Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: GM's vs Comps at random chess? I expect GM's to get spanked.

Author: Hasnain Mujtaba

Date: 19:16:50 12/19/99

Go up one level in this thread


Hi Charles

I don't know about any one-on-one Human vs Computer match-ups at chess variants,
but in an experiment conducted in 1997, a team of two different chess computers
(each with Elo > 2500) and an amateur human 'controller' (Elo 1900) defeated GM
Arthur Yusupov (at the time #31 in the world with Elo 2640) in an 8 games match
with a score of 5:3. The match was in "Shuffle Chess" -- a variant where you
start the game with the back rank pieces arranged at random.

The two chess programs ran in some k-best modes and the human controller would
then pick what he thought were the best moves from the two k-best lists.

The purpose of the experiment was to guage the combined strength of humans and
computers. On their own, the ametuer and the computers will lose to GMs at chess
variants because they don't have the tactical strength. But together, as this
experiment shows, they can do quite well.

You can read more about this experiment at
(http://www.minet.uni-jena.de/www/fakultaet/iam/personen/CA-Chessd.html).

Regards
Hasnain.

On December 19, 1999 at 19:36:08, Robert Hyatt wrote:

>On December 19, 1999 at 18:30:23, Charles Unruh wrote:
>
>>It's still chess, the same elements exist time, space, force, and development.
>>Still middlegame, opening, and endgame.  Kasparov himself made a statement about
>>how he would fair against other GM's in random chess, because he has been
>>accused so much of winning because of opening preparation.  He said that he
>>would be the best because random chess will favor the better tactical player,
>>and of course he claimed that the best tactical player was him :).  Kramnik
>>agreed with him though, it would be interesting to see at least 1 40/2 game in
>>random chess against a GM.  I have a bet going with an old bud from my tour in
>>Vietnam.
>
>
>This is a dangerous bet to make.  Computer evaluations are _not_ tailored to
>random chess positions.  IE a weakness at g3 is something else entirely when
>bishops are moved...  I have looked at this a bit, but it is difficult to
>handle.  IE some of the 'wild' variants on servers ought to be easy for the
>computer, but they aren't because the eval is so wrong...



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.