Author: Dirk Frickenschmidt
Date: 09:55:29 05/31/98
Hi all of you, We all know how good chess programs on fast hardware have become. The more I am fascinated to see GM positions which seem still to be beyond their understanding. One prominent example is a position from Kasparov-Gelfand, Amsterdam 1996. Here's the critical variation in FEN-format: r1b1r1k1/1p1nbp1p/1q1p2pB/p7/Pn1NP1P1/2N2B2/1PP4P/R2Q1R1K w - - 0 17 Kasparov now played Nf5: that looks a lot like a speculative sacrifice in the Tal tradition. It is very hard for black to find the best defending moves at the board, and in fact Gelfand made a mistake two moves later. I doubt that any of our programs would find Kasparov's move even overnight. Yet it seems to be correct, at least so far that white gets a lot of winning chances without more risk than ending up in a draw, and black has only a very narrow chance to escape the attack. Here's the complete game with rich CBM annotations by Bönsch/Dokhoian/Romero/Tischbierek: (some special characters originate from Chessbase fonts) Kasparov - Gelfand [B92] Amsterdam, 1996 1.e4 Dokhoian Romero Tischbierek 1...c5 2.Sf3 d6 3.d4 cxd4 4.Sxd4 Sf6 5.Sc3 a6 6.Le2 e5 7.Sb3 Le7 8.0-0 0-0 9.Kh1 Sc6 [9...Le6 10.f4 Dc7 11.Sd5 Lxd5 12.exd5 Sbd7 13.c4²; 9...b5!?] 10.f4 a5 A new and interesting idea, despite to loose a tempo. [10...exf4 11.Lxf4 Le6 12.De1 Se8 (12...Sg4 13.Lxg4 Lxg4 14.Dg3 Lh5 15.Sd5 Te8 16.Sxe7+ Dxe7 17.Tae1 De6= 0-1 Sutovskij,E-Gruenfeld,Y/Rishon le Zion (06) ;CBM 45 1994 (37)) 13.Td1 (13.Sd5!) 13...Lh4 14.Lg3 (14.Dd2 Se5 15.Sd4 Tc8 16.Sf5 Lxf5 17.exf5 Db6 18.Dc1± 1-0 Cseshkovsky,V-Skalik,P/Wisla BES (04) 1992 (25)) 14...Lxg3 15.Dxg3 Db6 16.Lg4 Lc4 17.Tf2 Se5 18.Tfd2 Td8 19.Le2 Le6 20.h3 Sf6= Dolmatov,S-Mithrakanth,P/Calcutta op (04) ;EXP 50 1996/ - (48)] 11.a4 Sb4 12.Lf3 Controlling d5. [12.Le3 Dc7 13.Lf3 Le6 14.Tf2 d5 15.exd5 exf4 Cabrilo,G-Kosic,D/JUG-chT Igalo ;EXT 94ch 1994/ - (48)] 12...Db6 [12...Ld7] 13.g4 [13.Te1 Dokhoian 13...Df2 14.Te2 Dh4 15.f5?! d5„] 13...exf4 14.Lxf4 Sd7 15.Sd4 g6 [15...Se5 16.Sf5 A) 16...Lxf5 17.exf5 (17.gxf5 Sxf3 18.Dxf3 Lf6 19.Tad1) 17...Sxf3 18.Dxf3 Dc6 19.Tad1 Tfd8 20.Td2 d5 21.Sb5±; B) 16...Te8÷ ] 16.Lh6 Te8 critical position: 17.Sf5!? Forcing matters 17...gxf5 18.gxf5 Lf6? [18...Kh8! 19.Tg1 Lf8 (19...Tg8? 20.Txg8+ Kxg8 21.Dg1+! Dxg1+ 22.Txg1+ Kh8 23.Lh5 Se5 24.Lg7+ Kg8 25.Lxe5+ Kf8 26.Lf4 Sxc2 27.Sd5+-) 20.Dd2 d5 (20...Sc6? 21.Dg5 Lxh6 22.Dxh6 Dd4 23.e5! Dxe5 24.Tae1+-; 20...Dd8!? 21.Lg5!?©) A) 21.Lxf8 Txf8 22.e5 Dc7 (22...Sxe5 23.Dg5 Sg6 24.fxg6 fxg6 25.Sxd5 Sxd5 26.Lxd5 Dxb2 27.Tae1©) ; B) 21.Le3 21...Lc5 22.Lxc5 Dxc5 23.exd5 (23.f6 Dd6 24.e5 Txe5 25.Dg2 Df8-+) B1) 23...De3 24.Dg2 Dh6 B1a) 25.Tae1 Dokhoian 25...Tf8 26.Se4 Sf6 27.Sd6 Ta6 28.Dg3 Sxc2 29.Te4 Txd6 (29...Sxe4 30.Sxf7+) 30.Dxd6 Sg4 31.Dg3 Sge3 32.De5+ f6 33.Dc7 Sxd5 34.Dg3±; B1b) 25.Tae1© ; B2) 23...Se5? 24.Dh6+-; B3) 23...Ta6 24.Se4 (24.Dg2 Df8µ) 24...Df8 25.c3 Se5 26.Le2©] 19.Tg1+ Kh8 20.Dd2 [20.e5!? Tischbierek] 20...Se5 Black has good squares for their pieces, but no comunication between queen and king side. [20...Dd8 21.Sb5 Se5 22.Sxd6 A) 22...Ta6 23.Sxe8!! Dxd2 (23...Dxe8 24.Dg2 Sxc2 25.Lg7++-) 24.Sxf6 Txf6 (24...Dxh2+ 25.Kxh2 Sxf3+ 26.Kh1 Txf6 27.Lg7++-) 25.Lxd2 Sxf3 26.Lc3+-; B) 22...Tg8 23.Lh5! Txg1+ (23...Ta6? 24.Sxf7+ Sxf7 25.Lxf7±) 24.Txg1 De7 (24...Ld7 Dokhoian 25.Dc3 26.Q:e5) 25.Dg2 Sg6 26.fxg6 fxg6 (26...hxg6 27.Lxg6+-) 27.Lxg6! hxg6 (27...Dxd6 28.Lf7+-) 28.Dxg6 Le6 29.Lf8!+-] 21.Dg2 Sg6 22.fxg6 fxg6 23.Tgf1! [23.Lg4 Romero; 23.Lh5?! Dokhoian 23...Tg8 24.Taf1 Ld4 (RR24...Le5!?) 25.Lg5 Le6÷] 23...Lg7?! Seems logical to propose this change, but black pawn d6 will be very difficult to defend. [23...Le5 defending d6 A) 24.Lg4 Romero xf7 24...Lxg4 25.Dxg4 Sxc2! A1) 26.Tf7 Dxb2 (26...Sxa1 27.Dd7 Db3 28.Txh7+ Kg8 29.Tg7+=; 26...Dxb2? 27.Tg1 RRDxc3 28.Ld2!+-) ; A2) 26.Sd5 26...Dxb2 27.Tab1 Da2 A2a) 28.Dh4 Sb4 (28...Tab8 29.Lf4) 29.Sc7 Tac8 30.Sxe8 Txe8 31.Tbc1²; A2b) 28.Txb7?! 28...Se3; B) 24.Tad1 24...Le6 (24...Ld7 Dokhoian 25.Lg4!? Lc6 26.Tf7 Lxc3 27.Lf5! Tg8 28.bxc3 Sd3!? 29.Dg5! Sf2+ 30.Kg2 Sxe4 31.Lxe4 Lxe4+ 32.Kg3 Dd8 33.Te7+-) 25.Lg4 B1) 25...Lxg4? 26.Dxg4 B1a) 26...Te7 27.Df3 (27.Td2±) 27...Tee8 28.Df7 Sxc2 29.Tf6 Tg8 30.Tfxd6 Dxb2 31.Td7+-; B1b) 26...Sxc2 27.Tf7 Dc6 28.Tg1! Tg8 (28...Lxc3? 29.Txh7+! Kxh7 30.Dxg6+ Kh8 31.Lg7++-) 29.Dh4 Lg7 30.Lxg7+ Txg7 31.Df6 Tag8 32.Txg6!!+-; B2) 25...Lc4 26.Tf3ƒ; 23...Dd8 Romero 24.Tad1] 24.Lxg7+ Kxg7 25.Tad1 [25.Lg4 Romero 25...Dc6 26.Df3 Lxg4 27.Df6+= Kh6? 28.Df4+ Kh5 29.Tg1; 25.Sb5?! Dokhoian 25...Ld7 26.Tad1 Lxb5 27.axb5 Dxb5÷] 25...Ld7 Avoiding the jump Nb5 [25...Le6 26.Sb5 Tad8 (26...Lc4 Romero 27.Sxd6 Lxf1 28.Sxe8+ Txe8 29.Txf1) 27.c3±; 25...Te5 Romero 26.Sb5!] 26.Dg3 Te5 [26...Sxc2 27.Txd6 Lc6 28.Tg1+-] 27.Lg2 Black has a lot of problems. Qf4 xf6 [27.Le2!? Dokhoian Be2-c4] 27...Sxc2?! [27...Tf8 28.Txf8 Kxf8 29.Tf1+ (29.Df4+ Romero) 29...Kg8 (29...Kg7 30.Dh4 g5 31.Dh5 Le8 32.Df3 h6 33.Df8+ Kh7 34.Tf6 Lg6 35.Txd6+-) 30.Dh4 Te8 31.Df6 (31.e5?! Dokhoian 31...Lc6) 31...Dd8 32.Dxd6 Lc6±; 27...Tae8 28.Dh4 Le6 29.Df6+ Kg8 30.Sb5 Txb5 31.axb5 Sxc2 32.Lh3+-] 28.Sd5 Dd8 [28...Dxb2 29.Df2 Lc6 30.Df6+ Kh6 31.Tf4 Sd4 32.Th4+ Th5 33.Txh5+ Kxh5 34.Sf4+ Kh6 35.Se6+-] 29.Df2 Sb4 30.Sb6! Lg4 31.Df7+ Kh8 32.Sxa8 Lxd1 33.Txd1 Dxa8 34.Df6+ Kg8 35.Txd6 Te8 [35...Sc6 36.Txc6 bxc6 37.Dxe5+-] 36.Td7 1-0 I am convinced that Kasparov's sacrifice is correct, because in 1997 the same sacrifice was played in a corresponcence chess game. The player commented this in a German chess magazine now (Schach 1998-6, page 76f). Although in this case black - after 22 days :-))) of anylizing - did not accept the sacrifice, white still won. Herschel-Hertel "50 years BdF" 1997 1.e4 c5 2.Sf3 d6 3.d4 cxd4 4.Sxd4 Sf6 5.Sc3 a6 6.Le2 e5 7.Sb3 Le7 8.0-0 0-0 9.Kh1 Sc6 10.f4 a5 11.a4 Sb4 12.Lf3 Db6 13.g4 exf4 14.Lxf4 Sd7 15.Sd4 Here Mr. Herschel comments that he analyzed the Kasparov-game and needed 20 days to find out that 15...g6 was not good for black. I translate a funny remark of his: "Although my chess programs judged the position wrong, probably not surprising anyone, without their help during anylysis I surely would have needed double as much time." ;-) The game continued: Se5 16.Sf5 Te8 17.Le2 Lxf5 18.gxf5 Lf6 19.Lb5 Ted8 20.Dd2 Tac8 21.Lg5 Lxg5 22.Dxg5 f6 23.Dg2 Dc7 24.Tf2 Df7 25.Td2 Dh5 26.Tad1 Dh4 27.Sb1 One more interesting position. Mr. Herrschel: "Only this move ... helps white to get decisive advance." None of his programs favoured this move: Fritz5 and Junior 4.6 did not even have it among the five best variations. Only Hiarcs counted the move as 3rd or fourth best move, but for wrong reasons judged from the mainline it expected. The game continued: Kf8 28.Sa3 Tc7 29.c3 Sa6 30.Sc2 Sc5 31.Te2+- Line Black lost by a formal mistake after already being in trouble soon after. I would be very how different programs judge a) Kasparovs Nf5! (or some of the critical variations after that) b) the move Nb1! in the second game from Herschel I will perhaps test some of them myself on a 200MMX/64Mb. I know some of you have much faster machines. Kind regards from Dirk
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.