cool endgame sac

[jdart]

Nataf-Chabanon, French League Top 16 2005:

[D] 2r2k2/6p1/1p2pp1p/p1P1R2P/2K5/1P6/P4PP1/8 w - -

bm Kb5. I can get Shredder 10 to play this but I had to jog it up a few moves and then back to fill up the hash tables. Seeing this wins requires a pretty deep search.

Discussed here:

http://www.chessgames.com/perl/chessgame?gid=1374757

[AdminX]

I found it easy to spot, so I am surprised that engines don't see it right away. Normally it's the other way around.

[Mike S.]

For Shredder 11 SE, it isn't easy either, at least not without any endgame tables:

Analysis by Shredder 11 SE (D945 3.4 GHz, 256 MB hash):

1.Rxe6 Rxc5+ 2.Kd4 Rxh5 3.g3 Rb5 4.Kc4 Rc5+ 5.Kd3 b5 6.Ra6 b4 7.Kd4 Rc2 8.Rxa5 Rxf2 9.Ra4 Rg2 10.Rxb4 Rxg3 11.Rb8+ Ke7 12.a4 h5 13.b4 g5 14.Rh8 Ra3
+/= (0.48) Depth: 1/5 00:00:00 0kN
=/+ (-0.63) Depth: 22/43 00:02:06 67277kN

After 3:30, I switched to multi-pv(3), but Kb5 was not in the top-3 after depth 19 was finished.


Analysis by Rybka 2.2 32-bit an-ftb:

1.Rxe6 Rxc5+ 2.Kd3 Rxh5 3.Rxb6 Rh2 4.g4 Rxf2 5.a4 Rg2 6.Rb5 Rg3+ 7.Kd2 Rxg4
= (0.25) Depth: 2 00:00:00
=/+ (-0.29) Depth: 23 00:03:19 16373kN, tb=347

Analysis by Rybka 2.3.2a 32-bit :

1.Rxe6 Rxc5+ 2.Kd3 Rxh5 3.Rxb6 Rh2 4.a4 Rxg2 5.Rb5 Rxf2 6.Rxa5 g5 7.b4 g4
= (0.03) Depth: 5 00:00:00
=/+ (-0.33) Depth: 22 00:02:22 11334kN, tb=1

I think the move Kb5 is a victim of pruning and probably not searched up to the nominal depths, each.

[Marc MP]

Nataf-Chabanon, French League Top 16 2005:

[D] 2r2k2/6p1/1p2pp1p/p1P1R2P/2K5/1P6/P4PP1/8 w - -

bm Kb5. I can get Shredder 10 to play this but I had to jog it up a few moves and then back to fill up the hash tables. Seeing this wins requires a pretty deep search.

Discussed here:

http://www.chessgames.com/perl/chessgame?gid=1374757

Not too many programs found it. So far only: Chess Tiger 2007.1, Pro Deo 1.6 and Delfi 5.4.

Code: Select all

Chess Tiger 2007.1

00:00:19.3	-0,66	21	16800713	Rxe6 Rxc5+ Kd4 Rxh5 Rxb6 Rh2 Ra6 Rxg2 Rxa5 Rxf2 b4 g5 b5 g4 b6 Rb2 Ra8+ Ke7 Kc5 g3 Rg8 g2 b7 Rxb7 Rxg2 

00:00:32.9	0,19	21	27837787	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Kd8 f3 e4 fxe4 e5 g4 g5 hxg6ep 

00:01:33.2	1,04	21	75856160	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb6 Rd8 Ka7 Rd7 Ka8 Rd8+ b8Q Rxb8+ Kxb8 

00:01:43.7	0,90	22	83434492	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb8 Kd6 Ka8 Rd8+ b8Q+ Rxb8+ Kxb8 Kc5 Kb7 Kb4 Kb6 a4 bxa4 Kxa4 Kc5 Ka3 Kd6 e4 Ke5 

00:02:18.1	0,96	23	105784999	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb8 Kd6 Ka8 Rd8+ b8Q+ Rxb8+ Kxb8 Kc5 Kb7 Kb4 Kb6 a4 bxa4 Kxa4 Kc5 Ka3 Kd6 e4 

00:02:59.3	0,96	24	133644474	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb8 Kd6 Ka8 Rd8+ b8Q+ Rxb8+ Kxb8 Kc5 Kb7 Kb4 Kb6 a4 bxa4 Kxa4 Kc5 Ka3 Kd6 e4 

00:04:13.9	0,96	25	184428805	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb8 Kd6 Ka8 Rd8+ b8Q+ Rxb8+ Kxb8 Kc5 Kb7 Kb4 Kb6 a4 bxa4 Kxa4 Kc5 Ka3 Kd6 e4 

Code: Select all

Delfi 5.4

20	+0.66	256.4M	4:48.16	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 e4 Kc7 Rd8 b8=Q Rxb8 Kxb8 Kd6 Kb7 Kc5 a3 Kb5 Ka7 e5 a4+ Kb4 Kb6

19	+0.66	172.1M	3:03.45	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 e4 Kc7 Rd8 b8=Q Rxb8 Kxb8 Kd6 Kb7 Kc5 Ka6 Kb4 a4 e5 Kb6

18	+0.09	71.5M	1:09.73	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Kd8 f3 e4 fxe4 e5 g3 g5 hxg6 Ke7 g7 Ke6

17	+0.20	51.9M	0:49.84	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Kd8 f3 e4 fxe4 e5 a4 g5 hxg6 Ke7 g7

17	-0.32	26.4M	0:24.38	Rxe6 Rxc5+ Kd3 Rxh5 Rxb6 Rh2 Rb8+ Kf7 Rb7+ Kg6 Ke3 Rxg2 Rb5 h5 Rxa5 h4 Kf3 h3 Rd5 f5

Code: Select all

ProDeo 1.6

00:01:53.3	-0,51	17	129175439	Rxe6 Rxc5+ Kd4 Rxh5 g3 a4 Rxb6 axb3 axb3 Ke7 Ke4 Kf8 Rb8+ Kf7 Kd4 Ke6 

00:02:42.1	-0,51	17	176132410	Kb5 

00:03:46.8	0,46	17	236589596	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb8 Kd6 Ka8 Rd8+ 

00:05:25.4	0,48	18	323738960	Kb5 fxe5 cxb6 Ke7 b7 Rb8 Kc6 Rd8 Kc7 Rd7+ Kb8 Kd6 Ka8 Rd8+ 

Code: Select all

Fruit 05/11/03

00:05:21.9	-0,62	23	241235765	Rxe6 Rxc5+ Kd4 Rxh5 g3 Rb5 Kc4 Kf7 Rc6 Rb4+ Kc3 h5 Rc7+ Kg6 Rb7 Kh6 a3 Re4 Kd3 Re6 Kd4 Rd6+ Kc4 Rd2 f3 Rg2 Rxb6 Rxg3 

Code: Select all

Twisted Logic

00:06:35.5	-0,48	24	249111804	Rxe6 Rxc5+ Kd4 Rxh5 Rxb6 Rh2 a4 Rxg2 Rb5 Rxf2 Rxa5 Rd2+ Ke4 Ke7 b4 Rb2 Rb5 Kd6 a5 Kc6 Rc5+ Kb7 b5 Rb4+ Ke3 g5 a6+ Kb6 Rd5 

Code: Select all

Spike 1.2

00:09:43.0	-0,71	20	395164186	Rxe6 Rxc5+ Kd4 Rxh5 Rxb6 Rh2 Rb8+ Kf7 Ke3 Rxg2 Rb7+ Kg6 Rb5 Rg5 Rxg5+ Kxg5 a3 Kf5 b4 axb4 axb4 Ke6 b5 Kd7 Ke4 h5 b6 

Code: Select all

SlowChess 2.1

[17] -0.88  10:12.8	Rxe6 Rxc5+ Kd3 Rxh5 Rxb6 Rh2 Ke3 Rxg2 Ra6 Rg5 Ra7 Re5+ Kd3 Rf5 Ke3 h5 Ra8+ Kf7 Ra7+ Kg6 Rc7 

Code: Select all

Glaurung 2.1

00:05:34.6	-0,66	25	232376407	Rxe6 Rxc5+ Kd4 Rxh5 Rxb6 Rh2 Rb5 Rxg2 Rxa5 Rxf2 Ra8+ Ke7 Ra7+ Ke6 Rxg7 Rxa2 b4 Rd2+ Kc5 Rd5+ Kc6 Rd1 Rh7 Rc1+ Kb6 Rh1 b5 Kd6 Rf7 Ke6 Ra7 f5 

Code: Select all

Bright 0.3a

00:22:00.3	-0,04	24	1296596472	Rxe6 Rxc5+ Kd4 Rxh5 Rxb6 Rh2 Ra6 Rxg2 Rxa5 Rxf2 a4 g5 b4 h5 b5 Rb2 Kc3 Rb1 Kc2 Rb4 Kc3 Re4 Ra8+ Ke7 a5 f5 a6 

[mercury]

Ktulu 8, Pharaon 3.5.1, Ruffian 1.0.5 find the move quite quickly.

[Terry McCracken]

I found it easy to spot, so I am surprised that engines don't see it right away. Normally it's the other way around.

I suspect the horizon effect comes into play here. Some programs with more R+P+K and promotion knowledge will find the solution faster.