Author: Chessfun
Date: 07:51:22 08/26/02
Go up one level in this thread
On August 26, 2002 at 10:39:15, Uri Blass wrote: >On August 26, 2002 at 10:09:30, Chessfun wrote: > >> >>You can talk about books and book learning but position learning can be just as >>extreme. In my lightning matches Gandalf_432h won 13 games out of a total played >>40 with the same book line that once it scored the first win it's position >>learning took over. This was against Fritz7.ctg, Deep Fritz 7 scores the >>position out of book as positive for black by 1/2 a point after a 2 min search, >>Gandalf had white. It also just took 6 also from Deep Fritz 7 using the same >>method, different opening.....position learning. >> >>Had it discovered this line in game 1 it would likely have taken 20 wins against >>any Fritz 7 version same goes for Gandalf 5.1. >> >>Sarah. > >I do not have Fritz7 but I am surprised to read it. >I expected that a professional program not to let things like it. > >I thought that book learning is about learning to avoid losing the same line >again and again. > >I think that it is not hard to prevent it and Fritz could simply remember the >moves in the lost game and decide not to play them(if there is no alternative in >the book it should search for the next best move and play it in case that it is >not losing). Fritz books typically mark the move as negative but at which point it decides to make a change couldn't tell you, otherwise it simply would have opened with d4 instead of e4. It should be remembered though that these games were run in tournament format IOW not 40 straight games against Gandalf. Therefore it may have switched opening moves in it's next white game but when it met Gandalf again was back in the same opening. Sarah.
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.