Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Has there been any discussion about CSS 6/2004 sensation article??

Author: Vincent Diepeveen

Date: 10:36:34 12/22/04

Go up one level in this thread


On December 22, 2004 at 02:04:10, Jouni Uski wrote:

>Hopp oder top – Steigern Tablebases die Spielstärke?
>Helmut Conrady untersucht in einem sehr breit angelegten Test die Auswirkungen
>der Tablebases auf die Spielstärke von Schachprogrammen. Die Ergebnisse sind –
>gerade auch für viele Programmierer – äußerst überraschend!
>
>Summary in english after reading article yesterday: after 1000s of games with
>different levels and engines tablebases seems to have always NEGATIVE effect
>on playing strength!! No doubt at all! Uri can easily forget implementing now..
>
>Jouni

For diep in blitz EGTB's are more important than at slower levels.

It matters at least 600 rating points in the endgames. The overall rating
improvement is very dependant upon the type of openingsbook used.

Note i did not read the article at all, but that in general it is very difficult
for outsiders of computerchess to draw conclusions about computerchess. I bet he
has put the egtbcache too low or is using too slow harddisks.

With 6 men you need at least 256MB egtb cache a processor.

EGTB's do not make the nps of diep slower in far endgames, it also in general
searches very deeply because of them. I do know that for several programs this
is the opposite. Drawing conclusions based upon using not very well optimized
software for EGTB's i cannot take very serious.

Shredder and Diep are of course programmers which have the habit to exchange a
lot of pieces, but neither of them will ever be able to estimate on their own
whether a KRPPKR or KRPKR is drawn or won, and those 2 endgames happen a lot.

Vincent



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.