Author: Uri Blass
Date: 23:32:57 12/15/00
Go up one level in this thread
On December 15, 2000 at 20:57:48, James T. Walker wrote: >On December 15, 2000 at 18:58:42, Uri Blass wrote: > >>On December 15, 2000 at 18:52:25, Uri Blass wrote: >> >>>On December 15, 2000 at 12:37:36, Laurence Chen wrote: >>> >>>>On December 14, 2000 at 20:31:56, James T. Walker wrote: >>>> >>>>>On December 14, 2000 at 13:18:47, Steve Maughan wrote: >>>>> >>>>>>I don't think I've ever seen such a wide margin diminish to defeat! >>>>><snip> >>>>> >>>>>I have only had Shredder for 4 days but I saw this type of behavior a lot at >>>>>blitz on an Athlon 900 Mhz. Even at G/30 this seems to happen. At G/60 on an >>>>>Athlon 900 it has not showed up yet. That's why I made the comment earlier >>>>>"Shredder a mudder". Refering to the fact that Shredder seems to like the >>>>>slower time controls with faster hardware. I suspect there are a lot of holes >>>>>in it's eval that can only be covered up with deeper searches. Just my opinion >>>>>based on 4 days. :-) >>>>>Jim >>>>I had Shredder 5 for a 3 days, and I came with the same conclusion. The >>>>evaluations of Shredder changes as it digs deeper, which means that one way to >>>>help Shredder is to use the fastest computer available, and slower time control. >>>>Regards, >>>>Laurence >>> >>>The evaluation of all programs is changed when they get deeper. >>>There are holes in the evaluations of all chess programs. >>> >>>If you find that the evaluation of Shredder is changed more than other program >>>then it still does not prove that there are bigger holes in Shredder's >>>evaluation because it is possible to explain it by other reasons. >>> >>>Uri >> >>I can add that the +4.68 was not a hole in Shredder's evaluation because >>Shredder had a clearly better position when it evaluated itself as +4.68 >> >>Uri > >Hello Uri, >Well without an indepth analysis I can only say I don't know of any other >programs I have that will lose a game after being up by +4.68. I have seen >several draw after that kind of score and it usually is because they believe >they can win with a minor piece and pawn when it is only a draw. It seems to me >that Shredder missed something big somewhere. Something most top programs would >not miss. >Jim 1)Other programs evaluate the +4.68 as +2 or +3 evaluation so it is possible that Shredder has bigger evaluation 2)I have a game when Gambit lost against gandalf after +3.42 score Here is the relevant position: [D]2r2rk1/R2q1ppp/4p3/8/1Kp5/4QB2/1P4PP/7R b - - 0 1 Gambit evaluated it as +3.42 for itself and it is clearly wrong. 3)I remember another game when Gambit had the same kind of evaluation against Gandalf and lost the game(I remember that black had passed pawns that simply fail to win) when the sides had queen and bishop. 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.