Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: books

Author: Moritz Berger

Date: 08:46:12 09/12/98

Go up one level in this thread


On September 12, 1998 at 11:01:36, blass uri wrote:


>I suspect that you use the same book that they used in paris
>It is a bad book.

Powerbook + Blitz learning against Genius is unlikely to work against any kind
of opponent at tournament time controls. Reason: Fritz won too many games with
weak lines, increasing confidence in playing them. It got punished for these
weak (learned) lines (which were good enough against Genius at blitz) by other
programs at tournament time controls.

Same effect if I play 100 crap games against Fritz with learning enabled: Next
time it will try the Orang Utan opening against an opponent (computer program)
of my choice and is more likely to lose, because my weak play increased its
confidence in those opening lines.

Easy to see how to cripple the opening book this way when applying learning
regardless of opponent's strength.

Only way out: Reset book weightings before each match and (optional) do book
learning from a database of Fritz games at the same time control against "known"
strong opponents.

>I am sure that they understood it after paris and I do not believe that they
>used this powerbook for the ssdf list.
>Do you use the same powerbook that they sell now?
>
>Uri

The commercial powerbook scored significantly better than the standard Fritz 5
book at the SSDF. It was something like about 40 ELO stronger, AFAIR, you can
look it up probably on the SSDF homepage somewhere ...

Moritz



This page took 0.01 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.