Computer Chess Club Archives


Search

Terms

Messages

Subject: The Problem with Opening Books for Engines

Author: Bob Durrett

Date: 11:31:47 11/19/02



The typical opening book, it seems to me, is designed to give the chess-playing
program the highest probability of success.  That may be it's weakness!

To improve the statistics, the designer of the opening book will naturally
include responses to the moves which are most likely to be played.

Similarly, to improve the statistics, the designer of the opening book will
provide responses to the moves which are considered to be the most challenging.
For example, an opening book for the engine playing the black side will have a
lot for 1.e4 but maybe nothing at all for 1.h3.

Generally, early queen sorties are regarded, by the "learned" people, as being
not worth preparing for.  It is assumed that the engine will easily find the
refutation without the help of an opening book.

But is that really true?  Or is it *merely* a universal but false assumption?

It occurred to me to use my CB8, with Fritz7 as an analysis engine running under
CB8, to see how well early queen sorties fared, historically, in human praxis.

I began with a query to identify all games in which White played a queen move on
his second move.  Amazingly, there were more than 1000 of these games in
Chessbase's Megabase 2002 database.  I then sorted the games by ELO.  Another
surprise!  There are many different openings of this type.

I'm now beginning to believe that it would be possible to produce an
anti-computer opening repertoire consisting solely of openings with early queen
moves.

Here's an example:

1.d4 f5 2.Qd3

If 2...d5 then maybe best is the line 3.c4 e6 4.g3 c6+/=, which might retain a
slight advantage for White if Black deviates from this line early enough.

If 2...d6 then either 3.e4 fxe4 4.Qxe4 Nf6 5.Qe3+/= or 3.Nf3 Nc6 4.e4 +/=

If 2...e6 then 3.e4 fxe4 4.Qxe4 Nf6 5.Qh4 b6 6.Bd3+/=

The question is:  would your engine find it's way through this analysis?

To find the best line, the engine must find the correct move early on.
Otherwise it could get into trouble.

That was just an example.

Maybe the way to defeat the current opening books is to find such lines.

I'm not at all convinced that modern chess programs are properly prepared for a
human opponent who is determined to bring out his/her queen early.

Bob D.



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.