Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF and the programmers............

Author: Ulrich Tuerke

Date: 03:12:14 03/18/98

Go up one level in this thread


On March 17, 1998 at 23:39:07, José Berdiñas Bonefua wrote:

> ...
>
>I do not share your point of sight.
>With a discreet engine , a good book tuning ( many previous games) and a
>good book "learning" , a program can be today the number 1 in SSDF  list
>; but surely never will win a WMCCC , because will lose with the
>"unknown" programs.
>I ask to you, So that serves us then the SSDF????.
>
>Any similarity with Fritz 5 is pure coincidence.!!!

IMO, this is 100% true.

Playing against each of 20 opponents 10 games would result in far
more reliable results than playing against 5 opponents 40 games,
though the total number of games being the same. There is a number
of sufficiently strong programs available with auto232 support (e.g.
Crafty, Crazy Bishop, CS-Tal, Comet, Patzer, Kallisto II, ...).

I think also that the book learning (=cooking) results would
"average out". A line producing good results against one opponent
would often turn out bad against another opponent.

The current state of the art of book learning is far too naive to
produce a real increase in playing level, but book learning can
however temporarily produce improvements against a limited number of
opponents.
Or does any high level player omit a line because of one negative
result ?
That's not the way the opening theory is being developed but whole
trees of variations have to be analyzed in order to establish a theory
line.

Regards, Uli

> ...



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.