Computer Chess Club Archives


Search

Terms

Messages

Subject: Honest ABE, part 2

Author: Dann Corbit

Date: 10:06:08 02/18/05


(A) (B)ook (E)xperiment -- what's a book worth?

I ran the contest on two machines.  On the Windows 2000 machine, I had a hundred
pop-ups about polyglot failures, so I do not trust those results at all.

On the other Windows 2003 AS machine (2.2 GHz), here is the result:
  Program                    Elo    +   -   Games   Score   Av.Op.  Draws
1 Polyglot fruit-book      : 2578   53  65   117    64.5 %   2474   21.4 %
2 Polyglot glaurung-book   : 2531   59  60   117    56.0 %   2489   16.2 %
3 Polyglot glaurung-no-book: 2453   63  57   117    41.0 %   2516   17.1 %
4 Polyglot fruit-no-book   : 2439   68  55   117    38.5 %   2520   13.7 %

I think a lot more games would be needed to achieve certainty, but it appears at
first glance that even an automatically generated book does provide some
benefit.

Both Fruit and Glaurung used the polyglot book, and I renamed the Glaurung book
so that when the polyglot book was removed no book would be in play.



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.