it, it will
always give the same move, because the limitation is then it's knowledge and
evaluation rather than the speed or time available.

We have only been going for 2 months or so, and naturally we don't have as many
games yet as we'd like because we started from nothing. But in another 3 months
most of our major engines will have over 1,000 games each and the error margins
acceptaby small. Once we reach that point, new engines including amateurs will
be able to be tested quickly even with the longer time control.



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.