Author: Dan Homan
Date: 11:49:51 07/28/99
Go up one level in this thread
On July 28, 1999 at 04:18:19, Bruce Moreland wrote: > >On July 27, 1999 at 21:13:03, Jeff Anderson wrote: > >>I won't try to list how enjoyable the engines were to play, I would probably >>find it is an intverted list of the strength list. > >This is a fascinating comment. > >bruce I think this is why people like personalities in programs like Chessmaster and PowerChess. I own PowerChess98 and find the King (who is supposed to adjust his strength to match yours) to be a very fun opponent. Making the computer play like a human opponent with gaps in its knowledge but a well defined style and plan seems like the best approach to weakening a program's play. A next step would be to have the program "learn" concepts as you use those concepts to beat it (Powerchess seems to do this). I am not sure how they do this in Powerchess, but I think the "learning" of the program is fairly well scripted. I am not sure how to do this in a more general way, except perhaps with TD learning. Different classes of player personalities (patzer, club, expert, etc...) could have different values for positional feature pre-set... Then with TD learning, those personalities could improve along with the play of the human. Of course, there should be a feature to reset or freeze the learning of a personality if it starts to become too strong. The only difficulty with TD learning is that it probably takes too many games to show a noticable improvement. This is only a guess, does anyone know how this might work? - Dan
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.