Mayhem NNUE - New NN engine

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Sylwy
Posts: 4467
Joined: Fri Apr 21, 2006 4:19 pm
Location: IASI - the historical capital of MOLDOVA
Full name: SilvianR

Re: Mayhem NNUE - New NN engine

Post by Sylwy »

A Finnish match has just begun. The old Hakkapeliitta TCEC v2 seems to be in good shape........... :wink:

Image
User avatar
Sylwy
Posts: 4467
Joined: Fri Apr 21, 2006 4:19 pm
Location: IASI - the historical capital of MOLDOVA
Full name: SilvianR

Re: Mayhem NNUE - New NN engine

Post by Sylwy »

WOW! Mayhem 0.50 NNUE won a game. :lol: It could still be a serious match.

[pgn]
[Event "NN Test51"]
[Site "ISR 3"]
[Date "2020.11.12"]
[Round "3"]
[White "Mayhem-0.50-x86-windows-modern-64bit"]
[Black "Hakkapeliitta TCEC v2 64-bit"]
[Result "1-0"]
[BlackElo "2200"]
[ECO "D27"]
[Opening "QGA"]
[Time "22:16:20"]
[Variation "Classical, Rubinstein, 8.Nc3"]
[WhiteElo "2200"]
[TimeControl "240+2"]
[Termination "adjudication"]
[PlyCount "158"]
[WhiteType "program"]
[BlackType "program"]

1. d4 d5 2. c4 e6 3. Nc3 Nf6 4. Nf3 dxc4 5. e3 a6 6. a4 c5 7. Bxc4 Nc6 8.
O-O Qb6 9. Qe2 {(Qd1-e2) +1.30/13 10} Bd6 {(Bf8-d6 d4xc5 Qb6xc5 b2-b3
Qc5-a5 Bc1-b2 O-O Rf1-d1 Rf8-d8 Ra1-c1 Qa5-c7 h2-h3 Nc6-a5 Nc3-a2 Qc7-e7
Bb2-e5 Na5xc4 Rc1xc4 b7-b5 Be5xd6 Rd8xd6 Rc4-d4) -0.64/22 12} 10. dxc5
{(d4xc5) +1.00/12 10} Qxc5 {(Qb6xc5 b2-b3) -0.64/24 11} 11. Rd1 {(Rf1-d1)
+1.23/12 10} Ne5 {(Nc6-e5 b2-b3 Ne5xc4 b3xc4 Qc5-c7 e3-e4 Nf6-d7 Qe2-d2
Bd6-e7 Bc1-a3 Be7xa3 Ra1xa3 O-O Qd2-d4 f7-f6 a4-a5 h7-h6 h2-h4 Nd7-c5 e4-e5
Bc8-d7 e5xf6 Rf8xf6 Ra3-a2 Ra8-f8 Nc3-e4 Nc5xe4 Qd4xe4) -0.55/24 12} 12. b3
{(b2-b3) +1.73/12 9} Nxc4 {(Ne5xc4 b3xc4 Qc5-c7 e3-e4 Nf6-d7 Qe2-d2 Bd6-e7
Bc1-a3 Be7xa3 Ra1xa3 O-O Qd2-d4 f7-f6 a4-a5 h7-h6 h2-h4 h6-h5 Qd4-d6 Qc7xd6
Rd1xd6 Nd7-c5 e4-e5 Bc8-d7 Nc3-e4 Nc5xe4 Rd6xd7) -0.60/25 37} 13. bxc4
{(b3xc4) +2.24/12 9} Qc7 {(Qc5-c7 e3-e4 Bd6-b4 Bc1-d2 Bb4-e7 e4-e5 Nf6-d7
Nc3-e4 O-O Bd2-f4 Nd7-c5 Ne4-d6 Bc8-d7 Qe2-c2 h7-h6 a4-a5 Ra8-d8 Bf4-g3
Bd7-c6 Nf3-d4 f7-f5 Nd4xc6 Qc7xc6 f2-f3 Be7-g5) -0.61/24 11} 14. Rxd6
{(Rd1xd6) +4.32/12 9} Qxd6 {(Qc7xd6 Bc1-a3 Qd6-c7 e3-e4 e6-e5 Qe2-b2 Bc8-d7
Nc3-d5 Nf6xd5 c4xd5 Ra8-c8 Qb2-b4 Bd7-g4 d5-d6 a6-a5 Qb4-b5+ Qc7-d7 Qb5xe5+
Ke8-f8 Nf3-d4 Rc8-e8 Qe5xa5 Re8xe4 Qa5-a8+ Re4-e8 Qa8xb7 h7-h5 Qb7xd7
Bg4xd7 a4-a5 Rh8-h6) -0.81/23 9} 15. Ba3 {(Bc1-a3) +5.18/13 8} Qc7 {(Qd6-c7
e3-e4) -0.65/23 8} 16. e4 {(e3-e4) +5.29/13 8} e5 {(e6-e5 Qe2-b2 Bc8-d7
Nc3-d5 Nf6xd5 c4xd5 f7-f6 Ra1-c1 Qc7-a5 Qb2xb7 Qa5-d8 Nf3-d2 Qd8-b8 Nd2-c4
Qb8xb7 Nc4-d6+ Ke8-f8 Nd6xb7+ Kf8-g8 Rc1-c7 Bd7xa4 d5-d6 Ba4-b5 d6-d7
Bb5xd7 Rc7xd7 h7-h6 Nb7-d6 Ra8-b8 f2-f3 Rb8-b3 Ba3-c5 Rb3-d3 h2-h3 Kg8-h7
Bc5-b6) -1.03/24 12} 17. Rc1 {(Ra1-c1) +5.06/12 8} Bg4 {(Bc8-g4 h2-h3
Bg4xf3 Qe2xf3 Qc7-c6 Nc3-d5 Nf6-d7 Qf3-g3 g7-g6 Ba3-b2 Ke8-f8 Bb2xe5 Nd7xe5
Qg3xe5 Rh8-g8 Nd5-f6 Kf8-g7 Nf6xg8+ Kg7xg8 Qe5-d5 Ra8-e8 Qd5xc6 b7xc6
Rc1-e1 Kg8-g7 e4-e5 a6-a5) -0.79/22 10} 18. h3 {(h2-h3) +5.16/13 8} Bxf3
{(Bg4xf3 Qe2xf3 Qc7-c6 Nc3-d5 Nf6-d7 Nd5-e7 Qc6-f6 Ne7-f5 g7-g6 Nf5-d6+
Ke8-f8 Qf3-g4 Ra8-d8 Nd6xb7+ Kf8-g7 Nb7xd8 Rh8xd8 c4-c5 Qf6-c6 a4-a5 Rd8-c8
Qg4-d1 Nd7-f6 f2-f3 Qc6-c7 Qd1-d6 Qc7xa5 c5-c6 Qa5-b6+ Kg1-h2) -0.95/24 11}
19. Qxf3 {(Qe2xf3) +5.18/13 8} Qc6 {(Qc7-c6 Nc3-d5) -0.79/24 7} 20. Nd5
{(Nc3-d5) +6.91/12 7} Nd7 {(Nf6-d7 Rc1-d1 g7-g6 a4-a5 f7-f6 Nd5-b6 Nd7xb6
Rd1-d6 Qc6xd6 Ba3xd6 Nb6-d7 Qf3-g4 O-O-O Bd6-e7 Rd8-e8 Qg4-e6 Rh8-g8 Qe6-f7
h7-h6 g2-g3 Kc8-c7 Qf7-e6 Kc7-c8 Qe6-d6 f6-f5 e4xf5 g6xf5 Kg1-g2) -1.46/23
7} 21. Rb1 {(Rc1-b1) +7.20/11 7} f6 {(f7-f6 a4-a5 g7-g6 Nd5-b6 Ra8-d8
Nb6xd7 Ke8xd7 Rb1-b6 Qc6-a4 Ba3-e7 Kd7-c8 Be7xd8 Rh8xd8 Qf3xf6 Rd8-d1+
Kg1-h2 Qa4xa5 Rb6-e6 Kc8-b8 Re6-e7 Qa5-e1 Qf6xe5+ Kb8-a7) -1.17/20 7} 22.
Ne7 {(Nd5-e7) +6.94/13 7} Qxa4 {(Qc6xa4 Rb1-b4 Qa4-a5 Rb4xb7 Ra8-b8 Rb7-a7
Rb8-b1+ Kg1-h2 Qa5-d2 Ne7-c6 Qd2-f4+ Qf3xf4 e5xf4 Ra7-a8+ Nd7-b8 Nc6xb8
Ke8-f7 Ra8-a7+ Kf7-g6 Nb8-c6 Rb1-f1 Nc6-e7+ Kg6-h5 Ba3-c5 Rf1-c1 Ne7-f5
Rc1xc4 Nf5xg7+ Kh5-h4 Ng7-f5+ Kh4-g5) -2.31/24 16} 23. Rxb7 {(Rb1xb7)
+5.76/12 7} Rb8 {(Ra8-b8 Rb7-c7 Rb8-b1+ Kg1-h2 Qa4-d1 Rc7-c8+ Ke8-f7 Qf3xd1
Rb1xd1 Rc8xh8 Rd1-d3 Rh8-d8 Nd7-c5 Rd8xd3 Nc5xd3 Kh2-g3 Kf7-e6 Ne7-d5 a6-a5
h3-h4 h7-h5 f2-f3 Nd3-e1 Ba3-f8 Ke6-f7 Bf8-d6 Ne1-c2 Nd5-c3 Nc2-e3 c4-c5
Ne3-c4 Nc3-d5 a5-a4 Nd5-b6 Nc4-b2) -2.84/23 6} 24. Rc7 {(Rb7-c7) +6.45/12
7} Rb1+ {(Rb8-b1+ Kg1-h2 Qa4-d1 Rc7-c8+ Ke8-f7 Qf3xd1 Rb1xd1 Rc8xh8 Rd1-d3
Rh8-d8 Nd7-c5 Rd8xd3 Nc5xd3 Kh2-g3 Kf7-e6 Ne7-d5 a6-a5 h3-h4 Nd3-e1 Ba3-f8
Ke6-f7 Bf8-e7 Ne1-c2 h4-h5 h7-h6 f2-f4 e5xf4+ Kg3xf4 Kf7-e6 Be7-c5 a5-a4
Bc5-b6) -2.96/25 7} 25. Kh2 {(Kg1-h2) +7.13/11 6} Qd1 {(Qa4-d1 Rc7-c8+
Ke8-f7 Qf3xd1 Rb1xd1 Rc8xh8 Rd1-d3 Rh8-d8 Nd7-c5 Rd8xd3 Nc5xd3 Kh2-g3
Kf7-e6 Ne7-d5 a6-a5 h3-h4 Nd3-e1 Ba3-f8 Ke6-f7 Bf8-d6 Ne1-c2 c4-c5 Nc2-d4
Nd5-c3 g7-g6 Bd6-c7 Nd4-c6 Kg3-f3 Nc6-d4+ Kf3-e3 Nd4-c2+ Ke3-d2 Nc2-d4
Nc3-d5 a5-a4 Bc7-d8 f6-f5 e4xf5 g6xf5) -3.21/28 5} 26. Rc8+ {(Rc7-c8+)
+8.26/13 6} Kf7 {(Ke8-f7 Qf3xd1 Rb1xd1 Rc8xh8 Rd1-d3 Rh8-d8 Nd7-c5 Rd8xd3
Nc5xd3 Kh2-g3 Kf7-e6) -3.21/7 0} 27. Rxh8 {(Rc8xh8) +10.17/12 6} Qxf3
{(Qd1xf3 g2xf3 Rb1-b3 Ba3-d6 Rb3xf3 Kh2-g2 Rf3-c3 Rh8-d8 Rc3xc4 Rd8xd7
Kf7-e6 Rd7-d8 Rc4xe4 Ne7-c6 Re4-c4 Nc6-a5 Rc4-a4 Na5-b3 Ke6-f5 Rd8-d7 g7-g5
Rd7xh7 Ra4-c4 Rh7-a7 Rc4-c3 Nb3-c5) -3.52/24 7} 28. gxf3 {(g2xf3) +13.63/14
6} Rb3 {(Rb1-b3 Ba3-d6 Rb3xf3 Kh2-g2 Rf3-c3 Rh8-d8 Rc3xc4 Rd8xd7 Kf7-e6
Rd7-d8 Rc4xe4 Ne7-c6 Re4-c4 Nc6-a5 Rc4-c2 Bd6-b4 e5-e4 Rd8-d6+ Ke6-f7
Rd6-c6 Rc2xc6 Na5xc6 Kf7-e6 Nc6-b8 Ke6-d5 Nb8xa6 f6-f5 Bb4-f8 g7-g6 Na6-b4+
Kd5-d4 Kg2-g3 g6-g5 Bf8-g7+ Kd4-c4) -3.71/25 6} 29. Bd6 {(Ba3-d6) +14.24/14
6} Rxf3 {(Rb3xf3 Kh2-g2 Rf3-c3 Rh8-d8 Rc3xc4 Rd8xd7 Kf7-e6 Rd7-d8 Rc4xe4
Ne7-c6 Re4-c4 Nc6-a5 Rc4-c2 Bd6-b4 e5-e4 Rd8-d6+ Ke6-f7 Rd6-c6 Rc2xc6
Na5xc6 Kf7-e6 Nc6-b8 f6-f5 Nb8xa6 f5-f4 Na6-c7+ Ke6-f5 Bb4-c3 f4-f3+ Kg2-g3
g7-g5 Nc7-d5 h7-h5 h3-h4 g5-g4 Bc3-b4) -3.85/26 5} 30. Kg2 {(Kh2-g2)
+15.69/14 6} Rc3 {(Rf3-c3 Rh8-d8 Rc3xc4 Rd8xd7 Kf7-e6 Rd7-d8 Rc4xe4 Ne7-c6
Re4-c4 Nc6-a5 Rc4-c2 Bd6-b4 e5-e4 Rd8-d6+ Ke6-f7 Rd6-c6 Rc2xc6 Na5xc6 g7-g5
Nc6-b8 a6-a5 Bb4xa5 f6-f5 Nb8-c6 f5-f4 Nc6-d4 Kf7-g6 Ba5-c7 Kg6-f6 h3-h4
h7-h6 h4xg5+ h6xg5 Nd4-c6 f4-f3+ Kg2-g3) -3.88/27 5} 31. Rd8 {(Rh8-d8)
+17.59/14 6} Rxc4 {(Rc3xc4 Rd8xd7 Kf7-e6 Rd7-d8 Rc4xe4 Ne7-c6 Re4-c4 Nc6-a5
Rc4-c2 Bd6-b4 e5-e4 Rd8-d6+ Ke6-f7 Rd6-c6 Rc2xc6 Na5xc6 g7-g5 Nc6-b8 a6-a5
Bb4xa5 f6-f5 Nb8-c6 f5-f4 Nc6-d4 Kf7-g6 Ba5-c7 Kg6-f6 h3-h4 h7-h6 h4xg5+
h6xg5 Nd4-c6 f4-f3+ Kg2-g3) -3.88/27 5} 32. Rxd7 {(Rd8xd7) +16.16/14 5} Ke6
{(Kf7-e6 Rd7-d8 Rc4xe4 Ne7-c6 Re4-c4 Nc6-a5 Rc4-c2) -3.88/26 5} 33. Ra7
{(Rd7-a7) +19.14/15 5} Kxd6 {(Ke6xd6 Ne7-f5+ Kd6-c5 Ra7xa6 g7-g6 Nf5-g3
Rc4-d4 Ra6xf6 Rd4-d7 Rf6-e6 Kc5-d4 h3-h4 Rd7-c7 Re6-d6+ Kd4-c3 Rd6-d5
Rc7-e7 h4-h5 g6xh5 Ng3xh5 Re7-e6 Nh5-g3 Re6-e8 Kg2-f3 Kc3-b3 Ng3-f5 Re8-e6
Rd5-c5 h7-h5 Kf3-e3) -3.87/24 6} 34. Nf5+ {(Ne7-f5+) +17.41/14 5} Kc5
{(Kd6-c5 Ra7xa6 g7-g6 Nf5-g3 Rc4-d4 Ra6xf6 Rd4-d7 Rf6-e6 Kc5-d4 h3-h4
Rd7-d8 Re6-e7 h7-h5 Re7-e6 Rd8-g8 Kg2-f3 Rg8-f8+ Kf3-e2 Rf8-g8 Re6-a6 g6-g5
h4xg5 Rg8xg5 Ra6-d6+ Kd4-c5 Rd6-h6 Rg5-g4 Rh6xh5 Kc5-d4 Rh5-h8) -3.95/25 4}
35. Rxa6 {(Ra7xa6) +18.39/14 5} g6 {(g7-g6 Nf5-g3 Rc4-d4 Ra6xf6 Rd4-d7
Rf6-e6 Kc5-d4 h3-h4 Rd7-d8 Re6-a6 Kd4-c3 Ra6-a7 h7-h5 Ra7-e7 Kc3-d4 Re7-e6
Rd8-g8 Kg2-f3 Rg8-f8+ Kf3-e2 Rf8-g8 Re6-a6 g6-g5 Ng3-f5+ Kd4xe4 Nf5-e7
Rg8-d8 h4xg5 Rd8-d7 Ra6-a4+ Rd7-d4) -3.88/22 6} 36. Ng3 {(Nf5-g3) +20.50/14
5} Rd4 {(Rc4-d4 Ra6xf6 Rd4-d8 Rf6-a6 Kc5-d4 h3-h4 Rd8-d7 Ra6-e6 Rd7-a7
Re6-d6+ Kd4-c4 Rd6-d5 Ra7-e7 Ng3-e2 Re7-f7 Rd5xe5 Kc4-d3 Ne2-g3 Kd3-d2)
-4.16/25 5} 37. Rxf6 {(Ra6xf6) +22.01/14 5} Rd8 {(Rd4-d8 Rf6-a6 Kc5-d4
h3-h4 Rd8-d7 Ra6-e6 Rd7-a7 Re6-d6+ Kd4-c3 h4-h5 g6xh5 Ng3xh5 Kc3-c2 Rd6-d5
Ra7-e7 Nh5-f6 Re7-g7+ Kg2-f3 Rg7-f7 Rd5-d6) -4.19/29 4} 38. Ra6 {(Rf6-a6)
+21.66/13 5} Kd4 {(Kc5-d4 h3-h4 Rd8-d7 Ra6-e6 Rd7-d8 Re6-e7 h7-h5 Re7-e6
Rd8-g8 Kg2-f3 Rg8-f8+ Kf3-e2 Rf8-g8 Re6-d6+ Kd4-c4 Ke2-e3 g6-g5 h4xg5
Rg8xg5 Rd6-h6 h5-h4 Rh6xh4 Rg5-g6 Rh4-h5 Rg6-e6 Rh5-h7 Re6-a6 Rh7-c7+
Kc4-b4 Ng3-f5 Ra6-a3+ Ke3-e2) -4.47/27 13} 39. Ra5 {(Ra6-a5) +23.63/13 5}
Rd6 {(Rd8-d6 h3-h4 Rd6-e6 h4-h5 g6xh5 Ra5-d5+ Kd4-c3 Ng3xh5 Kc3-b2 Kg2-f3
Kb2-c3 Nh5-g7 Re6-e7 Ng7-f5 Re7-e6 Rd5-d6 Re6-e8 Rd6-d7 h7-h5 Rd7-h7 Kc3-d2
Rh7xh5 Re8-f8 Rh5-h4 Rf8-f7 Rh4-h6 Rf7-d7) -4.50/24 3} 40. Ra3 {(Ra5-a3)
+23.07/13 5} Rd7 {(Rd6-d7 h3-h4 Kd4-c5 Ra3-a5+ Kc5-d4 Ra5-b5 h7-h5 Rb5-b6
Rd7-g7 Rb6-d6+ Kd4-c4 Rd6-d5 Rg7-e7) -4.56/25 3} 41. Ra4+ {(Ra3-a4+)
+23.46/13 4} Kc3 {(Kd4-c3 Ra4-a5 Kc3-d4 h3-h4 Rd7-e7 Ra5-d5+ Kd4-c4 h4-h5
g6xh5 Ng3xh5 h7-h6 Kg2-f3 Kc4-b3 Nh5-g3 Re7-e6 Ng3-f5 h6-h5 Kf3-e3 Re6-e8
Rd5-b5+ Kb3-c3 Rb5-d5) -4.58/24 3} 42. Ra5 {(Ra4-a5) +22.62/13 4} Re7
{(Rd7-e7 Ra5-d5 Re7-e6 h3-h4 Re6-e7 h4-h5 g6xh5 Ng3xh5 Kc3-c2 Nh5-f6
Re7-g7+ Kg2-h3 Rg7-f7 Rd5-d6 h7-h6 Kh3-g4 Rf7-g7+ Kg4-f5 Rg7-g5+ Kf5-e6
h6-h5 Rd6-d5 Rg5-g2 Nf6xh5 Kc2-b3 Ke6xe5 Rg2xf2 Nh5-f6) -4.61/26 3} 43. h4
{(h3-h4) +23.71/13 4} Kd2 {(Kc3-d2 h4-h5 g6xh5 Ng3xh5 Kd2-d3 Nh5-g3 Kd3-c3
Ng3-f5 Re7-e8 Kg2-f3 Kc3-d2 Ra5-d5+ Kd2-c3 Kf3-e3 Re8-e6 Nf5-h4 h7-h5
Nh4-f3 Re6-h6 Nf3xe5 h5-h4 Ne5-f7 Rh6-e6 Nf7-g5 Re6-b6 Rd5-c5+ Kc3-b3)
-4.65/24 3} 44. Nf1+ {(Ng3-f1+) +24.47/13 4} Kd3 {(Kd2-d3 f2-f3 Kd3-e2
Nf1-h2 h7-h5 Ra5-a6 Re7-g7 Ra6-e6 Ke2-e3 Re6xe5 Ke3-f4 Re5-e8 g6-g5 h4xg5
Rg7xg5+ Kg2-f2 Rg5-a5 Re8-f8+ Kf4-g5 f3-f4+ Kg5-h4 Nh2-f3+ Kh4-h3 Nf3-g5+
Kh3-g4) -3.96/22 3} 45. Ng3 {(Nf1-g3) +23.32/13 4} Kd2 {(Kd3-d2 h4-h5 g6xh5
Ng3xh5 Kd2-d3 Nh5-g3 Kd3-c3 Ng3-f5 Re7-e8 Kg2-f3 Kc3-d2 Ra5-d5+ Kd2-c3
Kf3-e3 Re8-e6 Nf5-h4 h7-h5 Nh4-f3) 0.00/61 4} 46. Ra2+ {(Ra5-a2+) +22.40/12
4} Ke1 {(Kd2-e1 h4-h5 g6xh5 Ng3xh5 Re7-f7 Ra2-a1+ Ke1-d2 Ra1-a5 Kd2-e1)
-4.44/25 3} 47. Ra1+ {(Ra2-a1+) +23.12/13 4} Kd2 {(Ke1-d2 h4-h5 g6xh5
Ng3xh5 Kd2-d3 Nh5-f6 Re7-c7 Ra1-a3+ Kd3-e2) -3.19/7 0} 48. Ra5 {(Ra1-a5)
+23.14/13 4} Ke1 {(Kd2-e1 h4-h5 g6xh5 Ng3xh5 Re7-f7 f2-f3 Rf7-e7 Kg2-g3
Re7-e6 Kg3-g4 Ke1-f2 Ra5-a2+ Kf2-g1 Kg4-f5 Re6-b6 Nh5-f6 Rb6-b3 Nf6xh7
Rb3xf3+ Kf5xe5 Rf3-f1 Nh7-f6 Rf1-f3 Ra2-a7 Kg1-f2 Ra7-g7 Kf2-e2 Rg7-d7
Ke2-e3) -4.56/26 9} 49. h5 {(h4-h5) +23.22/14 4} gxh5 {(g6xh5 Ng3xh5 Re7-f7
f2-f3 Rf7-e7 Kg2-g3 Ke1-e2 f3-f4 e5xf4+ Kg3xf4 Re7-f7+ Ra5-f5 Rf7-c7 Nh5-f6
Rc7-c2 Nf6xh7 Rc2-a2 Rf5-d5 Ke2-f2 Nh7-f6 Kf2-g2 Kf4-f5 Kg2-g3 e4-e5
Ra2-f2+ Kf5-e6 Rf2-e2 Rd5-d7 Kg3-f3) -4.67/26 2} 50. Ra1+ {(Ra5-a1+)
+22.41/14 4} Kd2 {(Ke1-d2 Ng3xh5 Kd2-d3 Nh5-f6 Re7-f7 Ra1-a3+ Kd3-e2
Ra3-a2+ Ke2-d3) -3.55/7 0} 51. Nxh5 {(Ng3xh5) +22.13/13 4} Rf7 {(Re7-f7
Ra1-a5 Kd2-e1 f2-f3 Rf7-e7 Kg2-g3 Re7-e6 Kg3-g4 Ke1-f2 Ra5-a2+ Kf2-e3
Ra2-a3+ Ke3-f2 Kg4-f5 Re6-e7 Nh5-f6 Kf2-g2 Ra3-c3 Re7-b7 Kf5xe5 h7-h5
Ke5-f4 h5-h4 Rc3-c2+ Kg2-h3 e4-e5 Rb7-b3 Rc2-c7) -4.63/27 3} 52. Ng3
{(Nh5-g3) +21.29/13 3} Rf8 {(Rf7-f8 Ra1-a5 Rf8-e8 Kg2-f3 Re8-f8+ Ng3-f5
Rf8-e8 Ra5-d5+ Kd2-c3 Nf5-e3 Re8-f8+ Kf3-e2 Rf8-e8 Rd5-d3+ Kc3-b4 Ne3-d5+
Kb4-c5 Nd5-f6 Re8-e6 Rd3-d5+ Kc5-c4 Nf6-g4 h7-h5 Ng4xe5+ Kc4-c3) -4.69/24
2} 53. Ra3 {(Ra1-a3) +21.13/13 3} Ke1 {(Kd2-e1 Ng3-f5 Rf8-c8 Ra3-a1+ Ke1-d2
Kg2-f3 h7-h5 Ra1-h1 Rc8-h8 Kf3-g2 Kd2-d3 Nf5-g3 h5-h4 Ng3-f5 Rh8-f8 Rh1xh4
Rf8-b8 Kg2-f3 Kd3-d2) -4.67/28 2} 54. Nf5 {(Ng3-f5) +20.77/13 3} Rc8
{(Rf8-c8 Ra3-a1+ Ke1-d2 Kg2-f3 h7-h5 Ra1-h1 Rc8-h8) -4.69/29 2} 55. Ra1+
{(Ra3-a1+) +22.29/12 3} Kd2 {(Ke1-d2 Kg2-f3 h7-h5 Ra1-h1 Rc8-h8 Kf3-g2
Kd2-d3 Nf5-g3 h5-h4 Ng3-f5 Rh8-f8 Rh1xh4 Rf8-b8 Kg2-f3 Kd3-d2 Rh4-h7
Kd2-e1) -4.69/28 2} 56. Ra5 {(Ra1-a5) +21.29/12 3} Ke1 {(Kd2-e1 Kg2-f3
Rc8-b8 Ra5-a7 h7-h5 Ra7-a1+ Ke1-d2 Ra1-h1 h5-h4 Rh1xh4 Rb8-f8 Rh4-h1 Rf8-a8
Rh1-h6 Ra8-b8) -4.75/26 2} 57. Kg3 {(Kg2-g3) +20.08/12 3} Ke2 {(Ke1-e2
Ra5-a2+ Ke2-e1 Kg3-f3 h7-h5 Ra2-a1+ Ke1-d2 Ra1-h1 Rc8-h8 Kf3-g2 Kd2-d3
Nf5-g3 h5-h4 Ng3-f5 Rh8-f8 Rh1xh4 Rf8-b8 Kg2-f3 Kd3-d2 Rh4-h7) -4.75/27 3}
58. Ra2+ {(Ra5-a2+) +20.24/12 3} Ke1 {(Ke2-e1 Kg3-f3 h7-h5 Ra2-a1+ Ke1-d2
Ra1-h1 Rc8-h8 Kf3-g2 Kd2-d3 Nf5-g3 h5-h4 Ng3-f5 Rh8-f8 Rh1xh4 Rf8-b8 Kg2-f3
Kd3-d2 Rh4-h7) -4.75/27 2} 59. Ra3 {(Ra2-a3) +20.66/12 3} h5 {(h7-h5 Kg3-f3
Rc8-h8 Kf3-e3 Ke1-d1 Ra3-d3+ Kd1-c2 Nf5-h4 Rh8-e8 Rd3-d5 Re8-e6 Nh4-f3
Re6-h6 Rd5xe5 h5-h4 Nf3-d4+ Kc2-c3 Nd4-f5 Rh6-h8 Re5-c5+ Kc3-b4 Rc5-d5
h4-h3) -4.65/23 2} 60. Kf3 {(Kg3-f3) +20.65/12 3} Rb8 {(Rc8-b8 Ra3-a1+
Ke1-d2 Ra1-h1 Rb8-h8 Kf3-g2 Kd2-d3 Nf5-g3 h5-h4 Ng3-f5 Rh8-f8 Rh1xh4 Rf8-b8
Kg2-f3 Kd3-d2) -4.75/26 2} 61. Ra1+ {(Ra3-a1+) +22.60/12 3} Kd2 {(Ke1-d2
Ra1-a5 Rb8-e8 Ra5-a6 Re8-b8 Ra6-h6 Rb8-b7 Rh6xh5 Rb7-f7 Rh5-h6 Rf7-a7
Rh6-d6+ Kd2-e1 Nf5-h4 Ra7-a3+ Kf3-g2 Ke1-e2 Nh4-g6 Ra3-a4 f2-f3 Ra4-a5
Rd6-d5 Ra5-a8 Rd5xe5 Ra8-a3 Ng6-f4+ Ke2-e3 Kg2-g3 Ke3-d4 Re5-d5+ Kd4-e3)
-5.33/27 2} 62. Ra2+ {(Ra1-a2+) +22.20/12 3} Ke1 {(Kd2-e1 Ra2-e2+ Ke1-d1
Nf5-e3+ Kd1-c1 Re2-c2+ Kc1-b1 Rc2-c6 Rb8-f8+ Ne3-f5 h5-h4 Rc6-h6 Kb1-c2
Rh6xh4 Kc2-d2 Rh4-h6 Rf8-a8 Rh6-d6+ Kd2-e1 Nf5-h4 Ra8-a3+ Kf3-g2 Ke1-e2
Nh4-g6 Ra3-a4 Rd6-d5 Ra4-c4 f2-f3 Rc4-c1 Ng6xe5 Rc1-e1 Rd5-d6 Re1-b1 Kg2-g3
Ke2-e3 Kg3-g4 Rb1-h1) -5.53/27 4} 63. Re2+ {(Ra2-e2+) +21.12/14 3} Kd1
{(Ke1-d1 Nf5-e3+ Kd1-c1 Re2-c2+ Kc1-b1 Rc2-c6 Rb8-f8+ Ne3-f5 h5-h4 Rc6-h6
Kb1-c2 Rh6xh4 Kc2-d2 Rh4-h6 Rf8-a8 Rh6-d6+ Kd2-e1 Nf5-h4 Ra8-a3+ Kf3-g2
Ke1-e2 Nh4-g6 Ra3-a4 Rd6-d5 Ra4-a7 Rd5xe5 Ke2-d3 Re5-e6 Ra7-a5 Kg2-f3
Kd3-d2 Kf3-f4 Kd2-e2 f2-f3 Ra5-a3 Ng6-e5 Ra3-b3) -5.73/29 2} 64. Ne3+
{(Nf5-e3+) +22.01/14 3} Kc1 {(Kd1-c1 Re2-c2+ Kc1-b1 Rc2-c6 Rb8-f8+ Ne3-f5
h5-h4 Rc6-h6 Kb1-c2 Rh6xh4 Kc2-d2 Rh4-h6 Rf8-a8 Rh6-d6+ Kd2-e1 Nf5-h4
Ra8-a3+ Kf3-g2 Ke1-e2 Nh4-g6 Ra3-a4 Rd6-d5 Ra4-a7 Rd5xe5 Ke2-e1 Re5-d5
Ra7-a2 Kg2-g3 Ra2-a6 Ng6-e5 Ke1-e2 f2-f3) -5.73/7 0} 65. Nc4 {(Ne3-c4)
+22.13/13 3} Rb3+ {(Rb8-b3+ Re2-e3 Rb3-b8 Re3-e1+ Kc1-c2 Re1-h1 Rb8-h8
Nc4xe5 Rh8-h6 Kf3-f4 Rh6-f6+ Kf4-e3 Rf6-h6 Ne5-f3 Kc2-c3 Nf3-d4 Rh6-h8
Nd4-f5 Kc3-c4 Ke3-f4 Kc4-d3 e4-e5 Kd3-e2 f2-f3 Rh8-c8 e5-e6 Rc8-c4+ Kf4-g3
Rc4-c3 Rh1-h2+ Ke2-d3 Rh2xh5) -5.78/22 2} 66. Kg2 {(Kf3-g2) +21.70/15 3}
Rc3 {(Rb3-c3 Re2-e1+ Kc1-c2 Nc4xe5 Rc3-c8 Re1-h1 Rc8-g8+ Kg2-f3 Rg8-f8+
Kf3-e3 Rf8-h8 Ke3-f3 Rh8-h6 Kf3-f4 Rh6-f6+ Kf4-e3 Rf6-h6 Ne5-f3 Kc2-c3
Nf3-d4 Rh6-h8 Nd4-f5 Kc3-c4 Ke3-f4 Kc4-d3 e4-e5 Kd3-e2 f2-f3 Rh8-c8 e5-e6
Rc8-c4+ Kf4-g3 Rc4-c3 Rh1-h2+ Ke2-d3 Rh2xh5) -5.78/22 2} 67. Nxe5 {(Nc4xe5)
+26.13/13 3} Kd1 {(Kc1-d1 Re2-b2 h5-h4 Rb2-b1+ Kd1-d2 Rb1-h1 Rc3-c5 Ne5-f3+
Kd2-d3 Rh1xh4 Rc5-c7 e4-e5 Rc7-g7+ Kg2-f1 Rg7-e7 Rh4-d4+ Kd3-c3 Rd4-d6
Re7-a7 e5-e6 Ra7-a1+ Kf1-e2 Ra1-a2+ Ke2-e3) -5.82/23 3} 68. Ra2 {(Re2-a2)
+24.94/13 3} h4 {(h5-h4 Ra2-a1+ Kd1-e2 Ra1-h1 Rc3-c5 Rh1-e1+ Ke2-d2 Ne5-f3+
Kd2-d3 Nf3xh4 Rc5-g5+ Kg2-f3 Rg5-h5 Nh4-f5 Kd3-c3 Re1-c1+ Kc3-d2 Rc1-c5
Rh5-h3+ Nf5-g3 Rh3-h8 Rc5-d5+) -5.85/23 2} 69. Ra1+ {(Ra2-a1+) +24.31/12 3}
Kd2 {(Kd1-d2 Ra1-h1 Rc3-c5 Ne5-f3+ Kd2-d3 Rh1xh4 Rc5-c7 e4-e5 Rc7-g7+
Kg2-f1 Rg7-g6 Rh4-d4+ Kd3-c2 Kf1-e2 Kc2-b3 Nf3-d2+ Kb3-b2 Nd2-e4 Kb2-b3
Ke2-f3 Rg6-g1 e5-e6 Rg1-e1 Rd4-d3+ Kb3-c2) -6.15/23 2} 70. Nf3+ {(Ne5-f3+)
+23.29/12 3} Kc2 {(Kd2-c2 Nf3xh4 Rc3-c5 Ra1-e1 Rc5-h5 Nh4-f5 Rh5-g5+ Kg2-f3
Kc2-c3 Re1-e3+ Kc3-d2 Kf3-f4 Rg5-g2 f2-f3 Rg2-g6 e4-e5 Rg6-e6 Re3-e4 Kd2-c3
Nf5-d6 Kc3-d2 Kf4-f5 Re6-h6 e5-e6 Rh6-h5+ Kf5-f6) -6.89/23 2} 71. Ng5
{(Nf3-g5) +24.53/12 3} Rc5 {(Rc3-c5 f2-f4 Rc5-c3 Ra1-a6 Rc3-g3+ Kg2-f2
Rg3-g4 Kf2-e3 Rg4-g1 Ra6-h6 Rg1-h1 Ke3-d4 Kc2-b3 f4-f5 Rh1-d1+ Kd4-e5
Rd1-g1 Ng5-f3 Rg1-a1 Rh6xh4 Ra1-a5+ Ke5-f6 Ra5-a6+ Kf6-f7 Ra6-a7+ Kf7-g6)
-6.15/20 2} 72. Nh3 {(Ng5-h3) +24.47/12 3} Kd3 {(Kc2-d3 Ra1-a4 Kd3-c2
Nh3-f4 Kc2-b1 Kg2-f3 Kb1-c2 Kf3-g4 Kc2-b3 Ra4-b4+ Kb3-a3 Rb4-b6 Rc5-c2)
-5.65/20 2} 73. Re1 {(Ra1-e1) +22.77/12 2} Re5 {(Rc5-e5 Nh3-f4+ Kd3-c4
Kg2-f3 Kc4-c3 Kf3-g4 h4-h3 Nf4xh3 Kc3-d4 Nh3-f4 Kd4-c4 Nf4-d5 Re5-e6 Kg4-f5
Re6-h6 Re1-c1+ Kc4-d4 Rc1-d1+ Kd4-c4) -6.32/22 2} 74. f4 {(f2-f4) +23.76/11
2} Rh5 {(Re5-h5 Nh3-g5 h4-h3+ Ng5xh3 Kd3-d2 Re1-h1 Rh5-a5 Nh3-g5 Kd2-e3
Rh1-h4 Ra5-a2+ Kg2-g3 Ra2-a1 Kg3-g4 Ra1-g1+ Kg4-f5 Rg1-b1 e4-e5 Rb1-b5
Rh4-h3+ Ke3-d4 Kf5-g4) -6.69/20 2} 75. Kf3 {(Kg2-f3) +27.42/13 2} Ra5
{(Rh5-a5 Re1-h1 Ra5-a3 Nh3-g5 Kd3-d2+ Kf3-g4 Ra3-a4 e4-e5 Kd2-e3 Rh1-h3+
Ke3-d4 Rh3xh4 Kd4-e3 Rh4-h3+ Ke3-d4) -7.43/22 2} 76. e5 {(e4-e5) +25.74/12
2} Kd2 {(Kd3-d2 Re1-e2+ Kd2-d1 e5-e6 Ra5-a3+ Re2-e3 Ra3-a8 e6-e7 Ra8-e8
Re3-e2 Kd1-c1) -8.20/22 3} 77. Re3 {(Re1-e3) +28.19/13 2} Ra7 {(Ra5-a7
e5-e6 Ra7-e7 Nh3-f2 h4-h3 Nf2xh3 Kd2-c2 Nh3-g5 Kc2-b2 Kf3-g4 Kb2-a2 Re3-h3
Re7-e8 Rh3-h7 Re8-g8 e6-e7 Ka2-a3 Kg4-f5 Rg8-a8 Kf5-f6 Ka3-b3 Kf6-f7 Ra8-a7
Kf7-f8 Ra7-a8+ e7-e8Q Ra8xe8+ Kf8xe8) -9.90/24 2} 78. e6 {(e5-e6) +28.07/13
2} Re7 {(Ra7-e7 Nh3-f2) -10.41/25 2} 79. f5 {(f4-f5) +28.15/13 2} Rg7
{(Re7-g7 Re3-e2+ Kd2-d3 e6-e7 Rg7-g3+ Kf3-f2 Rg3xh3 e7-e8Q Rh3-h2+ Kf2-g1
Rh2-h1+ Kg1xh1 Kd3-c3 Re2-h2 h4-h3 Rh2xh3+) -317.51/18 2 Arena
Adjudication} 1-0
[/pgn]
JohnWoe
Posts: 491
Joined: Sat Mar 02, 2013 11:31 pm

Re: Mayhem NNUE - New NN engine

Post by JohnWoe »

Sylwy wrote: Wed Nov 11, 2020 8:59 pm
JohnWoe wrote: Tue Nov 10, 2020 9:38 pm --bench should return at least 1,000,000 NPS. Then your cpu is strong enough.
Good enough ? :lol: For a test, of course.

Image
Yes! I recommend at least 1,000 Knps. On some older AMD CPUs Mayhem doesn't perform.

Nice to see others testing Mayhem too. I'm curious of the level. Since in my tests Mayhem performs better than Crafty 25.6 which is 2950 Elo.

With Sapeli I was struggling against 1900 Elo FairyMax. Now I'm in the big boys league. 60% score against 2953 Elo Crafty 25.6. :lol:
More strange is Mayhem is 1200 sloc. Sapeli is 1700 lines. I took 500 sloc out and gained 1000 Elo.

To be honest credits for the 1000 Elo boost go to the SF NNUE evaluation. Without it Mayhem is probably 1800 Elo. :D

Mayhem regularly manhandles Crafty 25.6 in Blitz games:

Code: Select all

Finished game 10 (Crafty-25.6 vs Mayhem 1.0): 0-1 {Black wins by adjudication}
Score of Mayhem 1.0 vs Crafty-25.6: 6 - 3 - 1  [0.650] 10
Elo difference: 107.54 +/- 268.38
Finished match
[pgn][Event "?"]
[Site "?"]
[Date "2020.11.13"]
[Round "7"]
[White "Mayhem 1.0"]
[Black "Crafty-25.6"]
[Result "1-0"]
[TimeControl "60/40"]
[ECO "A60"]
[Opening "Benoni defense"]
[PlyCount "207"]
[Annotator "1. +0,23 1... +0,14"]

1. d4 {+0,23/11 1,3}
{s}
1... Nf6 {+0,14/16 4}
{s}
2. c4 {+0,29/11 1,3}
{s}
2... c5 {-0,23/15 10}
{s}
3. d5 {+0,72/10 1,2}
{s}
3... e6 {-0,09/16 1,0}
{s}
4. Nc3 {+0,84/11 1,2}
{s}
4... exd5 {-0,15/15 4}
{s}
5. cxd5 {+0,78/11 1,2}
{s}
5... Bd6 {+0,16/14 6}
{s}
6. e4 {+1,09/10 1,1}
{s}
6... Qe7 {+0,40/15 1,3}
{s}
7. Bd3 {+1,49/10 1,1}
{s}
7... O-O {+0,52/16 9}
{s}
8. Nf3 {+1,52/10 1,1}
{s}
8... Ng4 {+0,51/16 1,0}
{s}
9. Nb5 {+2,10/9 1,0}
{s}
9... b6 {+0,41/16 7}
{s}
10. Nxd6 {+2,35/10 10}
{s}
10... Qxd6 {+0,54/17 5}
{s}
11. h3 {+2,43/10 9}
{s}
11... Ne5 {+0,48/18 4}
{s}
12. Nxe5 {+1,98/9 9}
{s}
12... Qxe5 {+0,48/16 4}
{s}
13. Qe2 {+2,10/10 9}
{s}
13... a5 {+0,28/17 7}
{s}
14. f4 {+2,29/10 9}
{s}
14... Qe7 {+0,44/18 1,5}
{s}
15. O-O {+1,69/10 8}
{s}
15... Re8 {+0,49/17 1,5}
{s}
16. e5 {+2,56/9 8}
{s}
16... Ba6 {+0,57/15 8}
{s}
17. a3 {+1,80/9 8}
{s}
17... d6 {+0,30/16 10}
{s}
18. Re1 {+1,16/10 7}
{s}
18... Qh4 {-0,05/17 5}
{s}
19. e6 {+1,58/9 7}
{s}
19... fxe6 {+0,02/18 5}
{s}
20. Bxa6 {+1,70/10 7}
{s}
20... Nxa6 {+0,00/18 3}
{s}
21. dxe6 {+1,05/10 7}
{s}
21... Nc7 {+0,00/18 4}
{s}
22. f5 {+1,37/9 6}
{s}
22... g6 {+0,00/17 4}
{s}
23. Bd2 {+0,96/9 6}
{s}
23... gxf5 {-0,30/14 6}
{s}
24. e7 {+1,16/9 6}
{s}
24... Ra7 {-0,17/14 1,4}
{s}
25. Rad1 {+2,04/8 6}
{s}
25... Qd4+ {+0,39/14 9}
{s}
26. Kh2 {+2,68/9 6}
{s}
26... d5 {+0,02/15 4}
{s}
27. Bg5 {+3,00/9 5}
{s}
27... Qg7 {+0,68/16 9}
{s}
28. Rd3 {+3,98/8 5}
{s}
28... h6 {+1,86/15 6}
{s}
29. Bf4 {+4,21/9 5}
{s}
29... Kh7 {+4,68/17 5}
{s}
30. Rg3 {+3,96/8 5}
{s}
30... Qf7 {+5,09/20 5}
{s}
31. Rb3 {+3,60/8 5}
{s}
31... b5 {+1,17/16 1,1}
{s}
32. Rg3 {+3,34/8 5}
{s}
32... d4 {+0,93/16 5}
{s}
33. Bd6 {+2,87/9 5}
{s}
33... Nd5 {+1,27/17 8}
{s}
34. Bxc5 {+2,81/9 5}
{s}
34... Raxe7 {+0,29/18 5}
{s}
35. Bxe7 {+1,82/10 5}
{s}
35... Rxe7 {+0,32/19 3}
{s}
36. Qd2 {+1,83/10 5}
{s}
36... Rxe1 {+0,28/19 4}
{s}
37. Qxe1 {+1,67/11 5}
{s}
37... Ne3 {+0,28/17 3}
{s}
38. Kg1 {+1,48/9 5}
{s}
38... Qa2 {+0,00/16 4}
{s}
39. Qc1 {+1,20/11 5}
{s}
39... Qc4 {+0,28/17 9}
{s}
40. Qb1 {+1,14/11 5}
{s}
40... Qe2 {+0,25/17 9}
{s}
41. a4 {+0,76/11 5}
{s}
41... bxa4 {-1,63/16 5}
{s}
42. Qc1 {+0,52/11 5}
{s}
42... Qd1+ {-1,36/22 6}
{s}
43. Qxd1 {+0,26/14 5}
{s}
43... Nxd1 {-0,01/23 5}
{s}
44. Rd3 {+0,39/15 5}
{s}
44... Nxb2 {+0,00/24 4}
{s}
45. Rxd4 {-0,14/14 5}
{s}
45... Kg6 {+0,71/24 3}
{s}
46. Rd5 {+2,47/13 5}
{s}
46... h5 {+1,13/20 3}
{s}
47. h4 {+3,76/13 5}
{s}
47... Nc4 {+0,89/20 5}
{s}
48. Rc5 {+3,89/13 5}
{s}
48... Nb2 {+2,20/21 1,9}
{s}
49. Rxa5 {+5,09/12 5}
{s}
49... Kf6 {+2,06/20 2,0}
{s}
50. Kf2 {+6,17/12 5}
{s}
50... Ke6 {+2,17/20 4}
{s}
51. Kf3 {+6,57/12 5}
{s}
51... Kf6 {+2,17/19 2,0}
{s}
52. Ra7 {+6,05/12 5}
{s}
52... Ke5 {+0,69/16 4}
{s}
53. Ra6 {+7,24/13 5}
{s}
53... Kd4 {+1,17/18 4}
{s}
54. Kf4 {+7,32/13 5}
{s}
54... Kc4 {+2,25/17 7}
{s}
55. Kg5 {+5,27/14 5}
{s}
55... Kb5 {+0,00/15 4}
{s}
56. Ra8 {+7,14/14 5}
{s}
56... Nc4 {+2,10/16 6}
{s}
57. Kxh5 {+6,00/13 5}
{s}
57... Nb6 {+2,18/15 2,8}
{s}
58. Ra7 {+10,18/13 5}
{s}
58... Nc8 {+2,32/15 1,7}
{s}
59. Rc7 {+9,90/13 5}
{s}
59... Nd6 {+2,32/14 3}
{s}
60. Kg6 {+9,93/12 5}
{s}
60... Nc4 {+2,32/12 4}
{s}
61. Rb7+ {+11,76/13 1,3}
{s}
61... Nb6 {+5,96/17 1,8}
{s}
62. h5 {+10,56/14 1,3}
{s}
62... a3 {+5,96/16 5}
{s}
63. Ra7 {+10,56/13 1,2}
{s}
63... Na4 {+6,59/18 7}
{s}
64. Re7 {+10,36/14 1,2}
{s}
64... a2 {+6,92/18 6}
{s}
65. Re1 {+11,45/13 1,2}
{s}
65... Nc5 {+8,60/18 8}
{s}
66. Ra1 {+13,89/13 1,1}
{s}
66... Nd7 {+9,41/18 2,1}
{s}
67. Rxa2 {+14,10/11 1,1}
{s}
67... Kb6 {+12,55/19 3}
{s}
68. Rb2+ {+14,30/10 1,1}
{s}
68... Kc7 {+12,22/18 9}
{s}
69. Rc2+ {+14,10/10 1,0}
{s}
69... Kd6 {+13,23/20 7}
{s}
70. Kxf5 {+14,30/11 10}
{s}
70... Nf8 {+15,91/17 1,4}
{s}
71. Rd2+ {+13,99/11 9}
{s}
71... Ke7 {+21,62/19 1,4}
{s}
72. Rb2 {+14,17/11 9}
{s}
72... Kd6 {+21,78/23 9}
{s}
73. Rb6+ {+13,80/11 9}
{s}
73... Kc5 {+22,08/24 2,2}
{s}
74. Rb1 {+14,35/11 9}
{s}
74... Kc6 {+15,36/20 7}
{s}
75. Re1 {+14,32/11 8}
{s}
75... Nd7 {+17,95/17 2,5}
{s}
76. Rc1+ {+14,23/10 8}
{s}
76... Kb7 {+16,38/17 1,0}
{s}
77. Rb1+ {+14,35/10 8}
{s}
77... Kc7 {+16,68/17 1,4}
{s}
78. g3 {+14,23/11 7}
{s}
78... Nc5 {+21,82/20 2,1}
{s}
79. Rc1 {+14,18/11 7}
{s}
79... Kd6 {+18,36/16 9}
{s}
80. Rd1+ {+13,97/11 7}
{s}
80... Ke7 {+18,49/17 9}
{s}
81. Rb1 {+14,12/11 7}
{s}
81... Na4 {+18,97/17 1,7}
{s}
82. Re1+ {+14,00/11 6}
{s}
82... Kd7 {+21,98/15 6}
{s}
83. Rd1+ {+14,14/10 6}
{s}
83... Ke8 {+21,98/16 4}
{s}
84. h6 {+14,55/11 6}
{s}
84... Ke7 {+327,49/16 4}
{s}
85. h7 {+14,54/12 6}
{s}
85... Nc3 {+327,51/16 1,7}
{s}
86. h8=Q {+14,66/11 6}
{s}
86... Nxd1 {+17,95/12 2,3}
{s}
87. Kg6 {+14,62/10 5}
{s}
87... Ke6 {+16,23/15 5}
{s}
88. Qd4 {+14,50/10 5}
{s}
88... Nf2 {+21,98/16 3}
{s}
89. Qc4+ {+14,53/9 5}
{s}
89... Kd6 {+21,92/15 2,6}
{s}
90. Qb4+ {+14,47/9 5}
{s}
90... Kc6 {+21,92/15 3}
{s}
91. Qc3+ {+14,47/9 5}
{s}
91... Kb5 {+21,98/17 1,9}
{s}
92. Qb3+ {+14,32/10 5}
{s}
92... Kc5 {+22,08/17 3}
{s}
93. Qe3+ {+14,36/10 5}
{s}
93... Kc4 {+327,47/18 2,9}
{s}
94. Qxf2 {+14,19/10 5}
{s}
94... Kc3 {+327,49/17 1,7}
{s}
95. Kf5 {+6,18/10 5}
{s}
95... Kd3 {+327,51/15 2,0}
{s}
96. Qc5 {+6,31/10 5}
{s}
96... Ke2 {+327,51/14 2,8}
{s}
97. g4 {+10,07/10 5}
{s}
97... Kd3 {+327,57/19 2,0}
{s}
98. g5 {+10,30/10 5}
{s}
98... Ke2 {+327,57/17 1,9}
{s}
99. Qc2+ {+104,85/9 2,8}
{s}
99... Ke3 {+327,57/17 1,8}
{s}
100. Qd1 {+104,85/7 2,0}
{s}
100... Kf2 {+327,59/10 0,5}
{s}
101. Kf4 {+104,85/6 0,4}
{s}
101... Kg2 {+327,61/10 0,2}
{s}
102. Qe2+ {+104,85/5 0,1}
{s}
102... Kg1 {+327,63/124 3}
{s}
103. Kg3 {+104,85/4}
{s}
103... Kh1 {+327,65/10}
{s}
104. Qf1#
{+104.85/3 0s, White mates} 1-0

[/pgn]
User avatar
Sylwy
Posts: 4467
Joined: Fri Apr 21, 2006 4:19 pm
Location: IASI - the historical capital of MOLDOVA
Full name: SilvianR

Re: Mayhem NNUE - New NN engine

Post by Sylwy »

The situation in my test after 60 games (TC=4'+2"):

Image

Hakkapeliitta TCEC v2 x64 has 3020 Elo points (CCRL Blitz).
Mayhem 0.50 x64 (SF)-NNUE seems to have between 2700-2800 Elo points. Let's see the final result.
JohnWoe
Posts: 491
Joined: Sat Mar 02, 2013 11:31 pm

Re: Mayhem NNUE - New NN engine

Post by JohnWoe »

Sylwy wrote: Fri Nov 13, 2020 8:37 pm The situation in my test after 60 games (TC=4'+2"):

Image

Hakkapeliitta TCEC v2 x64 has 3020 Elo points (CCRL Blitz).
Mayhem 0.50 x64 (SF)-NNUE seems to have between 2700-2800 Elo points. Let's see the final result.
Thanks!
3000 was too optimistic. My poor Mayhem :( I was hoping for ~50% score.
Mayhem usually improves in longer matches(against traditional TT engines). So hoping for strong finish.
Go Mayhem Go!
Daniel Anulliero
Posts: 759
Joined: Fri Jan 04, 2013 4:55 pm
Location: Nice

Re: Mayhem NNUE - New NN engine

Post by Daniel Anulliero »

JohnWoe wrote: Sat Nov 14, 2020 12:14 am
Sylwy wrote: Fri Nov 13, 2020 8:37 pm The situation in my test after 60 games (TC=4'+2"):

Image

Hakkapeliitta TCEC v2 x64 has 3020 Elo points (CCRL Blitz).
Mayhem 0.50 x64 (SF)-NNUE seems to have between 2700-2800 Elo points. Let's see the final result.
Thanks!
3000 was too optimistic. My poor Mayhem :( I was hoping for ~50% score.
Mayhem usually improves in longer matches(against traditional TT engines). So hoping for strong finish.
Go Mayhem Go!
So, YOUR engine Sapeli was around 1900 elo, then you've got 900 elo by adding sf NNUE eval in it right ?
So I'll add sf NNUE eval in my Isa engine , then I'll get . . a new sf , say , version 8 (around 3200 elo) ?
The eval is the heart of an engine, you said "working on the eval is boring" , can't understand this ...
Mayhem is strong but its not your Own , sorry ...
Chess computer is on the sad way now ...
Isa download :
JohnWoe
Posts: 491
Joined: Sat Mar 02, 2013 11:31 pm

Re: Mayhem NNUE - New NN engine

Post by JohnWoe »

Mayhem 1.1 released!
Release: https://github.com/SamuraiDangyo/mayhem ... s/tag/v1.1

Crafty spams my folder with log files. So I only test against Fruit now on. Unfortunately 3000 Elo was too optimistic. Proved also by Mr. Sylwy.
Mayhem 1.1 is +70 Elo over Fruit 2.1 in blitz. At least.

I used a book to make it fair for Fruit.
Highly scientific testing process. :lol:

Code: Select all

Score of Mayhem 1.1 vs Fruit 2.1: 10 - 5 - 6  [0.619] 21
Elo difference: 84.34 +/- 135.12
Finished matc
A nice game where Mayhem outplayed Fruit in equal endgame.

[pgn][Event "?"]
[Site "?"]
[Date "2020.11.18"]
[Round "11"]
[White "Mayhem 1.1"]
[Black "Fruit 2.1"]
[Result "1-0"]
[TimeControl "10+0.5"]
[ECO "C13"]
[Opening "French"]
[PlyCount "126"]
[Termination "adjudication"]
[Variation "Classical, Tartakower Variation"]
[Annotator "5. +0,39 5... -0,08"]

1. e4
{book}
1... e6
{book}
2. d4
{book}
2... d5
{book}
3. Nc3
{book}
3... Nf6
{book}
4. Bg5
{book}
4... Be7
{book}
5. e5 {+0,39/9 8}
{s}
5... Ne4 {-0,08/10 8}
{s}
6. Nxe4 {+0,69/10 8}
{s}
6... dxe4 {-0,25/10 8}
{s}
7. Bxe7 {+0,64/10 8}
{s}
7... Qxe7 {-0,06/9 5}
{s}
8. c3 {+0,88/9 8}
{s}
8... O-O {+0,00/10 6}
{s}
9. Qc2 {+0,88/10 8}
{s}
9... f5 {-0,17/10 7}
{s}
10. exf6 {+0,68/10 8}
{s}
10... Qxf6 {-0,16/10 7}
{s}
11. f3 {+0,75/9 8}
{s}
11... e3 {+0,20/9 5}
{s}
12. Bd3 {+1,02/9 7}
{s}
12... Kh8 {-0,09/10 2,3}
{s}
13. Ne2 {+1,20/9 7}
{s}
13... Qh4+ {-0,15/10 7}
{s}
14. g3 {+1,15/10 7}
{s}
14... Qh6 {-0,24/10 7}
{s}
15. O-O {+1,27/9 7}
{s}
15... Nd7 {-0,27/9 4}
{s}
16. Rae1 {+1,63/10 7}
{s}
16... Nb6 {-0,06/9 5}
{s}
17. f4 {+1,87/9 7}
{s}
17... Bd7 {-0,34/9 7}
{s}
18. Qc1 {+1,71/10 7}
{s}
18... Nd5 {-0,20/9 5}
{s}
19. c4 {+1,71/8 7}
{s}
19... Nf6 {+0,00/10 7}
{s}
20. Nc3 {+0,80/9 7}
{s}
20... Ng4 {+0,00/9 4}
{s}
21. Re2 {+0,54/8 7}
{s}
21... Rad8 {+0,24/8 5}
{s}
22. Nd1 {+0,93/9 7}
{s}
22... e5 {+0,03/9 9}
{s}
23. dxe5 {+0,02/9 7}
{s}
23... Nxe5 {+0,00/8 4}
{s}
24. Rxe3 {+0,51/9 7}
{s}
24... Ng4 {+0,16/9 7}
{s}
25. Re2 {+0,02/8 7}
{s}
25... Bc6 {+0,04/8 6}
{s}
26. Qc2 {-0,09/8 7}
{s}
26... Qd6 {+0,00/7 5}
{s}
27. Be4 {-0,09/9 6}
{s}
27... Qc5+ {-0,05/8 6}
{s}
28. Kg2 {-0,14/9 6}
{s}
28... Rd4 {+0,18/7 6}
{s}
29. Bxc6 {-0,05/8 6}
{s}
29... Qxc6+ {+0,36/7 4}
{s}
30. Kh3 {-0,26/8 6}
{s}
30... Rxc4 {+0,60/7 6}
{s}
31. Qd3 {-0,16/9 6}
{s}
31... Nf6 {+0,46/7 4}
{s}
32. Nc3 {-0,09/8 6}
{s}
32... Rc5 {+0,16/8 5}
{s}
33. Qf3 {+0,11/9 6}
{s}
33... Nd5 {+0,08/9 5}
{s}
34. Nxd5 {+0,51/9 6}
{s}
34... Rxd5 {+0,06/8 5}
{s}
35. Re7 {+0,51/8 6}
{s}
35... Qh6+ {-0,15/8 6}
{s}
36. Kg2 {+0,37/9 6}
{s}
36... Qd6 {-0,23/9 8}
{s}
37. Rfe1 {+0,27/9 6}
{s}
37... Rd2+ {-0,17/9 7}
{s}
38. R1e2 {+0,09/10 6}
{s}
38... Qxe7 {-0,19/10 8}
{s}
39. Rxd2 {+0,42/9 6}
{s}
39... c5 {-0,23/10 6}
{s}
40. Qe2 {+0,35/9 6}
{s}
40... Qf7 {-0,28/10 7}
{s}
41. b3 {+0,27/9 6}
{s}
41... Re8 {-0,34/9 3}
{s}
42. Qc4 {+0,23/9 6}
{s}
42... Qxc4 {-0,23/9 6}
{s}
43. bxc4 {+1,00/10 6}
{s}
43... Kg8 {-0,22/12 4}
{s}
44. Kf3 {+1,01/11 6}
{s}
44... Kf7 {-0,26/11 4}
{s}
45. Rd7+ {+2,51/12 6}
{s}
45... Re7 {-0,22/13 4}
{s}
46. Rxe7+ {+2,71/14 6}
{s}
46... Kxe7 {-0,12/11 2,5}
{s}
47. g4 {+2,08/13 6}
{s}
47... Kd6 {-0,17/12 7}
{s}
48. a4 {+2,07/15 6}
{s}
48... h5 {-0,23/12 5}
{s}
49. gxh5 {+4,37/13 6}
{s}
49... Ke6 {-0,69/14 6}
{s}
50. Ke4 {+4,52/15 6}
{s}
50... Kf6 {-0,84/14 4}
{s}
51. a5 {+4,85/16 6}
{s}
51... Ke6 {-1,39/14 4}
{s}
52. f5+ {+4,93/15 6}
{s}
52... Kf6 {-1,46/16 6}
{s}
53. Kf4 {+4,95/16 6}
{s}
53... Ke7 {-2,41/17 1,8}
{s}
54. Ke5 {+4,97/16 6}
{s}
54... Kf7 {-1,76/14 4}
{s}
55. Kd6 {+7,84/16 6}
{s}
55... Kf6 {-3,94/16 8}
{s}
56. Kc7 {+7,30/17 6}
{s}
56... b5 {-10,17/16 5}
{s}
57. axb6 {+7,27/16 6}
{s}
57... axb6 {-10,53/16 6}
{s}
58. Kxb6 {+7,15/16 5}
{s}
58... Kxf5 {-10,48/17 4}
{s}
59. Kxc5 {+6,97/16 5}
{s}
59... Kg5 {-10,58/16 7}
{s}
60. Kc6 {+7,15/16 5}
{s}
60... Kxh5 {-9,83/15 4}
{s}
61. c5 {+4,88/16 5}
{s}
61... Kg4 {-10,19/15 5}
{s}
62. Kd5 {+5,31/15 5}
{s}
62... Kh3 {-10,58/13 3}
{s}
63. c6 {+5,47/15 5}
{s}
63... Kxh2
{-10.61/12 0.36s, White wins by adjudication} 1-0

[/pgn]
Daniel Anulliero wrote: Sat Nov 14, 2020 7:13 am
JohnWoe wrote: Sat Nov 14, 2020 12:14 am
Sylwy wrote: Fri Nov 13, 2020 8:37 pm The situation in my test after 60 games (TC=4'+2"):

Image

Hakkapeliitta TCEC v2 x64 has 3020 Elo points (CCRL Blitz).
Mayhem 0.50 x64 (SF)-NNUE seems to have between 2700-2800 Elo points. Let's see the final result.
Thanks!
3000 was too optimistic. My poor Mayhem :( I was hoping for ~50% score.
Mayhem usually improves in longer matches(against traditional TT engines). So hoping for strong finish.
Go Mayhem Go!
So, YOUR engine Sapeli was around 1900 elo, then you've got 900 elo by adding sf NNUE eval in it right ?
So I'll add sf NNUE eval in my Isa engine , then I'll get . . a new sf , say , version 8 (around 3200 elo) ?
The eval is the heart of an engine, you said "working on the eval is boring" , can't understand this ...
Mayhem is strong but its not your Own , sorry ...
Chess computer is on the sad way now ...
Actually MY engine Sapeli was around 2300 Elo: See https://lichess.org/@/SapeliEngine
Too much lag. I won't play over TCP/IP ever.
But Mayhem is Sapeli. And GPLv3 like Stockfish. So no problem with some copy paste.
User avatar
Guenther
Posts: 4606
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Mayhem NNUE - New NN engine

Post by Guenther »

JohnWoe wrote: Wed Nov 18, 2020 2:18 pm Mayhem 1.1 released!
Release: https://github.com/SamuraiDangyo/mayhem ... s/tag/v1.1

Crafty spams my folder with log files.
A lot of Crafty options can be configured with its ini file, which should be named crafty.rc.
A simple 'log off' there would do it.

UCI was still in the abyss and also newer XB protocol versions when Crafty arrived, therefore XB engines relied heavily on config files in the past.
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy
User avatar
mvanthoor
Posts: 1784
Joined: Wed Jul 03, 2019 4:42 pm
Location: Netherlands
Full name: Marcel Vanthoor

Re: Mayhem NNUE - New NN engine

Post by mvanthoor »

JohnWoe wrote: Fri Nov 13, 2020 12:18 am Nice to see others testing Mayhem too. I'm curious of the level. Since in my tests Mayhem performs better than Crafty 25.6 which is 2950 Elo.

With Sapeli I was struggling against 1900 Elo FairyMax. Now I'm in the big boys league. 60% score against 2953 Elo Crafty 25.6. :lol:
More strange is Mayhem is 1200 sloc. Sapeli is 1700 lines. I took 500 sloc out and gained 1000 Elo.

To be honest credits for the 1000 Elo boost go to the SF NNUE evaluation. Without it Mayhem is probably 1800 Elo. :D
The fact that NNUE gains you 1000 Elo just means that your search is quite good, and your evaluation was junk. If I have this correct, Stockfish 12 has NNUE built in. Between 11 and 12, SF gained somewhere along the lines of 30-35 Elo. It means that its handcrafted evaluation was about as perfect as humans can make/tune it.

In short: if you write a decent search, *EVERYBODY* is in the "big boys" league as soon as they implement SF's NNUE. Why? Because a basic engine with no features is somewhere around 1600 Elo. (Such as my current engine.) Engines with some features, such as VICE, BBC, and yes, Sapeli, are somehwere around 2000 Elo. And as soon as you stick a NNUE on top of them, they gain 800-1000 Elo (more Elo if the search is faster).

So it's no wonder that a small engine with a few features + NNUE is at least 2950 as soon as you stick a NNUE on top of it.

That is the entire reason why so many people have an aversion against using the NNUE tech from another engine. It's a cheapskate way to rocket yourself into the big boys league. You can basically do it by cloning any +/- 2000 Elo engine you want, renaming it, and attaching someone else's NNUE to it.

Have fun; I'd have rather tested my engine against Sapeli, and then have it creep up the rating ladder version by version, and actually have an engine I at least know what it's doing.
Author of Rustic, an engine written in Rust.
Releases | Code | Docs | Progress | CCRL
Madeleine Birchfield
Posts: 512
Joined: Tue Sep 29, 2020 4:29 pm
Location: Dublin, Ireland
Full name: Madeleine Birchfield

Re: Mayhem NNUE - New NN engine

Post by Madeleine Birchfield »

mvanthoor wrote: Thu Nov 19, 2020 11:16 am
JohnWoe wrote: Fri Nov 13, 2020 12:18 am Nice to see others testing Mayhem too. I'm curious of the level. Since in my tests Mayhem performs better than Crafty 25.6 which is 2950 Elo.

With Sapeli I was struggling against 1900 Elo FairyMax. Now I'm in the big boys league. 60% score against 2953 Elo Crafty 25.6. :lol:
More strange is Mayhem is 1200 sloc. Sapeli is 1700 lines. I took 500 sloc out and gained 1000 Elo.

To be honest credits for the 1000 Elo boost go to the SF NNUE evaluation. Without it Mayhem is probably 1800 Elo. :D
The fact that NNUE gains you 1000 Elo just means that your search is quite good, and your evaluation was junk. If I have this correct, Stockfish 12 has NNUE built in. Between 11 and 12, SF gained somewhere along the lines of 30-35 Elo. It means that its handcrafted evaluation was about as perfect as humans can make/tune it.

In short: if you write a decent search, *EVERYBODY* is in the "big boys" league as soon as they implement SF's NNUE. Why? Because a basic engine with no features is somewhere around 1600 Elo. (Such as my current engine.) Engines with some features, such as VICE, BBC, and yes, Sapeli, are somehwere around 2000 Elo. And as soon as you stick a NNUE on top of them, they gain 800-1000 Elo (more Elo if the search is faster).

So it's no wonder that a small engine with a few features + NNUE is at least 2950 as soon as you stick a NNUE on top of it.

That is the entire reason why so many people have an aversion against using the NNUE tech from another engine. It's a cheapskate way to rocket yourself into the big boys league. You can basically do it by cloning any +/- 2000 Elo engine you want, renaming it, and attaching someone else's NNUE to it.

Have fun; I'd have rather tested my engine against Sapeli, and then have it creep up the rating ladder version by version, and actually have an engine I at least know what it's doing.
Agreed. If Sapeli were to go NNUE I would prefer it to be implemented from scratch, not copied directly from the Stockfish codebase, and trained on Sapeli evaluations. Do what the Komodo developers did.

Otherwise be honest and call the engine Mayhem Stockfish NNUE, like what the BBC author did. And TCEC would never accept such an engine in the first place if it ends up having a copy of Stockfish evaluation.