Any engines solve this one?

[peter]

[d]r1b2rk1/pp3ppp/1n2p3/3pP3/2P2B2/b1PB1qP1/P1Q1NP1P/1K1R3R w - -

1. Nd4 with at least +2 or equivalent winning score.

Great one!

Here's komodo with max. King Safety and Dynamism, Minimum Selectivity and Reductions, White Contempt 50:
24 threads of 3x12GHz CPU, no tbs used.

Engine: Komodo 13.02 64-bitM (32768 MB)
von Don Dailey, Larry Kaufman, Mark Lefler

Code: Select all

 12.00	 0:01 	+2.16--	1.Lc1 Sxc4 (6.860.692) 6465
...
 15.01	 0:05 	+1.47--	1.h4 Dxf2 (22.284.140) 3749
 15.06	 0:07 	+2.13 	1.Sd4 Dh5 2.g4 Dh3 3.Lf1 Dxg4 4.Le3 Ld7 5.f3 Dg6 6.Ld3 La4 7.Lxg6 Lxc2+ 8.Lxc2 Lc5 9.Thg1 Sxc4 10.Sxe6 Lxe3 11.Txg7+ Kh8 (27.771.355) 3776
...
 16.01	 0:10 	+1.90++	1.Sd4 Dh5 2.g4 (50.859.304) 5032
 16.01	 0:10 	+2.06 	1.Sd4 Dh5 2.g4 Dh3 3.Thg1 dxc4 4.Le4 Sd5 5.Lxd5 exd5 6.Lg5 Lc5 7.Tg3 Dxh2 8.e6 Lxd4 9.Th3 Dxh3 (51.229.556) 5060
...
 17.01	 0:11 	+1.41 	1.Sd4 Dh5 2.Le2 Dg6 3.Le3 Dxc2+ 4.Sxc2 Le7 5.cxd5 exd5 6.Sd4 Sa4 7.Kc2 Ld7 8.Lb5 Lxb5 9.Sxb5 Tfc8 10.Txd5 (87.165.794) 7402
 17.06	 0:12 	+2.90 	1.cxd5 Sxd5 2.Db3 Lc5 3.c4 g6 4.cxd5 exd5 5.Lh6 Te8 6.Sf4 Le6 7.Lc2 Dxb3+ 8.Lxb3 Lf5+ 9.Ka1 Txe5 10.Sxd5 Te6 11.Se3 Le4 (94.528.992) 7808
...
 19.01	 0:14 	+2.42--	1.cxd5 Sxd5 (140.909.611) 9920
...
 19.01	 0:14 	+3.25++	1.Sd4 Dh5 2.h4 (149.591.436) 10291
...
 19.01	 0:15 	+3.44 	1.Sd4 Dh5 2.h4 dxc4 3.Le4 Sd5 4.f3 f6 5.The1 Sxf4 6.gxf4 Kh8 7.Da4 Le7 8.Tg1 Df7 9.Dxc4 f5 10.Sxf5 exf5 11.Ld5 De8 12.Db3 Lc5 13.Tg2 Dh5 14.e6 (173.745.434) 11154
...
 23.01	 1:39 	+2.49 	1.Sd4 Dh5 2.g4 Dh4 3.Sf3 Dd8 4.Lg5 Dc7 5.Lf6 Sxc4 6.Lxc4 Dxc4 7.Sg5 Db5+ 8.Ka1 Lb2+ 9.Dxb2 Dxb2+ 10.Kxb2 gxf6 11.exf6 Ld7 12.Td3 Lc6 13.Te1 h6 14.Sf3 (2.014.589.251) 20256
 24.01	 2:02 	+2.43--	1.Sd4 Dh5 (2.525.764.809) 20561
...
 24.01	 3:37 	+1.77 	1.Sd4 Dh5 2.Le2 Dg6 3.cxd5 Sxd5 4.Ld3 Dh5 5.c4 Sb4 6.De2 Dxe2 7.Lxe2 Sc6 8.Sxc6 bxc6 9.Kc2 Lc5 10.Le3 Lxe3 11.fxe3 c5 12.Td6 Tb8 13.Kc3 g6 14.Ld3 (4.906.891.672) 22568
 24.02	 3:37 	+1.95 	1.cxd5 Sxd5 2.Sd4 Dh5 3.c4 Sb4 4.De2 Dxe2 5.Lxe2 Sc6 6.Sxc6 bxc6 7.Kc2 Lc5 8.Le3 Lxe3 9.fxe3 c5 10.Td6 Tb8 11.Kc3 g6 12.Ld3 Lb7 13.Tb1 Lf3 14.Tb5 (4.906.958.502) 22568
 25.01	 3:53 	+1.89--	1.cxd5 Sxd5 (5.261.593.859) 22504
 25.02	 3:54 	+2.01++	1.Sd4 Dh5 2.g4 (5.271.527.455) 22504
...
 25.01	 4:17 	+1.93 	1.Sd4 Dh5 2.cxd5 Sxd5 3.c4 Sb4 4.De2 Dxe2 5.Lxe2 Sc6 6.Sxc6 bxc6 7.Kc2 Lc5 8.Le3 Lxe3 9.fxe3 c5 10.Td6 Tb8 11.Ld3 Lb7 12.Tb1 g6 13.Tb5 Lf3 14.Kc3 (5.778.975.898) 22416
 26.01	 5:41 	+1.87--	1.Sd4 Dh5 (7.588.843.265) 22190
 26.01	 7:14 	+1.91++	1.Sd4 Dh5 2.g4 (9.620.914.275) 22131
...
 26.01	 8:51 	+3.30 	1.Sd4 Dh5 2.g4 Dh4 3.Sf3 Dd8 4.Lxh7+ Kh8 5.c5 Ld7 6.Sg5 La4 7.Dd3 De8 8.Lg6 Db5+ 9.Dxb5 Lxb5 10.cxb6 axb6 11.Lxf7 Lc4 12.Le3 Le7 13.Lxe6 Lxg5 14.Lxd5 (11.827.485.667) 22233

[Eelco de Groot]

After forcing Qxg4 white is going to win. But black is not going to play Qxg4 on the previous ply, but instead Qh4.
[d]r1b2rk1/pp3ppp/1n2p3/3pP3/2PN1Bq1/b1PB4/P1Q2P1P/1K1R3R w - - acd 36; acs 207; bm Qd2; c3 "Qd2"; ce 218; pm Qd2; pv Qd2 Qh5 Rhg1 g6 Bg5 Nxc4 Qf4 f6 Bxf6 Rf7 h4 Bd7 Nf3 Bf8 Rg5 Bh6 Rxh5 Bxf4 Bxg6 Rxf6 Bxh7+ Kf7 exf6 Kxf6 Bd3 Rc8 Nd4 Be8 Rh8 Nd6 Ne2 Be5 Rh6+ Ke7 Rh7+ Bf7 f4 Bxc3 Nxc3 Rxc3 Kb2 Rc8 h5 a5 Bg6 a4 Bxf7 Nxf7 Rg1 Rc4;

Yes, but does not mean that CyberNezh has a wrong PV, if it recognizes somehow that Qh4 also does not cut it. If you have very speculative eval, such things happen. Not knowing anything about CyberNezh so that is just my speculation also. Kaissa eventally sees Qh4 not good I think (depth 50) and then also switches back to simple Qxg4:

[d]r1b2rk1/pp3ppp/1n2p3/3pP2q/2PN1B2/b1PB2P1/P1Q2P1P/1K1R3R w - -

Engine: Kaissa IV NoContempt (512 MB)
by T. Romstad, M. Costalba, J. Kiiski, G. Linscott

51 874:37 +2.03 2.g4 Qxg4 3.Qd2 Qh5 4.Rhg1 f6 5.cxd5 Nxd5
6.Be2 Qf7 7.Bh6 g6 8.Nb5 Be7 9.Nd6 Bxd6
10.exd6 Rd8 11.h4 Ne7 12.h5 Nf5
13.hxg6 hxg6 14.Qc2 Kh7 15.Bc1 (76.520.744.450) 1458

51 874:37 +0.47 2.Be2 Qh3 3.Nf3 Bd7 4.Ng5 Qh6 5.Nxf7 Qg6
6.Qxg6 hxg6 7.Nd6 Bxd6 8.exd6 Nxc4
9.Bxc4 dxc4 10.Rhe1 g5 11.Be3 Rf5
12.Rd4 Rc8 13.Rg4 Rc6 14.Rd1 Kf7
15.Kb2 (76.520.744.450) 1458

51 874:37 0.00 2.Rdg1 Nxc4 3.Be2 Qg6 4.Bd3 (76.520.744.450) 1458

51 874:37 0.00 2.Nb5 Bc5 3.cxd5 Nxd5 4.Be2 Qg6
5.Bd3 (76.520.744.450) 1458

51 874:37 0.00 2.cxd5 Nxd5 3.Be2 Qg6 4.Bd3 (76.520.744.450) 1458

51 874:37 0.00 2.c5 Nc4 3.Qb3 Bxc5 4.Be2 Qg6+ 5.Bd3 f5
6.exf6 Qxf6 7.Bxc4 dxc4 8.Qxc4 b6
9.Rhe1 Qf7 10.Kb2 a5 11.a4 Re8
12.Nb5 Bb7 13.Qd3 Bc6 14.Nc7 Bxa4
15.Nxa8 (76.520.744.450) 1458
but, it just takes far too long

Code: Select all

 49	537:56 	 0.00 	2.Be2 Qg6 3.Bd3 Qh5 (46.616.529.592) 1444
 49	537:56 	 0.00 	2.Rdg1 Nxc4 3.Be2 Qg6 4.Bd3 (46.616.529.592) 1444
 49	537:56 	 0.00 	2.g4 Qh4 3.Nf3 Qxg4 4.Rhg1 Qxf3 5.Bh6 g6 6.Bxg6 Qxd1+ 7.Qxd1 fxg6 8.h4 Bd7 9.h5 Be8 10.cxd5 Nxd5 11.Qb3 Be7 12.Bxf8 Kxf8 13.Qxb7 Rd8 14.c4 Nf4 15.c5 (46.616.529.592) 1444
 49	537:56 	 0.00 	2.Nb5 Bc5 3.cxd5 Nxd5 4.Be2 Qg6 5.Bd3 (46.616.529.592) 1444
 49	537:56 	 0.00 	2.cxd5 Nxd5 3.Be2 Qh3 4.Bf1 Qh5 (46.616.529.592) 1444
 49	537:56 	 0.00 	2.c5 Nc4 3.Qb3 Bxc5 4.Be2 Qg6+ 5.Bd3 Qh5 (46.616.529.592) 1444
 
 
 50	670:31 	+1.28 	2.g4 Qxg4 3.Qd2 Nxc4 4.Bxc4 dxc4 5.Rhg1 Qh5 6.Rg5 Qh4 7.Nf3 Qh6 8.Rdg1 g6 9.R5g3 Qg7 10.Bh6 Qh8 11.Qf4 Be7 12.Bxf8 Kxf8 13.Rh3 Qg7 14.Ng5 Bxg5 15.Qxg5 (58.173.322.426) 1445
 50	670:31 	 0.00 	2.Be2 Qg6 3.Bd3 Qh5 4.Be2 (58.173.322.426) 1445
 50	670:31 	 0.00 	2.Rdg1 Nxc4 3.Be2 Qg6 4.Bd3 (58.173.322.426) 1445
 50	670:31 	 0.00 	2.Nb5 Bc5 3.cxd5 Nxd5 4.Be2 Qg6 5.Bd3 (58.173.322.426) 1445
 50	670:31 	 0.00 	2.cxd5 Nxd5 3.Be2 Qh3 4.Bf1 Qg4 5.Be2 Qg6 (58.173.322.426) 1445
 50	670:31 	 0.00 	2.c5 Nc4 3.Qb3 Bxc5 4.Be2 Qg6+ 5.Bd3 f5 6.exf6 Qxf6 7.Bxc4 dxc4 8.Qxc4 b6 9.Rhe1 Qf7 10.Kb2 a5 11.a4 Re8 12.Nb5 Bb7 13.Qd3 Bc6 14.Nc7 Bxa4 15.Nxa8 (58.173.322.426) 1445
 
 

Probably you could try this speculative testposition with old Rebel Q3 or Q5T personalities also

[Dann Corbit]

The right score for this position with the right plan is about +900 at depth 40

[peter]

Hi Eelco

Yes, but does not mean that CyberNezh has a wrong PV, if it recognizes somehow that Qh4 also does not cut it.

White's clearly winning after 1.Nd4 Qh5 2.g4, with 2...Qh4 as well as with 2...Qxg4.

2...Qh4 3.Nf3 and 3...Qxg4 is refuted by 4.Rhg1.

After some Forward- Backward of this line and 2...Qh4 3.Nf3:

r1b2rk1/pp3ppp/1n2p3/3pP3/2P2BPq/b1PB1N2/P1Q2P1P/1K1R3R b - - 0 1

Analysis by Cfish 010719 64 POPCNT NUMA:

3...Qd8 4.Bxh7+ Kh8 5.c5 Bxc5 6.Ng5 Nc4 7.Ka1 Ba3 8.Rb1 Nxe5 9.Bg8 f5 10.Bxe6 Be7 11.Bxe5 Bxg5 12.gxf5 Bxe6 13.fxe6 Qe7 14.Rhe1 Rac8 15.f4 Bf6 16.Bxf6 Qxf6 17.Re5 Kg8 18.Rxb7 Rc4 19.e7 Re8 20.Qd3 Rxf4 21.Qxd5+ Qf7 22.Qxf7+ Kxf7 23.Rxa7 Rf2 24.a4 g5 25.a5 Rxh2 26.Rf5+ Ke6 27.Rxg5 Rc2 28.a6 Rxc3 29.Kb2 Re3 30.Rb7 Kf6 31.a7 Kxg5 32.Rb8
+- (3.54) Depth: 37/70 00:01:11 1666MN

[Dicaste]

Best move is not Nd4. It's Bxh7!!

[Dann Corbit]

[d]r1b2rk1/pp3ppp/1n2p3/3pP3/2P2B2/b1PB1qP1/P1Q1NP1P/1K1R3R w - - bm Nd4; acd 49; ce 889; pv Nd4 Qh5 g4 Qh4 Nf3 Qxg4 Rhg1 Qg6 Bxg6 fxg6 Rg4 Be7 Nh4 Bxh4 Rxh4 dxc4 Rd6 Bd7 Qd2 Ba4 Rxe6 Bd7 Qd6 Rad8 Bg5 Nc8 Qd5 Bc6 Qxc4 Bd5 Qc7 Bxe6 Bxd8 h6 f4 Rf7 Qb8 Rf8 Bc7 Nd6 Qxa7 Nb5 Qb6 Nxc7 Qxc7 g5 fxg5 hxg5 Ra4 Rf1+ Kb2 Rf2+ Kc1 Rxh2 Ra8+ Kh7;

[pgn] So you can scroll through the plan: [Event "?"] [Site "?"] [Date "????.??.??"] [Round "?"] [White "?"] [Black "?"] [Result "*"] [FEN "r1b2rk1/pp3ppp/1n2p3/3pP3/2P2B2/b1PB1qP1/P1Q1NP1P/1K1R3R w - - 0 1"] [Setup "1"] 1. Nd4 Qh5 2. g4 Qh4 3. Nf3 Qxg4 4. Rhg1 Qg6 5. Bxg6 fxg6 6. Rg4 Be7 7. Nh4 Bxh4 8. Rxh4 dxc4 9. Rd6 Bd7 10. Qd2 Ba4 11. Rxe6 Bd7 12. Qd6 Rad8 13. Bg5 Nc8 14. Qd5 Bc6 15. Qxc4 Bd5 16. Qc7 Bxe6 17. Bxd8 h6 18. f4 Rf7 19. Qb8 Rf8 20. Bc7 Nd6 21. Qxa7 Nb5 22. Qb6 Nxc7 23. Qxc7 g5 24. fxg5 hxg5 25. Ra4 Rf1+ 26. Kb2 Rf2+ 27. Kc1 Rxh2 28. Ra8+ Kh7 * [/pgn]

[Dann Corbit]

After 3. Nf3, there is a better defense for black than Qxg4 which is Qd8.
So the more realistic plan is:
[d]r1b2rk1/pp3ppp/1n2p3/3pP3/2P2B2/b1PB1qP1/P1Q1NP1P/1K1R3R w - - bm Nd4; acd 50; ce 413; pv Nd4 Qh5 g4 Qh4 Nf3 Qd8 Bxh7+ Kh8 c5 Bxc5 Ng5 Nc4 Ka1 Ba3 Rb1 Nxe5 Bg8 f5 Bxe6 Be7 Bxe5 Bxg5 gxf5 Bxe6 fxe6 Qe7 Qg6 Bf6 Rhe1 Qxe6 Rxb7 Qe8 Rxg7 Qxg6 Rxg6 Bxe5 Rxe5 Kh7 Rg2 Kh6 Rxd5 Rf3 Kb2 Rc8 Rd4 Kh5 c4 Rb8+ Kc2 Rb4 Kd1 Rf6 f4 Ra4 Rd8 Rh6 c5 Rxf4 Rd5+ Kh4 Rd6 Rhf6;

[Dann Corbit]

[pgn][Event "?"] [Site "?"] [Date "????.??.??"] [Round "?"] [White "?"] [Black "?"] [Result "*"] [FEN "r1b2rk1/pp3ppp/1n2p3/3pP3/2P2B2/b1PB1qP1/P1Q1NP1P/1K1R3R w - - 0 1"] [Setup "1"] 1. Nd4 Qh5 2. g4 Qh4 3. Nf3 Qd8 4. Bxh7+ Kh8 5. c5 Bxc5 6. Ng5 Nc4 7. Ka1 Ba3 8. Rb1 Nxe5 9. Bg8 f5 10. Bxe6 Be7 11. Bxe5 Bxg5 12. gxf5 Bxe6 13. fxe6 Qe7 14. Qg6 Bf6 15. Rhe1 Qxe6 16. Rxb7 Qe8 17. Rxg7 Qxg6 18. Rxg6 Bxe5 19. Rxe5 Kh7 20. Rg2 Kh6 21. Rxd5 Rf3 22. Kb2 Rc8 23. Rd4 Kh5 24. c4 Rb8+ 25. Kc2 Rb4 26. Kd1 Rf6 27. f4 Ra4 28. Rd8 Rh6 29. c5 Rxf4 30. Rd5+ Kh4 31. Rd6 Rhf6 * [/pgn]

[OneTrickPony]

Fantastic position!
I think the main problem for Stockfish is that it doesn't see Bxh7 and Rxg7 sac from far away. Even if you click the moves it takes a while to realize this is winning. Leela is clueless a well even if you show it both Bxh7 and Rxg7.
Would you mind sharing the opening moves? I think it starts:
1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Qc2 O-O 5.e4 d5 6.e5 Ne4 7.Bd3 c5 8.Nf3 cxd4 9.Nxd4 Nd7 10.Bf4 Qh4 11.g3 *

but I have trouble deciphering how black queen ends up on f3 with white managing to 0-0-0 in the meantime.

[todd]

Fantastic position!
I think the main problem for Stockfish is that it doesn't see Bxh7 and Rxg7 sac from far away. Even if you click the moves it takes a while to realize this is winning. Leela is clueless a well even if you show it both Bxh7 and Rxg7.
Would you mind sharing the opening moves? I think it starts:
1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Qc2 O-O 5.e4 d5 6.e5 Ne4 7.Bd3 c5 8.Nf3 cxd4 9.Nxd4 Nd7 10.Bf4 Qh4 11.g3 *

but I have trouble deciphering how black queen ends up on f3 with white managing to 0-0-0 in the meantime.

This winning shot does indeed occur in the line you gave, which is a bit trendy now.

After your moves, black plays 11... Qh3?! (Black might still survive in this line, but heroic defense will be required, and as seen in this thread, even Stockfish is sometimes not up for it. Qh5 is a more reliable way to equalize) and then after 12. O-O-O Nxc3 (Bxc3 is also very difficult for black after 13. bxc3 Nb6 14. Bf1 Qh5 15. f3!)

1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.Qc2 O-O 5.e4 d5 6.e5 Ne4 7.Bd3 c5 8.Nf3 cxd4 9.Nxd4 Nd7 10.Bf4 Qh4 11.g3 Qh3 12. O-O-O Nxc3 13. bxc3 Ba3+ 14. Kb1 Nb6 15. Bf1 Qh5 16. Be2 Qg6 17. Bd3 Qh5

And now 18. g4 wins, as demonstrated in this thread.

But the puzzle can be made a bit harder by "taking back" one move for each side (Qg4-f3 and Nd4-e2). Perhaps this is what was done.