Crafty 25.3 Release

[Steve Brown]

Stockfish 11 64-bit 3499
Crafty 25.3 64-bit 2929

Your results sound fake to me. Sorry.

I found out what was hampering Stockfish 11. I opened the debug window in Arena and saw that on startup of the Stockfish engine, Arena was sending it the command

setoption name UCI_LimitStrength value true

even though there is no menu option in Arena for it to do that. The command "UCI_Elo" was not sent, but apparently there is a default value that the engine uses when UCI_LimitStrength is set to true, and that value is well below maximal strength. I solved that problem by sending the engine the command

setoption name UCI_LimitStrength value false

Then I had it play the black pieces in a game against Crafty, and Stockfish 11 won in 68 moves.

So, my estimation of Crafty was elevated by Stockfish being attenuated, and now I am inclined to use Stockfish 11, instead of Crafty, to play against a human-like opponent, at whatever elo value I input to the engine. This would not have come about without your skepticism. Here is a pgn of the game output by Arena.

[pgn]
[Event "Computer chess game"]
[Site "WIN8BOX"]
[Date "2025.04.28"]
[Round "1"]
[White "Crafty 25.3"]
[Black "Stockfish 11"]
[Result "0-1"]
[BlackElo "3000"]
[ECO "D01"]
[Opening "Richter-Veresov"]
[Time "03:31:56"]
[Variation "3...h6"]
[WhiteElo "3000"]
[TimeControl "0+10"]
[Termination "normal"]
[PlyCount "136"]
[WhiteType "program"]
[BlackType "program"]

1. d4 d5 {(d7-d5 c2-c4) -0.20/24 10} 2. Nc3 Nf6 {(Ng8-f6 Bc1-g5 h7-h6
Bg5xf6 e7xf6 e2-e3 c7-c6 Ng1-f3 Bf8-d6 Bf1-d3 Nb8-a6 a2-a3 O-O O-O Na6-c7
Nc3-e2 Bc8-e6 h2-h3 Rf8-e8 c2-c3 f6-f5 Nf3-e5 g7-g6 Ne2-f4 Bd6xe5 d4xe5)
+0.27/24 9} 3. Bg5 h6 {(h7-h6) +0.55/25 10} 4. Bf4 a6 {(a7-a6 e2-e3)
+0.51/25 10} 5. e3 Bg4 {(Bc8-g4 Qd1-d2) +0.45/24 10} 6. Nf3 e6 {(e7-e6
Bf1-e2) +0.41/25 9} 7. Bd3 c5 {(c7-c5 d4xc5) +0.78/24 10} 8. dxc5 Nc6
{(Nb8-c6 Qd1-d2) +0.48/24 10} 9. Qd2 Bxf3 {(Bg4xf3) +0.83/23 9} 10. gxf3
Bxc5 {(Bf8xc5 Bf4-g3 Bc5-e7 O-O-O Qd8-a5 Kc1-b1 Ra8-c8 Rh1-g1 Be7-b4 a2-a3
Bb4xc3 Qd2xc3 Qa5xc3 b2xc3 Ke8-e7 c3-c4 d5xc4 Bd3xc4 Rh8-d8 Bc4-e2 Rd8xd1+
Rg1xd1 Rc8-d8 Kb1-b2 Rd8xd1 Be2xd1 Nf6-h5 f3-f4 Nh5xg3 h2xg3) +0.65/24 9}
11. O-O-O Qa5 {(Qd8-a5 Kc1-b1 Bc5-b4 a2-a3 Bb4xc3 Qd2xc3 Qa5xc3 b2xc3 b7-b5
e3-e4 Ke8-e7 Rh1-g1 g7-g5 Bf4-g3 Ra8-c8 h2-h4 Nc6-a5 h4xg5 h6xg5 Bg3-e5
Rh8-h3 f3-f4 g5xf4 Be5xf6+ Ke7xf6 e4xd5 e6xd5 Rg1-g4 Rh3-f3) +0.99/26 9}
12. Kb1 Bb4 {(Bc5-b4) +0.96/27 9} 13. a3 Bxc3 {(Bb4xc3 Qd2xc3 Qa5xc3 b2xc3
Nf6-h5 Bf4-d6 Ke8-d7 Bd6-g3 Nh5xg3 h2xg3 Nc6-a5 c3-c4 d5xc4 Bd3xc4+ Kd7-e7
Bc4-e2 Ra8-c8 f3-f4 Rc8-c5 Rd1-d2 Na5-c4 Be2xc4 Rc5xc4 Rh1-d1 Rc4-c7 g3-g4
b7-b5 f4-f5 e6xf5 g4xf5 Ke7-f6 Rd2-d6+ Kf6xf5 Rd6xa6) +1.00/29 9} 14. Qxc3
Qxc3 {(Qa5xc3 b2xc3) +0.82/28 10} 15. bxc3 Nh5 {(Nf6-h5 Bf4-d6) +0.68/27 9}
16. Bd6 Rd8 {(Ra8-d8) +0.77/27 9} 17. Bb4 Rc8 {(Rd8-c8 Kb1-b2) +0.71/28 10}
18. h4 Nf6 {(Nh5-f6 a3-a4 Nc6xb4 c3xb4 d5-d4 e3xd4 Nf6-d5 Kb1-b2 Nd5xb4
Bd3-e4 Ke8-e7 c2-c3 Nb4-d5 Be4xd5 e6xd5 Kb2-c2 Rh8-e8 Rd1-b1 Ke7-f6 Rh1-e1
Re8xe1 Rb1xe1 Rc8-c7 a4-a5 Rc7-c4 Kc2-d3 Rc4-c8) +0.69/25 9} 19. f4 Ne4
{(Nf6-e4 Bd3xe4) +0.73/26 9} 20. Bxe4 dxe4 {(d5xe4 Kb1-b2) +0.55/30 10} 21.
Rhg1 g6 {(g7-g6 Rg1-h1 a6-a5 Bb4-c5 b7-b6 Bc5-d4 Nc6xd4 c3xd4 Ke8-e7 h4-h5
Ke7-f6 Kb1-b2 Rc8-c6 d4-d5 e6xd5 Rd1xd5 Rh8-c8 Rh1-c1 Kf6-g7 h5xg6 f7xg6
Rd5-d4 h6-h5 c2-c3 h5-h4 Rd4xe4 Kg7-f6 Re4-d4 h4-h3 Rc1-h1 Rc6xc3 Rh1xh3
Rc3-c2+ Kb2-b3) +0.63/27 10} 22. Kb2 Ne7 {(Nc6-e7 Bb4-d6 Rh8-g8 Bd6xe7
Ke8xe7 Rd1-d4 Rc8-c7 Rd4xe4 Rg8-c8 Re4-b4 Ke7-f6 a3-a4 Rc7xc3 Rg1-c1 Rc3-c7
c2-c3 Kf6-f5 f2-f3 Kf5-f6 Rb4-d4 Rc7-c5 Rd4-b4 Rc5-h5 Rc1-h1 Rh5-d5 Kb2-c2
Rd5-c5 Rb4-b3 Rc8-c7 Rh1-a1 b7-b6 Kc2-d2 Rc7-c6 Kd2-c2 a6-a5 Ra1-h1)
+0.93/30 9} 23. Rd6 Nf5 {(Ne7-f5 Rd6-b6) +0.96/28 9} 24. Rb6 Rc7 {(Rc8-c7
h4-h5 g6xh5 c3-c4 Nf5-e7 Bb4-a5 h5-h4 Rg1-h1 O-O Rh1xh4 Rf8-c8 Kb2-b3
Kg8-h7 Rh4-h1 Ne7-f5 Rh1-h2 Rc7xc4 Rb6xb7 Kh7-g6 Rh2-g2+ Kg6-f6 Rb7-d7
e6-e5 f4xe5+ Kf6-e6 Rd7-a7 Rc4xc2 Ra7xa6+ Ke6-d5 e5-e6 Nf5xe3 f2xe3 Rc2xg2
e6xf7) +0.89/29 9} 25. Rd1 Nxh4 {(Nf5xh4 Bb4-a5) +0.98/29 9} 26. Kb3 Nf5
{(Nh4-f5 c3-c4 h6-h5 Bb4-a5 O-O Rb6xe6 Rc7-c5 Ba5-b4 f7xe6 Bb4xc5 Rf8-f7
Rd1-h1 h5-h4 Bc5-b6 Rf7-h7 Rh1-h3 e6-e5 Bb6-d8 e5xf4 e3xf4 Rh7-d7 Bd8-g5
Rd7-d2 Bg5xh4 Nf5xh4 Rh3xh4 Rd2xf2 Kb3-c3 e4-e3 Kc3-d3 e3-e2 Kd3-d2 e2-e1Q+
Kd2xe1 Rf2xc2 c4-c5 Rc2xc5 Ke1-e2 Rc5-c4) +1.55/29 9} 27. c4 h5 {(h6-h5
Bb4-a5 O-O Rb6xe6 Rc7-c5 Ba5-b6 f7xe6 Bb6xc5 Rf8-f7 Bc5-d4 Kg8-f8 Bd4-e5
Kf8-e8 Kb3-b4 Rf7-h7 Rd1-h1 Ke8-d7 Kb4-c5 Kd7-c8 Kc5-b4 Rh7-f7 Kb4-c3 h5-h4
Rh1-h3 Kc8-d7 a3-a4 Kd7-c6 Rh3-h1 Kc6-c5 Rh1-h3 Kc5-b6) +1.13/29 9} 28. Bc3
O-O {(O-O Bc3-e5 Rc7-c6 Rb6xb7 Rf8-c8 Rd1-d7 Nf5-h6 Rd7-d4 Nh6-g4 Rd4-d7
Ng4xe5 f4xe5 Kg8-g7 Rd7xf7+ Kg7-h6 Rb7-e7 a6-a5 c4-c5 h5-h4 Rf7-f4 g6-g5
Rf4xe4 h4-h3 Re4-d4 Rc6xc5 Re7xe6+ Kh6-h5 Re6-e7 Rc8-h8 Rd4-d1 h3-h2 Rd1-h1
Kh5-g4 f2-f4 Rh8-b8+ Kb3-a4 Rc5-c4+ Ka4xa5 Rc4xc2 Re7-g7 Rc2-b2 Ka5-a6
Rb8-a8+ Rg7-a7 Ra8xa7+ Ka6xa7 g5xf4 e3xf4 Kg4xf4) +1.59/28 9} 29. Be5 Rc6
{(Rc7-c6 Rb6xb7) +1.44/31 10} 30. Rxb7 Rfc8 {(Rf8-c8 Rd1-d7) +1.50/29 10}
31. Rdd7 Nh6 {(Nf5-h6) +1.38/28 9} 32. Bf6 Rxc4 {(Rc6xc4) +1.68/28 9} 33.
c3 a5 {(a6-a5 a3-a4) +0.93/27 9} 34. Kb2 a4 {(a5-a4 Bf6-g5) +1.23/28 9} 35.
Ra7 R4c6 {(Rc4-c6 Rd7-b7) +1.49/27 9} 36. Rdb7 Rc4 {(Rc6-c4) +1.16/27 9}
37. Rd7 R4c5 {(Rc4-c5 Ra7xa4) +1.33/28 10} 38. Rxa4 Ng4 {(Nh6-g4) +1.80/29
9} 39. Bh4 g5 {(g6-g5 Bh4xg5) +1.51/27 9} 40. Bxg5 Nxf2 {(Ng4xf2 Rd7-d2
Nf2-g4 Ra4xe4 Kg8-g7 Bg5-h4 Rc5xc3 a3-a4 Kg7-g6 Rd2-g2 Kg6-f5 Re4-b4 Rc3xe3
Rb4-b5+ Kf5xf4 Rb5xh5 f7-f6 Kb2-b1 Re3-a3 a4-a5 e6-e5 Bh4xf6 Ng4xf6 Rg2-f2+
Kf4-g3 Rf2xf6 Ra3xa5 Rf6-b6 Kg3-g4 Rh5-h6 e5-e4 Rh6-e6 Kg4-f4) +1.61/30 9}
41. Bf6 Rb5+ {(Rc5-b5+ Kb2-a2) +2.69/26 9} 42. Ka2 Ng4 {(Nf2-g4 Bf6-e7)
+2.32/26 9} 43. Be5 Rd5 {(Rb5-d5) +2.95/28 9} 44. Rxd5 exd5 {(e6xd5 Ra4-a5
f7-f6 Be5-d4 Kg8-f7 Ra5xd5 Kf7-g6 Rd5-b5 h5-h4 Rb5-b1 h4-h3 Rb1-h1 h3-h2
Ka2-b3 Kg6-h5 Bd4xf6 Ng4xf6 Rh1xh2+ Kh5-g4 c3-c4 Kg4-f3 Rh2-h3+ Kf3-g2
Rh3-h6 Nf6-g4 Rh6-e6 Kg2-f3 f4-f5 Kf3xe3 a3-a4 Ke3-d4 a4-a5 e4-e3 a5-a6
Ng4-f2 Re6-d6+ Kd4-e5 Rd6-e6+ Ke5-f4 Kb3-c3 Nf2-e4+ Kc3-d3 Ne4-c5+ Kd3-e2
Nc5xe6 f5xe6 Rc8xc4 a6-a7) +2.86/30 9} 45. Ra5 f6 {(f7-f6 Be5-d4) +3.11/30
10} 46. Bd4 Kf7 {(Kg8-f7 Ra5xd5 Kf7-g6 Rd5-b5 h5-h4 Rb5-b1 h4-h3 Rb1-h1
h3-h2 Ka2-b3 Kg6-h5 c3-c4 Kh5-h4 c4-c5 Kh4-g3 Kb3-c4 Kg3-g2 Rh1-e1 h2-h1Q
Re1xh1 Kg2xh1 Kc4-d5 f6-f5 Kd5-e6 Kh1-g2 Ke6-d7 Rc8-a8 c5-c6 Kg2-f3 c6-c7
Ng4xe3 c7-c8Q Ra8xc8 Kd7xc8 Kf3xf4 Kc8-d7 Ne3-c4 Kd7-c6 e4-e3 Bd4-c3
Kf4-e4) +3.24/32 9} 47. c4 h4 {(h5-h4) +3.91/28 9} 48. cxd5 h3 {(h4-h3
d5-d6 Rc8-h8 d6-d7 h3-h2 Ra5-h5 Rh8-d8 a3-a4 Rd8xd7 Ka2-b3 Kf7-g6 Rh5xh2
Ng4xh2 a4-a5 Nh2-f3 Bd4-b6 Rd7-d1 Kb3-c4 Rd1-b1 Bb6-a7 Nf3-d2+ Kc4-c3
Nd2-f1 Ba7-d4 Nf1-g3 Kc3-c4 Ng3-f5 Bd4-b6 Rb1-a1 Kc4-d5 Ra1-a4 Kd5-c6
Ra4-a3 Kc6-d5 Nf5xe3+ Kd5xe4 Ne3-c2 Ke4-d5 Kg6-f5) +4.48/30 9} 49. d6 Rh8
{(Rc8-h8) +5.12/30 9} 50. Ra7+ Ke6 {(Kf7-e6) +8.25/28 9} 51. d7 h2 {(h3-h2)
+11.76/26 10} 52. d8=Q Rxd8 {(Rh8xd8) +16.27/24 9} 53. Rh7 Rxd4 {(Rd8xd4)
+28.37/27 9} 54. exd4 e3 {(e4-e3 f4-f5+ Ke6xf5 Ka2-b2 e3-e2 Rh7-e7 h2-h1Q
Re7xe2 Qh1-d1 Re2-c2 Qd1xd4+ Kb2-b3 Ng4-e3 Rc2-c3 Qd4-d1+ Kb3-a2 Qd1-d5+
Rc3-b3 Kf5-f4 a3-a4 Qd5-d4 Rb3xe3 Kf4xe3 Ka2-b3 Ke3-f4 a4-a5 Kf4-f5 Kb3-c2
Qd4-c5+ Kc2-d3 Qc5-a3+ Kd3-e2 Qa3xa5 Ke2-d3 Qa5-b4) +60.56/29 9} 55. Rh8
Kf5 {(Ke6-f5) +147.92/31 9} 56. Kb2 e2 {(e3-e2 Rh8-e8 h2-h1Q Re8xe2 Qh1-d1
Re2-c2 Ng4-e3 Rc2-c3 Qd1xd4 Kb2-b3 Kf5xf4 a3-a4 Qd4-d1+ Kb3-a3 Qd1-b1
Rc3-c5 Ne3-c2+ Rc5xc2 Qb1xc2 Ka3-b4 Qc2-b2+ Kb4-c4 Qb2-a3 Kc4-b5 Qa3-b3+
Kb5-c6 Qb3xa4+ Kc6-d6 Kf4-f5 Kd6-c5 Kf5-e5 Kc5-b6 Ke5-d6 Kb6-b7 Qa4-b4+
Kb7-c8 Qb4-b2 Kc8-d8 Qb2-b8+) +M20/41 9} 57. Re8 h1=Q {(h2-h1Q Re8xe2
Qh1-d1 Re2-c2 Ng4-e3 Rc2-c3 Qd1xd4 Kb2-b3 Kf5xf4 a3-a4 Qd4-d1+ Kb3-a3
Qd1-b1 a4-a5 Ne3-c2+ Rc3xc2 Qb1xc2 a5-a6 Qc2-d3+ Ka3-a4 Qd3xa6+ Ka4-b4
Qa6-b6+ Kb4-c3 Kf4-e3 Kc3-c4 Qb6-a5 Kc4-b3 Ke3-d3 Kb3-b2 Qa5-b5+ Kb2-c1
Qb5-b6 Kc1-d1 Qb6-b1+) +M18/46 9} 58. Rxe2 Qd1 {(Qh1-d1 Re2-c2 Ng4-e3
Rc2-h2 Qd1xd4+ Kb2-a2 Qd4-c4+ Ka2-b1 Qc4-d3+ Kb1-a1 Qd3xa3+ Ka1-b1 Ne3-c4
Rh2-h5+ Kf5xf4 Rh5-h4+ Kf4-g3 Rh4xc4 Qa3-b3+ Kb1-a1 Qb3xc4 Ka1-b2 Kg3-f4
Kb2-a3 Kf4-e3 Ka3-b2 Ke3-d3 Kb2-a3 Qc4-b5 Ka3-a2 Kd3-c3 Ka2-a3 Qb5-b3+)
+M17/46 9} 59. Rc2 Ne3 {(Ng4-e3 Rc2-h2 Qd1xd4+ Kb2-a2 Qd4-c3 Rh2-f2 Ne3-c4
a3-a4 Nc4-d2 Rf2xd2 Qc3xd2+ Ka2-b3 Qd2-a5 Kb3-c4 Kf5xf4 Kc4-b3 Kf4-e4
Kb3-c4 Ke4-e3 Kc4-b3 Ke3-d3 Kb3-a2 Kd3-c2 Ka2-a3 Qa5-c3+ Ka3-a2 Qc3-b2+)
+M14/50 9} 60. Rc3 Ke4 {(Kf5-e4 Kb2-a2 Qd1-e2+ Ka2-b3 Ke4xd4 Rc3-c1 Ne3-c4
Rc1-c2 Qe2-d3+ Kb3-b4 Qd3xa3+ Kb4-b5 Qa3-c5+ Kb5-a4 Qc5-a5+ Ka4-b3 Qa5-a3+)
+M9/59 9} 61. Rc8 Kd3 {(Ke4-d3 Rc8-c3+ Kd3xd4 Rc3xe3 Qd1-d2+ Kb2-b3 Qd2xe3+
Kb3-b4 Qe3-f3 a3-a4 Qf3-b7+ Kb4-a3 Kd4-c3 a4-a5 Qb7-b3+) +M8/62 9} 62. Rc3+
Kxd4 {(Kd3xd4 Rc3xe3 Qd1-d2+ Kb2-b3 Qd2xe3+ Kb3-b4 Qe3-f3 a3-a4 Qf3-b7+
Kb4-a3 Kd4-c3 a4-a5 Qb7-b3+) +M7/99 9} 63. Rxe3 Qd2+ {(Qd1-d2+ Kb2-b3
Qd2xe3+ Kb3-b4 Qe3-f3 a3-a4 Qf3-b7+ Kb4-a3 Kd4-c3 a4-a5 Qb7-b3+) +M6/153 9}
64. Kb3 Qxe3+ {(Qd2xe3+ Kb3-b4 Qe3-f3 a3-a4 Qf3-b7+ Kb4-a3 Kd4-c3 a4-a5
Qb7-b3+) +M5/245 2} 65. Kb2 Qc3+ {(Qe3-c3+ Kb2-b1 Qc3-b3+ Kb1-c1 Kd4-d3
f4-f5 Qb3-c2+) +M4/245 0} 66. Ka2 Qc2+ {(Qc3-c2+ Ka2-a1 Kd4-c3 a3-a4
Qc2-b2+) +M3/245 0} 67. Ka1 Kc3 {(Kd4-c3 a3-a4 Qc2-b2+) +M2/245 0} 68. a4
Qb2# {(Qc2-b2+) +M1/245 0} 0-1
[/pgn]

[Graham Banks]

Glad you found out what the problem was.

[chesskobra]

@Steve Brown - You know that Stockfish 16 has both NN and hand crafted evaluation, and you can choose one or the other, although it is hard to say what is 'human like' about any of Crafty or SF 11 or SF 16.

[Steve Brown]

@Steve Brown - You know that Stockfish 16 has both NN and hand crafted evaluation, and you can choose one or the other, although it is hard to say what is 'human like' about any of Crafty or SF 11 or SF 16.

I would expect an engine programmed to play according to positional principles learned by humans playing chess to be more like a human in the way it plays chess than an engine programmed to play according to positional principles learned by a neural net playing millions of games without input of any human knowledge. I would expect that to be more true at shallower search depths, where positional knowledge outweighs tactical calculation. Clearly, neural net knowledge imparts greater strength to engines than human knowledge alone. It would be interesting if neural net chess knowledge could be translated into language, a set of principles humans could read. As it is, humans can only tap into that knowledge by the example of how chess engines play.

[AndrewGrant]

I have found Crafty to be a very strong non-neural net chess engine. It beats Stockfish 11, the last version that did not use a neural net, every time at 15 seconds per move. I believe that a chess engine programmed with human chess knowlege is more likely to play human-like moves, and I would like to use Crafty with the elo feature enabled. I would appreciate any help you can provide.

That claim simply cannot be true?

Aside from your reply being of no help whatsoever, on what basis do you accuse me of lying? I set up several games between Crafty 25.3 and Stockfish 11 on the Arena chess GUI, and Crafty won all games impressively. Feel free to try that yourself. Note that the current version of Stockfish is 17, and version 11 was the current version several years ago, before Stockfish began using a neural network in its evaluation. Crafty is a direct descendent of Cray Blitz, which won several world computer championships.

I really would like help getting the elo feature of Crafty to work.

Seems like my "no help whatsoever" reply helped prompt you to dig a little deeper, which revealed there was indeed an issue.

Also, just for human interactions, note the question mark in my comment.

[Steve Brown]

I have found Crafty to be a very strong non-neural net chess engine. It beats Stockfish 11, the last version that did not use a neural net, every time at 15 seconds per move. I believe that a chess engine programmed with human chess knowlege is more likely to play human-like moves, and I would like to use Crafty with the elo feature enabled. I would appreciate any help you can provide.

That claim simply cannot be true?

Aside from your reply being of no help whatsoever, on what basis do you accuse me of lying? I set up several games between Crafty 25.3 and Stockfish 11 on the Arena chess GUI, and Crafty won all games impressively. Feel free to try that yourself. Note that the current version of Stockfish is 17, and version 11 was the current version several years ago, before Stockfish began using a neural network in its evaluation. Crafty is a direct descendent of Cray Blitz, which won several world computer championships.

I really would like help getting the elo feature of Crafty to work.

Seems like my "no help whatsoever" reply helped prompt you to dig a little deeper, which revealed there was indeed an issue.

Also, just for human interactions, note the question mark in my comment.

I just realized there was more than one person expressing skepticism about the results I got playing Crafty 25.3 against Stockfish 11, which turned out to be skewed because of a UCI_LimitStrength command being sent to Stockfish by Arena. So, my reply to Graham Banks was addressed as much to you as to him. I wrote "This [solution to the problem] would not have come about without your skepticism." I replied as I did because I read your reply as an implication that I was lying.

[Peter Berger]

...

[Peter Berger]

I would expect an engine programmed to play according to positional principles learned by humans playing chess to be more like a human in the way it plays chess than an engine programmed to play according to positional principles learned by a neural net playing millions of games without input of any human knowledge.

I consider the original topic to be closed/solved by now.

But this remark of yours is way more interesting to me anyway. How come you think this way? Just going by the games played for me personally lc0 running on a lowly CPU is definitely closest to what I can get from the computerchess camp to watching a human grandmaster think.

For me this also feels very logical. We don't really know how humans think about chess exactly. We only know that the procedure is certainly way more complicated than what classical evaluation functions of chess programs emulate. Even very mediocre chessplayers recognize exceptions from rules, that can't be programmed easily. So I'd personally rather spend some hours watching lc0 for your purposes.

[Steve Brown]

I would expect an engine programmed to play according to positional principles learned by humans playing chess to be more like a human in the way it plays chess than an engine programmed to play according to positional principles learned by a neural net playing millions of games without input of any human knowledge.

I consider the original topic to be closed/solved by now.

But this remark of yours is way more interesting to me anyway. How come you think this way? Just going by the games played for me personally lc0 running on a lowly CPU is definitely closest to what I can get from the computerchess camp to watching a human grandmaster think.

For me this also feels very logical. We don't really know how humans think about chess exactly. We only know that the procedure is certainly way more complicated than what classical evaluation functions of chess programs emulate. Even very mediocre chessplayers recognize exceptions from rules, that can't be programmed easily. So I'd personally rather spend some hours watching lc0 for your purposes.

I think that way because my impression of non-neural net engines is that their evaluation functions are carefully crafted by chess masters, and in some cases grandmasters, to implement as much chess knowledge as possible, such as control the center, develop all your pieces, support your pieces, don't bring the queen out early, protect your king, connect rooks, put rooks on open files, create a pawn chain, avoid doubled pawns, try to double the opponent's pawns, try to have passed pawns, secure knight outposts, put bishops on long open diagonals, put rook on seventh rank, attack the opponent's king, etc. Those are all principles learned by humans from experience. A neural net likely learns those principles but also learns hidden principles that we can only deduce from observing how they play. Humans no longer compete against engines because engines do not play like humans. They seem to possess a superior alien intelligence, based on knowledge acquired from playing millions of games, far more than any human can play or study. So I have the impression that a non-neural net engine plays more like a human. I do not assert that as a truth but rather as my personal preference in my selection of what engines to play against. Stockfish 17 is a monster that will crush you at its best level, and that is not exactly fun. I am not a master but an intermediate player, and I enjoyed playing the black pieces in the following game against Stockfish 11 set to play at elo 1500, allowed to think 10 seconds per move on 4 cores, while I had unlimited time per move.

[pgn]
[Event "Computer chess game"]
[Site "WIN8BOX"]
[Date "2025.04.28"]
[Round "2"]
[White "Stockfish 11"]
[Black "Steve"]
[Result "0-1"]
[BlackElo "1500"]
[Time "12:07:35"]
[WhiteElo "1500"]
[TimeControl "0+10"]
[Termination "normal"]
[PlyCount "72"]
[WhiteType "program"]
[BlackType "computer"]

1. e4 e5 2. Nf3 Nc6 {(Nb8-c6 Bf1-b5 a7-a6 Bb5-a4 Ng8-f6 d2-d3 b7-b5 Ba4-b3
Bf8-c5 Nb1-c3 O-O Bc1-g5 h7-h6 Bg5xf6 Qd8xf6 Nc3-d5 Qf6-d6 O-O Bc8-b7 c2-c3
Rf8-e8 Bb3-c2 Nc6-e7 d3-d4 e5xd4 c3xd4 Bc5-b6 Nd5xb6 Qd6xb6 d4-d5 Ne7-g6)
-0.09/22 9} 3. Bc4 Nf6 4. Nc3 {(Nf3-g5 d7-d5) +0.37/22 10} d6 5. d3
{(Nf3-g5 d6-d5 e4xd5 Nc6-a5 Bc4-b5+ c7-c6 d5xc6 b7xc6 Bb5-d3 Nf6-g4 Ng5-f3
f7-f5 h2-h3 e5-e4 h3xg4 e4xd3 c2xd3 Bf8-e7 g4xf5 O-O d3-d4 Bc8xf5 d2-d3
Bf5-g4 Bc1-e3 Ra8-b8 Qd1-e2 Bg4xf3 g2xf3 Qd8-b6 Ra1-b1 h7-h6) +1.44/22 10}
Na5 6. O-O {(Bc4-d5 Bf8-e7 a2-a3 Na5-c6 O-O Nc6-d4 Nf3xd4 e5xd4 Nc3-e2
Nf6xd5 e4xd5 Be7-f6 Ne2-g3 O-O Bc1-d2 Bc8-d7 f2-f4 Rf8-e8 Qd1-f3 c7-c5
Ng3-e4 Bd7-f5 Ra1-e1 Bf5xe4 d3xe4) +0.64/22 10} Nxc4 7. dxc4 {(d3xc4 Bf8-e7
h2-h3 O-O a2-a4 Bc8-e6 Qd1-e2 c7-c6 Rf1-d1 Qd8-c7 Nf3-h4 h7-h6 Nh4-f5
Be6xf5 e4xf5 Ra8-d8 Bc1-e3 b7-b6 g2-g4 d6-d5 g4-g5 h6xg5 Be3xg5 Qc7-d7
Bg5xf6 Be7xf6 Qe2-h5) +0.07/21 9} Bg4 8. b3 {(Qd1-e2 Nf6-d7 h2-h3 Bg4-h5
Rf1-d1 b7-b6 Nc3-d5 Bf8-e7 b2-b4 a7-a5 a2-a4 O-O b4-b5 Bh5-g6 Bc1-a3 Ra8-b8
Kg1-h2 f7-f5 e4xf5 Rf8xf5 Ba3-b2 Nd7-c5) +0.50/22 10} Be7 9. Bg5 {(c4-c5
O-O) +0.22/22 9} h6 10. Bh4 {(Bg5xf6 Be7xf6 h2-h3 Bg4-e6 Nc3-d5 O-O Qd1-d3
a7-a5 Ra1-d1 b7-b6 a2-a3 Bf6-e7 b3-b4 a5xb4 a3xb4 f7-f5 Rd1-a1 f5xe4 Qd3xe4
Ra8xa1 Rf1xa1 Be6-f5 Qe4-e2 c7-c6 Nd5xe7+ Qd8xe7 Ra1-a6 Qe7-b7 c4-c5 b6xc5)
+0.27/23 10} O-O 11. Bxf6 {(h2-h3 Bg4-e6 Bh4xf6 Be7xf6 Nc3-d5 g7-g6 Qd1-e2
Bf6-g7 Ra1-d1 b7-b6 b3-b4 c7-c6 Nd5-e3 Qd8-c7 Qe2-d3 Ra8-d8 b4-b5 Kg8-h7
b5xc6 Qc7xc6 Ne3-d5 f7-f5 Rf1-e1 f5xe4 Qd3xe4 Be6xd5 Qe4xd5 Qc6xd5 Rd1xd5)
-0.02/22 9} Bxf6 12. h3 {(Nc3-d5 c7-c6 Nd5xf6+ Qd8xf6 Rf1-e1 Ra8-d8 Re1-e3
Qf6-e7 h2-h3 Bg4-e6 Qd1-e2 f7-f5 c4-c5 f5xe4 c5xd6 Qe7xd6 Re3xe4 Be6-d5
Ra1-d1 Qd6-c5 Re4-e3 Bd5xf3 g2xf3 Rd8xd1+ Qe2xd1 Rf8-f5 Kg1-g2 Rf5-h5
Re3-d3 Rh5-g5+ Kg2-f1 Qc5-b6 Kf1-e2) -0.17/24 9} Bh5 13. Nd5 {(Qd1-d3
c7-c6) +0.06/24 10} a5 14. Kh2 {(g2-g4 Bh5-g6) +0.23/25 10} Bg6 15. Qe2
{(Qd1-d3 c7-c6 Nd5-c3 Bf6-e7 g2-g4 Qd8-c7 Ra1-d1 Qc7-c8 a2-a4 Qc8-e6 Rf1-g1
Ra8-d8 Qd3-e3 Rf8-e8 Rg1-e1 Rd8-a8 Re1-g1) 0.00/25 9} c6 16. Nxf6+
{(Nd5xf6+ Qd8xf6 Rf1-e1 a5-a4 Ra1-d1 Rf8-c8 Qe2-d2 a4xb3 a2xb3 Rc8-d8
Kh2-g1 Qf6-e7 Qd2-b4 Ra8-a2 Qb4-c3 Rd8-e8 Qc3-d2 Re8-d8) 0.00/25 10} Qxf6
17. Rae1 {(Rf1-e1 a5-a4 Ra1-d1 a4xb3 a2xb3 Bg6-h5 Rd1-d3 Rf8-d8 Kh2-g1
Qf6-e7 Re1-d1 Ra8-a2 Rd3-d2 Qe7-f6 Qe2-e3 Bh5xf3 Qe3xf3 Qf6xf3 g2xf3 g7-g6
c4-c5 d6-d5 e4xd5 Rd8xd5 Rd2xd5) +0.13/24 9} Qe6 18. c3 {(a2-a4 Ra8-d8
Qe2-d2 f7-f5 Qd2xa5 f5xe4 Nf3-d2 d6-d5 Qa5-b4 e4-e3 f2xe3 Bg6xc2 Qb4xb7
Bc2-d3 Rf1xf8+ Rd8xf8 c4xd5 c6xd5 e3-e4 Rf8-f2 Qb7-a8+ Kg8-h7 e4xd5 Qe6-g6
d5-d6 Qg6xd6 Nd2-c4 Qd6-b4 Re1-d1 Qb4xb3 Rd1-d2 Rf2xd2 Nc4xd2 Qb3-b2)
+0.15/23 9} f5 19. a4 {(e4xf5 Bg6xf5 Qe2-e3 Ra8-e8 Kh2-g1 Rf8-f6 Nf3-h4
Bf5-h7 Nh4-f3) 0.00/24 9} fxe4 20. Nh4 {(Nf3-d2 Ra8-e8 Nd2xe4 d6-d5 c4xd5
c6xd5 Ne4-d2 Qe6-f6 Qe2-e3 d5-d4 Qe3-g3 d4-d3 f2-f3 h6-h5 Nd2-e4 Bg6xe4
Re1xe4 Re8-e6 Rf1-d1 Rf8-d8 Rd1-d2 b7-b6 Re4-e1 h5-h4 Qg3-f2 Qf6-f4+ Kh2-g1
Re6-g6 Qf2-e3) -0.75/24 10} Bh7 21. c5 {(g2-g3 Rf8-f6) -2.57/24 9} d5 22.
Kh1 {(Qe2-g4 Qe6-e8 g2-g3 Qe8-e7 Kh2-g1 Rf8-f6 b3-b4 Ra8-f8 Nh4-g2 a5xb4
c3xb4 Bh7-f5 Qg4-h5 Qe7-e6 f2-f4 e4xf3/ep Qh5xf3 Bf5xh3 Qf3-d1 Bh3-g4
Rf1xf6 Rf8xf6 Qd1-b1 Rf6-f3 Re1-e3) -4.59/23 9} g5 23. f3 {(Nh4-f5 Bh7xf5
Qe2-e3 Qe6-f6 Kh1-g1 d5-d4 Qe3-d2 Bf5-g6 Re1-d1 d4-d3 Rf1-e1 Kg8-g7 Rd1-b1
Qf6-f4 Re1-e3 Rf8-f7 Qd2-e1 h6-h5 Rb1-d1 g5-g4 Re3-g3 g4xh3 g2xh3 Ra8-f8
Rd1-d2 h5-h4 Rg3-g4 Qf4-f3) -7.46/22 10} gxh4 24. Kg1 {(Kh1-g1 Rf8-f4)
-8.32/23 10} exf3 25. Rxf3 {(g2xf3 Qe6xh3) -8.30/22 9} e4 26. Rf2 {(Rf3-f2
Qe6-e7 b3-b4 Rf8xf2 Qe2xf2 Ra8-f8 Qf2-d4 a5xb4 c3xb4 Qe7-g7 Qd4xg7+ Kg8xg7
a4-a5 d5-d4 Re1-f1 e4-e3 Rf1xf8 Kg7xf8 Kg1-f1 d4-d3 Kf1-e1 Bh7-e4 Ke1-d1
Be4-d5) -8.29/22 9} Bf5 27. Ref1 {(Qe2-h5 Kg8-h7 Qh5xh4 Qe6-e5 Re1-e3
Bf5-g6 Rf2-d2 Ra8-d8 Rd2-f2 Rf8xf2 Qh4xf2 Rd8-g8 Qf2-d2 Bg6-h5 Qd2-f2
Bh5-f3 Re3xf3 e4xf3 Qf2xf3 Rg8-g3 Qf3-f8 Rg3-g7 Qf8-f2) -7.32/23 9} Qg6 28.
Kh1 {(Qe2-e3 Bf5-e6 Kg1-h2 Qg6-g7 g2-g4 h4xg3/ep+ Qe3xg3 Rf8xf2+ Rf1xf2
Qg7xg3+ Kh2xg3 Kg8-g7 Rf2-f4 Kg7-g6 Rf4-f2 Kg6-g5 Kg3-h2 Ra8-e8 Kh2-g3
Re8-g8 Kg3-h2 Rg8-g7 Kh2-g3 Kg5-h5+ Kg3-h2 Kh5-g6 Rf2-g2+ Kg6-f6) -7.52/21
9} Rae8 29. Qb2 {(Qe2-e3 Bf5-e6 Kh1-g1 Rf8xf2 Rf1xf2 Kg8-h7 b3-b4 a5xb4
c3xb4 Qg6-g7 Rf2-f4 Re8-g8 Rf4-f2 Qg7-a1+ Rf2-f1 Qa1-e5 Qe3-f4 Qe5-g7
Rf1-f2 Be6xh3 Qf4-f7 e4-e3 Qf7xg7+ Rg8xg7 Kg1-f1 e3xf2 g2xh3 Rg7-f7 b4-b5)
-8.16/21 10} e3 30. Rf4 {(Rf2-f4 e3-e2 Rf1-e1 Bf5-d3 Qb2-c1 Rf8xf4 Qc1xf4
Qg6-g3 Qf4-g4+ Qg3xg4 h3xg4 Re8-f8 Re1xe2 Rf8-f1+ Kh1-h2 Bd3xe2 b3-b4
Be2xg4 b4xa5 Rf1-c1 a5-a6 b7xa6 c3-c4 d5xc4 a4-a5 c4-c3 g2-g3 Rc1-c2+
Kh2-g1 h4xg3 Kg1-f1 Bg4-f3 Kf1-e1 Rc2-c1+) -M17/42 9} Qg3 31. c4 {(Qb2-e2
Bf5-g6) -20.79/26 10} e2 32. Qc1 {(Qb2-a1 e2xf1Q+ Qa1xf1 Re8-e1 Qf1xe1
Qg3xe1+ Kh1-h2 Qe1-g3+ Kh2-g1 Qg3xf4 b3-b4 Bf5-e4 c4xd5 Qf4-f1+ Kg1-h2
Qf1xg2+) -M8/56 10} exf1=Q+ 33. Rxf1 {(Qc1xf1 Re8-e1 Qf1xe1 Qg3xe1+ Kh1-h2
Qe1-g3+ Kh2-g1 Qg3xf4 b3-b4 Bf5-e4 c4xd5 Qf4-f1+ Kg1-h2 Qf1xg2+) -M7/75 9}
Be4 34. Rf3 {(Rf1-f3 Be4xf3 g2xf3 Rf8xf3 Qc1-g1 Re8-e1 Qg1xe1 Qg3xe1+
Kh1-h2 Rf3-f2+) -M5/240 10} Bxf3 35. Qf1 {(g2xf3 Rf8xf3 Qc1-g1 Re8-e1
Qg1xe1 Qg3xe1+ Kh1-h2 Rf3-f2+) -M4/245} Bxg2+ 36. Kg1 {(Qf1xg2 Re8-e1+
Qg2-f1 Re1xf1+) -M2/245} Rxf1# 0-1
[/pgn]