Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Blitz or long time matches to evaluate engines?

Author: George Sobala

Date: 10:07:02 12/07/04

Go up one level in this thread


On December 07, 2004 at 12:55:07, Bernardo Wesler wrote:

>Why, if so, is it better to evaluate engines in long times or blitz games?
>Beyond that, why 2 min or 5 min or 15 min/game and why 1h or 2 hr etc games?
>Thx in advance.

I suppose you should evaluate with blitz if you plan to use for blitz!

In a crude sense, short games test search depth more than anything else, whereas
at longer time controls chess "understanding" starts to play a larger role. (The
effective benefit in searching 17 rather than 15 ply is much smaller than the
benefit of searching 7 rather than 5, the "understanding" of the resulting
position is therefore more important.)  This is a facile and over simplistic
explanantion, but has a grain of truth.

At short time controls with no increment, many games will be decided by simple
blunders in drawish long end games in desperate time trouble. The engines may
not make such simple mistakes if they had a bit of time to think. If you are
seeing simply which engine is best at 3+0 then fine - but you are not really
testing overall chess ability.

Even at 15'/game you can end up considerably over-stretching the end-game
ability of an engine e.g. compared to 10'+6"



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.