Author: James T. Walker
Date: 06:47:12 12/31/98
Go up one level in this thread
On December 31, 1998 at 08:55:03, Harald Faber wrote: >On December 31, 1998 at 07:38:05, Inmann Werner wrote: > > >>Can anybody tell me how book learning works? >>The principle should be, mark good moves as good and bad moves as bad. >>But how can a program determine, what was a good or bad move? >>Werner > >Reduce the playing preference when you lose and raise when you win. >Other step of learning is a change in eval, e.g. from +2.20 to -1.50 then there >went s.th. wrong... >>>>>>>>>>>>>>>>>>>>>>> Hello, This "Book Learning" thing seems over simple to me. I wish a programmer would get on here and explain the real fine points to me also. I'm bothered by the fine points of the whole thing. For instance, when I play Junior vs Nimzo, Junior's book settings change in an attempt to only use winning lines (I assume). This sounds good but what happens when I now take Junior and play against Fritz ? Just because certain lines did not work good against Nimzo does not mean they are not good against Fritz does it ? After 84 games against Nimzo, Junior was left playing only C4. The first 10 games it played against Fritz 5.32 it got beat up. Last nite I "Cleared" the book settings for Junior and started an auto232 match with Fritz 5.32 at Game/60. This morning the score for Junior was 3 wins and 3 draws - no loses! So it seems to me for book learning to work, you need to play a lot of games against one opponent. Then if you change opponents, you need to reset the values to zero and start again. Is this the way it should work ?? Jim Walker
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.