Author: Uri Blass
Date: 06:42:39 08/25/00
Go up one level in this thread
On August 25, 2000 at 09:18:13, Robert Hyatt wrote: >On August 25, 2000 at 04:08:14, Uri Blass wrote: > >>I thought that the fact that nimzo got mate in 11 out of book against SOS >>and a winning position against shredder was a result of impressive opening >>preperation of alexander kure but I found that all the positions in the book >>were from the history of human-human games and not novelties of alexander. >> >>The fact that it is possible to build a better opening book by using only >>history games of humans without inventing something new is disappointing. >> >>Uri > > >You overlook the main point. Just because humans have played the move doesn't >matter. Someone has to tell the program to follow that variation. _that_ >takes a huge amount of time. I think that programmers should at least tell their program not to follow lines when the evaluation is bad out of book or lines when statistics tell that humans won quickly. In the games of nimzo I found that the opponents (Shredder and Sos) played a line when the experience of human-human games showed a fast win for white. Shredder found a better defence and drew the game but the position was bad for shredder. It was easy to guess the last book moves of nimzo in the game nimzo-shredder because statistics show that white had 100% in humans games and there was only one move that white played in these games(the move of nimzo) I think that programs should usually not go for these lines when history say that the opponent has 100% score. Exceptions can be only when the evaluation gives a very big advantage to the loser side in the human games or if it is a prepared line when the games of the humans was analyzed before. Uri
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.