Zevra v2.0 r146 beta

[Sergio Martinez]

40 moves in 2 minutes.

there are time loses. Zevra spent all time without move (appears engine crashes)

I make update binaries, you can check it

Thank you very much

[sovaz1997]

Working?

[Kotlov]

код стал почище, да

[sovaz1997]

код стал почище, да

Правильно вы сказали - лучше удалить все эвристики и настроить по-новой. Я сейчас стараюсь не добавлять множество различных вещей, а концентрируюсь всего на нескольких и пытаюсь улучшать их. Также, сортировка ходов: крайне важная вещь, без нее вообще добрая половина эвристик работать не будет))

По сути, в движке есть сейчас LMR, Null Move, Futility pruning и aspiration window (окно стремления). Само собой, для сортировки киллеры (их 4), история. Продления на шахах, конечно. SEE тоже добавил и использую для того, чтобы плохие взятия рассматривать в самом конце, а в поиске покоя вообще отбрасывать. Оценочная функция простая, улучшать ее пока не планирую, т. к. много ЭЛО это не даст. Автоматический тюнинг - возможно. Ну и цель такая - увеличивать силу игры без сильного увеличения кода.

[Sergio Martinez]

Working?

It seems that now it works well

[sovaz1997]

Zevra v2.0 r172 Gauntlet 60s+0.6s 64mb: https://sites.google.com/view/scet-testing/zevra

PGN: https://drive.google.com/drive/folders/ ... sp=sharing

Results:

Code: Select all

 # PLAYER                      :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W    D     L  D(%)  OppAvg
 1 GreKo 2018.08               :  2814.9   32.5  1606.0    1700    94     100  1552  108    40     6  2304.1
 2 Asymptote v0.3              :  2542.1   43.6   195.5     300    65     100   166   59    75    20  2431.6
 3 Zevra v2.0 r172             :  2431.6   24.0   929.5    1800    52      90   799  261   740    15  2430.8
 4 Galjoen 0.39.2              :  2415.6   17.9  1104.0    1700    65     100   938  332   430    20  2304.1
 5 Isa 2.0.64                  :  2346.5   43.5   114.5     300    38      97    86   57   157    19  2431.6
 6 Zevra v1.8.6 r672 popcnt    :  2303.5   16.5  1715.0    3700    46     100  1174  682  1644    19  2390.0
 7 CT800 V1.31 64 bit          :  2266.2   17.6   763.5    1700    45      86   595  337   768    20  2304.1
 8 Baislicka                   :  2253.4   18.3   654.0    1400    47      64   501  306   593    22  2276.8
 9 Zevra v1.8.5 r664           :  2250.0   ----  1304.0    3500    37     100   981  646  1873    18  2390.0
10 Hedgehog 1.9 64-bit         :  2199.8   18.2   614.0    1700    36     ---   419  390   891    23  2304.1

White advantage = 28.96 +/- 3.64
Draw rate (equal opponents) = 21.79 % +/- 0.50

Code: Select all

 1) Zevra v2.0 r172          2431.6 :   1800 (+799,=261,-740),  51.6 %

   vs.                             :  games (   +,   =,   -),   (%) :    Diff,    SD, CFS (%)
   GreKo 2018.08                   :    300 (  16,  22, 262),   9.0 :  -383.3,  17.3,    0.0
   Asymptote v0.3                  :    300 (  75,  59, 166),  34.8 :  -110.5,  18.3,    0.0
   Galjoen 0.39.2                  :    300 ( 141,  43, 116),  54.2 :   +16.0,  12.3,   90.4
   Isa 2.0.64                      :    300 ( 157,  57,  86),  61.8 :   +85.1,  18.4,  100.0
   CT800 V1.31 64 bit              :    300 ( 199,  33,  68),  71.8 :  +165.5,  12.4,  100.0
   Hedgehog 1.9 64-bit             :    300 ( 211,  47,  42),  78.2 :  +231.8,  12.8,  100.0
 2) Zevra v1.8.6 r672 popcnt 2303.4 :   3500 (+1174,=682,-1644),  43.3 %

   vs.                             :  games (    +,   =,    -),   (%) :    Diff,    SD, CFS (%)
   GreKo 2018.08                   :    700 (   12,  48,  640),   5.1 :  -501.0,  18.7,    0.0
   Galjoen 0.39.2                  :    700 (  162, 155,  383),  34.2 :  -115.3,   9.7,    0.0
   CT800 V1.31 64 bit              :    700 (  313, 142,  245),  54.9 :   +37.4,   9.4,  100.0
   Baislicka                       :    700 (  319, 157,  224),  56.8 :   +50.0,   9.2,  100.0
   Hedgehog 1.9 64-bit             :    700 (  368, 180,  152),  65.4 :  +104.7,   9.6,  100.0

3) Zevra v1.8.5 r664        2250.0 :   3500 (+981,=646,-1873),  37.3 %

   vs.                             :  games (   +,   =,    -),   (%) :    Diff,    SD, CFS (%)
   GreKo 2018.08                   :    700 (  12,  38,  650),   4.4 :  -554.4,  18.7,    0.0
   Galjoen 0.39.2                  :    700 ( 127, 134,  439),  27.7 :  -168.8,   9.7,    0.0
   CT800 V1.31 64 bit              :    700 ( 256, 162,  282),  48.1 :   -16.1,   9.0,    3.7
   Baislicka                       :    700 ( 274, 149,  277),  49.8 :    -3.4,   9.6,   36.2
   Hedgehog 1.9 64-bit             :    700 ( 312, 163,  225),  56.2 :   +51.2,   9.7,  100.0

[Guenther]

40 moves in 2 minutes.

there are time loses. Zevra spent all time without move (appears engine crashes)

I make update binaries, you can check it

Oleg, thanks for the new release and especially for the no popcount binary.

[tpoppins]

Оценочная функция простая, улучшать ее пока не планирую, т. к. много ЭЛО это не даст.

Pardon me for butting in, Oleg, but that should be "Эло", not "ЭЛО", since it's a person's name rather than an acronym.

Is r172 the final release or, barring unexpected discovery of some bug, do you expect further updates in the nearest future? I would rather not start a test only to abort it the next day because a new version came out.

[sovaz1997]

Hello! No, there is no new version in the plans (a new one will appear after a significant increase). 2 people before the upgrade had an error due to which the engine could crashed. Now, all of this version works stably on both Linux and Windows. At the moment, no errors were found.

It's just funny historical game (very old Zevra plays it):

[pgn] [Event "?"] [Site "?"] [Date "2017.02.24"] [Round "?"] [White "zevra"] [Black "Zevra_v1_3_1_r340_popcnt"] [Result "1-0"] [ECO "B00"] [Opening "KP"] [PlyCount "139"] [TimeControl "1/move"] [Variation "Nimzovich defense"] 1. e4 {0.00/8 0.91s} Nc6 {-0.03/8 0.91s} 2. Nf3 {+0.36/7 0.91s} e5 {-0.02/7 0.91s} 3. Nc3 {+0.51/7 0.91s} Nf6 {-0.01/7 0.91s} 4. Bc4 {+0.06/7 0.91s} Nxe4 {+0.02/7 0.91s} 5. Nxe4 {+0.41/7 0.91s} d5 {0.00/7 0.91s} 6. Bxd5 {+0.42/7 0.91s} Qxd5 {+0.11/7 0.91s} 7. d3 {-0.19/6 0.91s} Bb4+ {+0.11/6 0.91s} 8. c3 {-0.28/6 0.91s} f5 {+0.12/7 0.91s} 9. Neg5 {-0.19/7 0.91s} Bc5 {+0.11/6 0.91s} 10. b4 {-0.28/6 0.91s} Be7 {+0.12/6 0.91s} 11. O-O {-0.30/6 0.91s} h6 {+0.15/6 0.91s} 12. c4 {-0.51/6 0.91s} Qd6 {+1.06/7 0.91s} 13. Nh3 {-0.79/7 0.91s} Nxb4 {+1.09/7 0.91s} 14. Ba3 {-0.81/7 0.91s} c5 {+1.12/7 0.91s} 15. Bxb4 {-0.77/7 0.91s} cxb4 {+0.21/7 0.91s} 16. Re1 {-0.27/7 0.91s} Bf6 {+0.21/7 0.91s} 17. Nxe5 {-0.63/7 0.91s} O-O {+0.23/7 0.91s} 18. Qb3 {-0.63/7 0.91s} Be6 {+0.24/7 0.91s} 19. Qb2 {-0.48/7 0.91s} Rfe8 {+0.22/7 0.91s} 20. f4 {-0.46/7 0.91s} Rad8 {+0.21/6 0.91s} 21. Rad1 {-0.45/7 0.91s} Qd4+ {+0.22/6 0.91s} 22. Qxd4 {-0.47/7 0.91s} Rxd4 {+0.22/8 0.91s} 23. g3 {-0.44/7 0.91s} Rdd8 {+0.22/7 0.91s} 24. Nf2 {-0.42/7 0.91s} Kh7 {+0.22/7 0.91s} 25. h4 {-0.20/7 0.91s} Kg8 {+0.22/7 0.91s} 26. h5 {-0.22/7 0.91s} Kh7 {+0.22/7 0.91s} 27. Ra1 {-0.14/7 0.91s} Rd4 {+0.22/7 0.91s} 28. a3 {+0.23/7 0.91s} bxa3 {+0.21/8 0.91s} 29. Rxa3 {+0.17/7 0.91s} a6 {+0.21/7 0.91s} 30. Raa1 {+0.14/7 0.91s} Rd6 {+0.21/7 0.91s} 31. Rab1 {+0.20/7 0.91s} Bc8 {+0.21/7 0.91s} 32. Nf7 {+0.25/7 0.91s} Rxe1+ {+0.21/8 0.91s} 33. Rxe1 {+0.21/8 0.91s} Re6 {+0.20/9 0.91s} 34. Rxe6 {+0.39/9 0.91s} Bxe6 {+0.20/10 0.91s} 35. Nd6 {+0.21/10 0.91s} Be7 {+0.20/9 0.91s} 36. Nxb7 {+0.23/10 0.91s} Bf7 {+0.18/10 0.91s} 37. Na5 {+0.33/9 0.91s} Be8 {+0.19/10 0.91s} 38. d4 {+0.33/9 0.91s} Bf6 {+0.18/9 0.91s} 39. d5 {+0.26/10 0.91s} Bxh5 {+0.17/9 0.91s} 40. Nd3 {+0.47/9 0.91s} Be2 {+0.17/9 0.91s} 41. Ne5 {+0.41/10 0.91s} Be7 {+0.14/9 0.91s} 42. Nb7 {+1.46/9 0.91s} Bf6 {-2.06/9 0.91s} 43. d6 {+1.60/9 0.91s} a5 {-2.07/9 0.91s} 44. Nxa5 {+2.52/9 0.91s} Bd1 {-3.06/9 0.91s} 45. d7 {+3.19/9 0.91s} Bc2 {-3.07/9 0.91s} 46. Nac6 {+3.32/9 0.91s} Bb1 {-3.12/10 0.91s} 47. d8=R {+3.61/9 0.91s} Bxd8 {-3.14/10 0.91s} 48. Nxd8 {+4.29/10 0.91s} Bc2 {-3.19/10 0.91s} 49. c5 {+4.55/10 0.91s} Ba4 {-4.18/10 0.91s} 50. Ne6 {+5.28/9 0.91s} Bb3 {-4.15/10 0.91s} 51. Nd4 {+5.45/10 0.91s} Bd5 {-5.37/10 0.91s} 52. Nxf5 {+5.84/9 0.91s} Ba2 {-4.37/9 0.91s} 53. c6 {+6.41/9 0.91s} Be6 {-6.42/10 0.91s} 54. Ne7 {+6.74/9 0.91s} g6 {-7.37/10 0.91s} 55. N5xg6 {+8.01/9 0.91s} Kg7 {-7.41/10 0.91s} 56. f5 {+7.94/10 0.91s} Bc4 {-7.43/10 0.91s} 57. c7 {+8.21/10 0.91s} Ba6 {-7.43/10 0.91s} 58. c8=N {+8.34/9 0.91s} Bd3 {-7.35/9 0.91s} 59. Nd6 {+8.44/9 0.91s} Bb1 {-7.35/8 0.91s} 60. Nf4 {+8.78/9 0.91s} Ba2 {-7.39/8 0.91s} 61. Nh5+ {+9.26/9 0.91s} Kh8 {-8.34/8 0.91s} 62. f6 {+9.84/9 0.91s} Kh7 {-9.62/9 0.91s} 63. Ng7 {+10.24/9 0.91s} h5 {-9.61/9 0.91s} 64. Ngf5 {+10.80/9 0.91s} Be6 {-10.62/9 0.91s} 65. f7 {+11.82/9 0.91s} Bxf7 {-10.63/9 0.91s} 66. Nxf7 {+11.82/9 0.91s} h4 {-M12/9 0.61s} 67. g4 {+M9/9 0.91s} h3 {-M10/7 0.090s} 68. g5 {+M5/9 0.91s} h2+ {-M6/4 0.052s} 69. Kf2 {+M3/11 0.91s} h1=Q {-M4/3 0.051s} 70. g6# {+M1/10 0.91s, White mates} 1-0 [/pgn]

[tpoppins]

What a fine game! Until around move 40 I thought Black was going to win with his two bishops. And what an end, too. You don't often see a chess engine with a sense of humor.

Thank you for sharing.