Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: an example how users - not programmers - use tests

Author: Uri Blass

Date: 13:00:38 06/20/04

Go up one level in this thread


On June 20, 2004 at 15:56:00, Steve Glanzfeld wrote:

>On June 20, 2004 at 15:34:39, David Dahlem wrote:
>
>>The best way, and only way, in my opinion, to test engine strength is in actual
>>game play. The engine that plays "better" moves than the opponent, not
>>necessarily the "best" move, will determine engine strength more accurately than
>>all test suites ever created.
>
>But not everbody is intereted in the gameplay strength (only).
>
>Some chessplayers might be interested to find the best engines for analysis
>rather than for computer vs. computer games. I think, my example above was not
>uncommon. - And: In analysis, it is desirable that the engines finds decisive
>moves, if existing, as often and as quickly as possible, not just good playable
>moves. In analysis, engines have to go for absolute chess truth, not just for
>moves sufficient to win "somehow"...
>
>What is the best way to test ANALYSIS QUALITY in your opinion?
>
>(Simply taking the best gameplay engines, can't be the answer obviously, because
>the cleverest people have repeatedly explained that results from test positions'
>analysis have nothing to do with the real gameplay strength of an engine. So
>that can't be the reversed case either :)))
>
>Steve

The best engine for analysis can be different based on the positions that you
choose to analyze

Engines like yace that can learn from previous analysis may be better than an
engine with similiar strength that does not do it because yace after going
forward and backward may give you better information than another engine that
cannot learn from previous analysis.

Uri



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.