Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Opponent-modeling in computer chess

Author: GuyHaworth

Date: 16:07:43 07/15/05

Go up one level in this thread



Check out Jeroen Donkers Ph.D. thesis 'Nosce Hostem', i.e. 'Know thine enemy'.

Jeroen also tried out Probabilistic Opponent Modelling (PrOM), where the
opponent is Opponent(i) with Probability p(i), a probability that can be refined
by Bayesian techniques.  I think he found that this was not a good use of a
finite amount of computing time compared with simpler approaches to minimax.


While Jeroen was doing his Ph.D., I independently created an opponent model
specific to endgames with computed EGTs.  This model can be used in a PrOM to
calculate the 'optimum move', and the probabilities p(i) can be changed by
Bayesian analysis. This goes beyond Jansen's work on the endgame by proposing a
whole spectrum of opponent models.

Articles in ICGA_J, and (with Rafael Andrist) in the ACG (Graz) book and (to
come) in the Journal of Theoretical Computer Science (Elsevier).


g



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.