Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka website updated

Author: Ross Boyd

Date: 13:35:27 12/11/05

Go up one level in this thread


On December 11, 2005 at 16:06:48, Joseph Ciarrochi wrote:

>On December 11, 2005 at 15:54:52, Arnold Gove wrote:
>
>>http://www.rybkachess.com/
>
>i am amazed rybka dominates with "no endgame knowledge", as the website states.
>Can anybody estimate how much rybka will improve with endgame knowledge?

I doubt Rybka has "no endgame knowledge" but clearly it is incomplete in some
areas. So Rybka will improve... but by how much you ask...

Keep in mind that many of Rybka's games don't even last until the endgame ;-)

However, think of it in terms of how many extra points it would score in a 100
game match. Therefore, how often do you see it draw when it should win _OR_ lose
when it should draw?

If this happens in 5 out of every 100 games... that means the new Rybka will
score ~2.5% more. In ELO terms against an equally matched engine this translates
to approximately 20  Elo points.

I'm only guessing that Rybka's endgame bugs occur in 5 out of every 100 games...
 hopefully it happens much more and Rybka is going to get even stronger!

Ross



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.