Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A Great New [?] Idea for Testing Chess Computers

Author: Bob Durrett

Date: 18:08:59 10/23/02

Go up one level in this thread


On October 23, 2002 at 05:32:41, Murray wrote:

>>
>>You have a really neat idea there!  [A way to identify the most "human-like"
>>chess computer] Select fify or 100 excellent games played between the top GMs in
>>the last year or two.  Then let each chess computer have plenty of time [at
>>least 5 minutes per move] to deliberate over each of the positions in those
>>games.  After all is said and done, then the Chess Computer which predicted the
>>largest number of GM moves "wins" and is declared some sort of "champ."
>>Ideally, find a sponsor with lots of money!  [Preferably several million
>>dollars.]
>>
>>Bob D.
>
>This idea has been tried before and the conclusion has always been the same: The
>computers often find alternate moves that are as good as or better than the GM
>moves. Sometimes the computers play weaker moves than the GM. The conclusion?
>There isn't any conclusion!
>
>Murray Cash

Hold on there!  There HAS to be a conclusion!

Your idea suggests that maybe the computer moves "proven" to be better [in some
sense] than human GM moves should be simply removed from the analysis or maybe
treated differently.  They shouldn't occur very often in the near future.

Alternatively, maybe those computer moves, PROVEN to be better, would be really
worth studying.  They may be little nuggets of gold!

Proving that a given computer move is better than any GM move [in the same
position] may be an ambitious undertaking, but perhaps do-able.

In the future, maybe it will be the rule rather than the exception that computer
moves are consistently better than GM moves [unless they are the same move].
But, for now, I suspect that most of the GM moves will be equal or better than
the computer moves.

Agree?

Bob D.



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.