Could Gaviota have won against Shredder, in TCEC?

[JuLieN]

Gaviota just qualified splendidly by drawing Shredder in TCEC. But many observers wonder if Gaviota could actually have won the game, with a handset of more accurate moves. A typical example occurred in this position:

[d]8/6k1/3q1pn1/8/2p2PPp/2pr4/P5RK/2Q2R2 b - - 3 50
Shredder, with white, just played Qc1, and Gaviota will answer with the much debated Kh7 move. Could this be improved?

StockFish has Kh7 at -0.96 (so still a good position for black), but this is only its... 15th choice (!!!), far from it's #1 choice: Kf7 (-4.08):

Code: Select all

Stockfish  130131 64bit: 
-4.08/26 50...Kf7 51.Rgf2 Qd7 52.Rg2 Qd4 53.g5 Rd2 54.Qb1 Rxg2+ 55.Kxg2 Nxf4+ 56.Rxf4 Qxf4 57.Qh7+ Ke6 58.Qg8+ Ke7 59.Qg7+ Kd6 60.Qf8+ Kd5 61.Qf7+ Kd6
-3.87/26 50...Qd4 51.g5 fxg5 52.Rxg5 Rd2+ 53.Rg2 h3 54.Kxh3 Kf7 55.Rxd2 cxd2 56.Qb1 Kf6 57.Kh2 Qe3 58.Qd1 c3 59.Qa4 Qd3 60.Qc6+ Kf7 61.Qb7+ Ne7 62.Qb3+ Kf6 63.Qb6+ Kf5 64.Qc5+ Nd5 65.Qc8+ Ke4 66.Qe8+ Kd4 67.Qh8+
-2.30/26 50...h3 51.Rgf2 Qd4 52.Qe1 Rd2 53.Kxh3 Qd3+ 54.Kh2 Kf7 55.Rg2 Qd4 56.g5 Nxf4 57.g6+ Nxg6 58.Qg3 Qh4+ 59.Qh3 Rxg2+ 60.Kxg2 Nf4+ 61.Rxf4 Qxf4 62.Qxc3 Qc7 63.Kf3 f5 64.Ke3 Qb6+ 65.Kf4 Qd6+ 66.Ke3 Qc5+
-2.30/26 50...Qd7 51.Qe1 h3 52.Rc2 Qd4 53.Rcf2 Rd2 54.Kxh3 Qd3+ 55.Kh2 Kf7 56.Rg2 Qd5 57.Rff2 Nh4 58.Rg3 Nf3+
-2.30/26 50...Qd8 51.Qe1 h3 52.Rc2 Qd4 53.Rcf2 Rd2 54.Kxh3 Qd3+
-2.18/26 50...Ne7 51.Qb1 Qd7 52.g5 Qh3+ 53.Kg1 Qe3+ 54.Rgf2 h3 55.Qb5 Rd2 56.gxf6+ Kxf6 57.Qe5+ Qxe5 58.fxe5+ Ke6 59.Re1 Nd5 60.a3 Rd3 61.Rfe2 Nf4 62.Rc2 Rg3+ 63.Kh1 Nd3 64.Rg1 Re3 65.Kh2 Ne1 66.Rxe1 Rxe1 67.Rxc3
-1.93/26 50...Rd2 51.Qxc3 Rd4 52.Kh1 Nxf4 53.Rgf2 Qd5+ 54.Qf3 c3 55.Qxd5 Nxd5 56.Ra1 Ne3 57.Re2 Rd3 58.Rc1 Kg6 59.a4 c2 60.Rexc2 Nxc2 61.Rxc2 Kg5 62.a5
-1.69/26 50...Qe6 51.a4 Qd5 52.a5 h3 53.Rgf2 Qxa5 54.Qb1 Qa7 55.Qb4 Qc7 56.Qb5 Rd2 57.Kxh3 Rxf2 58.Rxf2 Nxf4+ 59.Kh4 Nd3 60.Re2 Ne5 61.Kh3 Nf3 62.Qb1 Nd4 63.Re8 Qf7 64.Re3
-1.49/26 50...Qd5 51.Qe1 h3 52.Rgf2 Qd6 53.Qe4 Rd2 54.Kh1 c2 55.Qb7+ Kh6 56.g5+ fxg5 57.fxg5+ Kxg5 58.Rg1+ Kh6 59.Qf3 Rxf2 60.Qxf2 Qd5+ 61.Kh2 Qe5+ 62.Kh1 Qe4+ 63.Kh2 Kh5 64.Qe1 Qf4+ 65.Kxh3 Qf5+ 66.Kg3 Ne5 67.Qe2+ Kg6 68.Rf1
-1.49/26 50...Qb6 51.Qe1 h3 52.Rgf2 Qd6 53.Qe4 Rd2 54.Kh1 c2 55.Qb7+ Kh6 56.g5+ fxg5 57.fxg5+ Kxg5 58.Rg1+ Kh6 59.Qf3 Rxf2 60.Qxf2 Qd5+ 61.Kh2 Qe5+ 62.Kh1 Qe4+ 63.Kh2 Kh5 64.Qe1 Qf4+ 65.Kxh3 Qf5+ 66.Kg3 Ne5 67.Qe2+ Kg6 68.Rf1
-1.49/26 50...Qc6 51.Qe1 h3 52.Rc2 Qd5 53.Rcf2 Qd6 54.Qe4 Rd2 55.Kh1 c2 56.Qb7+ Kh6 57.g5+ fxg5 58.fxg5+ Kxg5 59.Rg1+ Kh6 60.Qf3 Rxf2 61.Qxf2 Qd5+ 62.Kh2 Qe5+ 63.Kh1 Qe4+ 64.Kh2 Kh5 65.Qe1 Qf4+ 66.Kxh3 Qf5+ 67.Kg3 Ne5 68.Qe2+ Kg6 69.Rf1
-1.09/26 50...Rd4 51.Rgf2 Kh7 52.Qe3 Rd2 53.a4 Nxf4 54.Kh1 Rxf2 55.Rxf2 Nd3 56.Rc2 Kg7 57.a5 Qd5+ 58.Kh2 Ne5 59.Qxc3 Nf3+ 60.Kh3 Ng5+ 61.Kh2 Qd1 62.Kg2 Ne4 63.Qb2
-1.05/26 50...Kf8 51.a4 h3 52.Rgf2 Ne7 53.Qe1 Nd5 54.Qe4 Ne3 55.Rg1 c2 56.Re2 c3 57.Kh1 Qd7 58.a5 c1=Q 59.Rxc1 Rd1+ 60.Re1 Rxc1 61.Rxc1 Nxg4 62.Qa8+ Kf7 63.Rg1 h2 64.Rc1 Qd2 65.Qb7+ Kg6 66.Qe4+ Kh6 67.Qc2 Qxf4 68.a6 Qf3+
-1.05/26 50...Qe7 51.g5 f5 52.Qb1 h3 53.Rgf2 Qe3 54.Qb7+ Ne7 55.Qc7 Rd2 56.Rxd2 cxd2 57.Qxc4 d1=R 58.Rxd1 Qf2+ 59.Kxh3 Qf3+ 60.Kh2 Qxd1 61.Qc3+ Kf7 62.Qe5 Qc2+ 63.Kg3 Qd3+ 64.Kf2 Qa6 65.Ke1 Qa7 66.Kd1 Qa6
-0.96/26 50...Kh7 51.a4 Rd4 52.Rgf2 Rd2 53.Qxc3 Qxf4+ 54.Kg1 Qxg4+ 55.Kh1 Rd3 56.Qe1 Rg3 57.Rh2 c3 58.Rxf6 Kg7 59.Rff2 Qxa4 60.Rhg2 Qc6 61.Kg1 Qd6 62.Rxg3
-0.80/26 50...Qb4 51.g5 f5 52.Re2 Qd6 53.a4 Qd4 54.Rff2 Kf7 55.a5 Qd5 56.Qf1 Qxa5 57.Ra2 Qc7 58.Qa1 Qc5 59.Rae2 Qc6 60.Ra2 Qd6 61.Qb1 Ne7 62.Qe1
-0.60/26 50...Kg8 51.a4 Qd5 52.Qb1 Kf7 53.Rgf2 Qe6 54.Qb7+ Ne7 55.Qb5 Kg7 56.Qb7 Kf8 57.Qb8+ Kf7 58.Qb5 Kg7 59.Qb7 Kf8 60.Qb8+ Kf7 61.Qb5
-0.40/26 50...Kh8 51.a4 Qc5 52.Qe1 h3 53.Re2 Kg7 54.Qf2 Qc7 55.Kg1 Qa5 56.Kh2 Kh7 57.Kg1 Qc7 58.Re4
-0.36/26 50...Qc7 51.g5 fxg5 52.Rxg5 Qd7 53.Qe1 Kf6 54.Qe4 Rd2+ 55.Rg2 Qd3 56.Qxd3 Rxd3 57.a4 Ne7 58.Rc1 Nd5 59.a5 Rd2 60.a6 h3 61.Kxh3 Nxf4+ 62.Kg3 Nxg2 63.Rxc3 Ra2 64.Rxc4 Ne3 65.Rc6+ Ke5
-0.52/26 50...Qc5 51.g5 f5 52.a4 Qb4 53.Ra2 Qd6 54.Raf2 Ne7 55.Re1 Nd5 56.Re8 Rd4 57.Kg1 Rxf4 58.Rxf4

Here's the whole game:

[pgn]
[Event "nTCEC - Stage 1 - Season 1"]
[Site "http://www.tcec-chess.net"]
[Date "2013.02.18"]
[Round "7.8"]
[White "Shredder 12.0"]
[Black "Gaviota 0.86b3"]
[Result "1/2-1/2"]
[Variant "normal"]

1. d4 d6 2. Nf3 g6 3. g3 Bg7 4. Bg2 Nd7 5. O-O e5 6. c4 Ne7 7. Nc3 O-O 8. e4
exd4 9. Nxd4 Re8 10. b3 Nc6 11. Be3 Nde5 12. Rc1 Ng4 13. Nxc6 bxc6 14. Bd2 Ne5
15. Be3 Ng4 16. Bd2 Ne5 17. Na4 g5 18. h3 Rb8 19. Be3 c5 20. Nc3 g4 21. hxg4
Bxg4 22. f3 Be6 23. Qc2 h5 24. Kh1 h4 25. g4 Ng6 26. Qd2 a5 27. Kh2 Rb4 28. Kh3
Qd7 29. Qd1 a4 30. Nxa4 Ra8 31. Rf2 Raxa4 32. bxa4 Rxa4 33. Rb1 Bxc4 34. Bf1
Bxf1+ 35. Rxf1 Bd4 36. Bxd4 Rxd4 37. Qc1 Kg7 38. Qg5 d5 39. Rf2 c4 40. Re1 c3
41. Kg2 Qe6 42. Kh1 Qe5 43. Qc1 Rd3 44. exd5 Qxd5 45. Ref1 f6 46. Kh2 c5 47. f4
Qd7 48. Rg2 c4 49. Qe1 Qd6 50. Qc1 Kh7 51. a4 Rd4 52. Rgf2 Rd2 53. Qxc3 Qxf4+
54. Kg1 Qxg4+ 55. Kh1 Rd3 56. Qe1 f5 57. Rh2 Qd4 58. Qf2 Qf6 59. Rg1 f4 60. Qc5
c3 61. a5 Kh6 62. a6 Qd8 63. Qc6 Rg3 64. Rc1 Qe7 65. Ra1 Re3 66. Rg2 Rh3+
67. Rh2 1/2-1/2
[/pgn]

There were other inaccuracies later, that drove this game to a draw, so I'll try an auto play with Stockfish starting with 50. ..., Kf7, and see if it leads to a black victory.

[muxecoid]

The positional sacrifice on a4 is awesome for sure. Shredder's eval jumped to 1.2 and than quickly fell down.

Black's Qd4 with Rd2 to follow looks unstoppable. Black gets pawns on c3 and d2 and can re-position his king to a safe place preparing for final push.

[JuLieN]

Here is the result of this (very) long analysis (usually 20-30mn/move on my Core i7 MacBook Pro) by Stockfish:

[pgn]

[Event "Shredder - Gaviota TCEC game"]
[Site "TCEC"]
[Round "1"]
[White "Stockfish"]
[Black "Stockfish"]
[Result "0-1"]
[FEN "8/6k1/3q1pn1/8/2p2PPp/2pr4/P5RK/2Q2R2 b - - 3 50"]
[SetUp "1"]

50...Kf7 51.a4 Qd4 52.g5 fxg5 53.fxg5+ Kg8 54.Rf6 Qe5+ 55.Kg1 Kg7 56.Qf1 Qc5+ 57.Qf2 Rd1+ 58.Kh2 Qc7+ 59.Kh3 Ne5 60.Qf4 Rh1+ 61.Rh2 Rxh2+ 62.Kxh2 Ng4+ 63.Kg1 Qxf4 64.Rxf4 Ne3 65.Re4 c2 66.Rxe3 c1=Q+ 67.Kf2
0-1
[/pgn]

... with a mate in 12 moves after Kf2.

For those interested, here's the whole analysis, with all the lines from Stockfish.

[lucasart]

Here is the result of this (very) long analysis (usually 20-30mn/move on my Core i7 MacBook Pro) by Stockfish:

[pgn]

[Event "Shredder - Gaviota TCEC game"]
[Site "TCEC"]
[Round "1"]
[White "Stockfish"]
[Black "Stockfish"]
[Result "0-1"]
[FEN "8/6k1/3q1pn1/8/2p2PPp/2pr4/P5RK/2Q2R2 b - - 3 50"]
[SetUp "1"]

50...Kf7 51.a4 Qd4 52.g5 fxg5 53.fxg5+ Kg8 54.Rf6 Qe5+ 55.Kg1 Kg7 56.Qf1 Qc5+ 57.Qf2 Rd1+ 58.Kh2 Qc7+ 59.Kh3 Ne5 60.Qf4 Rh1+ 61.Rh2 Rxh2+ 62.Kxh2 Ng4+ 63.Kg1 Qxf4 64.Rxf4 Ne3 65.Re4 c2 66.Rxe3 c1=Q+ 67.Kf2
0-1
[/pgn]

... with a mate in 12 moves after Kf2.

For those interested, here's the whole analysis, with all the lines from Stockfish.

Interesting.

I submitted the problem to DiscoCheck. It never considered either Gaviota's Kh7, or Stockfish's Kf7, prefering instead Qd4!?
[d]8/6k1/3q1pn1/8/2p2PPp/2pr4/P5RK/2Q2R2 b - - 3 50

Code: Select all

info score cp 92 depth 7 nodes 29502 time 27 pv d6d4 f4f5 d4e5 h2h1 h4h3 g2c2 e5e4 h1h2 g6e5
info score cp 125 depth 8 nodes 59763 time 46 pv d6d4 g4g5 d3d2 g5f6 g7f6 c1e1 d2g2 h2g2 g6f4 g2h2 d4d2 e1d2 c3d2 f1f4 f6e5
info score cp 125 depth 9 nodes 76345 time 55 pv d6d4 g4g5 d3d2 g5f6 g7f6 c1e1 d2g2 h2g2 g6f4 g2h2 d4d2 e1d2 c3d2 f1f4 f6e5
info score cp 120 depth 10 nodes 113328 time 77 pv d6d4 g4g5 d3d2 g5f6 g7f6 c1e1 d2g2 h2g2 g6f4 g2f3 c3c2 e1h4 f6e5
info score cp 124 depth 11 nodes 184174 time 117 pv d6d4 g4g5 d3d2 g5f6 g7f6 c1e1 d2g2 h2g2 g6f4 g2f3 d4g1
info score cp 140 depth 12 nodes 270141 time 165 lowerbound
info score cp 126 depth 12 nodes 444834 time 258 pv d6d4 g2f2 d3d2 c1e1 d2f2 e1f2 d4f2 f1f2 g7h7
info score cp 94 depth 13 nodes 1366838 time 759 upperbound
info score cp 78 depth 13 nodes 1612977 time 895 upperbound
info score cp 76 depth 13 nodes 1874417 time 1038 pv d6d4 c1e1 d3d2 f1f2 g6f4 f2d2 c3d2 g2d2 d4c3
info score cp 87 depth 14 nodes 2232984 time 1226 pv d6d4 c1e1 d3h3 h2h3 g6f4 h3h2 h4h3 e1e7 g7g6 g2f2 d4f2 f1f2 f4d5
info score cp 103 depth 15 nodes 3318224 time 1799 lowerbound
info score cp 99 depth 15 nodes 3624138 time 1963 pv d6d7 a2a4 h4h3 g2f2 d7g4 h2h1 g4f5 c1a3 f5a5 f4f5 a5d5 h1g1 g6e5 a3e7 d5f7
info score cp 88 depth 16 nodes 7201288 time 3816 pv d6d7 a2a4 h4h3 g2f2 d7g4 f1g1 g4h4 c1f1 d3d2 g1g3
info score cp 104 depth 17 nodes 13280726 time 7039 lowerbound
info score cp 120 depth 17 nodes 14567536 time 7703 lowerbound
info score cp 149 depth 17 nodes 20559286 time 10662 pv d6d4 g4g5 f6g5 g2g5 d3d2 g5g2 g7f6 c1e1 g6f4 e1h4 f6e5 g2d2 d4d2 f1f2 e5e6
info score cp 159 depth 18 nodes 25106501 time 12929 pv d6d4 g4g5 f6g5 g2g5 d3d2 g5g2 g7f6 c1e1 d2g2 h2g2 g6f4 g2f3 f6f5 e1e5 f5e5 f1c1
info score cp 143 depth 19 nodes 30835899 time 15743 upperbound
info score cp 175 depth 19 nodes 35412848 time 18044 lowerbound
info score cp 149 depth 19 nodes 49025897 time 24634 pv d6d4 g4g5 f6g5 g2g5 d3d2 g5g2 d2g2 h2g2 d4d5 g2h3 d5e6 h3g2 h4h3 g2f2 e6b6 f2f3 b6c6 f3e3 c6c5 e3f3 g6h4 f3e2 c5h5 e2e1 h4g2 e1f2 h5h4 f2e2 h4e7 e2f2 e7a7 f2f3 g2h4 f3g3 a7a3
info score cp 162 depth 20 nodes 84340604 time 41333 pv d6d4 g4g5 f6g5 g2g5 d4f4 f1f4 d3d2 g5g2 d2c2
info score cp 178 depth 21 nodes 133541683 time 64667 lowerbound
info score cp 194 depth 21 nodes 157128032 time 75900 lowerbound
info score cp 196 depth 21 nodes 191450524 time 91937 pv d6d4 g4g5 d3d2 g5f6 g7f6 a2a4 g6f4 h2h1 f6e7 c1e1 e7d7 g2f2 d4d5 h1h2 f4d3 f2d2 d5d6 h2h1 c3d2 f1f7 d7d8 e1h4 d8e8
info score cp 195 depth 22 nodes 283888149 time 134083 pv d6d4 g4g5 d3d2 g5f6 g7f6 a2a4 g6f4 h2h1 f6e7 c1e1 e7d7 g2f2 d4d5 h1h2 d5c6

Is this Qd4 also a potentially winning move ? Or another drawish one like Kh7 ?

[JuLieN]

Is this Qd4 also a potentially winning move ? Or another drawish one like Kh7 ?

Qd4 seems to be an even better move than Kf7. Kf7 was SF's best move at depth 26 (see first post), with Qd4 not much behind it... But I currently have an analysis at depth 31 and Qd4 is now SF's fist choice (with -6.22) just before Kf7 (-6.06).
I'll do the same analysis with Qd4 as the one I did with Kf7. This is an interesting position because it shows that you just need a few inaccuracies to turn a victory into a draw.

[lucasart]

Is this Qd4 also a potentially winning move ? Or another drawish one like Kh7 ?

Qd4 seems to be an even better move than Kf7. Kf7 was SF's best move at depth 26 (see first post), with Qd4 not much behind it... But I currently have an analysis at depth 31 and Qd4 is now SF's fist choice (with -6.22) just before Kf7 (-6.06).
I'll do the same analysis with Qd4 as the one I did with Kf7. This is an interesting position because it shows that you just need a few inaccuracies to turn a victory into a draw.

Thank you for the analysis. So Qd4 is also a winning move. Critter 1.6 seems to like Qd4 also:

Code: Select all

info multipv 1 depth 23 seldepth 72 nodes 5945051412 time 440799 nps 13486989 hashfull 1000 tbhits 0 score cp 341 pv d6d4 g4g5 f6g5 g2g5 d3d2 g5g2 h4h3 h2h3 g7f6 a2a3 d4d3 h3h2 d2g2 h2g2 g6h4 g2h2 h4f3 f1f3 d3d2 c1d2 c3d2 f3f1 c4c3 f1d1 c3c2 d1d2 c2c1q d2f2 f6f5 a3a4 c1c5 h2g3 c5a3 f2f3 a3a4 g3g2 a4d4 g2h3 d4d1 h3h4 d1e1 h4h3 e1h1 h3g3 h1g1 g3h4 g1h2 

[JuLieN]

For those interested, here's Stockfish's analysis at depth 38*.

Code: Select all

Stockfish  130131 64bit: 
-6.62/38 50...Kf7 51.a4 Qd4 52.a5 h3 53.Rgf2 Rd2 54.Qe1 Nxf4 55.a6 Qd6 56.Rxd2 cxd2 57.Qg3 c3 58.a7 Qd5 59.Ra1 Qa8 60.g5 c2 61.Qb3+ Ne6 62.g6+ Kg7 63.Qxh3 Nf8 64.Qg2 Qxg2+ 65.Kxg2 c1=Q 66.a8=Q d1=Q 67.Rxc1 Qxc1 68.Qb7+ Kxg6 69.Qe4+ Kh6 70.Qh4+ Kg7 71.Qe4 Qg5+ 72.Kf1 Qb5+ 73.Kf2 Qc5+
-6.30/38 50...Qd4 51.g5 fxg5 52.Rxg5 Rd2+ 53.Rg2 h3 54.Kxh3 Kf7 55.Kh2 Nxf4 56.Rxd2 cxd2 57.Qd1 Ke7 58.Kg3 c3 59.Rxf4 Qe3+ 60.Kg4 Qe1 61.Rf1 Qxd1+ 62.Rxd1 c2 63.Rxd2 c1=Q 64.Re2+ Kd6 65.Kf3 Qf1+ 66.Rf2 Qh3+ 67.Ke4 Qc3
-4.84/38 50...h3 51.Rgf2 Qd4 52.Qe1 Rd2 53.Kxh3 Qd3+ 54.Kh2 Kf7 55.Rg2 Qd5 56.Rgf2 Nxf4 57.Qe3 Qg5 58.Rxd2 cxd2 59.Kg3 c3 60.Qa7+ Kg6 61.Rh1 Nh5+ 62.Rxh5 Qxg4+ 63.Kxg4 d1=Q+ 64.Kg3 Qxh5 65.Qa8 Qe5+ 66.Kf3 Qe6 67.Kf2 f5 68.Qa4 Qd6 69.Kf3 Qd5+ 70.Kf4 Qd2+ 71.Kf3 Qd3+ 72.Kf4
-3.79/38 50...Ne7 51.Qb1 Qd7 52.g5 Qh3+ 53.Kg1 Qe3+ 54.Rgf2 h3 55.Kh1 fxg5 56.fxg5 Qe4+ 57.Kh2 Qe5+ 58.Kh1 Qd5+ 59.Kh2 Qe5+ 60.Kh1 Qd5+ 61.Kh2 Qe5+
-3.07/38 50...Qb6 51.a4 Qd4
-2.54/38 50...Qd5 51.Qe1 h3 52.Rgf2 Rd2 53.a4 Qd4 54.a5 Qxf4+ 55.Kh1 Qd4 56.a6 Qd5+ 57.Kg1 Rd3 58.Qb1 h2+ 59.Rxh2 Rg3+ 60.Kf2 Qe5 61.Qb7+ Ne7 62.Qxe7+ Qxe7 63.Kxg3 Qd6+ 64.Kg2 Qxa6 65.Rf2 Qb7+ 66.Kg3 Qe4
-2.54/38 50...Qd7 51.Qb1 Qc7 52.Qc1 Qd8 53.a4 Qd4 54.a5 Rd2
-2.54/38 50...Qd8 51.Qb1 Qc7 52.Qc1 Qd6 53.a4 Qd4 54.Qe1
-1.89/38 50...Qe6 51.a4 Qd5 52.Qe1 Qd6 53.Rgf2 Rd2 54.Kh1 Qd5+ 55.Kg1 Qd4 56.Kh1 Kf7 57.a5 Qd5+ 58.Kh2 Qf3 59.g5 Nxf4 60.gxf6 Qg3+ 61.Kh1 Qh3+ 62.Kg1 Ne2+ 63.Qxe2 Rxe2 64.Rxe2 Qg4+ 65.Rg2 Qd4+ 66.Kh2 Qd6+ 67.Kg1 Qd4+ 68.Kh2 Qd6+ 69.Kg1 Qd4+
-1.57/38 50...Kh7 51.a4 Qc7 52.Rc2 Qd7 53.Qe1 Qxg4 54.Rg2 Qf5 55.Kg1 Kh6 56.a5 h3 57.Rgf2 Nh4 58.Kh1 Qxa5 59.Qe6 Qd5+ 60.Qxd5 Rxd5 61.Rc2 Rd3 62.Rfc1 Nf5 63.Rxc3 Ne3 64.Kh2
-0.92/38 50...Kf8 51.a4 h3 52.Rgf2 Ne7 53.Qe1 Nd5 54.Qe4 Ne3 55.Rg1 c2 56.Re2 c3 57.Ree1 Kg7 58.Rc1 Qd8 59.a5 Rd2+ 60.Kg3 Rd4 61.Qxe3 Rd3 62.Qxd3 Qxd3+ 63.Kh2 f5 64.gxf5+ Kf6 65.Rgf1 Qd2+ 66.Kxh3 Kxf5 67.Kg3 Qe2 68.Rf3 Qd2 69.Rff1 Qe2 70.Rf3 Qd2 71.Rff1
-0.60/38 50...Qc6 51.Qe1 h3 52.Rgf2 Qd5 53.a4 Rd2 54.a5 Qxa5 55.Qe3 Rd3 56.Qe6 Qc7 57.Kh1 Qb7+ 58.Kh2 Qc7 59.Kh1 Qb7+ 60.Kh2 Qc7
-0.48/38 50...Rd4 51.Rgf2 Nxf4 52.Rxf4
-0.32/37 50...Rd2 51.Qxc3 Rd4 52.Rgf2 Nxf4 53.Kh1 Qd5+ 54.Qf3 c3 55.Qxd5 Nxd5 56.Ra1 Rd3 57.a4 Nb4 58.a5 c2 59.Kh2 Ra3 60.Rc1 Rxa5 61.Kh3 Kg6 62.Kxh4 Rc5 63.Rd2 Rc3 64.Rd6 Kg7 65.g5 fxg5+ 66.Kxg5 Rc5+ 67.Kf4 Na2 68.Rxc2
-0.32/37 50...Rd5 51.Rgf2 Rd4 52.Qxc3 Nxf4 53.Kh1 Qd5+ 54.Qf3 c3 55.Qxd5 Nxd5 56.Ra1 Rd3 57.a4 Nb4 58.a5 c2 59.Kh2 Ra3 60.Rc1 Rxa5 61.Kh3 Kg6 62.Kxh4 Rc5 63.Rd2 Rc3 64.Rd6 Kg7 65.g5 fxg5+ 66.Kxg5 Rc5+ 67.Kf4 Na2
-0.28/37 50...Kh6 51.a4 Kg7 52.a5 Rd4 53.Rgf2 c2 54.Kh1 Qd5+ 55.Kg1 Qc5
-0.28/37 50...Kh8 51.a4 Kg7 52.a5 Rd4 53.Rgf2 c2 54.Kh1 Qd5+ 55.Kg1 Qc5
-0.28/37 50...Kg8 51.a4 Kg7 52.a5 Rd4 53.Rgf2 c2 54.Kh1 Qd5+ 55.Kg1 Qc5
-0.20/37 50...Qa6 51.g5 f5 52.Re2 Kg8 53.a4 Qxa4 54.Qb1 Qb3 55.Qa2 Nf8 56.Ra1 Rd4 57.Qa7 Rd2 58.Qe3 Qc2 59.Re1 Qe4 60.Qxe4 fxe4 61.Kh3 e3 62.Rxe3 c2 63.Re4 Rd1 64.Rxc4 Rxe1 65.Rxc2 Rh1+ 66.Rh2 Rg1 67.Rg2 Rf1
-0.20/37 50...Qe7 51.g5 f5 52.a4 h3 53.Rgf2 Nh4 54.Qe1 Qxe1 55.Rxe1 Nf3+ 56.Rxf3 Rxf3 57.Rc1 Rxf4 58.Kxh3 Rf3+ 59.Kh4 Rd3 60.a5 Kg6

Qd4 was leading between depth 28 and 33, but then Kf7 became SF's first choice again.

* Don't try that on a Windows machine: you need a long uptime**.
** Troll inside.

[Adam Hair]

For those interested, here's Stockfish's analysis at depth 38*.

* Don't try that on a Windows machine: you need a long uptime**.
** Troll inside.

kiss my apple