Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: automated book generation in chess/checkers

Author: Paul Byrne

Date: 11:01:27 12/13/01

Go up one level in this thread


On December 12, 2001 at 18:31:14, martin fierz wrote:

>aloha,
>
>there was a discussion about this, deep down in a thread about who is the best
>chess programmer, so i decided share our experience here:
>
>thomas lincke (lincke_at_inf.ethz.ch) has generated an opening book for my
>checkers program with a technique called dropout expansion. he wrote a paper on
>this, i think. for checkers it works very well, my book is very close to what
>you expect from reading human opening books. the book has 1.1 million scored
>positions in, which gives me about 200'000 book moves.
>
>tom also used crafty as an engine to try and build a chess opening book. while i
>do not know all details, i know he had at least a million nodes, and the book
>was (if you ask me, 2200 rated) completely useless.
>
>the big difference in performance for the two games is that checkers has a very
>high percentage of losers in every position. chess on the other hand has lots of
>positions where you can make nearly any move. [...]

Sounds very similar to how my G2K builds books.  It expands the book by making
a series of book moves to reach a non-book position, evaluates all moves there,
then backs up the resulting scores.  It helps somewhat to allow a wider range
of scores in choosing a book move when expanding than when playing, but it
obviously still misses deep ideas.

The second approach to expanding the book is to take it's own games from ICC
against strong players (plus anything it lost) and expand along the lines
played -- this sometimes helps it see things it would never catch otherwise.

How effective is it?  In regular chess, pretty dismal.  I've abondoned it and
gone back to my old book, which isn't great, but can at least play the
more common openings.

On the other hand, G2K also plays a lot of other variants on ICC, and in
some of those -- atomic and wild 5, for example -- the auto generated book
seems to be pretty effective.  This matches what you have seen in checkers:
the variants the book works well in are very tactical, with few reaonable
moves from many positions.  (In atomic, for example, after 1. Nf3, black only
has 3 responses that don't lose immediately).

The second situation I use the automatic book generator is some of the
variants where I don't have a supply of high quality games and don't know
enough about to make a simple book myself.  Here, at least the book can't
hurt -- at worst, it plays the moves it would have played anyway, and saves
a little time on the clock.  Usually the process of backing up scores within
the book improves its play somewhat.

-paul



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.