Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Stockfish+NNUEsv +80 Elo vs Stockfish 17jul !!
- - By MrKris (***) Date 2020-07-28 21:29 Upvotes 1
G/90"+1" , Nooman Low Draw Rate v1, all, colors reversed
https://github.com/nodchip/Stockfish/releases
https://www.comp.nus.edu.sg/~sergio-v/nnue/
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=33299

                          Elo +-          Score         Games
1. Sf-NNUE_sv728-06    :  80  23  213.5/348  61.4%  +109 -30 =209
2. Stockfish_20071720  :   0  22  134.5/348  38.6%      draws 60%


Note the very low for CPU engines draw rate - thanks to Nooman & the newest SV NNUE net!
(The below ( )'s are the round #'s in the .pgn.)

Stockfish won only 30 games and in 16 of those Stockfish+NNUE won the colors reversed game.

Stockfish+NNUE wins:
[Event "SfNNUEsv - Stockfish , 90+1 NoomanLDR"]
[Site "2700X 7th x2 | 2Gh Sy-6 50mrOff c=0"]
[Date "2020.07.28"]
[Round "57"]
[White "Stockfish_20071720"]
[Black "Sf-NNUE_sv728-06"]
[Result "1-0"]
[ECO "B42"]
[Opening "Sicilian"]
[Variation "Kan, Polugaevsky Variation, 1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 a6 5.Bd3 Bc5"]
[TimeControl "90+1"]
[Termination "adjudication"]
[PlyCount "72"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} c5 {book} 2. Nf3 {book} e6 {book} 3. d4 {book} cxd4 {book} 4.
Nxd4 {book} a6 {book} 5. Bd3 {book} Bc5 {book} 6. Nb3 {book} Be7 {book} 7.
Qg4 {book} g6 {book} 8. Qe2 {book} d6 {book --Colors reversed was a draw;
the shortest win.--} 9. c4 {+0.71/25 0 12s} Nc6 {-0.36/25 0 13s} 10. O-O
{+0.50/25 0 10s} Ne5 {-0.37/22 0 1.4s} 11. Nc3 {+0.51/24 0 5.0s} Nf6
{-0.51/23 0 2.4s} 12. Bg5 {+0.62/21 0 1.7s} h6 {-0.27/20 0 1.6s} 13. Be3
{+0.80/23 0 3.2s} Nxd3 {-0.22/22 0 1.9s} 14. Qxd3 {+0.76/20 0 1.6s} Ng4
{-0.24/24 0 2.4s} 15. Rad1 {+0.64/22 0 2.0s} Nxe3 {-0.16/23 0 1.6s} 16.
Qxe3 {+0.19/29 0 3.2s} Qc7 {0.0024 3.7s} 17. f4 {+0.52/26 0 1.4s !
--Sacrificing a pawn.--} Qxc4 {+0.22/23 0 2.6s} 18. f5 {+0.81/25 0 1.5s}
Bd7 {0.0023 1.9s} 19. Qg3 {+1.36/24 0 1.6s} Rg8 {-0.37/27 0 13s} 20. f6
{+3.04/25 0 1.6s} Bf8 {-0.28/23 0 1.6s} 21. Rc1 {+3.00/25 0 2.0s} Qb4
{-1.33/26 0 5.5s --Finally realizing that there is a problem.--} 22. Qe3
{+2.54/32 0 2.8s} g5 {-2.71/30 0 14s} 23. Nd4 {+3.51/27 0 1.5s} e5
{-2.91/25 0 3.2s} 24. Nb3 {+4.09/30 0 2.2s} Be6 {-3.98/26 0 5.8s} 25. Nd5
{+4.48/28 0 1.9s} Bxd5 {-3.85/23 0 0.95s} 26. exd5 {+5.14/27 0 2.5s} g4
{-5.21/28 0 11s} 27. Rc7 {+5.34/27 0 1.6s} g3 {-5.38/27 0 6.3s} 28. Rfc1
{+5.63/28 0 1.5s} gxh2+ {-3.76/20 0 1.0s} 29. Kh1 {+6.04/27 0 1.9s --Using
the 2nd rank enemy pawn as K cover.--} Qg4 {-6.92/23 0 3.2s} 30. Qf2
{+6.31/28 0 1.8s} Rg5 {-7.32/25 0 6.6s} 31. Qc2 {+6.47/28 0 1.7s} Bg7
{-9.42/25 0 6.0s} 32. fxg7 {+7.09/28 0 1.7s} Rxg7 {-6.80/18 0 0.15s} 33.
Na5 {+7.35/27 0 1.7s} Kf8 {-9.02/24 0 1.9s} 34. Nxb7 {+8.16/27 0 1.8s} Rg6
{-6.00/17 0 0.15s} 35. Rf1 {+8.41/28 0 1.7s} Kg8 {-11.78/22 0 2.1s} 36.
Rcxf7 {+8.78/28 0 1.7s} Rc8 {-12.15/22 0 1.8s, White wins by adjudication} 1-0

[Event "SfNNUEsv - Stockfish , 90+1 NoomanLDR"]
[Site "2700X 7th x2 | 2Gh Sy-6 50mrOff c=0"]
[Date "2020.07.27"]
[Round "10"]
[White "Sf-NNUE_sv728-06"]
[Black "Stockfish_20071720"]
[Result "1-0"]
[ECO "C41"]
[Opening "Philidor"]
[Variation "3...exd4 4.Nxd4 Nf6 5.Nc3 Be7"]
[TimeControl "90+1"]
[Termination "adjudication"]
[FEN "r5k1/3R1pp1/7p/p2Q4/1p3P2/1Pq3P1/P1P4P/1K1Rr3 b - - 40 47"]

47... Rxd1+ {-1.71/27 0 1.0s} 48. Qxd1 {+3.54/8 0 0s} g6 {-1.74/28 0 1.7s} 49. Qd3 {+3.77/26 0
5.6s} Re8 {-1.82/27 0 3.5s} 50. Qxc3 {+4.05/25 0 1.5s} bxc3 {-2.20/28 0
1.7s} 51. Rd1 {+3.20/27 0 3.7s} a4 {-2.13/28 0 1.9s} 52. a3 {+4.30/26 0
1.3s} axb3 {-2.28/29 0 0.96s} 53. cxb3 {+4.50/25 0 1.1s} Re2 {-2.24/26 0
0.83s} 54. Kc1 {+4.23/26 0 1.8s} Ra2 {-2.82/28 0 2.6s} 55. Rd3 {+5.12/25 0
1.4s} Rxa3 {-2.89/24 0 1.1s} 56. Rxc3 {+5.13/25 0 1.4s} Ra2 {-2.96/27 0
2.9s} 57. Rc2 {+5.37/26 0 1.3s} Ra7 {-2.81/26 0 0.80s} 58. Kb2 {+6.57/25 0
1.9s} Rb7 {-3.42/26 0 0.92s} 59. Rc5 {+6.91/25 0 1.4s} Kf8 {-3.62/27 0
1.4s} 60. Kc3 {+7.36/24 0 1.2s} Ke7 {-3.72/27 0 0.82s} 61. b4 {+7.81/24 0
1.3s} Kd6 {-3.91/24 0 0.95s} 62. Kc4 {+8.31/23 0 0.93s} Ra7 {-3.98/25 0
0.58s} 63. Rd5+ {+8.31/22 0 1.1s} Kc6 {-3.94/26 0 2.3s} 64. Ra5 {+9.41/24 0
1.1s} Re7 {-4.48/27 0 1.8s} 65. Ra6+ {+9.83/22 0 0.93s} Kb7 {-4.52/22 0
0.22s} 66. Rf6 {+10.34/24 0 1.7s} Rc7+ {-4.99/24 0 1.4s} 67. Kd5 {+11.68/21
0 1.1s} Rd7+ {-5.00/22 0 0.77s} 68. Rd6 {+11.78/21 0 0.80s} Re7 {-5.01/22 0
0.73s} 69. f5 {+12.26/22 0 1.2s} gxf5 {-5.13/23 0 1.0s} 70. Rxh6 {+12.64/23
0 0.71s} Kc8 {-7.39/24 0 1.0s} 71. Rf6 {+12.96/25 0 0.74s} Rb7 {-8.86/25 0
2.4s} 72. Kc6 {+13.59/26 0 2.0s} Rc7+ {-7.78/23 0 0.99s} 73. Kd6 {+13.77/24
0 0.73s} Kd8 {-10.18/25 0 2.0s} 74. Rxf5 {+14.01/24 0 0.78s} Rd7+
{-10.53/21 0 0.69s, White wins by adjudication} 1-0


[Event "SfNNUEsv - Stockfish , 90+1 NoomanLDR"]
[Site "2700X 7th x2 | 2Gh Sy-6 50mrOff c=0"]
[Date "2020.07.28"]
[Round "172"]
[White "Sf-NNUE_sv728-06"]
[Black "Stockfish_20071720"]
[Result "1-0"]
[BlackElo "2400"]
[ECO "B15"]
[Opening "Caro-Kann"]
[Variation "Gurgenidze: 4.e5 Bg7 5.f4 h5"]
[WhiteElo "2400"]
[TimeControl "90+1"]
[Termination "normal"]
[FEN "k2r1b1r/1p1qpp2/p1p3pn/P3P3/1PpP1PPN/2P5/5B2/1R1QK1R1 w - - 0 24"]

24. d5 {+1.83/26 0 2.3s --The first of 2 pawn sacrifices for a won ending. Fram a
Slav, the colors reversed was a draw.--} Qxd5 {-1.14/26 0 2.6s} 25. Qxd5
{+2.29/23 0 1.5s} cxd5 {-1.67/29 0 4.1s} 26. b5 {+2.44/27 0 4.9s} Rb8
{-2.33/33 0 24s} 27. g5 {+2.52/26 0 1.1s} Nf5 {-1.99/31 0 1.6s} 28. Nxf5
{+2.93/27 0 1.9s} gxf5 {-2.27/30 0 3.1s} 29. e6 {+2.91/30 0 3.9s !! --The
2nd pawn sacrifice.--} fxe6 {-2.69/27 0 2.7s} 30. Bd4 {+3.16/28 0 3.2s} Rg8
{-2.93/33 0 11s} 31. Kd2 {+3.49/28 0 7.1s} Bg7 {-1.72/21 0 0.44s} 32. Ke3
{+3.78/28 0 3.6s} e5 {-2.69/29 0 6.3s --Returning 1 but it does not
help.--} 33. fxe5 {+4.27/28 0 1.6s} e6 {-2.79/25 0 0.44s} 34. Rh1 {+5.09/25
0 1.4s} Bf8 {-3.08/25 0 0.98s} 35. Rh7 {+5.31/29 0 1.4s} Rg7 {-3.19/29 0
1.5s} 36. Rxg7 {+7.43/31 0 3.1s} Bxg7 {-3.74/27 0 0.65s} 37. Kf4 {+8.24/27
0 1.5s} axb5 {-3.83/27 0 0.85s} 38. Rxb5 {+8.50/27 0 1.4s} b6 {-3.98/24 0
0.38s} 39. axb6 {+9.87/26 0 1.4s} Rb7 {-6.81/28 0 2.1s} 40. Rb1 {+12.16/26
0 1.6s} Kb8 {-6.25/21 0 0.20s} 41. Rh1 {+12.64/28 0 1.6s} Rd7 {-4.97/23 0
1.0s} 42. g6 {+13.37/25 0 1.7s} Kb7 {-11.10/24 0 2.5s} 43. Kg5 {+14.24/25 0
5.8s --White's plan from the start - decisive K invasion.--} f4 {-11.85/23
0 1.6s, White wins by adjudication} (43. .. Re7 {For example} 44. Rh7 Rd7
45. Rxg7 Rxg7 46. Kf6 {winning.}) 1-0

Stockfish wins:
[Event "SfNNUEsv - Stockfish , 90+1 NoomanLDR"]
[Site "2700X 7th x2 | 2Gh Sy-6 50mrOff c=0"]
[Date "2020.07.28"]
[Round "97"]
[White "Stockfish_20071720"]
[Black "Sf-NNUE_sv728-06"]
[Result "1-0"]
[BlackElo "2400"]
[ECO "C11"]
[Opening "French"]
[Variation "Steinitz, 6.Nf3"]
[WhiteElo "2400"]
[TimeControl "90+1"]
[Termination "normal"]
[PlyCount "72"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} e6 {book} 2. d4 {book} d5 {book} 3. Nc3 {book} Nf6 {book} 4.
e5 {book} Nfd7 {book} 5. f4 {book} c5 {book} 6. Nf3 {book} Be7 {book} 7.
Be3 {book} O-O {book} 8. Qd2 {book} b6 {book} 9. h4 {book} Ba6 {book} 10.
O-O-O {book --The 2nd shortest win, colors rev. was a draw.--} c4 {-0.60/24 0 11s} 11. f5 {+0.75/28 0 9.8s} b5 {-0.61/22 0
1.4s} 12. Ne2 {+0.56/29 0 7.0s} b4 {-0.65/24 0 3.4s} 13. Qe1 {+0.76/28 0
11s} Nc6 {-0.42/22 0 4.8s} 14. Qg3 {0.0029 2.8s} Kh8 {0.0022 2.0s} 15. fxe6
{+1.09/32 0 3.5s} fxe6 {0.0024 3.3s} 16. b3 {+1.61/23 0 1.8s} Qa5 {0.0027
1.8s} 17. Kb1 {+1.45/24 0 2.9s} cxb3 {0.00142 2.9s} 18. axb3 {+1.52/28 0
6.5s} Nb6 {0.0029 1.6s} 19. Ng5 {+1.49/25 0 1.5s} Nc4 {0.0032 2.1s} 20. Bc1
{+1.33/27 0 1.9s} Na3+ {0.0026 4.3s} 21. Bxa3 {+1.61/30 0 2.2s} bxa3
{-1.11/27 0 12s} 22. Nc3 {+1.41/29 0 3.5s} Nd8 {0.0027 6.2s ??} (22. .. Bc4
{! ----Black should offer 2 minor pieces to equalize, S'NNUEgek27} 23. Nxe6
(23. Nxh7 {??} Kxh7 24. h5 Kh8 25. h6 g5 {-+}) 23. .. Rf7 24. Ka1 Nb4 {90''
S'NNUEgek27 0.00 =} 25. Rc1 Nd3 26. cxd3 Bxb3 27. Nc5 Bxc5 28. dxc5 d4 29.
Ne4 Rb7 30. Qe1 Qa4 31. Nd6 Rb4 32. Be2 Be6 33. Bd1 Rb2 34. Rc4 Qa6 35. Bg4
Bxg4 36. e6 Re2 37. Qg3 h5 38. Re1 Rxe6 39. Rxe6 Bxe6 40. Qg6 Rf8 41. Qxe6
Rf1+ 42. Ka2 Rf2+ {drawing.----}) 23. Bxa6 {+3.43/28 0 2.3s} Qxa6 {-1.60/26
0 3.5s} 24. Nxh7 {+3.51/28 0 1.7s} Kxh7 {-3.27/27 0 4.9s} 25. h5 {+4.12/29
0 2.5s} Qb7 {-4.12/32 0 20s} 26. Qg6+ {+4.04/28 0 1.4s} Kg8 {-4.21/26 0
2.3s} 27. h6 {+4.12/30 0 1.8s} Bb4 {-4.13/25 0 4.4s} 28. Rdf1 {+4.59/30 0
2.1s} Qe7 {-4.37/28 0 5.1s} 29. Ne2 {+5.16/27 0 1.4s} Rxf1+ {-6.46/25 0
9.5s} 30. Rxf1 {+5.35/29 0 1.6s} a5 {-6.10/24 0 0.83s} 31. Nf4 {+5.59/26 0
1.6s} a2+ {-6.26/19 0 1.2s} 32. Kb2 {+5.94/26 0 1.6s} a4 {-8.16/22 0 2.7s}
33. h7+ {+8.21/24 0 1.8s} Kh8 {-8.23/22 0 0.12s} 34. Qh5 {+8.33/25 0 1.7s}
a1=Q+ {-9.59/28 0 2.0s} 35. Rxa1 {+10.03/25 0 1.7s} a3+ {-10.56/26 0 1.2s}
36. Ka2 {+11.43/25 0 1.7s} Qf7 {-12.51/18 0 0.29s, White wins by
adjudication} 1-0

[Event "SfNNUEsv - Stockfish , 90+1 NoomanLDR"]
[Site "2700X 7th x2 | 2Gh Sy-6 50mrOff c=0"]
[Date "2020.07.27"]
[Round "2"]
[White "Stockfish_20071720"]
[Black "Sf-NNUE_sv728-06"]
[Result "1-0"]
[ECO "E99"]
[Opening "King's Indian"]
[Variation "Mar del Plata, 10.f3 f5 11.Be3 f4 12.Bf2 g5"]
[TimeControl "90+1"]
[Termination "adjudication"]
[FEN "r1b2q1k/2p2nr1/p1Pp4/Pp1Pp1b1/1P2PpPp/2N2P1P/1K2B3/R1N2Q1R w - - 44 70"]

70. Nxb5 {+2.60/24 0 0.29s} axb5 {-2.32/20 0 0.20s} 71. Bxb5
{+2.78/29 0 0.47s} Rg8 {-2.19/21 0 0.41s} 72. Na2 {+2.58/30 0 0.96s} Bd8
{-2.42/27 0 1.7s} 73. Be2 {+3.05/34 0 2.3s} Ng5 {-2.43/20 0 1.3s} 74. Nc3
{+2.75/28 0 1.5s} Qe7 {-3.04/25 0 1.6s} 75. Qf2 {+3.09/25 0 0.55s} Qh7
{-3.25/23 0 1.3s} 76. b5 {+3.29/34 0 0.58s} Rf8 {-3.37/22 0 1.3s} 77. Qg2
{+3.21/26 0 0.32s} Bf6 {-3.62/19 0 0.70s} 78. Rhb1 {+3.22/26 0 0.88s} Bd8
{-3.44/20 0 0.59s} 79. Kc2 {+3.43/25 0 0.59s} Qe7 {-3.84/19 0 1.7s} 80. Ra4
{+3.38/26 0 0.58s} Kg8 {-3.93/18 0 0.33s} 81. Bc4 {+3.38/29 0 0.78s} Qg7
{-3.84/18 0 0.65s} 82. Be2 {+3.38/29 0 0.88s} Re8 {-4.02/20 0 1.8s} 83. b6
{+4.16/27 0 3.0s} cxb6 {-3.67/18 0 0.13s} 84. axb6 {+3.98/24 0 0.59s} Rxa4
{-3.19/17 0 0.28s} 85. Nxa4 {+4.26/20 0 0.39s} Bxb6 {-3.33/20 0 0.32s} 86.
Nxb6 {+5.61/20 0 0.57s} Qa7 {-3.05/19 0 0.32s} 87. Qf2 {+5.97/22 0 0.54s}
Qa2+ {-5.46/21 0 3.7s} 88. Rb2 {+5.94/24 0 0.67s} Qa3 {-5.57/20 0 0.41s}
89. Nc4 {+7.32/21 0 1.9s} Qa4+ {-5.92/20 0 0.78s} 90. Kb1 {+7.94/22 0
0.32s} Rf8 {-5.64/21 0 0.61s} 91. Qxh4 {+8.75/19 0 0.48s} Nxf3 {-4.87/18 0
0.84s} 92. Qh6 {+8.88/19 0 0.63s} Nd4 {-5.78/19 0 0.61s} 93. Qg6+ {+9.82/19
0 0.58s} Kh8 {-6.95/6 0 0s} 94. Qh6+ {+10.29/24 0 1.5s} Kg8 {-7.34/9 0 0s}
95. Qg6+ {+10.33/23 0 0.63s} Kh8 {-7.34/8 0 0s} 96. Qxd6 {+10.54/23 0
0.56s} Nxe2 {-8.86/23 0 3.2s} 97. Qxf8+ {+10.77/22 0 0.92s} Kh7 {-8.86/7 0
0s} 98. Qe7+ {+11.08/22 0 2.7s} Kg8 {-8.92/21 0 0.71s} 99. Rb4 {+11.12/23 0
1.7s} Qd1+ {-7.54/19 0 1.1s} 100. Ka2 {+11.59/20 0 0.65s} Qc2+ {-10.68/23 0
4.5s} 101. Ka3 {+11.71/22 0 0.51s} Nc1 {-9.59/19 0 0.34s} 102. Nb2
{+11.66/22 0 0.54s} Qb1 {-11.25/21 0 2.2s} 103. Qd8+ {+11.89/20 0 0.71s}
Kg7 {-5.95/18 0 0.75s} 104. Ka4 {+12.04/21 0 0.56s} Qa1+ {-11.26/19 0 1.5s}
105. Kb5 {+11.66/9 0 0.001s} Ba6+ {-12.22/18 0 1.1s, White wins by
adjudication} 1-0
Attachment: Snnue-Sf.pgn - Games (947k)
Parent - - By cma6 (****) Date 2020-07-30 20:36
Kris:
These are impressive results for SF NNUE. What was rate of play and how reliable would these results be for correspondence analysis, where 5 minutes/ply would be a typical analysis time?

Also, how do you pick the best NNUE net to use?
Parent - - By MrKris (***) Date 2020-07-31 00:11 Upvotes 1

> What was rate of play


The first thing in my post says: G/90"+1" - this is the old fashioned standard: " = seconds (' = minutes), and in my game headers it says 90|1 - this is CCC's standard seconds-base|seconds-increment.

> how reliable would these results be for correspondence analysis, where 5 minutes/ply would be a typical analysis time?


The longest data I know of is G/25'+8", 1-core, CEGT: http://www.cegt.net/25plus8Rating/Purelist/rangliste.html
But that is so still just:
S'NNUE vs. Houdini 6.0 x64 - 3337 100 game: +42 = 57 - 1 , 70.5% (1 in 100 loss, 42 wins)
It uses the first strong net Gekkehenker 20200627, current nets will be stronger, see below.

> Also, how do you pick the best NNUE net to use?


Definitely the most recent (probably not 'experimental') here:
https://www.comp.nus.edu.sg/~sergio-v/nnue/ The author has a good reputation with Lc0 nets also.
This series began with the Gek-27 at CEGT and is 'reinforcing', or something NN technical, every 5 hours.
Parent - - By cma6 (****) Date 2020-07-31 00:24 Edited 2020-07-31 00:26
Kris, thanks for explanation. At about 26 minutes a 40-move game, this is a lot faster than even a moderately deep correspondence analysis, but it is still a strong indication that SF NNUE may be the #1 engine already, perhaps better than lco?!

Still, looking at Sergio's graph of ELO ratings, I wondering if perhaps the strongest net is not necessarily the latest (non-experimental) but rather the one with highest ELO, e.g., 20200728-2138?
Parent - - By MrKris (***) Date 2020-07-31 04:52 Upvotes 1

> Still, looking at Sergio's graph of ELO ratings, I wondering if perhaps the strongest net is not necessarily the latest (non-experimental) but rather the one with highest ELO, e.g., 20200728-2138?


I did not notice the graph until after I wrote the above, I think it was put up recently.

I found this at https://abrok.eu/stockfish/comment :
30583  sf-tester
2020-07-30 19:56:36  sf-nn_20072905 / Author: mstembera
1k.epd (1000 pos)

20200731-0111.nnue 909 / 90.90 Top / New
20200728-2138.nnue 899 / 89.90 (Best rating in sergios rating table)
20200730-1934.nnue 889 / 88.90 New
20200730-2333.nnue 888 / 88.80 New
stockfish_20071720 879 / 87.90 (last dev, 2x faster, no nnue)
20200729-2214.nnue 872 / 87.20 (the current nnue)
nn.bin/nnue------- 866 / 86.60 (gk from June 27th)
20200720-1017.nnue 858 / 85.80 (sergios first nnue)
-c157e0a5755b.nnue 855 / 85.50

They were discussing the graph recently.
mstembera is a Sf contributor, I think.
Parent - By cma6 (****) Date 2020-07-31 15:53
Kris, this is a useful post. I see that the SF update page is a good place to check for knowledgeable comments.
Parent - By MrKris (***) Date 2020-07-31 06:09 Edited 2020-07-31 06:13
On Discord the author something about the peak one you have ...728-2138 "or any of the next 5" at the time "about the same." And "problems are not always the best indicator of strength."

I do not know why it is still being run, though I do not know anything about NN.

The link to the unoffical Stockfish Discord 'server' (with NNUE section) is at: https://github.com/nodchip/Stockfish/releases
 
Parent - By MrKris (***) Date 2020-08-01 08:28 Upvotes 1
Peak Elo graph vs. recent SV NNUE net
G/90"+1", R7 2700X 7-threads each x 2 concurrently
see details in games header

                      Elo   +  -     games  score  LOS       data
1. Sf-NNUE_sv731-0821   2  14 14  176.5/348  51%  60.5%  +49 =255 -44
2. Sf-NNUE_sv728-2138  -2  14 14  171.5/348  49%  39.4%  draws 73%


After the book white attacked on the kingside and won quickly, reversed white attacked on the queenside and it was quickly drawn.
[Event "90|1 x2 , SVnnue new vs peak"]
[Site "2700X 7th ea. 2Gh , c0 , NoomanLDR"]
[Date "2020.07.31"]
[Round "1"]
[White "Sf-NNUE_sv728-21"]
[Black "Sf-NNUE_sv731-08"]
[Result "1-0"]
[ECO "E87"]
[Opening "King's Indian"]
[Variation "Sämisch, Orthodox, Bronstein, 9.g3"]
[TimeControl "90+1"]
[Termination "adjudication"]
[PlyCount "68"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {book} Nf6 {book} 2. c4 {book} g6 {book} 3. Nc3 {book} Bg7 {book} 4.
e4 {book} d6 {book} 5. f3 {book} O-O {book} 6. Be3 {book} e5 {book} 7. d5
{book} Nh5 {book} 8. Qd2 {book} Qh4+ {book} 9. g3 {book} Nxg3 {book} 10.
Qf2 {book} Nxf1 {book} 11. Qxh4 {book} Nxe3 {book} 12. Ke2 {book} Nxc4
{book} 13. Nh3 {+1.12/25 0 11s} Nxb2 {-0.80/24 0 17s} 14. Ng5 {+1.03/25 0 0.020s}
( 14. Nb5 {sv731-00821 1/2-1/2 sv728-2138} Na6 15. Rhc1 Bxh3 16. Qxh3 Nc5 17. Rxc5 dxc5 18. Nxc7 Rac8 19. Nb5 Rcd8
20. Rb1 Na4 21. Kf1 c4 22. Nxa7 f5 23. Nb5 fxe4 24. Qe6+ Kh8 25. d6 Nc5 26.
Qxc4 b6 27. a4 Rxf3+ 28. Kg1 Rdf8 29. a5 Bh6 30. Nc7 Be3+ 31. Kg2 R8f5 32.
Kh1 Nd3 33. Ne6 Rf1+ 34. Kg2 R1f2+ 35. Kh1 Rf1+ {1/2-1/2} )
h6 {-0.58/23 0 3.1s} 15. Rab1 {+0.91/24 0 0.70s} Nc4 {-0.56/22 0
3.1s} 16. Rbg1 {+1.08/24 0 0.26s} Re8 {-1.17/25 0 24s} 17. Nxf7 {+1.60/22 0
1.7s} Kxf7 {-1.73/23 0 1.2s} 18. Qg3 {+3.08/26 0 1.5s} g5 {-2.23/24 0 2.4s}
19. h4 {+3.23/24 0 1.3s} g4 {-2.82/25 0 3.4s} 20. fxg4 {+3.05/24 0 0.98s}
Kg8 {-3.73/27 0 18s} 21. g5 {+4.06/28 0 0.002s} h5 {-3.71/25 0 1.1s} 22.
Qf3 {+3.78/28 0 5.8s} Rf8 {-3.48/28 0 1.8s} 23. Qxh5 {+4.31/24 0 1.1s} Nd7
{-3.86/23 0 0.001s} 24. Ke1 {+4.50/23 0 3.0s} Nb2 {-4.08/27 0 15s} 25. Rg3
{+4.99/22 0 5.0s} b6 {-5.31/25 0 9.5s} 26. Qe2 {+5.98/22 0 7.2s} Nc5
{-5.69/25 0 0.014s} 27. Qxb2 {+6.01/22 0 2.5s} Rf4 {-6.34/22 0 0.80s} 28.
Qe2 {+6.67/22 0 3.0s} Ba6 {-4.97/19 0 2.5s} 29. Qh5 {+6.89/22 0 1.9s} Nxe4
{-6.45/20 0 0.071s} 30. Rf3 {+6.72/21 0 2.1s} Rxf3 {-7.09/20 0 1.1s} 31.
Qxf3 {+8.55/22 0 2.9s} Nc5 {-8.67/19 0 0.22s} 32. Ne4 {+9.40/21 0 2.6s}
Nd3+ {-9.77/21 0 1.2s} 33. Kd2 {+10.08/22 0 4.4s} Nf4 {-10.60/19 0 0.024s}
34. h5 {+11.22/23 0 3.1s} Bd3 {-11.32/19 0 1.3s, White wins by
adjudication} 1-0

The situation becomes extrodinarily complex.
[Event "90|1 x2 , SVnnue new vs peak"]
[Site "2700X 7th ea. 2Gh , c0 , NoomanLDR"]
[Date "2020.07.31"]
[Round "1"]
[White "Sf-NNUE_sv728-21"]
[Black "Sf-NNUE_sv731-08"]
[Result "0-1"]
[BlackElo "2400"]
[ECO "D36"]
[Opening "QGD"]
[Variation "Exchange, Main Line, 9.Nge2 Re8"]
[WhiteElo "2400"]
[TimeControl "90+1"]
[Termination "normal"]
[PlyCount "71"]
[WhiteType "human"]
[BlackType "human"]

1. d4 {book} d5 {book} 2. c4 {book} e6 {book} 3. Nc3 {book} Nf6 {book} 4.
cxd5 {book} exd5 {book} 5. Bg5 {book} c6 {book} 6. e3 {book} Be7 {book} 7.
Bd3 {book} O-O {book} 8. Qc2 {book} Nbd7 {book} 9. Nge2 {book} Re8 {book}
10. O-O-O {book} b5 {book} 11. h4 {-0.48/24 0 12s} h6 {+0.43/23 0 11s} 12.
Bf4 {-0.43/23 0 9.6s} Qa5 {+0.69/22 0 5.0s} 13. g4 {-0.43/22 0 5.4s} b4
{+0.70/20 0 1.8s} 14. Na4 {-0.58/20 0 0.001s} Ne4 {+0.83/20 0 2.2s} 15. g5
{-0.89/23 0 3.0s} Ba6 {+0.27/22 0 3.9s} (15. .. Nxf2 {Stockfish soon starts
to give 0.00 to every move for both sides} 16. Bh7+ Kh8 17. gxh6 g6 (17. ..
Nxd1 {??} 18. hxg7+ Kxg7 19. Rg1+ {M12, soon the exchange down} Kh8 20. Bh6
Bf6 21. Qf5 Nxe3 22. Bxe3 b3 23. Nac3 bxa2 24. Qh5 {2nd Q with check
24...a1Q+? 25.Bb1 #} Bg7 25. Bf5+ Kg8 26. Rxg7+ Kxg7 {white the double
exchange and 2 pawns down} 27. Bh6+ Kh8 28. Bg5+ Kg7 29. Qh7+ Kf8 30. Qh8#)
18. Bxg6 fxg6 19. Qxg6 Rg8 20. Qf7 Qd8 21. Rdg1 Rxg1+ 22. Rxg1 Qf8 23. Qg6
Ba6 24. h7 Nh3 25. Bh6 Nxg1 {gets a lot for it Q but Stockfish is still
saying 0.00} 26. Bxf8 Nxe2+ 27. Kd2 Rxf8 28. Qxc6 Rf2 29. Qxa6 {Q+N vs.
R+B+N+N} Nxd4+ 30. Kc1 Nf5 31. Qxa7 Nf6 32. Nc5 Rf1+ 33. Kc2 Nxe3+ 34. Kb3
Nf5 35. Ne6 Kxh7 36. Qc7 Re1 37. Qf4 Kg6 38. h5+ Nxh5 39. Qg4+ Kf6 40. Nc7
Re5 41. Qg8 Nd4+ 42. Ka4 Nf4 43. Qh8+ Kf5 44. Qh7+ Kg5 (44. .. Kg4 45. Qg7+
Kf5) 45. Qg8+ Kf5 46. Qh7+ Kg5 47. Qg7+ Ng6 48. Qg8 Kf6 49. Nxd5+ Kf5 50.
Nxb4 Bxb4 {1 min. still 0.00} 51. Qc8+ Ne6 52. Kxb4) 16. gxh6 {-0.26/23 0
0.001s} c5 {+0.46/25 0 12s}
[FEN "r3r1k1/p2nbpp1/b6P/q1pp4/Np1PnB1P/3BP3/PPQ1NP2/2KR3R w - - 0 17"]

17. Rhg1 {-0.14/22 0 2.2s ?? losing instead of drawing} (17. dxc5 {Stockfish -0.10
=+ 2 mins., but soon it starts to say 0.00 for every move for both sides
again} Bxd3 18. Rxd3 Rac8 (18. .. Nxf2 19. Rg1 Nxd3+ 20. Qxd3 Bf6 21. Qf5
Re6 22. Nd4 b3 (22. .. Qxa4 {??} 23. Nxe6) 23. axb3 Nxc5 24. Nxc5 Qxc5+ 25.
Kb1 Ra6 26. Be5 Qe7 27. Bxf6 Rxf6 28. Rxg7+ Kf8 29. Rh7 Qxe3 30. Qxf6 Qe1+
31. Kc2 Rc8+ 32. Kd3 Qd1+ 33. Ke3 Re8+ 34. Kf2 Qe1+ 35. Kg2 Qe4+ 36. Kf2
Qe3+) 19. Rxd5 Ndxc5 20. Nxc5 Qxa2 21. b3 Qa1+ 22. Qb1 Qa3+ 23. Qb2 Rxc5+
24. Rxc5 Bxc5) 17. .. Bxd3 {+1.00/22 0 1.3s} 18. Rxd3 {-0.99/25 0 17s} Rac8
{+1.12/26 0 0.019s} 19. Rxg7+ {-1.25/22 0 5.3s} Kh8 {+2.05/20 0 2.3s} 20.
Kb1 {-1.44/22 0 8.1s} cxd4 {+2.36/26 0 0.031s} 21. Qb3 {-1.51/22 0 2.4s}
Qa6 {+2.69/23 0 1.4s} 22. exd4 {-1.64/25 0 1.1s} Bxh4 {+2.89/22 0 1.5s} 23.
Be3 {-1.56/26 0 14s} Bxf2 {+2.99/23 0 2.3s} 24. Bc1 {-1.71/21 0 2.1s} Bh4
{+3.48/22 0 0.18s} 25. a3 {-1.92/23 0 7.8s} Nf2 {+4.92/21 0 3.9s} 26. Rf3
{-4.67/22 0 5.6s} Qxe2 {+6.31/27 0 6.3s} 27. Rg1 {-6.97/22 0 1.7s} Rg8
{+7.65/20 0 2.4s} 28. Rxg8+ {-8.01/20 0 0.24s} Kxg8 {+7.64/21 0 2.7s} 29.
Bf4 {-7.52/22 0 0.50s} bxa3 {+8.09/22 0 2.8s} 30. Qxa3 {-8.58/23 0 5.7s}
Ne4 {+8.86/23 0 0.79s} 31. Ka2 {-9.46/22 0 2.3s} Bf6 {+9.13/22 0 1.1s} 32.
Qd3 {-9.20/21 0 0.59s} Qxd3 {+9.41/24 0 2.0s} 33. Rxd3 {-9.44/20 0 0.28s}
Rc2 {+9.80/21 0 1.8s} 34. Be3 {-9.41/22 0 2.2s} Nb6 {+10.21/22 0 2.9s} 35.
Kb1 {-10.49/25 0 3.5s} Rg2 {+10.66/24 0 2.4s} 36. Nc5 {-10.78/23 0 0.019s,
Black wins by adjudication} 0-1
Attachment: SVnnue_new-peak.pgn - Games (860k)
Parent - By MrKris (***) Date 2020-08-01 22:05
Said on Discord: The latest,
  20200801-1515.bin   2020-08-01 19:35   20M 
seems strongest yet.
https://www.comp.nus.edu.sg/~sergio-v/nnue/
Parent - - By Sesse (****) Date 2020-08-05 21:55
Forgive me the confusion… is this Stockfish version number really right? Ie., a version from 2007? (It can't, really, since there's no 17th month…)
Parent - By sarona Date 2020-08-06 00:15 Edited 2020-08-06 01:04 Upvotes 1
It's one of the July 17,2020 Stockfish builds on https://abrok.eu/stockfish/.  Either Vizvezdenec's (bench 4578298) or mstembera's (
bench 4541608).

Stockfish 20071720 = Stockfish 20(year) 07(month) 17(day) 20(hour).  I believe the last two digits signify the hour released.
Up Topic The Rybka Lounge / Computer Chess / Stockfish+NNUEsv +80 Elo vs Stockfish 17jul !!

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill