Computer Chess Club Archives


Search

Terms

Messages

Subject: Horizon problems. ( Imagine you were back in the 80s )

Author: Georg v. Zimmermann

Date: 07:21:19 07/19/00


Hi,

I have huge problems with the horizon effect in my program (i.e. it sacks pawns
to push a queen loss far away).

It is my understanding that these problems have been solved in modern chess
programs mainly because they reach high enough ply depths.

Now imagine you were back in those times of much slower hardware : what can you
do to battle this problem ?


===

What I tried so far is

I. big extensions on checks and captures
This even worsened the problem, and I don't understand why.

II. detection of useless sacks
If a sack just lost material but didn't improve the positional eval old killer
moves (ply 2, opps turn) get extended to see whether the original threat was
really eliminated. This didn't work probably cause it is so complicated I had
too many bugs in it.

===

Some background:

I am playing around with a program that plays crazyhouse and bughouse, chess
variants where you can drop the pieces you captured from your opponent or your
partner captured from his opponent instead of a move. This increases the number
of possible moves by a lot (average maybe 33 in normal chess to over 110 in
crazyhouse/bughouse). This means you need an about 50 times faster computer to
reach 7 ply than in normal chess. Together with the nature of this game (games
are decided by king attacks) it makes for an exiting challence. When was the
last time you beat your program at 1 0 lightning/bullet time control ? :-)

http://bughouse.net/info/index.htm#howtoplay  ( Rules )
http://home.att.net/~cferrante/chess/index.html ( Bughouse Problems )



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.