(Very?) hard test position

[smirobth]

[d] 8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - - 0 36 bm 36...Bg7

36...Bg7 is the only move that doesn't lose. However programs can get this move for the wrong reason (many programs see several moves as more or less equivalent) so the best way to use this position as a test is in show top two moves mode. There should be a moderate to large difference (at least 1/4 pawn) between 36...Bg7, which is good enough to hold the draw, and all other moves, which lose. Do any programs solve this?

[Eelco de Groot]

[d] 8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - - 0 36 bm 36...Bg7

36...Bg7 is the only move that doesn't lose. However programs can get this move for the wrong reason (many programs see several moves as more or less equivalent) so the best way to use this position as a test is in show top two moves mode. There should be a moderate to large difference (at least 1/4 pawn) between 36...Bg7, which is good enough to hold the draw, and all other moves, which lose. Do any programs solve this?

Hello Robin,

Toga plays 1...Bf6. Could you maybe give some details why Bg7 is the best move? Is this a position from one of your own games?

Regards, Eelco


8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - -

Engine: Toga 1.3 "Checkov" 2c (256 MB, Athlon 2009 MHz)
gemaakt door Thomas Gaksch and Fabien Letouzey

1/03 0:00 +0.39 1...cxb5 2.axb5 Lxc3 (6)

1/03 0:00 +0.53 1...Lxc3 2.bxc6 (9)

2/09 0:00 +0.07 1...Lxc3 2.bxc6 bxc6 3.a5 Le5 (264)

2/09 0:00 +0.39 1...cxb5 2.axb5 Lxc3 (300)

3/11 0:00 +0.22 1...cxb5 2.axb5 Lxc3 3.Ld4 (1.869)

4/11 0:00 +0.09 1...cxb5 2.axb5 Lxc3 3.Ld4 Lxd4+
4.Kxd4 b6 (2.539)

5/12 0:00 +0.04 1...cxb5 2.cxb5 Lxc3 3.a5 Le5 4.Ld4 (5.356)

6/14 0:00 -0.42 1...cxb5 2.cxb5 Kd5 3.Ld4 Lc7 4.Lf6 Ld6 (13.347)

7/15 0:00 -0.47 1...cxb5 2.cxb5 Kd5 3.Ld4 Lc7 4.Lf6 Kc4
5.Ld4 (28.727)

8/18 0:00 -0.79 1...cxb5 2.cxb5 Lf6 3.Ld4 Lg5+ 4.Ke2 Ld8
5.c4 Lf6 6.Lxf6 Kxf6 (59.770)

9/20 0:00 -0.79 1...cxb5 2.cxb5 Lf6 3.Ld4 Ld8 4.c4 La5
5.c5 Kd5 6.c6 bxc6 (121.991)

10/31 0:00 -1.77 1...cxb5 2.cxb5 Lf6 3.a5 Kd7 4.a6 bxa6
5.bxa6 Lg5+ 6.Kd4 Kc6 7.Lc5 Lf6+
8.Kc4 Le5 (385.867)

10/31 0:00 -1.43 1...Kd7 2.a5 Lf6 3.bxc6+ bxc6 4.Kf4 Ke6
5.Ld4 Le7 6.a6 Lf8 7.Ke3 Lh6+ 8.Ke2 (492.624)

11/33 0:00 -1.38 1...Kd7 2.a5 Kc8 3.a6 bxa6 4.bxa6 Kb8
5.Lc5 Kc8 6.Ld4 Ld6 7.c5 Lb8 (711.443)

12/33 0:00 -1.43 1...Kd7 2.a5 Kc8 3.a6 bxa6 4.bxa6 Kb8
5.Lc5 Kc8 6.Ld4 Ld6 7.Ke2 Kb8 8.Ke3 (841.745)

13/33 0:01 -1.42 1...Kd7 2.a5 Kc8 3.a6 bxa6 4.bxa6 Kb8
5.Lc5 Kc8 6.Ld4 Ld6 7.c5 Lb8 8.Lf6 Kd7 (1.092.854) 965

14/33 0:01 -1.39 1...Kd7 2.a5 Kc8 3.a6 bxa6 4.bxa6 Kb8
5.Lc5 Kc8 6.Ld4 Ld6 7.Kd2 Kb8 8.Lb6 Le7
9.Ld4 (1.476.249) 963

15/33 0:02 -1.45 1...Kd7 2.a5 Kc8 3.a6 bxa6 4.bxa6 Kb8
5.Lc5 Kc8 6.Ld4 Ld6 7.c5 Lc7 8.Lf6 Kb8
9.c4 Ka7 (2.196.909) 955

15/37 0:03 -1.36 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Le7 7.Ld4 Kb8 8.Lb6 Lf6
9.Ke3 Le5 10.Ld4 Ld6 11.Kd2 (3.277.243) 956

16/37 0:04 -1.40 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Ld8 7.Lf4 Le7 8.Le5 Lc5
9.Ld4 Ld6 10.Lb6 Kb8 11.Kd2 Le5 (4.032.855) 960

17/44 0:07 -1.34 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Ld8 7.Lf4 Le7 8.Le5 Lc5
9.Ld4 Ld6 10.Lb6 Kb8 11.Kd2 Le5
12.Lc5 (6.719.909) 952

18/45 0:12 -1.34 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Ld8 7.Lf4 Le7 8.Le5 Lc5
9.Ld4 Ld6 10.Lb6 Le7 11.Kd2 Lg5+
12.Kc2 Lf6 13.Lc5 (11.619.198) 957

19/45 0:16 -1.33 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Ld8 7.Lf4 Le7 8.Le5 Lc5
9.Ld4 Ld6 10.Lb6 Kb8 11.Kd2 Le5
12.Lc5 Lf6 13.Ld4 (16.280.002) 958

20/49 0:24 -1.34 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Ld8 7.Lf4 Le7 8.Le5 Lc5
9.Ld4 Ld6 10.Lb6 Kb8 11.Kd2 c5
12.Ke2 Le7 13.Kd1 Ld6 (23.529.730) 953

21/52 0:48 -1.32 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Kd2 Kc7 8.Lc5 Kb8
9.Kc2 Kc7 10.Kb2 Lg5 11.Kb3 Lf6
12.Kb4 Lg5 13.Ld4 Le7+ 14.c5 Kb8 (46.159.722) 960

22/55 1:18 -1.32 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Kd2 Kc7 8.Lc5 Kb8
9.Kc2 Kc7 10.Kb2 Lg5 11.Kb3 Lf6
12.Kb4 Lg5 13.Ld4 Le7+ 14.c5 Kb8 (75.766.353) 966

23/56 2:40 -1.32 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Kd2 Kc7 8.Lc5 Kb8
9.Kc2 Kc7 10.Kb2 Lg5 11.Kb3 Lf6
12.Kb4 Lg5 13.Ld4 Kb8 14.a7+ Kb7 (152.831.804) 951

24/58 4:19 -1.34 1...Lf6 2.bxc6 bxc6 3.a5 Lg5+ 4.Ke2 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Kd2 Le7 8.Lf4 Lc5
9.Ke2 Le7 10.Le5 Lc5 11.Ke1 La7
12.Kf1 Kd7 13.Ke2 Lb6 14.Ld4 c5 (247.753.403) 954

25/65 13:31 -1.34 1...Lf6 2.bxc6 Lg5+ 3.Ke2 bxc6 4.a5 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Kd2 Le7 8.Lf4 Lc5
9.Ke2 Le7 10.Kd1 Lc5 11.Ke1 Kd7
12.Le5 La7 13.Ke2 Lb6 14.Ld4 c5 (769.152.024) 948

26/67 24:06 -1.34 1...Lf6 2.bxc6 Lg5+ 3.Ke2 bxc6 4.a5 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Ld4 Le7 8.Le5 Lc5
9.Ke1 La7 10.Lf4 Lc5 11.Ke2 Le7
12.Kd2 Lc5 13.Ke1 Kd7 14.Le5 Kc8 (1.396.698.596) 965

27/72 44:29 -1.36 1...Lf6 2.bxc6 Lg5+ 3.Ke2 bxc6 4.a5 Kd7
5.a6 Kc8 6.Le3 Lf6 7.Ld4 Le7 8.Le5 Lc5
9.Ke1 Le7 10.Kd2 Lc5 11.Ke2 Le7
12.Ke3 Lc5+ 13.Ld4 Ld6 14.Lb6 Le5 (2.599.011.761) 973

[smirobth]

[d] 8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - - 0 36 bm 36...Bg7

36...Bg7 is the only move that doesn't lose. However programs can get this move for the wrong reason (many programs see several moves as more or less equivalent) so the best way to use this position as a test is in show top two moves mode. There should be a moderate to large difference (at least 1/4 pawn) between 36...Bg7, which is good enough to hold the draw, and all other moves, which lose. Do any programs solve this?

Hello Robin,

Toga plays 1...Bf6. Could you maybe give some details why Bg7 is the best move? Is this a position from one of your own games?

Regards, Eelco

Hi Eelco,

It is essential that White's king be denied the f4 square. After 1...Bf6 White can and should play Kf4 right away. After 1...Bg7! 2.Kf4?? Bh6 is mate, so White will have to play a different move, after which Black drives back White's king with Bh6+. The position is from one of my postal games, which is not yet completed. My opponent did not play Bg7, and is now clearly losing.

[PauloSoare]

[d] 8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - - 0 36 bm 36...Bg7

36...Bg7 is the only move that doesn't lose. However programs can get this move for the wrong reason (many programs see several moves as more or less equivalent) so the best way to use this position as a test is in show top two moves mode. There should be a moderate to large difference (at least 1/4 pawn) between 36...Bg7, which is good enough to hold the draw, and all other moves, which lose. Do any programs solve this?

Luck? I think that it´s not luck, because Kd7 and Bf6 have more than
+1.00.

Deep Fritz 10 with one thread.

New game
8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - - 0 1

Analysis by Deep Fritz 10:

1...Bf6 2.a5
± (0.91) Depth: 9/20 00:00:00 131kN
1...Bf6 2.a5 Bg5+
± (0.91) Depth: 9/20 00:00:00 137kN
1...Bf6 2.a5
± (1.19) Depth: 10/20 00:00:00 255kN
1...Bf6 2.a5 Bg5+ 3.Ke2
± (1.19) Depth: 10/23 00:00:00 277kN
1...Kd7
± (1.18) Depth: 10/23 00:00:00 390kN
1...Kd7
± (1.03) Depth: 10/23 00:00:00 392kN
1...Kd7 2.a5 Kc8 3.Bd4 Bd6 4.bxc6 bxc6 5.a6 Kb8 6.a7+ Ka8 7.c5
± (0.86) Depth: 10/23 00:00:00 403kN
1...Kd7 2.a5 Kc8 3.Bd4 Bd6 4.Ke2 Bc7 5.a6 bxa6 6.bxa6 Kb8 7.a7+ Ka8
± (0.92) Depth: 11/21 00:00:00 469kN
1...Kd7 2.a5 Kc8 3.Bd4 Bd6 4.Kd2 Bc7 5.a6 bxa6 6.bxa6 Bd6 7.a7 Kb7
± (0.86) Depth: 12/23 00:00:00 635kN
1...Kd7 2.a5 Kc8 3.Kd2 f4 4.gxf4 Bxf4+ 5.Be3 Be5 6.a6
± (0.71) Depth: 13/25 00:00:01 1240kN
1...Kd7 2.a5 Kc8 3.Kd2 f4 4.gxf4 Bxf4+ 5.Be3 Be5 6.Kc2 Kb8 7.Kb3 Kc7 8.Bd4 Bf4 9.Kb4
± (0.76) Depth: 14/27 00:00:02 1903kN
1...Kd7 2.a5 Kc8 3.Kd2 Kb8 4.Bd4 Bc7 5.Ke3 Kc8
± (0.78) Depth: 15/28 00:00:03 3605kN
1...Kd7 2.a5 Kc8 3.Ba7 Bc7 4.a6 bxa6 5.bxa6 Be5 6.Kd2 Bc7 7.Bd4 f4 8.gxf4 Bxf4+ 9.Be3 Be5 10.a7 Kb7
± (0.77) Depth: 16/26 00:00:05 5293kN
1...Kd7 2.a5 Kc8 3.Ba7 Bc7 4.a6 bxa6 5.bxa6 Be5 6.Kd2 Bc7 7.Be3 Kb8 8.Ke2 Be5 9.a7+ Kb7 10.Bd4 Bc7
± (0.76) Depth: 17/29 00:00:07 7625kN
1...Kd7 2.a5 Kc8 3.Ba7 Bc7 4.a6 bxa6 5.bxa6 Be5 6.Bd4 Bd6 7.Bb6
± (0.76) Depth: 18/34 00:00:11 11844kN
1...Kd7 2.a5 Kc8 3.Ba7 Bc7 4.a6 bxa6 5.bxa6 Be5 6.Bd4 Bd6 7.Bb6
± (0.76) Depth: 19/34 00:00:18 19618kN
1...Kd7 2.a5 Kc8 3.Ba7 Bc7 4.a6 bxa6 5.bxa6 Be5 6.Bd4 Bd6 7.Bb6 Be5 8.Ba7
± (0.76) Depth: 20/37 00:00:32 35731kN
1...Kd7 2.a5
± (1.04) Depth: 21/42 00:00:48 53946kN
1...Kd7 2.a5 Kc8
± (1.04) Depth: 21/55 00:00:56 62157kN
1...Bg7
± (1.04) Depth: 21/55 00:01:46 120mN
1...Bg7
± (0.88) Depth: 21/55 00:02:04 141mN
1...Bg7 2.a5 Bh6+ 3.Ke2 Kd7 4.Be3 Bg7 5.Bd4 Bf8
± (0.78) Depth: 22/52 00:04:47 336mN
1...Bg7 2.Kd2 Bh6+ 3.Be3 Bg7 4.a5 Kd7 5.Bb6
± (0.83) Depth: 23/53 00:09:10 649mN
1...Bg7 2.Kd2 Bh6+ 3.Be3 Bg7 4.a5 Kd7 5.Bb6 Kc8
± (0.83) Depth: 24/53 00:18:24 1318mN

(Soares, Residencia 04.06.2007)

[smirobth]

36...Bg7 is the only move that doesn't lose. However programs can get this move for the wrong reason (many programs see several moves as more or less equivalent) so the best way to use this position as a test is in show top two moves mode. There should be a moderate to large difference (at least 1/4 pawn) between 36...Bg7, which is good enough to hold the draw, and all other moves, which lose. Do any programs solve this?

Luck? I think that it´s not luck, because Kd7 and Bf6 have more than +1.00.

I thought you were right.

Analysis by Deep Fritz 10 (4 threads, depth 23, show top two moves):

1. +/- (0.86): 36...Bg7 37.bxc6 bxc6 38.a5 Bh6+ 39.Kd4 Bg7+ 40.Kc5 Kd7 41.Kb4 Bf8+ 42.Ka4 Be7 43.a6 Kc8 44.Ba7 Bd6 45.Ka5 f4 46.gxf4 Bc7+ 47.Bb6 Bxf4 48.Ba7 Bc7+
2. +/- (1.35): 36...Bf6 37.Kf4 cxb5 38.cxb5 Bxc3 39.Kg5 Bd2+ 40.Kh5 f4 41.Kxg4 fxg3 42.fxg3 Kd6 43.a5 Kd7 44.Bd4 Kc7 45.Be5+ Kc8 46.Bf4 Be1 47.Kf5 e3 48.Bxe3 Bxg3

Then I let it run a little longer and both Bg7 and Bf6 came out the same:

Analysis by Deep Fritz 10 (depth 25):

1. +/- (0.77): 36...Bf6 37.bxc6 Bg5+ 38.Kd4 bxc6 39.a5 Bf6+ 40.Ke3 Bg5+ 41.Ke2 Kd7
2. +/- (0.77): 36...Bg7 37.a5 Bh6+ 38.Ke2 Kd7 39.Bd4 Bf8 40.Kd2 Be7 41.Kc2 Kc8 42.Kb3 Kc7 43.Be3 Bf8 44.Ka4 Kc8 45.Bd4 Bd6 46.Kb3 Kb8

[PauloSoare]

36...Bg7 is the only move that doesn't lose. However programs can get this move for the wrong reason (many programs see several moves as more or less equivalent) so the best way to use this position as a test is in show top two moves mode. There should be a moderate to large difference (at least 1/4 pawn) between 36...Bg7, which is good enough to hold the draw, and all other moves, which lose. Do any programs solve this?

Luck? I think that it´s not luck, because Kd7 and Bf6 have more than +1.00.

I thought you were right.

Analysis by Deep Fritz 10 (4 threads, depth 23, show top two moves):

1. +/- (0.86): 36...Bg7 37.bxc6 bxc6 38.a5 Bh6+ 39.Kd4 Bg7+ 40.Kc5 Kd7 41.Kb4 Bf8+ 42.Ka4 Be7 43.a6 Kc8 44.Ba7 Bd6 45.Ka5 f4 46.gxf4 Bc7+ 47.Bb6 Bxf4 48.Ba7 Bc7+
2. +/- (1.35): 36...Bf6 37.Kf4 cxb5 38.cxb5 Bxc3 39.Kg5 Bd2+ 40.Kh5 f4 41.Kxg4 fxg3 42.fxg3 Kd6 43.a5 Kd7 44.Bd4 Kc7 45.Be5+ Kc8 46.Bf4 Be1 47.Kf5 e3 48.Bxe3 Bxg3

Then I let it run a little longer and both Bg7 and Bf6 came out the same:

Analysis by Deep Fritz 10 (depth 25):

1. +/- (0.77): 36...Bf6 37.bxc6 Bg5+ 38.Kd4 bxc6 39.a5 Bf6+ 40.Ke3 Bg5+ 41.Ke2 Kd7
2. +/- (0.77): 36...Bg7 37.a5 Bh6+ 38.Ke2 Kd7 39.Bd4 Bf8 40.Kd2 Be7 41.Kc2 Kc8 42.Kb3 Kc7 43.Be3 Bf8 44.Ka4 Kc8 45.Bd4 Bd6 46.Kb3 Kb8

Can position learning helped Deep Fritz 10 here? I think it´s possible, because before the depth 24 1. Bf6 score was around +1,65.

Depth 26
New game
8/1p6/1Bp1k3/1P2bp2/P1P1p1p1/2P1K1P1/5P2/8 b - - 0 1

Analysis by Deep Fritz 10:

1. ± (0.87): 1...Be5-g7 2.b5xc6 b7xc6 3.a4-a5 Bg7-h6+ 4.Ke3-e2 Ke6-d7 5.a5-a6 Kd7-c8 6.Bb6-e3 Bh6-f8 7.Be3-f4 Bf8-e7 8.Bf4-e5 Be7-c5 9.Ke2-e1 Bc5-b6 10.Be5-d6 c6-c5 11.Ke1-e2 Bb6-a7 12.Ke2-e3
2. +- (1.84): 1...Be5-f6 2.Ke3-f4 c6xb5 3.c4xb5 Bf6xc3 4.Kf4-g5 e4-e3 5.f2xe3 Bc3-f6+ 6.Kg5-h5 Bf6-e5 7.a4-a5 Be5xg3 8.a5-a6 b7xa6 9.b5xa6 Bg3-b8 10.Kh5-h4 Ke6-f6 11.a6-a7 Bb8xa7

(Soares, Residencia 04.06.2007)