Computer Chess Club Archives


Search

Terms

Messages

Subject: The end of the opening books ?

Author: Georg v. Zimmermann

Date: 08:48:32 05/21/01


Hi,

what started as an attemp to get my program ( crazyhouse chess ) to play a
reasonable opening when its opponent tries to kick it out of book (1.g3 or so)
made me think about opening books in general.
IMHO every program can get stronger when the author tries to make it play good
openings even without opening book, that will help in the middle game as well.

I believe that the ammount of memory taken up by "perfect knowledge" (Endgame
Tablebases) will ever increase while the ammount of memory taken up by "human
knowledge" (Opening books created by humans) will decrease.

- Programs have become so strong that often times humans do not know better
which move is best in a certain position.
- Programs often get out of book in a losing line.
- Programs often get out of book in a line they do not understand.
- Learning features become better and better, so programs will "learn" openings
over time anyway.
- A single program plays many more games than a GM in a given time period, so to
learn from its own games which are far more predicatble and contain far less
blunders which change the outcome of the game completely makes more sense.

This might not be true for very small expert written opening books, but I think
it holds true for the Crafty style books.
What do you think ?

Georg



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.