Author: Jon Dart
Date: 17:03:08 01/02/01
Go up one level in this thread
On January 02, 2001 at 16:10:53, Christian Söderström wrote: > >Something like that definately works. Isn't a float a bit >overkill though? (No pun intentended) You could use an integer, of course. Floats are handy because you are going to be making small adjustments to the values in the course of learning. If you try to adjust an integer score of 3 down by 10%, nothing happens. If you adjust 3.0 down by 10% you get 2.7. So with integers you have to be careful if the scores are small or approach zero. > >Anyway I always figured one of the big points of adding >learning was that the program would avoid playing openings >which it plays poorly long-term, not just those that result >in bad positions a few moves out of book. For instance, >Mint often plays the french with black now, and does very >poorly (especially against humans), however it doesn't >understand that it's doing badly (and in fact it might >objectively be fine) until it sees a material-winning >or mating attack 30 moves in the game. > >So isn't it possible to supplement your idea by also >learning from the end-result of a game? > Learning from the end result is possible but there's the problem that you can win with a bad opening against a poor opponent, or lose with a good opening against a very good opponent. Re avoiding generally bad openings: I also use a small file with manually tuned weights when I build my book. So certain openings are never played. Several variations of the French are in this category. --Jon
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.