Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How people could detect if a game was cooked?

Author: Robert Hyatt

Date: 11:08:44 09/16/02

Go up one level in this thread


On September 16, 2002 at 13:05:27, Rolf Tueschen wrote:

>On September 16, 2002 at 11:43:24, Vincent Diepeveen wrote:
>
>>Diep never considers taking on d5 here. It says Rfd1 white up 0.117
>>after 10 minutes of search still.
>
>You must not answer me directly but perhaps you could comment on the question at
>any appropriate occasion. What would you say if you, as always participating in
>tournaments, observed such a game against Shredder who then played Nxd5. But
>your Diep and also Crafty BTW do NEVER even consider Nxd5 but play Rd1. Would
>you say that this is a clear example of a cook against SHredder in special? I
>ask because I'm not so familiar with all such tricks being known to the
>participants. Not long ago another operator of Rebel was proud to use other
>tricks, allowed tricks as he said, against Shredder. Are such tricks part of the
>tournament practice? And if yes, would you say that in _testings_ such cooks
>should be left out? Or is the golden rule the stupid "data is data, no matter
>where or how from"?
>
>Rolf Tueschen


I would not say that.  For reasons I have given _many_ times.  There are
two opportunities for variability in a chess program.  Parallel search is the
best known example, but it is not the _only_ one.  If the opponent takes more
or less time to move, you get more or less time to "ponder".  This pre-loads the
hash table with information that may (or may not) change the search result.

I've seen more than one or two non-reproducible moves.  With position learning
further compounding this, whether crafty will or won't play the move is itself
open to error...




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.