Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rebel Anti Grandmaster style

Author: Ed Schröder

Date: 00:09:45 09/20/98

Go up one level in this thread


>I was just looking at the Rebel page, reviewing upcoming features of Rebel10,
>and I couldn't help but wonder, what is the Anti GM setting. Is this for real.
>Is it a revolutionary change in way a chess program searches, or is it just
>something that is just a little trick of programming that human GM's with some
>time against this program will able to exploit and beat just like they have
>done many times before?

I already answered this question a few times here. The anti-GM setting
changes Rebel's playing style when the program thinks there is a need
for, meaning depending on the type of the position and the previous moves
of the game in progress.

In most cases (say 75%) this means Rebel decides not to use the
algorithm and Rebel plays just like Rebel.

In the remaining 25% a special piece of software is added to the evaluation.
This may result in different moves, but not always. I have seen games
which were highly influenced by anti-GM and games where the effect was
simply zero.

About revolutionary: I think the basic idea certainly is not new. I am sure
many other programs have such options. Speaking for Rebel only, through
the years playing games against grandmasters I noticed a pattern best
defined as the struggle for the initiative. I found out that this was the main
reason (the loss of the initiative) why Rebel lost games against grandmasters.

anti-GM is a first attempt to deal with this "specific" (grandmaster) problem.
I have given it a name. In my opinion the only right name, anti-grandmaster
play as it is meant to fix a huge (Rebel related!) problem when facing
strong human players like grandmasters.

Revolutionary? That's not up to me to decide. I am (generally) very pleased
with what I have seen sofar but all in respect to the above meaning playing
against grandmasters. Not being a GM (or IM) it's a worthless feature as
mostly you and I will lose anyway with or without the setting :-)

>Please don't take this the wrong way - I know that Rebel is incredibly strong
>(I have Rebel 7,8 & 9), but I'm just curious as to what the real potential of this
>feature is, especially since it is currently Rebel10 and Fritz5 (Junior), that
>are beating up on the top humans.  Are the top programs all that different?

I don't know. Other programs do not get much chances to proof their
strength against grandmasters. In this respect I wonder about the status
of Ferret playing GM Morovic this month.

- Ed -


>Thank you programmers, that take a stab at answering this...

>mrslug - the inkompetent chess software addict!




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.