Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: For all intense and purposes Kramnik is correct.

Author: Mogens Larsen

Date: 06:33:32 04/10/02

Go up one level in this thread


On April 10, 2002 at 08:58:36, Hristo wrote:

>What is that supposed to mean?
>The entire "conjecture" that DB is faster and therefore a lot better than the
>rest of the COMPS is somehow not convincing. If nothing else the following
>problem arises ...
>Lets use ELO points as the relative chess strength for COMPS.
>If 2*nps = +100 ELO points.
>The modern computer, at 1M nps, are 200 times slower than DB.
>DB performed, during the match with Kasparov, at ELO 2900. (not sure about that)
>Can we acertain that the relative strength of the computers today is between
>2000-2200 ELO points? If, at the same time, we accept that the eval used by DB
>was superior to all modern COMPS, then the relative strength of the modern
>computers should be even lower. Is this correct?
>Anyway, the information is so scarce and inconclusive that making any
>conjectures based on this information is dubious and fraudulent. Dr. Hyatt
>desrves a lot(!) of recognition, but when it comes to chess (the game) I would
>rather listen to Kramnik.

You miss the point I was trying to make, I think. I was talking about the actual
moves played and their significance, not speculating on the importance of
knowledge or speed differences. The only problem I commented on was the
importance of Kramnik's discovery, ie. an unknown number of objectively better
moves found by Fritz compared to Deep Blue. That doesn't prove anything at all
IMO. Not only is it possible to question the term "objectively better", even
when he's the WC and 2nd ranked player in the World, but also the actual
importance of the method. At least in the sense of establishing a strength
evaluation.

Finding a better move here and there doesn't mean that Fritz wouldn't do much
worse, because it doesn't tell you about the likelihood of comitting blunders or
reaching similar positions, where it excels. Of course it doesn't tell you the
opposite either. It might do better, it might not. Kramnik's investigation
doesn't prove anything either way. It's like determining the strength of a
program based on a secret testsuit. An example would be to find a game, or
several games, where one side played a number of "objectively better" moves and
testing a certain number of engines. Would the best "move solver" be the
strongest program? I don't think you can make that conclusion.

BTW, even as a complete patzer, I'm able to find objectively better moves than
2200 ELO engines. Especially in the endgame. Does that make me a 2200+ ELO
player? Not in a million years :-)).

Regards,
Mogens



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.