Author: blass uri
Date: 09:32:48 04/01/00
Go up one level in this thread
On April 01, 2000 at 10:50:46, Harald Faber wrote: >On March 31, 2000 at 11:20:39, Shanti wrote: > >>Hi >> >>Shredder 4 - Fritz 6 0-1 >> >>Shredder out-played Fritz 6 in the middle game and won a knight for a pawn. AT >>this point I left the hall while I was sure the game is over. >>When I returned Shredder was lost!!!! It seems that in the endgame (2R+N+pawns >>vs. 2R+pawns) Shredder went for Fritz king side pawns and let Fritz get 2 >>passing pawns on the queen side. > >First I thought this is the losing position, where white plays Nf5 with rook >exchange but loss of the white a-pawn. But try against Tiger, it shows 41.Rb7! >This is a very fine move which keeps the advantage and cares for the connected >pawns. Main line is: > >41.Rb7 Ra2 42.Ng4! a5 43.h4! (creates a free pawn for white) gxh4 44.gxh4 e3 >45.Nxe3 Ra4 46.Kg3 Ra1 47.h5 Rh1 48.Kg4 a4 49.Kg5 a3 50.Ra7 with a clearly won >position. After 12min Tiger changes to 50.Nf5 +7.xx. >It is probably worth trying the other top programs, I will try Rebel and Junior >on it. > >BTW I have some games too where Shredder already evaluates +6 or +7 and ends in >a draw, but these cases are rare. I think the losing blunder was later. I think 49.h4 was a bad move and white could do better by Nd4+ Kc5 Nc2 but it is not easy to know the losing blunder or the missing win blunder and probably white could win also after 49.h4. White had alternatives(for example it could play 51.Ra4 instead of 51.Nd5+) It is also not clear that the queen endgame was lost and maybe shredder could draw by perpetual check later. This game only proves that programs sometimes do stupid blunders that even I know to avoid. It seems that humans should go for positions with many passed pawns against shredder. Against other programs different ideas can help to win(other programs can play better the endgame in this case and do not play stupid blunders like h4 but they probably have other problems). I hope that this game is going to convince the people who believe that program deserve 2700 fide rating and expected Deep Junior to win Xie 5:1 to understand that they are wrong. I hope humans are going to start to play against the opponent weakness and in this case we are going to see the performance of programs against humans go down to 2300-2400. Unfortunately most humans are not motivated to do it. 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.