Author: Mathieu Pagé
Date: 16:18:10 07/14/05
Go up one level in this thread
Hi Jay, >There are some old papers by Carmel and Markovitch about opponent modeling in >games. Here's a starting point. > > http://citeseer.ist.psu.edu/carmel95opponent.html > Thanks, I'll look at this. >Opponent modelling is part of agent modeling in general, which there is >literature on. Most agent modeling findings have little application to games, >but you may want to scan around to see if you can glean some ideas. If you want >to do original research, you should learn about agent modeling so that you can >place your research in its broader context. > >One way to look at opponent modeling is this: The performance program has a >general game model, "in this situation, do that." When nothing is known about >the opponent, rely on the general model. An opponent model is a set of tweaks to >the general model. When more is known about the opponent, larger tweaks can be >justified. An opponent can be an individual player (Crafty) or a class of >players (humans vs. computers). The opponent model may be hierarchical: If we >know nothing about the opponent, use the general model; if we know the opponent >is a computer, use the computer tweaks to the model; if we know the opponent is >Shredder, use the Shredder tweaks to the computer model. You seem to think of it as I do. I think it need to be explored and I will probably in the future. When I finish implementing all the other classical algorithms. Thanks for the link and the hint about agent modeling. Mathieu Pagé
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.