Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: An algorithm for KBN vs. K

Author: Dieter Buerssner

Date: 18:21:23 09/04/04

Go up one level in this thread


Not an exact answer to your questions, but perhaps still helpful. From a post of
mine in the Winboard forum:


Perhaps not trivial, but I think not so hard either. The following will work.
Score KBNK for the winner side as:

score = c1 - c2*edge_dist(losing_K) - c3*corner_dist(losing_k) -
c4*dist(winning_k,losing_k) - c5*dist(winning_k, knight).

edge_dist is the minimum of 4 distances to the edges of the board. corner_dist
minimum of the 2 distances to the correct colored corner. Probably not even all
of the terms are needed. I tried that formula (motivated by the thread in the
German forum mentioned) with a maximum mate distance position in KBNK (mate in
33) and with game in one minute, it worked vs an EGTB equipped opponent. I used
c1=1000, c2=16, c3=8, c4=3 and c5=1 (it was my first guess, I did not try
anything else). Probably a small centeralization bonus for the winning K could
also help (so, that it will not try to keep the distance small to the losing K
by going too close to the edge).

Regards,
Dieter




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.