Author: James T. Walker
Date: 10:18:12 01/17/01
Go up one level in this thread
On January 17, 2001 at 13:07:59, Uri Blass wrote: >On January 17, 2001 at 12:58:22, James T. Walker wrote: > >>On January 17, 2001 at 11:29:33, Mogens Larsen wrote: >> >>>On January 17, 2001 at 11:22:39, James T. Walker wrote: >>> >>>>Does anyone know if the SSDF is allowing "duplicate" games in their testing? >>> >>>I think they do allow doubles and so they should. Repeating winning lines is an >>>obvious thing to do and if the opponent can't diverge at some point then it's >>>just too bad. >>> >>>>Junior 6 is the worst program I know of for trying to duplicate won games. I >>>>would like to know if it is inflating it's rating with this method. >>> >>>Is Junior different from other ChessBase programs in this area? I thought that >>>the book learning was identical. Deep Fritz won't suffer on this account AFAIK. >>> >>>Mogens. >> >>It is not the book learning but the agressive use of it. Fritz does not try to >>repeat won games forever. They all use book learning to avoid repeating games >>that are lost. This is understandable. However Junior will repeat the same >>opening as white as long as it continues to win. I have seen Junior win 14 >>straight games as white Vs Fritz when the Fritz learning feature was broken as >>black. This problem was fixed in a patch by Chessbase. I have watched Junior 6 >>play the same opening vs CM8K as white 5 times in a row before I just gave up >>and stopped playing them against each other. Junior takes it to the extreme. >>Jim > >I remmeber that Fritz5 did it a long time ago in the ssdf games against Rebel. > >It is known that Chessbase programs try to repeat the same winning opening again >and again and I did not know that chessbase fixed it in Fritz6. > >I think that saying it about Junior is not fair because Amir ban is not >responsible for the learning function of Junior. > >Chessbase is responsible for it and they used the learning algorithm of Junior6 >for more programs including Fritz5(I do not know about Fritz6). > >Uri Hello Uri, What Chessbase fixed was the broken part of the learning by Fritz. Fritz at one time for about 2-3 months did not learn as Black. When I played a match of Junior vs Fritz I saw the 14 consecutive wins by Junior as white vs Fritz as black. At that time Fritz was learning with white but not black. I told Chessbase about it and they came out with another patch that fixed it. I have played Fritz vs CM8K and Fritz did not repeat won games as white. As I said above Junior played the same opening 5 consecutive times as white after winning the first one vs CM8K. All 5 games were identical through 38 moves. At that point Junior had a very nice advantage. So they did not act the same. Junior seems much more agressive in using it's "learned knowledge". That's why I asked if the SSDF is allowing duplicate games now. If they are I would sure be interested in some statistics on games duplicated by Junior and other programs. Jim
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.