Chess Engine for analysis that has human like lines

[menniepals]

Hello, I have been using Stockfish Dev Builds for analysis. Stockfish a lot of times will give mind numbing lines. Are there other strong engines that may give human like lines or lines humans can easily understand. The majority of strong engines will come up with electronic gibberish lines. I don't know if Schooner 2.2 could do a better job. Are there other engines you can suggest?

[dkappe]

Hello, I have been using Stockfish Dev Builds for analysis. Stockfish a lot of times will give mind numbing lines. Are there other strong engines that may give human like lines or lines humans can easily understand. The majority of strong engines will come up with electronic gibberish lines. I don't know if Schooner 2.2 could do a better job. Are there other engines you can suggest?

If you are willing to step back to SF13 or CFish13, my Harmon net is trained on 1m FIDE 2300+ games results plus naive material qsearch. Substantially weaker than default SF nets, but is still plenty strong and definitely has a different style of play. https://www.patreon.com/posts/harmon-nnue-44549655

[menniepals]

Will definitely give it a try. Thanks!

[matejst]

I could recommend Orion 0.8 by David Carteau for analysis. Very interesting engine. I hope that the project is not over.

[carldaman]

I can vouch for Harmon, too, as well as the (commercially available) Komodo Human style.

[Archimedes]

If you are willing to step back to SF13 or CFish13, my Harmon net is trained on 1m FIDE 2300+ games results plus naive material qsearch. Substantially weaker than default SF nets, but is still plenty strong and definitely has a different style of play. https://www.patreon.com/posts/harmon-nnue-44549655

Harmon net can also be used with Fire, Igel, Mayhem and RubiChess.

[Rubinus]

If you are willing to step back to SF13 or CFish13, my Harmon net is trained on 1m FIDE 2300+ games results plus naive material qsearch. Substantially weaker than default SF nets, but is still plenty strong and definitely has a different style of play. https://www.patreon.com/posts/harmon-nnue-44549655

It doesn't seem that much weaker. It was only twelve games against Stockfish 14.1, but the result is 0:1=11. Better than Ethereal and maybe even Dragon would have achieved.
Why can't this be used with 14, has the network format changed?
15+10, 6cpu (bmi), the only decisive game here:
[pgn][Event "15m+10s"]
[Site "?"]
[Date "2021.11.10"]
[Round "1"]
[White "Stockfish 14.1"]
[Black "Stockfish 13-Harmon_NN"]
[Result "1-0"]
[ECO "C30"]
[Annotator "0.00;0.00"]
[PlyCount "179"]
[TimeControl "900+10"]

{Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz W=38.3 plies; 6 565kN/s; 3 380 245
TBAs; C30-39-LC0.ctg B=34.8 plies; 8 503kN/s; 3 722 652 TBAs; C30-39-LC0.ctg}
1. e4 {B 0} e5 {B 0} 2. f4 {B 0} Bc5 {B 0} 3. Nf3 {B 0} d6 {B 0} 4. c3 {B 0}
Nc6 {B 0} 5. Bb5 {B 0} Bb6 {B 0} 6. d4 {B 0} Bd7 {B 0} 7. O-O {B 0} Nf6 {B 0}
8. fxe5 {B 0} Nxe5 {B 0 Oba mimo knihovnu} 9. Bd3 {0.00/98 164} h6 {0.00/38 52
(Jxd3)} 10. Bc2 {0.63/34 12 (a4)} O-O {0.13/38 35 (Jg6)} 11. a4 {0.69/36 27
(Kh1)} a5 {0.05/36 19 (a6)} 12. Kh1 {0.70/34 18} Ng6 {0.00/34 19} 13. Be3 {0.
98/37 55} c5 {0.21/40 106 (c6)} 14. Bg1 {0.73/36 39 (d5)} Re8 {0.00/38 11} 15.
Nfd2 {0.85/36 32 (e5)} Rc8 {0.01/36 86} 16. Na3 {1.06/36 30} cxd4 {0.46/41 258}
17. cxd4 {1.11/35 33} Bc6 {0.39/36 10} 18. Qb1 {0.87/38 52 (Ve1)} Nh4 {0.00/34
14 (Vc7)} 19. Nb5 {1.60/32 18} Nh5 {0.27/36 33 (Jh7)} 20. Be3 {1.70/32 19} f5 {
0.00/33 16 (Jf6)} 21. Qe1 {2.29/33 20 (g3)} Kh8 {1.00/41 160 (d5)} 22. Bd1 {3.
07/36 27} Bxe4 {1.24/37 15} 23. Nxe4 {3.03/40 26} Rxe4 {1.11/40 14} 24. Nc3 {
2.98/41 19} Re6 {0.84/40 17} 25. Bxh5 {3.00/42 19} Bxd4 {1.10/44 25} 26. Nd5 {
3.00/43 26} Bxe3 {1.17/42 14} 27. g3 {3.27/43 36} Rc5 {1.09/43 12} 28. Nxe3 {
2.99/46 37} f4 {1.01/40 26} 29. Rxf4 {3.03/50 102} Rxh5 {1.13/41 13} 30. gxh4 {
2.93/50 14} Qe7 {1.07/38 13} 31. Qd1 {3.06/49 35 (Va3)} Rxh4 {1.35/44 66} 32.
Nd5 {3.00/46 18} Qg5 {1.36/43 22} 33. Rxh4 {3.12/46 30} Qxh4 {1.62/44 25} 34.
Qf3 {3.09/46 18} Re5 {1.60/43 21} 35. Nc3 {3.10/47 25 (Vg1)} Rh5 {1.20/37 13}
36. Qg2 {3.10/47 45} Qf4 {1.33/37 16 (Vf5)} 37. Rg1 {3.42/38 16} Qe5 {1.44/40
18} 38. Rf1 {3.25/39 23} Kg8 {1.57/36 14} 39. Rf3 {3.24/39 26} Rg5 {1.54/37 15
(d5)} 40. Qf1 {3.39/39 25 (Df2)} Qe8 {1.52/36 9} 41. Rg3 {3.70/35 18} Re5 {1.
59/35 14 (Vc5)} 42. Qf2 {3.66/36 20} Qc6+ {1.52/36 20} 43. Kg1 {3.67/37 17} Qc4
{1.58/36 15 (Dc7)} 44. h3 {3.79/37 31} Qe6 {1.62/38 48} 45. Qd4 {3.75/42 49}
Qf6 {1.62/34 6 (Df5)} 46. Rg4 {3.68/40 22 (Kg2)} Kh8 {1.69/36 8 (Df3)} 47. Kg2
{3.79/36 13} Qe6 {1.69/35 6} 48. Qc4 {3.87/37 22} Qe7 {1.69/32 6 (De8)} 49. Rf4
{4.00/35 36} Qg5+ {1.76/34 10 (d5)} 50. Kh2 {4.31/36 14} Kh7 {1.77/43 7} 51.
Qd3+ {4.22/35 11} Kg8 {1.77/41 7} 52. Qd4 {4.03/38 21} Kh7 {1.83/38 15 (Ve6)}
53. Re4 {4.55/34 11} Rf5 {1.65/38 8} 54. Rg4 {4.58/34 6} Qe7 {1.86/39 10} 55.
Rf4 {4.59/34 7} Re5 {1.85/37 10 (Vc5)} 56. Qd3+ {4.62/34 12} Kh8 {1.67/32 6}
57. Rf5 {4.64/35 10 (Dd4)} Qe8 {1.76/34 10} 58. Rxe5 {4.73/36 7 (Vf3)} Qxe5+ {
1.92/33 13} 59. Kh1 {4.71/39 7 (Kg2)} Qf4 {1.89/36 27 (Dg5)} 60. b3 {4.87/33
10 (Kg2)} Qc1+ {2.03/34 13} 61. Kg2 {4.91/35 8} Qb2+ {2.11/32 9} 62. Ne2 {5.03/
34 10} Qe5 {2.11/36 6} 63. Nc3 {5.01/34 8} Qg5+ {2.19/36 28} 64. Kf2 {5.17/31
10 (Kf1)} Qe5 {2.19/37 4 (Df4+)} 65. Qe3 {5.28/31 9 (Kf1)} Qh2+ {2.27/34 14
(Df6+)} 66. Ke1 {5.40/33 9} Qb2 {2.19/34 19 (Dh1+)} 67. Ne2 {5.38/32 8} Qb1+ {
2.56/33 26} 68. Kf2 {5.61/32 9} Qf5+ {2.57/29 8} 69. Qf3 {5.66/35 8} Qc5+ {2.
63/30 4} 70. Ke1 {5.79/34 8} Qb4+ {2.64/29 2 (Dc7)} 71. Kd1 {5.97/34 14} Kh7 {
2.87/33 22 (Db6)} 72. Qd5 {6.14/33 10 (Kc2)} Kh8 {2.86/30 7 (b6)} 73. Kc2 {6.
09/34 15 (Df3)} Qe1 {2.84/28 2 (Db6)} 74. Qf3 {6.25/37 8} Kh7 {3.09/29 8 (d5)}
75. Nf4 {6.31/27 8 (Jc3)} Kh8 {3.09/33 9 (Db4)} 76. Qxb7 {6.85/29 12 (Je2)}
Qf2+ {3.56/30 15} 77. Kd3 {6.89/30 9} Qf1+ {3.98/32 30 (Dg3+)} 78. Ne2 {6.97/
31 9} Qf5+ {3.97/28 3} 79. Kd2 {7.47/32 24} Qg5+ {4.18/24 2} 80. Kd1 {7.62/30
10} Kh7 {4.42/30 12 (h5)} 81. Qb5 {7.76/43 7 (De4+)} Qe3 {4.48/30 2} 82. Qxa5 {
7.86/41 7 (Df5+)} Qxb3+ {3.94/23 8} 83. Kd2 {7.85/47 20} Qa2+ {6.58/27 32} 84.
Kd3 {7.88/46 7} Qb1+ {6.50/24 5} 85. Ke3 {7.89/49 10} Qa2 {7.43/28 19 (d5)} 86.
Qf5+ {8.05/40 7 (Db5)} g6 {6.17/22 2} 87. Nc3 {8.12/42 8 (Dd7+)} Qg2 {7.66/24 7
} 88. Qf7+ {8.47/44 15 (Dd7+)} Kh8 {7.75/24 1} 89. Qf3 {8.72/49 18 (De8+)} Qg5+
{6.62/24 7} 90. Kd3 {8.86/24 5 (Df4)} 1-0
[/pgn]

[dkappe]

It doesn't seem that much weaker. It was only twelve games against Stockfish 14.1, but the result is 0:1=11. Better than Ethereal and maybe even Dragon would have achieved.
Why can't this be used with 14, has the network format changed?

Well, if you wait a while, Stefan Pohl will appear and tell you how you can only test with openings that never appeared in the 1m human games. It stands to reason that Harmon would perform better on “normal” openings. Curious how it would do in Rebel’s gambit world.

The SF net architecture has changed twice since SF13. Not sure if the newer SF can still work with the old nets.

[Rubinus]

Well, if you wait a while, Stefan Pohl will appear and tell you how you can only test with openings that never appeared in the 1m human games. It stands to reason that Harmon would perform better on “normal” openings. Curious how it would do in Rebel’s gambit world.

The SF net architecture has changed twice since SF13. Not sure if the newer SF can still work with the old nets.

It's just that I don't so much test engines as I do my, as yet private, library for the king gambit. And I use a variety of engines to do it. Because the old games, though many of them nice, are basically ballast - at least in terms of the best possible library for a chess engine.

[dkappe]

It's just that I don't so much test engines as I do my, as yet private, library for the king gambit. And I use a variety of engines to do it. Because the old games, though many of them nice, are basically ballast - at least in terms of the best possible library for a chess engine.

Sounds interesting. Is this for human consumption?