Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Proposal: New testing methods for SSDF (1)

Author: Mats Winther

Date: 10:20:44 04/13/98

Go up one level in this thread


Thank you for your testing proposal and those questions about the
autoplayer mystery. I too am bewildered by this and would like to
know more.

Concerning your view of the Fritz module: (discussed at length
earlier)
Has there been any doubt that Fritz5 is one of the top programs?
The argument is whether it is the best or not.

Apart from this, since you infer that Fritz5 is good positionally,
what do you then mean by "knowledgebased programs"? I thought that
Hiarcs and Rebel had more knowledge and therefore play better
positionally than Fritz5 whilst the latter compensates this with
longer variations. So then Fritz5 should play less good positional
chess than the others. I can't draw any other conclusion. When
we talk about "good chess" we don't really refer to the end result.

In fact, Fritz5 sometimes (not always!) definitely plays bad chess
that is positionally weak - typical for engines with less knowledge.
You are quite right in saying that this occurs to other engines too.
I have positionally horrid games by Hiarcs too. But this should
in theory occur more often with Fritz5 and it does too in practice,
I would say.

We live in the positional era of chess when "good chess" is
defined as positionally refined chess. Then it is obvious that
Fritz5 plays less good chess than certain others (drawn from a
large amount of games).
But I take it that you regard tactical trickery as "good chess" too?
But isn't it better to use the traditional definition of "good chess".


Mats Winther




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.