Author: Matthew White
Date: 14:14:36 05/29/03
Hi all,
We all know that computers are weakest at opening and endgame play. I
realize that there are opening books to help this and endgame tablebases to give
the computer perfect play once the game is reduced to a few pieces per side.
What about endgame positions with known theoretical results, but too deep of a
solution to be calculated by brute force? Most programs today transfer into the
endgame "hoping" that they are winning or that they can salvage a draw.
Has anyone experimented with an "endgame book"? There are hundreds
(thousands?) of endgame positions that have been exhaustively analyzed but which
may never be a part of endgame tablebases given the current state of the art. I
hear GM's talk about "book endgames" all of the time. They have memorized
endgame positions, why should a computer be forced to re-invent the wheel every
time?
True, an endgame book wouldn't enumerate every possible position on the
board and tell the outcome given a position, but isn't it true that some
positions are thousands of times more likely to appear in a game than others? I
am not sure how an endgame book would be loaded. Maybe it would be placed into
the hash table at the beginning of the game.
Just an idea.
Regards,
Matt
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.