Part of the problem is Stockfish extending the PVs with Rb8 so much, for instance depths 22/42 with the 42 somewhere in the PV. And if the PV is then not so stable, lots of FHs and FL means a lot of time lost. Not finding the Mate in 7 is then probably due to many things: the quiet moves, White capturing a piece with temporarily promoting, the promotion is not singular because White can also underpromote, and slightly less extension of checks if the check is not singular (but this is minor I think). Maybe more causes, I have not checked nullmove pruning.Uri Blass wrote:Jeroen wrote:This is the craziest game of the match. At move 22 Stockfish indicates a crushing +5.44 advantage, but still loses:
[pgn][Event "K1101 - asmFish 220517 2x"]
[Site "I5-4200M @2500Mhz (Fritzmark4"]
[Date "2017.06.23"]
[Round "67"]
[White "Komodo 11.01 64-bit"]
[Black "asmFishW_2017-05-22_base"]
[Result "1-0"]
[ECO "E99"]
[Annotator "0.00;0.00"]
[PlyCount "117"]
[EventDate "2017.??.??"]
[TimeControl "180+2"]
{Int@l(R) Core(TM) i5-4200M CPU î 2.50GHz 2493 MHz W=23.7 plies; 2.741kN/s
B=24.5 plies; 3.319kN/s} 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6.
Be2 e5 7. O-O Nc6 8. d5 Ne7 9. Ne1 Nd7 10. Be3 f5 11. f3 f4 12. Bf2 g5 13. Rc1
Rf6 14. b4 Rh6 15. c5 Qe8 16. Nb5 Qh5 17. h4 {Both last book move} Ng6 {
0.00/26 36} 18. Nxc7 {0.00/23 16 (g3)} Nxh4 {-4.28/23 8} 19. Bxh4 {-0.61/28 29}
Qxh4 {-4.84/21 1} 20. c6 {-2.18/22 23 (Qc2)} Nf6 {-5.96/22 8} 21. Qd3 {-3.05/
20 8} Rb8 {-5.33/19 2} 22. Bd1 {-3.53/21 13 (b5)} bxc6 {-5.44/23 17 (Nh5)} 23.
dxc6 {-3.15/20 12} Rxb4 {-3.56/25 36 (Kh8)} 24. Rc4 {-3.14/20 6} Rb8 {-3.97/25
9} 25. Nd5 {-1.43/21 11 (Rc1)} Nxd5 {-4.45/21 4} 26. Qxd5+ {-1.09/22 8 (Bb3)}
Be6 {-1.82/26 20} 27. Qd3 {-1.91/21 3} Kh8 {-2.34/25 3 (g4)} 28. Rc3 {-0.80/20
3 (c7)} Re8 {-2.67/24 6} 29. Bb3 {-0.34/22 10 (c7)} Bc8 {-2.18/26 12 (d5)} 30.
c7 {0.00/20 3} a5 {-2.13/24 1 (g4)} 31. Bc4 {0.00/23 4 (Rc6)} Bf6 {-1.49/29 11}
32. Qd5 {0.00/26 5 (Ba6)} Bg7 {0.00/25 16 (Rf8)} 33. Qxa5 {0.40/25 15 (Qd3)}
Bf6 {-1.91/23 7} 34. Qa4 {1.76/22 6 (Ba6)} Rf8 {1.17/26 9 (Qh5)} 35. Ba6 {
1.56/21 2} Be6 {2.10/25 7} 36. c8=Q {1.70/22 3} Bxc8 {2.40/23 2} 37. Bxc8 {
1.90/23 5} Bd8 {2.14/24 5} 38. Qb5 {1.72/24 6} d5 {2.08/22 1} 39. Rc6 {1.88/21
3} Rxc6 {2.54/25 10} 40. Qxc6 {2.73/22 10} dxe4 {2.52/24 7} 41. Bf5 {2.48/23 7
(Bh3)} e3 {2.48/25 11} 42. Nd3 {2.39/21 7} Qh5 {2.50/25 4 (g4)} 43. Rb1 {
2.91/21 4} Qe8 {2.68/23 0} 44. Qxe8 {3.01/23 7 (Qe4)} Rxe8 {2.64/24 4} 45. Rb8
{3.19/22 5 (Rb7)} Rg8 {2.62/22 4 (Kg8)} 46. Kf1 {3.24/23 4} Bc7 {3.30/24 7} 47.
Rb7 {3.13/23 3} Rf8 {3.10/22 1 (Rg7)} 48. Be4 {3.62/23 3 (Rxc7)} Rc8 {4.33/24 4
} 49. a4 {3.70/25 5 (Nb4)} Ba5 {3.20/20 2} 50. Rxh7+ {3.63/27 5 (Ra7)} Kg8 {
2.88/1 0} 51. Ra7 {3.93/26 2 (Rh5)} Bb6 {3.30/23 5} 52. Ra8 {5.03/30 9 (Rb7)}
Rxa8 {3.54/27 2} 53. Bxa8 {5.03/31 1} Bc7 {3.82/32 3 (Kf8)} 54. Nb2 {5.11/31 4
(Bd5+)} Kf7 {5.46/26 2 (Kg7)} 55. Nc4 {5.16/30 2} Ke8 {6.15/28 2 (Ke6)} 56. a5
{5.34/30 5} Kd8 {6.77/27 2 (Ke7)} 57. a6 {5.52/26 2} Bb8 {6.97/31 2} 58. Na5 {
5.75/27 2} Kd7 {7.89/27 2 (Bc7)} 59. Nc6 {6.98/25 1 (Be4)} 1-0[/pgn]
Stockfish missed a simple mate in 7 when other program find the mate in 7 clearly faster.
[D]r1b3k1/ppN3bp/2Pp1n1r/3Pp1p1/1P2Pp1q/3Q1P2/P3B1P1/2R1NRK1 b - - 2 21
Stockfish_17062123_x64_modern:
1/1 00:00 277 1k +7.40 21. ... Rb8
2/2 00:00 358 2k +8.45 21. ... Rb8 22.a3
3/3 00:00 399 2k +8.45 21. ... Rb8 22.a3 bxc6
4/4 00:00 541 3k +8.06 21. ... Rb8 22.b5 bxc6 23.dxc6
5/5 00:00 2k 10k +5.97 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Bxe6 24.dxe6
6/7 00:00 3k 15k +6.55 21. ... Rb8 22.Ne6 bxc6 23.Nxg7 Kxg7 24.dxc6
7/9 00:00 5k 24k +5.68 21. ... Rb8 22.Ne6 bxc6 23.Nxg7 Kxg7 24.dxc6 Ng8
8/9 00:00 7k 35k +6.22 21. ... Rb8 22.Ne6 Ne8 23.b5 bxc6 24.Nxg7 Kxg7 25.bxc6
9/10 00:00 14k 64k +6.73 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Ne8 24.Nc2 Bb7 25.Qa3 Ra8
10/13 00:00 43k 185k +5.54 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Ne8 24.Qa3 Ra8 25.Rxc8 Rxc8 26.Qxa7 Bf6
11/15 00:00 71k 289k +5.52 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Bd7 24.Rxd6 Bf8 25.Nxf8 Rxf8 26.Qc3 g4 27.Rxd7 Nxd7 28.fxg4
12/21 00:00 188k 576k +5.35 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Ne8 24.Qa3 a6 25.Rxc8 Rxc8 26.Nxg5 Qxg5 27.Bxa6 Rb8 28.Kf2 Nc7 29.Ke2
13/21 00:00 351k 870k +5.43 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Qh2+ 24.Kf2 Bxe6 25.dxe6 Qh4+ 26.Kg1 Bf8 27.e7 Bxe7 28.Qc4+ d5 29.exd5 Rxb4 30.d6+ Rxc4 31.Bxc4+ Kf8 32.dxe7+ Kxe7 33.Nd3
14/25 00:00 588k 1,146k +5.38 21. ... Rb8 22.Ne6 Bf8 23.cxb7 Bxb7 24.Rc7 Bc8 25.Nxg5 Qxg5 26.Qc3 Bd7 27.Rxd7 Nxd7 28.Qc7 Qe7 29.Nd3 Qh4
15/25 00:00 1,518k 1,886k +5.02 21. ... Rb8 22.Ne6 Bf8 23.cxb7 Bxb7 24.b5 Nd7 25.Rc7 Nc5 26.Nxc5 dxc5 27.Rxb7 Rxb7 28.d6 Bxd6 29.Qd5+ Rf7 30.Nd3 Kg7 31.Nxe5
16/25 00:01 3,297k 2,288k +4.84 21. ... Rb8 22.Ne6 b5 23.a4 bxa4 24.b5 Bxe6 25.dxe6 Bf8 26.Qc4 d5 27.exd5 Bd6 28.b6 axb6 29.c7 Rc8 30.Qa6
17/30 00:02 5,111k 2,544k +4.72 21. ... Qh1+ 22.Kf2 Qh4+ 23.Kg1 Rb8 24.Ne6 bxc6 25.Rxc6 Bxe6 26.dxe6 Bf8 27.Nc2 Re8 28.Rc7 Rxe6 29.Rxa7 g4 30.Qc4 Ne8 31.fxg4 Kh8 32.Qc8 Bg7 33.g5
18/30 00:02 5,302k 2,569k +4.61 21. ... Qh1+ 22.Kf2 Qh4+ 23.Kg1 Rb8 24.Ne6 bxc6 25.Rxc6 Bxe6 26.dxe6 Bf8 27.Nc2 Re8 28.Rc7 Rxe6 29.Rxa7 Kh8 30.Rf7 Bg7 31.Qb5 Ne8 32.Qb8
19/35 00:02 7,576k 2,678k +4.65 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Bxe6 24.dxe6 d5 25.Nc2 dxe4 26.Qc4 Re8 27.e7+ Kh8 28.Rxf6 Qh2+ 29.Kf2 e3+ 30.Ke1 Rxf6 31.Kd1 Rxe7 32.Qc8+ Bf8 33.Qg4 Rc6 34.Bd3 Rc3 35.Qf5 Qh6 36.Kc1 Rcc7 37.Qh3
20/42+ 00:03 8,376k 2,728k +4.72 21. ... Rb8
20/42+ 00:03 8,400k 2,727k +4.79 21. ... Rb8
20/42+ 00:03 8,477k 2,722k +4.90 21. ... Rb8
20/42+ 00:03 8,659k 2,708k +5.06 21. ... Rb8
20/42+ 00:03 8,791k 2,698k +5.27 21. ... Rb8
20/42- 00:03 8,903k 2,693k +5.10 21. ... Rb8 22.Ne6
20/42 00:03 9,611k 2,713k +5.25 21. ... Rb8 22.Ne6 bxc6 23.Rxc6 Bxe6 24.dxe6 d5 25.Nc2 dxe4 26.Qc4 Re8 27.e7+ Kh8 28.Rxf6 Qh2+ 29.Kf2 e3+ 30.Ke1 Rxf6 31.Kd1 Rxe7 32.Kc1 Rf8 33.Rd1 Qxg2 34.Ne1 Qh3 35.Qc5 Ref7 36.Bc4 Rd7 37.Rxd7 Qxd7 38.Nd3 Rd8 39.Qa5
21/42+ 00:03 10,307k 2,743k +5.32 21. ... Rb8
21/42+ 00:03 10,785k 2,760k +5.39 21. ... Rb8
21/42+ 00:04 12,598k 2,761k +5.50 21. ... Rb8
21/42+ 00:04 12,973k 2,773k +5.65 21. ... Rb8
21/42+ 00:05 14,390k 2,820k +5.86 21. ... Rb8
21/42- 00:05 15,477k 2,814k +5.69 21. ... Rb8 22.Ne6
21/42- 00:05 16,354k 2,834k +5.32 21. ... Rb8 22.Ne6
21/42 00:06 18,149k 2,850k +4.87 21. ... Rb8 22.Ne6 Bh8 23.b5 bxc6 24.bxc6 g4 25.fxg4 Qh2+ 26.Kf2 Ba6 27.Qf3 Qh4+ 28.Kg1 Bxe2 29.Qh3 Qxh3 30.gxh3 Bxf1 31.g5 Rh5 32.Kxf1 Nxe4 33.c7 Rc8
22/42+ 00:06 18,624k 2,849k +4.94 21. ... Rb8
22/42+ 00:06 18,728k 2,848k +5.01 21. ... Rb8
22/42+ 00:06 18,822k 2,849k +5.12 21. ... Rb8
22/42+ 00:06 18,893k 2,850k +5.27 21. ... Rb8
22/42+ 00:06 19,053k 2,852k +5.48 21. ... Rb8
22/42+ 00:07 22,804k 2,888k +5.77 21. ... Rb8
22/42 00:10 32,101k 2,934k +5.58 21. ... Rb8 22.Nb5 bxc6 23.Nxd6 Bf8 24.Rxc6 Bd7 25.a3 Be7 26.Qc4 g4 27.fxg4 Qh2+ 28.Kf2 Qg3+ 29.Kg1 Bxd6 30.Rxd6 Bb5 31.Rxf6 Bxc4 32.Rxh6 Bxe2 33.Rh3 Qxg4 34.Rf2 Rc8 35.Kh2
23/42- 00:12 35,835k 2,949k +5.51 21. ... Rb8 22.Nb5
23/42- 00:13 39,828k 2,956k +5.43 21. ... Rb8 22.Nb5
23/42- 00:14 41,907k 2,976k +5.33 21. ... Rb8 22.Nb5
23/42- 00:14 42,956k 2,972k +5.17 21. ... Rb8 22.Nb5
23/42- 00:15 46,035k 2,979k +4.96 21. ... Rb8 22.Nb5
23/42+ 00:15 46,105k 2,979k +5.24 21. ... Rb8
23/42+ 00:15 47,175k 2,983k +5.61 21. ... Rb8
23/43 00:19 59,251k 2,967k +5.44 21. ... Rb8 22.cxb7 Rxb7 23.Ne6 Bd7 24.a3 Bh8 25.Nxg5 Qxg5 26.Qa6 Rb8 27.Qxd6 Rd8 28.Bb5 Qh4 29.Qe7 Re8 30.Qxd7 Nxd7 31.Bxd7 Rd8 32.Rc8 Bf6 33.Rxd8+ Bxd8 34.Bh3 Qg3 35.Kh1 Rxh3+ 36.gxh3
24/43- 00:25 74,677k 2,920k +5.37 21. ... Rb8 22.Bd1
24/43- 00:27 79,484k 2,911k +5.29 21. ... Rb8 22.Bd1
24/43+ 00:28 82,479k 2,908k +5.81 21. ... Qh1+
24/43+ 00:28 82,492k 2,908k +5.97 21. ... Qh1+
24/43+ 00:28 82,558k 2,908k +6.18 21. ... Qh1+
24/43+ 00:28 82,608k 2,909k +6.46 21. ... Qh1+
24/43+ 00:28 82,663k 2,909k +6.83 21. ... Qh1+
24/43+ 00:28 82,758k 2,909k +7.31 21. ... Qh1+
24/43+ 00:28 82,869k 2,909k +7.94 21. ... Qh1+
24/43+ 00:28 82,995k 2,910k +8.74 21. ... Qh1+
24/43+ 00:28 83,175k 2,911k +9.75 21. ... Qh1+
24/43+ 00:28 83,359k 2,912k +11.04 21. ... Qh1+
24/43+ 00:28 83,512k 2,912k +12.68 21. ... Qh1+
24/43+ 00:28 83,701k 2,914k +14.74 21. ... Qh1+
24/43+ 00:28 83,787k 2,915k +17.33 21. ... Qh1+
24/43+ 00:28 83,831k 2,915k +20.59 21. ... Qh1+
24/43+ 00:28 83,861k 2,916k +24.68 21. ... Qh1+
24/43+ 00:28 83,886k 2,916k +29.81 21. ... Qh1+
24/43+ 00:28 83,914k 2,916k +36.25 21. ... Qh1+
24/43+ 00:28 83,934k 2,916k +44.31 21. ... Qh1+
24/43+ 00:28 83,966k 2,917k +54.41 21. ... Qh1+
24/43+ 00:28 83,986k 2,917k +67.06 21. ... Qh1+
24/43+ 00:28 84,009k 2,917k +82.89 21. ... Qh1+
24/43+ 00:28 84,039k 2,917k +102.69 21. ... Qh1+
24/43+ 00:28 84,100k 2,918k +127.46 21. ... Qh1+
24/43 00:28 84,321k 2,919k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
25/43 00:28 84,373k 2,919k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
26/43 00:28 84,514k 2,919k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
27/43 00:28 84,630k 2,920k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
28/43 00:29 84,839k 2,920k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
29/43 00:29 84,996k 2,921k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
30/43 00:29 85,354k 2,921k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
31/43 00:29 85,660k 2,920k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
32/43 00:29 85,907k 2,920k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
33/43 00:29 86,402k 2,921k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
34/43 00:29 86,820k 2,923k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
35/43 00:29 87,328k 2,925k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
36/43 00:29 87,841k 2,928k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
37/43 00:30 88,841k 2,933k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
38/43 00:30 89,486k 2,936k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
39/43 00:30 90,225k 2,940k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
40/43 00:30 91,051k 2,944k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
41/43 00:31 92,038k 2,949k +M7 21. ... Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2 24.Ne8 Nxe8 25.cxb7 Qh4 26.g3 Qxg3+ 27.Ng2 Qxg2+
There is a slight inconsistency though in Stockfish that also does not help; we don't do singular extensions in rootNode. It's not entirely clear why but one reason could be if you extend the PV move and it Fails Low which is the main reason for extending it, you can not drop out of the node quickly to the (grand-)parent node to try something better there. You can only fall back on the other moves in the root. In that sense the root node is even more of an ALL node than the other PV nodes. So probably a good thing not to SE here. And yet we do, on the other hand, do LMR in the Root as in other PvNodes. This breaks us up in this particular case, at least relatively. Not doing LMR helps, to some extent:
r1b3k1/ppN3bp/2Pp1n1r/3Pp1p1/1P2Pp1q/3Q1P2/P3B1P1/2R1NRK1 b - -
Engine: Sf20170524_047 HT MOD (512 MB)
by T. Romstad, M. Costalba, J. Kiiski, G. Linscott
20/32 0:01 -5.85 21...Rb8 22.b5 Bf8 23.Ne6 bxc6
24.bxc6 Be7 25.Bd1 Ne8 26.Bb3 Rb6
27.Rc2 a5 28.Qe2 a4 29.Bxa4 g4
30.fxg4 Bxe6 31.dxe6 (18.105.199) 10841
21/32 0:02 -4.53 21...Rb8 22.Ne6 Bf8 23.Nxf8 Kxf8
24.b5 Ra8 25.Nc2 Kg8 26.Nb4 Kh8
27.a4 a5 28.Na2 b6 29.Qa3 Ra7 30.Bc4 Ng8
31.Qc3 g4 32.fxg4 Bxg4 (30.838.525) 10847
22/32 0:04 -4.60++ 21...Rb8 (50.356.515) 10806
22/32 0:05 -4.68++ 21...Rb8 (56.396.272) 11071
22/32 0:05 -4.79++ 21...Rb8 (58.995.575) 11097
22/32 0:05 -4.94++ 21...Rb8 (61.362.053) 11124
22/32 0:05 -5.15++ 21...Rb8 (64.367.969) 11153
22/32 0:06 -5.43++ 21...Rb8 (78.238.053) 11262
22/32 0:12 -5.80++ 21...Rb8 (138.790.394) 11503
22/32 0:12 -M16++ 21...Qh2+ (147.416.841) 11513
22/32 0:12 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (147.425.529) 11513
23/32 0:12 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (147.447.211) 11513
24/32 0:12 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (147.472.847) 11513
25/32 0:12 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (147.510.772) 11513
26/32 0:12 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (147.553.706) 11514
27/32 0:12 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (147.608.167) 11513
.
.
.
51/32 0:18 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (208.203.393) 11456
52/32 0:20 -M7 21...Qh2+ 22.Kf2 Qg3+ 23.Kg1 Rh2
24.cxb7 Qh4 25.bxc8Q+ Rxc8 26.g3 Qxg3+
27.Ng2 Qxg2+ (233.310.540) 11460
best move: Qh4-h2 time: 0:20.484 min n/s: 11.460.949 nodes: 233.310.540
Note there also is lots of variability from one run to the next, so I can't be sure this really helps. Twelve seconds is better than any other thing I have tried though.