Author: blass uri
Date: 13:42:00 03/30/00
Go up one level in this thread
On March 30, 2000 at 15:59:40, Dave Gomboc wrote: >On March 30, 2000 at 02:32:46, blass uri wrote: > >>On March 29, 2000 at 23:42:32, Lawrence S. Tamarkin wrote: >> >>>I think you have over-simplified your opinion about the different opening books >>>that go into making the books for the programs. I think a lot of thought goes >>>into making the best available book, especially for a program like Fritz6 or >>>Junior6 which has large books, (over 120 mbs), and also has 'Tournament use' >>>turned off at default, because the program is designed to aim for positons from >>>the book that favour its best play. Also, I specifically bought Junior6, >>>because the opening book was prepared by GM Boris Alterman. >> >>I also thought that GM will do a better book and I see that I was wrong because >>when I look at Junior's games I see that Junior does not get a better position >>in most of the games against computers. >> >>I suspect that alterman did not work seriously for improving Junior's book and >>the only reason that they gave a GM to work for Junior's book is not to improve >>Junior's book but to tell customers that the book was prepared by a GM. >> >>It also seems that the book is based on games and there is no case when Junior >>plays a move that was not played in more than one game if you optimize book for >>playing and I think that it is a mistake. >> >>I read that in one of the games of Junior against zarkov one of the moves(Qxb2) >>was marked by ? but Junior did not find a move that was played in more than one >>game so it had to calculate and finally played the bad move Qxb2. >> >>I do not understand it because if alterman analyzed the relevant position and >>found that Qxb2 is bad then Junior should play the move that alterman believes >>that is best without caring about number of games. >> >>Uri > >Deep Junior seems to have a parameter to set this with, Uri. You can tell it to >ignore the prepared book when a line has been played too infrequently, but the >cutoff for how many times is infrequent is adjustable. > >Dave The point is that I understood that if you click on optimize book then the result is that it ignores preperations if the line was not played more than once(this is what happened in the game Zarkov-Junior6a from mclane's tournament). I expect optimize book to give the best results and it is not logical always to ignore preperations when there are no games or only one game. I think that the prepared book should have also information for every move about the minimal number of games that you need to play this move, so it can let you in part of the cases to play a move that is not based on previous games. 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.