Topalov - Gelfand evaluation by Strelka

[Dirt]

[d]3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w KQkq -

In this position, where Gelfand resigned, Strelka 2.0 B initially sees the correct Nxh6+, with a strong advantage for White. For no good reason it quickly changes to Nf6+ with an advantage to Black.

Surprisingly, Rybka 1.0 Beta 32-bit does even worse!

[genorb]

For me Rybka 1.0 beta 32bits behave has expected for this position (32MB for hash). Here is the analysis under Shredder GUI

3.00 0:00 +8.56 1.Cxh6+ (114) 2
4.00 0:00 +8.84 1.Cxh6+ Rf8 (363) 4
5.00 0:00 +6.45 1.Cxh6+ Rf8 2.Cxf5 (1.117) 12
6.00 0:00 +6.90 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 (2.809) 22
7.00 0:00 +6.90 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe6 (5.825) 31
8.00 0:00 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 (19.016) 49
9.00 0:00 +7.23 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (35.507) 62
10.00 0:00 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (77.626) 95
11.01 0:01 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (172.111) 134
12.01 0:02 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (333.021) 164
13.01 0:03 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (585.572) 170
14.01 0:12 +7.84 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.c4 g6 4.Dh4 Te8 5.Txe8+ Rxe8 (2.127.020) 177
15.01 0:37 +8.15 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Dg5 g6 4.c4 Tc8 5.cxd5 Cxd4 (5.058.141) 136
16.01 1:27 +9.32 1.Cxh6+ Rf8 2.Cxf5 Fxf5 (9.869.298) 116

[Uri Blass]

[d]3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w KQkq -

In this position, where Gelfand resigned, Strelka 2.0 B initially sees the correct Nxh6+, with a strong advantage for White. For no good reason it quickly changes to Nf6+ with an advantage to Black.

Surprisingly, Rybka 1.0 Beta 32-bit does even worse!

There are missing details.
1)How much hash did you use?
2)At what depth do you see the problem?
3)Do you give rybka all the game when you analyze and if yes what is the game?

Strelka 2.0B with 512 mbytes hash after some minutes of search has no problems when it gets the position without previous moves

New game
[d]3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w - - 0 1

Analysis by Strelka 2.0 B:

1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5
+- (6.96) Depth: 1 00:00:00
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5
+- (6.96) Depth: 2 00:00:00
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.h3-h4
+- (7.03) Depth: 3 00:00:00
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e6
+- (6.71) Depth: 4 00:00:00
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e6 4.f4-f5
+- (6.83) Depth: 5 00:00:00
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Nc6-e7 4.Qg3-h4 Bf5-e6 5.Qh4-h8+ Ne7-g8
+- (7.06) Depth: 6 00:00:00 6kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e6 4.f4-f5 Be6xf5 5.Bf3xd5
+- (7.20) Depth: 7 00:00:00 15kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e6 4.Qg3-g5 f7-f6 5.Qg5-h5 Be6-f7 6.Qh5-h8+ Bf7-g8
+- (7.09) Depth: 8 00:00:00 31kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e6 4.Bf3-g4 Be6xg4 5.Qg3xg4 Nc6-e7 6.Qg4-h4 f7-f6 7.Qh4-h8+ Ne7-g8
+- (7.43) Depth: 9 00:00:00 68kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e4 4.Bf3xe4 d5xe4 5.Re1xe4 d6-d5 6.Re4-e1 g7-g6 7.f4-f5 g6xf5
+- (7.64) Depth: 10 00:00:00 155kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e4 4.Bf3xe4 d5xe4 5.Re1xe4 d6-d5 6.Re4-e1 Kf8-g8 7.Kg1-f2 Rd8-c8 8.Qg3-d3
+- (7.79) Depth: 11 00:00:02 262kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e4 4.Bf3xe4 d5xe4 5.Re1xe4 d6-d5 6.Re4-e2 Kf8-g8 7.f4-f5 Rd8-c8 8.f5-f6 g7-g6
+- (7.86) Depth: 12 00:00:03 529kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e4 4.Bf3xe4 d5xe4 5.Re1xe4 d6-d5 6.Re4-e2 Kf8-g8 7.f4-f5 Rd8-c8 8.f5-f6 g7-g6 9.Qg3-d6
+- (7.96) Depth: 13 00:00:04 1117kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Bh5-f3 Bf5-e4 4.Bf3xe4 d5xe4 5.Re1xe4 d6-d5 6.Re4-e2 Kf8-g8 7.f4-f5 Rd8-c8 8.f5-f6 g7-g6 9.Qg3-d6 Rc8-d8
+- (8.06) Depth: 14 00:00:05 2156kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Qg3-g5 Bf5-d7 4.f4-f5 Rd8-e8 5.Re1xe8+ Bd7xe8 6.f5-f6 g7xf6 7.Qg5xf6 b7-b5 8.Qf6xd6+ Nc6-e7 9.Bh5-f3 Be8-c6 10.g2-g4 f7-f5 11.Qd6-f6+ Kf8-e8 12.g4xf5
+- (9.18) Depth: 15 00:00:08 9087kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Qg3-g5 g7-g6 4.Qg5-f6 Kf8-g8 5.Bh5-f3 Rd8-f8 6.Bf3xd5 Nc6-d8 7.Re1-e7 Bf5-e6 8.Bd5xe6 Nd8xe6 9.d4-d5 Ne6-c5 10.Qf6xd6
+- (9.23) Depth: 16 00:00:10 12901kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Qg3-g5 Bf5-d7 4.f4-f5 Rd8-e8 5.Re1xe8+ Bd7xe8 6.f5-f6 g7xf6 7.Qg5xf6 Nc6-e7 8.Qf6-h6+ Kf8-g8 9.Qh6xd6 Kg8-f8 10.Qd6-b8 a7-a5 11.Qb8xb7 Be8-c6 12.Qb7-c7
+- (9.32) Depth: 17 00:00:17 21335kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Qg3-g5 Bf5-d7 4.f4-f5 Rd8-e8 5.Re1xe8+ Bd7xe8 6.f5-f6 g7xf6 7.Qg5xf6 Nc6-e7 8.Qf6-h6+ Kf8-g8 9.Qh6xd6 Kg8-f8 10.Qd6-b8 a7-a5 11.Qb8xb7 Be8-c6 12.Qb7-c7 a5-a4
+- (9.41) Depth: 18 00:00:31 37178kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Qg3-g5 Bf5-d7 4.f4-f5 Rd8-e8 5.Re1xe8+ Bd7xe8 6.f5-f6 g7xf6 7.Qg5xf6 Nc6-e7 8.Qf6-h6+ Kf8-g8 9.Qh6xd6 Kg8-f8 10.Qd6-b8 b7-b6 11.Qb8xa7 b6-b5 12.Kg1-f2 f7-f6 13.Bh5-g4
+- (9.53) Depth: 19 00:00:56 68877kN
1.Ng4xh6+ Kg8-f8 2.Nh6xf5 Bd3xf5 3.Qg3-g5 Bf5-e4 4.Re1xe4 d5xe4 5.Qg5-f5 g7-g6 6.Bh5xg6 Rd8-d7 7.Qf5xd7 f7xg6 8.Qd7xd6+ Nc6-e7 9.Kg1-f2 b7-b5 10.Qd6-b8+ Kf8-f7 11.Qb8xb5 Ne7-f5 12.Qb5-b7+ Kf7-e6 13.Qb7xe4+ Ke6-d6 14.Qe4-b7
+- (10.98) Depth: 20 00:02:55 222044kN

(, 17.01.2008)

Uri

[Dirt]

For me Rybka 1.0 beta 32bits behave has expected for this position (32MB for hash). Here is the analysis under Shredder GUI

3.00 0:00 +8.56 1.Cxh6+ (114) 2
4.00 0:00 +8.84 1.Cxh6+ Rf8 (363) 4
5.00 0:00 +6.45 1.Cxh6+ Rf8 2.Cxf5 (1.117) 12
6.00 0:00 +6.90 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 (2.809) 22
7.00 0:00 +6.90 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe6 (5.825) 31
8.00 0:00 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 (19.016) 49
9.00 0:00 +7.23 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (35.507) 62
10.00 0:00 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (77.626) 95
11.01 0:01 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (172.111) 134
12.01 0:02 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (333.021) 164
13.01 0:03 +7.29 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Ff3 Fe4 4.Fxe4 dxe4 5.Txe4 d5 (585.572) 170
14.01 0:12 +7.84 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.c4 g6 4.Dh4 Te8 5.Txe8+ Rxe8 (2.127.020) 177
15.01 0:37 +8.15 1.Cxh6+ Rf8 2.Cxf5 Fxf5 3.Dg5 g6 4.c4 Tc8 5.cxd5 Cxd4 (5.058.141) 136
16.01 1:27 +9.32 1.Cxh6+ Rf8 2.Cxf5 Fxf5 (9.869.298) 116

That's interesting. Rybka just crashes on my computer, and I've never had that problem on any other position.

[Steve Maughan]

[d]3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w KQkq -

This epd is invalid as it has castling rights for both Black and White. Maybe that's giving the engines / GUIs a problem.

Steve

[Dann Corbit]

[d]3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w KQkq -

In this position, where Gelfand resigned, Strelka 2.0 B initially sees the correct Nxh6+, with a strong advantage for White. For no good reason it quickly changes to Nf6+ with an advantage to Black.

Surprisingly, Rybka 1.0 Beta 32-bit does even worse!

Code: Select all

Analysis from e:\nxh6.epd   
1/17/2008 3:20:41 PM Level: 410 Seconds
Analyzing engine: Rybkav2.3.2a.w32

1) Nxh6+; c0 "Topalov - 
    Searching move: Ng4xh6
    Best move (Rybkav2.3.2a.w32): Ng4xh6
    identical moves! Found in: 00:00
      5	00:00	         744	6.863	+6.56	Ng4xh6+ Kg8f8 Nh6xf5
      6	00:00	       1.331	12.278	+6.57	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3
      7	00:00	       3.253	23.458	+6.85	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6
      8	00:00	       5.308	31.418	+6.66	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6 Bf3g4
      9	00:01	       9.995	43.368	+6.67	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6 Bf3g4 Be6xg4 Qg3xg4 Nc6e7
     10	00:01	      39.544	66.273	+7.13	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e1 Rd8c8
     11	00:01	      69.229	68.625	+7.32	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e1 f7f6 f4f5
     12	00:02	     142.788	70.840	+7.38	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e1 f7f6 f4f5 Rd8d7
     13	00:04	     231.741	69.631	+7.37	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e1 f7f6 f4f5 Rd8d7
     14	00:13	     912.048	68.460	+8.28	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5h7 Qg5xd5 Rd8d7 Bh5xf7 Rd7xf7 Qd5xd6+ Rf7e7 Re1xe7 Nc6xe7
     15	00:24	   1.524.842	66.218	+8.46	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5c2 Qg5xd5 g7g6 Qd5b5 Rd8d7 d4d5 Nc6b8 Qb5e2 Kf8g7
     16	00:58	   3.773.988	66.790	+8.82	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 g7g6 Qg5f6 Kf8g8 Bh5f3 Rd8f8 Bf3xd5 Nc6d8 Qf6xd6 a7a5
     17	01:14	   4.759.769	66.212	+8.85	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 g7g6 Qg5f6 Kf8g8 Bh5f3 Rd8f8 Bf3xd5 Nc6d8 Qf6xd6 a7a5
     18	01:47	   6.864.504	65.847	+8.99	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 g7g6 Qg5f6 Kf8g8 Bh5f3 Rd8f8 Bf3xd5 Nc6d8 Qf6xd6 a7a5
     19	03:37	  15.303.179	72.484	+9.17	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 g7g6 Qg5f6 Kf8g8 Bh5f3 Rd8f8 Bf3xd5 Nc6d8 Qf6xd6 a7a5
   1/17/2008 3:27:35 PM, Time for this analysis: 00:06:51, Rated time: 00:00

1 of 1 matching moves
1/17/2008 3:27:35 PM, Total time: 12:06:54 AM
Rated time: 00:00 = 0 Seconds

--------------------------------------------------------------------------------

Analysis from e:\nxh6.epd   
1/17/2008 3:20:41 PM Level: Infinite
Analyzing engine: Strelka_2_0

1) Nxh6+; c0 "Topalov - 
    Searching move: Ng4xh6
    Best move (Strelka_2_0): Ng4xh6
    identical moves! Found in: 00:00
      2	00:00	         118	7.375	+6.96	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5
      3	00:00	         393	24.562	+7.03	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 h3h4
      4	00:00	       1.172	73.250	+6.71	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6
      5	00:00	       2.614	163.375	+6.83	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6 f4f5
      6	00:00	       5.893	368.312	+7.06	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Nc6e7 Qg3h4 Bf5e6 Qh4h8+ Ne7g8
      7	00:00	      14.656	458.000	+7.20	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6 f4f5 Be6xf5 Bf3xd5
      8	00:00	      31.157	662.914	+7.09	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6 Qg3g5 f7f6 Qg5h5 Be6f7 Qh5h8+ Bf7g8
      9	00:01	      67.670	856.582	+7.43	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e6 Bf3g4 Be6xg4 Qg3xg4 Nc6e7 Qg4h4 f7f6 Qh4h8+ Ne7g8
     10	00:01	     154.862	1.098.312	+7.64	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e1 g7g6 f4f5 g6xf5
     11	00:01	     261.778	1.195.333	+7.79	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e1 Kf8g8 Kg1f2 Rd8c8 Qg3d3
     12	00:01	     528.903	1.299.515	+7.86	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e2 Kf8g8 f4f5 Rd8c8 f5f6 g7g6
     13	00:01	   1.117.063	1.347.482	+7.96	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e2 Kf8g8 f4f5 Rd8c8 f5f6 g7g6 Qg3d6
     14	00:02	   2.155.803	1.393.537	+8.06	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Bh5f3 Bf5e4 Bf3xe4 d5xe4 Re1xe4 d6d5 Re4e2 Kf8g8 f4f5 Rd8c8 f5f6 g7g6 Qg3d6 Rc8d8
     15	00:07	   9.086.929	1.421.832	+9.18	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5d7 f4f5 Rd8e8 Re1xe8+ Bd7xe8 f5f6 g7xf6 Qg5xf6 b7b5 Qf6xd6+ Nc6e7 Bh5f3 Be8c6 g2g4 f7f5 Qd6f6+ Kf8e8 g4xf5
     16	00:10	  12.900.678	1.418.592	+9.23	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 g7g6 Qg5f6 Kf8g8 Bh5f3 Rd8f8 Bf3xd5 Nc6d8 Re1e7 Bf5e6 Bd5xe6 Nd8xe6 d4d5 Ne6c5 Qf6xd6
     17	00:17	  21.334.502	1.426.675	+9.32	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5d7 f4f5 Rd8e8 Re1xe8+ Bd7xe8 f5f6 g7xf6 Qg5xf6 Nc6e7 Qf6h6+ Kf8g8 Qh6xd6 Kg8f8 Qd6b8 a7a5 Qb8xb7 Be8c6 Qb7c7
     18	00:30	  37.177.933	1.411.248	+9.41	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5d7 f4f5 Rd8e8 Re1xe8+ Bd7xe8 f5f6 g7xf6 Qg5xf6 Nc6e7 Qf6h6+ Kf8g8 Qh6xd6 Kg8f8 Qd6b8 a7a5 Qb8xb7 Be8c6 Qb7c7 a5a4
     19	00:54	  68.877.121	1.424.729	+9.53	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5d7 f4f5 Rd8e8 Re1xe8+ Bd7xe8 f5f6 g7xf6 Qg5xf6 Nc6e7 Qf6h6+ Kf8g8 Qh6xd6 Kg8f8 Qd6b8 b7b6 Qb8xa7 b6b5 Kg1f2 f7f6 Bh5g4
     20	02:47	 222.043.678	1.394.028	+10.98	Ng4xh6+ Kg8f8 Nh6xf5 Bd3xf5 Qg3g5 Bf5e4 Re1xe4 d5xe4 Qg5f5 g7g6 Bh5xg6 Rd8d7 Qf5xd7 f7xg6 Qd7xd6+ Nc6e7 Kg1f2 b7b5 Qd6b8+ Kf8f7 Qb8xb5 Ne7f5 Qb5b7+ Kf7e6 Qb7xe4+ Ke6d6 Qe4b7
   1/17/2008 3:30:30 PM, Time for this analysis: 00:02:48, Rated time: 00:00

1 of 1 matching moves
1/17/2008 3:30:31 PM, Total time: 12:09:50 AM
Rated time: 00:00 = 0 Seconds

[Ben Lau]

n this position, where Gelfand resigned, Strelka 2.0 B initially sees the correct Nxh6+, with a strong advantage for White. For no good reason it quickly changes to Nf6+ with an advantage to Black.
Surprisingly, Rybka 1.0 Beta 32-bit does even worse!

You should check what Gui you are using now. As an engine under different GUIs will have different behavior.

Under chessbase GUI it works well. But it will make some odd mistakes under some other guis.

[Dann Corbit]

n this position, where Gelfand resigned, Strelka 2.0 B initially sees the correct Nxh6+, with a strong advantage for White. For no good reason it quickly changes to Nf6+ with an advantage to Black.
Surprisingly, Rybka 1.0 Beta 32-bit does even worse!

You should check what Gui you are using now. As an engine under different GUIs will have different behavior.

Under chessbase GUI it works well. But it will make some odd mistakes under some other guis.

I used Arena, but I also corrected the castling rights. I guess that is what was causing the problem.
Try this version with the software that was giving you trouble:
3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w - - bm Nxh6+; c0 "Topalov - Gelfand";

I guess that the real problem was the invalid castling rights throwing some programs and/or interfaces for a loop

[Dirt]

I used Arena, but I also corrected the castling rights. I guess that is what was causing the problem.
Try this version with the software that was giving you trouble:
3r2k1/pp3pp1/2np3p/3p1q1B/3P1PN1/2Pb2QP/P5P1/4R1K1 w - - bm Nxh6+; c0 "Topalov - Gelfand";

I guess that the real problem was the invalid castling rights throwing some programs and/or interfaces for a loop

That seems to be right. To follow the game online I had set up the position in WinBoardF to some point after castling had occurred, but WinBoardF appears to always leave castling rights unchanged after an edit. This seems to be a new behavior, probably done to support fairy chess. The earlier versions of Winboard know better than to do this.

Thanks to everyone who looked at this, especially Steve who first noted the bad FEN.