Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka 1.0 Beta 32-bit (40'/40) After 650 games (Match vs Pro Deo 1.1

Author: Ryan B.

Date: 02:04:31 12/24/05

Go up one level in this thread


On December 24, 2005 at 04:24:20, Heinz van Kempen wrote:

>>The endgame knowledge in Rybka is very good.  The claim of no endgame knowledge
>>was just easier than explaining the distance to mate bug.  I think fixing the
>>bug will help some but not as much as people think.
>>
>>Ryan
>
>Hi Ryan,
>
>my impression from watching games is also that the endgame knowledge is already
>good. So I am basing on the claim by Vasik that here many things have to be
>improved still.
>
>About improvement I do not have a concrete guess, I only start to ask myself
>generally where the limits are. ELO 3200? No limit at all?
>
>Best Regards
>Heinz


I think elo of computer programs can be a little deceptive because a small
advantage in computer chess gets bigger in computer vs computer matches because
computer programs do not adapt to weaknesses and passed games vs a given
opponent naturally like humans do.  It is very interesting to me that a program
can be better than a top GM in a 6 game match however I feel I have better
chances of adapting to that program and stealing a game vs the computer than I
do of getting a game vs the top GM.  I think the jumps in computer chess
improvement are very exciting and it will be very interesting to see where it
all is a year from now.  I think Rybka does make some good progress in computer
vs human play also, something I think got ignored in computer chess too often
while in the catch shredder era.

Ryan



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.