Eelco de Groot wrote: ↑Wed May 22, 2019 8:13 pm
An easy way to bring down the losses would be to include an opening book. You'd just copy Brainfish and maybe hide the the immediate moves.
Nah, Eman doesn't have a need for that, as it has a built-in "experience file", that learns from games as it plays. It would be trivial for the author to inflate Eman's results by just prividing it with one that has already played plenty of games.
But Eman is distributed with an empty file, and no book (it also implements book learning... and double bin book support, at that!), so all the praise comes from Eman's search and evaluation (the author explains it on
his site, which... truly, it sounds like snake oil, until you use it.)
It plays very differently than all other Stockfish derivatives, and while I haven't seen yet the aggressivity that is being talked about (maybe because I have the slowest hardware in the entire site?) it seems its "traversing of the search tree" is allowing to play very solid moves very fast. Maybe it's indeed because it's finding the opponent's winning moves "from a mile away" with its "Coherence" evaluation, and avoiding playing into them.
I just drew this game against the strongest hardware I've ever faced:
White: Eman 3.72 64-bit : 2268 kN/s 1 Processors with 4 Cores, Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz, 3 GB of RAM - Running InfinityChess (4.0.2.64)
Black: stockfish_19050219_x64_modern_CTG : 66851 kN/s 1 Processors with 32 Cores, AMD Ryzen Threadripper 2990WX 32-Core Processor, 31.87 GB of RAM - Running InfinityChess (4.0.2.64)
[pgn][Event "Rapid 12' + 2'', Rated, Friday 24 May Rapid Engine Tour 12+2 Rounds 11 Starts 1"]
[Site "InfinityChess"]
[Date "2019.05.24 "]
[Round "7"]
[White "Vytron"]
[Black "MR-AJ"]
[WhiteElo "2303"]
[BlackElo "2414"]
[ECO "D43"]
[Result "1/2-1/2"]
1. d4 d5 2. c4 e6 3. Nc3 Nf6 4. Nf3 c6 5. Bg5 h6 6. Bxf6
Qxf6 7. e3 Nd7 8. Bd3 dxc4 9. Bxc4 Bd6 10. O-O Qe7 11. Ne4
O-O 12. Nxd6 Qxd6 13. Qe2 b6 14. Rfd1 Rd8 15. Rac1 a5
16. Bd3 Bb7 17. Nd2 Rac8 18. Ne4 Qb8 19. h3 c5 20. dxc5
Bxe4 21. Bxe4 Rxc5 22. a3 g6 23. Rxc5 Nxc5 24. Bf3 Rxd1+
25. Qxd1 Qc7 26. g3 Qd7 27. Qxd7 Nxd7 28. Kf1 Kf8 29. Ke1
Ke7 30. Kd2 Kd6 31. Kc3 Kc5 32. b4+ axb4+ 33. axb4+ Kd6
34. g4 f5 35. gxf5 exf5 36. h4 Ne5 37. Bd1 g5 38. hxg5 hxg5
39. Kd4 g4 40. Be2 Ke6 41. Bd1 Nc6+ 42. Kc4 Ne5+ 43. Kd4
Nc6+ 44. Kc4 Kd6 45. Kb5 Kc7 46. Be2 Ne5 47. Ka6 Kc6
48. Bf1 Kc7 49. Be2 Kc6 50. Bb5+ Kd5 51. Kxb6 Ke4 52. Kc7
Kf3 53. Kd6 Nf7+ 54. Ke6 Kxf2 55. Kxf7 Kxe3 56. Bf1 Kf2
57. b5 Kxf1 58. b6 g3 1/2-1/2[/pgn]
InfinityChess doesn't let you copy Depths, but I was reaching some Depths 30s against some Depth 54s and holding my own.
Here's me beating someone that had 16Cores.
[pgn][Event "Rapid 12' + 2'', Rated, Thursday 23 May Rapid Engine Tour 12+2 Rounds 11 Starts"]
[Site "InfinityChess"]
[Date "2019.05.24 "]
[Round "8"]
[White "Vytron"]
[Black "CrapCleaner"]
[WhiteElo "2262"]
[BlackElo "2437"]
[ECO "E10"]
[Result "1-0"]
1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Nc3 Be7 5. Bf4 O-O 6. e3
Nbd7 7. c5 c6 8. Bd3 b6 9. b4 a5 10. a3 Ba6 11. O-O Qc8
12. h3 Qb7 13. Qc2 Bxd3 14. Qxd3 axb4 15. axb4 Rxa1
16. Rxa1 Ra8 17. Qd1 h6 18. Rxa8+ Qxa8 19. Qa4 Qxa4
20. Nxa4 Bd8 21. Nxb6 Bxb6 22. cxb6 Nxb6 23. Ne5 Nc4
24. Nxc6 Ne4 25. g4 Kf8 26. Ne5 g5 27. Bh2 f6 28. Nd3 Ke7
29. Kf1 Nc3 30. Ke1 Kf7 31. Bb8 Kf8 32. Nc1 Kf7 33. Ba7 Nb5
34. Bc5 Kg6 35. Ne2 h5 36. gxh5+ Kxh5 37. Be7 Kg6 38. Kd1
Kf7 39. Bc5 e5 40. Kc2 Nbd6 41. Nc3 Ke6 42. b5 exd4
43. exd4 Nb7 44. b6 f5 45. Ne2 Nd8 46. f4 g4 47. h4 Ne3+
48. Kd2 Nf1+ 49. Ke1 Ne3 50. h5 Nb7 51. Kf2 Nd1+ 52. Kg3
Kf7 53. Kh4 Kf6 54. Bf8 Ne3 1-0[/pgn]
I believe there's no way I could have done that with any other engine.
So what's my reference point? 2013, back then I had this same hardware, and was using the strongest thing of that time (Ii always use the strongest thing
), and everyone was kicking my ass, so badly, that I quit.
I had the same hardware, but after 6 years these guys have improved in hardware and software in a way that is hard to grasp. I'm still impressed with reading 66000kN/s (even though Leelas play as strong with 15kN/s...) it only makes sense that my results only get worse against them, much worse, but instead, I can draw, and beat them...
And who are the people I struggle with, that beat me the most? People using Eman themselves
People continue to say Eman is a very strong chess engine, what if it's the stongest one out there yet nobody talks about it?
Your beliefs create your reality, so be careful what you wish for.