Interesting Test Position

[Marc MP]

[d]r1b2rk1/1p3pp1/p1n4p/2bBP3/3N3B/1R6/q1PQ2PP/5RK1 w - - 0 1

Rybka 3 finds it (Rg3!, see: http://rybkaforum.net/cgi-bin/rybkaforu ... l?tid=6601, post by Uri Blass). Will another engine find it?

[Marc MP]

I let Chess Tiger 2007.1 ponder on it for a few hours on my P4. It did eventually find the move, but it doesn't shows a positive score after more than 3 hours (it seems to focus on a perpetual check):

Code: Select all

...
00:04:53.7	-0,60	20	195385901	Bxc6 bxc6 Bf2 Bxd4 Bxd4 Be6 Rb6 Rab8 Rxc6 Rfc8 Rxc8+ Rxc8 Bc3 Qc4 Rf3 Qc5+ Qf2 Qxf2+ Kxf2 Bd5 Rg3 a5 e6 

00:05:57.5	-0,54	20	232262996	Rg3 Nxd4 Kh1 Qxc2 Bf6 Qh7 Rxg7+ Qxg7 Bxg7 Kxg7 Rf6 Nf5 Qc2 Be3 Rxf5 Bxf5 Qxf5 Ra7 Bb3 

00:11:05.6	-0,32	21	412212086	Rg3 Nxd4 Kh1 Qxc2 Bf6 Qh7 Rxg7+ Qxg7 Bxg7 Kxg7 Rf6 Nf5 Qc2 Be7 Rxf5 Bxf5 Qxf5 Ra7 Bb3
 
00:20:09.4	-0,41	22	741069289	Rg3 Nxd4 Kh1 Qxc2 Bf6 Qh7 Rxg7+ Qxg7 Bxg7 Kxg7 Qc3 Ba7 e6 Bxe6 Bxe6 fxe6 Qg3+ Kh7 Qc7+ Kg6 Qg3+ Kh7 Qc7+ Kg6 Qg3+ Kh7 Qc7+ Kg6 

02:51:39.0	-0,27	23	2017319578	Rg3 Nxd4 Kh1 Qxc2 Bf6 Qxd2 Rxg7+ Kh8 Rg6+ Kh7 Rg7+ Kh8 Rg6+ Kh7 Rg7+ Kh8 Rg6+ Kh7 Rg7+ Kh8 

[kgburcham]

[d] r1b2rk1/1p3pp1/p1n4p/2bBP3/3N3B/1R6/q1PQ2PP/5RK1 w - - 0 1


Zappa Mexico II

16/51 2:16 0.00 1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rf7+ (285.662.680) 2089
16/51 2:32 0.00 1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rf7+ (317.734.987) 2088
17/53 3:19 0.00 1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rf7+ (414.194.656) 2075
17/53 3:48 0.00 1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rf7+ (473.426.815) 2072
18/55 5:43 0.00 1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rf7+ (708.589.991) 2063
18/55 6:48 0.00 1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rf7+ (842.888.189) 2061

[Marc MP]

Hi Kg,

At the end of this post (actually that was the one I wanted to post a link to, in my first message): http://rybkaforum.net/cgi-bin/rybkaforu ... 3#pid97483

Uri hints at a winning line:

After 18.Rg3 Nxd4(forced) 19.Kh1 Qxc2(forced) 20.Qxc2 Nxc2(forced) 21.Bf6 g5(forced) 22.Rf4
It seems to me that white is winning(I did not analyze to be sure).

That yields this position:

[d]r1b2rk1/1p3p2/p4B1p/2bBP1p1/5R2/6R1/2n3PP/7K b - - 1 1

Chess Tiger seems to agree after 15min here:

Code: Select all

...
00:04:39.0	-1,14	20	146678920	Rd8 Bxd8 Nd4 Bxf7+ Kf8 Rf6 Kg7 Rc3 Bd7 Rxc5 Rxd8 Ba2 Bc6 Ra5 Nb5 Rf7+ 

00:07:18.9	-1,21	21	235011083	Rd8 Bxd8 Nd4 Bxf7+ Kf8 Rf6 Kg7 Rc3 Bd7 Ba5 Be7 Rf2 Bd8 Bb4 Be6 

00:12:43.5	-1,18	22	413067991	Rd8 Bxd8 Nd4 Bxf7+ Kh7 Rf6 Kg7 Rc3 Bd7 Rxc5 Rxd8 Rd5 Nc6 Rfd6 Bg4 Rxd8 Nxd8 Rxd8 Kxf7 Rb8 b5 

[Marc MP]

So actually it is not a good test position since engines may choose Rg3 in order do to a perpetual. Better is:

[d]r1b2rk1/1p3pp1/p6p/2bBP3/3n3B/6R1/2qQ2PP/5R1K w - - 0 1

With Qxc2 as the best move.

[kgburcham]

With enough time Zappa would pass through the perpetual and show White winning as does Hiarcs below. It takes Hiarcs 15 minutes on my quad to go +.

[d] r1b2rk1/1p3pp1/p1n4p/2bBP3/3N3B/1R6/q1PQ2PP/5RK1 w - - 0 1

Deep Hiarcs Paderborn 2007:

1.Rg3
= (-0.24) Depth: 16/42 00:01:53 172mN
1.Rg3
= (-0.01) Depth: 16/42 00:01:55 174mN
1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rg8+ Kh7 8.Rg7+ Kh8
= (0.00) Depth: 16/44 00:01:59 181mN
1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rg8+ Kh7 8.Rg7+ Kh8
= (0.00) Depth: 17/49 00:03:06 286mN
1.Rg3 Nxd4 2.Kh1 Qxc2 3.Bf6 Qxd2 4.Rxg7+ Kh8 5.Rxf7+ Kg8 6.Rg7+ Kh8 7.Rg8+ Kh7 8.Rg7+ Kh8
= (0.00) Depth: 18/53 00:05:57 551mN
1.Rg3
= (0.25) Depth: 19/53 00:15:01 1398mN
1.Rg3 Nxd4 2.Kh1 Qxc2 3.Qxc2 Nxc2 4.Bf6 g5 5.Rf4 Rd8 6.Bxd8 Nd4 7.Bxf7+ Kh7 8.Rf6 Nc6
+/- (0.88) Depth: 19/53 00:17:34 1631mN
1.Rg3 Nxd4 2.Kh1 Qxc2 3.Qxc2 Nxc2 4.Bf6 g5 5.Rf4 Rd8 6.Bxd8 Nd4 7.Bxf7+ Kg7 8.Bd5 Nf5 9.Bf6+
+/- (1.13) Depth: 20/53 00:31:50 2943mN

[Mike S.]

Similar to this position, I was considering to play Rb3-g3 earlier in the variation:

[d]rnb2rk1/1p3pp1/p6p/2bBP3/3N3B/1R6/q1PQ2PP/4K2R w K - 0 17

After some backwards analysis, Rybka 232a agrees:

Analysis by Rybka 2.3.2a 32-bit (after depth 15, but with hash filled from analysis in later positions):

1. = (0.00): 17.Rg3 Qxd5 18.Rxg7+ Kxg7 19.Bf6+ Kg6 20.Qd3+ Kh5 21.h3 Qxg2 22.Qd1+ Kg6 23.Qd3+ Kh5 24.Qd1+ Kg6 25.Qd3+ Kh5 26.Qd1+ Kg6 27.Qd3+ Kh5 28.Qd1+ Kg6 29.Qd3+ Kh5 30.Qd1+ Kg6 31.Qd3+ Kh5 32.Qd1+ Kg6
2. = (0.00): 17.0-0 Nc6 18.Rg3 Nxd4 19.Kh1 Qxc2 20.Bf6 Qxd2 21.Rxg7+ Kh8 22.Rxf7+ Kg8 23.Rg7+ Kh8 24.Rg8+ Kh7 25.Rg7+ Kh8 26.Rg8+ Kh7 27.Rg7+ Kh8 28.Rg8+ Kh7 29.Rg7+ Kh8 30.Rg8+ Kh7 31.Rg7+ Kh8 32.Rg8+ Kh7
3. =/+ (-0.68): 17.Qc3 Bxd4 18.Qxd4 Qxc2 19.0-0 Nc6 20.Qe3 Be6 21.Bxe6 fxe6 22.Rc1 Qf5 23.Rxb7 Qxe5 24.Qxe5 Nxe5 25.Rcc7 Rf7 26.Rxf7 Nxf7 27.Re7 e5 28.Bg3 Kf8 29.Re6 a5 30.Bxe5


[Event "?"]
[Site "?"]
[Date "2008.08.20"]
[Round "?"]
[White "P.P.Analysis, 12...Nd5"]
[Black "(incomplete)"]
[Result "*"]
[ECO "B97"]
[PlyCount "53"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Qb6 8. Qd2
Qxb2 9. Rb1 Qa3 10. e5 h6 11. Bh4 dxe5 12. fxe5 Nd5 13. Nxd5 exd5 14. Bc4 Bc5
15. Bxd5 O-O 16. Rb3 Qxa2 (16... Qa4 17. Rg3 Qxd4 18. Qxd4 Bxd4 19. Bf6 {
is very similar to the other continuation, but the a2 pawn remains (see 20.Bf6)
.}) 17. Rg3 (17. O-O Nc6 18. Kh1 Qa5 19. Qxa5 Nxa5) 17... Qa1+ (17... Qxd5 18.
Rxg7+ Kxg7 19. Bf6+ Kg8 20. Qxh6 Qe4+ 21. Kf1 Qh7 22. Qg5+ Qg6 23. Qh4 Qh7 24.
Qg3+ Qg6 25. Qh4 {repetition}) 18. Ke2 Qxd4 (18... Qxh1 $4 19. Qxh6 Qe1+ 20.
Kxe1 Bb4+ 21. Kf1 Bg4 22. Rxg4 g6 23. Rxg6#) 19. Qxd4 Bxd4 20. Bf6 g5 21. h4
Re8 (21... g4 22. Bxb7) 22. hxg5 Rxe5+ 23. Kd3 Rxd5 24. gxh6+ Kh7 25. Rg7+ Kh8
26. Rg8+ Kh7 27. Rg7+ {repetition} (27. Rh8+ Kg6 28. Bxd4 Nc6 29. h7 Bf5+ 30.
Ke3 Rxh8 31. Bxh8 Rd8 32. Bc3 Bxc2 33. h8=Q Rxh8 34. Rxh8 {Who's better here?})
*

It would be interesting to see how the Rybka 3 versions default/human/dynamic rank 17.Rg3 and 17.O-O, in multi-pv. 17.Rg3 threats Qxh6 immediatly (and Bxa2). Of course, 17...Qa1+ defends easily but 18.Ke2 attacks the queen again, and 18...Qxh1 is forbidden:

[d]rnb2rk1/1p3pp1/p6p/2bBP3/3N3B/6R1/2PQK1PP/7q w - - 0 19

Analysis by ChestUCI Ver.5.1:

19.Qxh6 Qe1+ 20.Kxe1 Bb4+ 21.Kf1 Bg4 22.Rxg4 g6 23.Rxg6#
+- (#5) Depth: 5 00:00:00
+- (#5) Depth: 5/5 00:00:00

That's why I like 17.Rg3 even if 17.O-O should turn out to be slightly better, objectively (which I am not sure about).