Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To build a book or not?

Author: Mogens Larsen

Date: 03:03:35 07/16/02

Go up one level in this thread


On July 16, 2002 at 05:37:46, Russell Reagan wrote:

>One question came to mind when hearing all of the opening book talk. Gian-Carlo
>said that he made his own opening book for Sjeng, and it backfired a few times
>at WCCC.
>
>For myself, I have some chess knowledge, but I'm no better than average. For
>someone like me, would I be better off attempting to make my own book from
>studying MCO and other opening manuals, or to just add in some special opening
>code to help it play better in the opening?
>
>I have a feeling that I would probably hinder my engine more than help it, but
>maybe I don't realize how badly engines play the opening without their books
>when facing top competition. Please advise.

There are several ways of doing it. Some prefer to make their engine capable of
autogenerating a book from a pgn file. That's a good intermediate solution, but
generally not enough. Mainly because its indiscriminate of the strengths and
weaknesses of the program in question. Therefore it's a good idea to guide it
past the more obvious of the dubious lines. Like Crafty for example. If you add
learning to the equation then it's another question altogether.

The other method, which I think it's a good idea in your case, is a small
handtyped book to start with. It doesn't have to be especially broad or deep.
That approach might give you some information about your engine; preferred
variations out of book and their validity; development of pieces and castling in
the opening. Then you can start expanding until you have a repetoire book. After
that you can add an autogenerated extension book and maybe learning.

Regards,
Mogens



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.