Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Simple solution to question of GM strength computer programs

Author: Graham Laight

Date: 03:44:47 06/20/01

Go up one level in this thread


On June 19, 2001 at 21:07:23, Bill Gletsos wrote:

>Simply have a single computer program play against 12 different GM's at long
>time controls.
>Once as white and once as black.
>You could calculate its rating as if it would appear on the FIDE rating list.
>If it aceived a rating at least a rating of 2500 and also a TPR of 2600 over
>those 24 games then you would have an excellent argument that the program is of
>GM calibre.
>
>Provided all 12 GMs above played each other as well then a human meeting the
>above criteria would earn the GM title. In fact for the average human he needs a
>2600 TPR over 27 games since a valid GM norm can only be obtained in tournaments
>of at least 9 rounds and most need 3 GM norms.
>Lets relax the condition about the GM's playing amongst themselves because its
>not going to happen.
>
>Currently at long time controls no program has a TPR of 2600 over 24 games.

What about Dortmund last year?

Not quite the conditions you stipulate - but Deep Junior managed to obtain a
rating of 2700. Moreover - IMO, Deep Fritz is significantly better than Deep
Junior.

-g

>FIDE dont need to apporve any of this. Just takes some enterprising chess
>programmer to fulfil the above.



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.