Testposition for Glaurung

[Eelco de Groot]

[d]8/2k5/1p5R/p7/3K1PP1/8/r6P/8 w - -

In this postion advancing pawn g4 should win I think, but Glaurung prefers to push h2-h4. In spite of astronomical evaluations which are entirely my own fault, Glaurung Mjolnir managed to actually lose from this position It was such a promising game up to this point

I'm trying to improve the passed pawn eval here a bit but it is not yet correct. The weight for passed pawns was 125 which may be over the top also.


[Event "12 Minutes/Game + 6 Seconds/Move"]
[Site "Engine Match"]
[Date "2008.04.26"]
[Round "2"]
[White "Glaurung 080422 Mjolnir 004 Build 3"]
[Black "Strelka 2.0 B"]
[Result "0-1"]

1. c4 {book 0s} e5 {book 0s} 2. Nc3 {book 0s} Nc6 {book 0s}
3. g3 {book 0s} g6 {book 0s} 4. Bg2 {book 0s} Bg7 {book 0s}
5. e4 {book 0s} d6 {book 0s} 6. d3 {book 0s} f5 {book 0s}
7. Nf3 {-0.29/14 1:04m} Nf6 {-0.26/15 21s} 8. O-O {-0.29/14
29s} O-O {-0.27/15 34s} 9. Qb3 {+0.03/14 36s} fxe4
{-0.29/15 57s} 10. dxe4 {0.00/15 29s} Na5 {-0.32/15 56s}
11. Qa4 {-0.13/14 33s} b6 {-0.16/14 21s} 12. Nd5 {-0.15/14
1:17m} Bg4 {-0.12/13 18s} 13. Bg5 {-0.17/14 21s} Qd7
{-0.15/15 1:21m} 14. Qxd7 {+0.03/15 23s} Bxd7 {-0.12/16
31s} 15. Nxc7 {-0.17/14 33s} Rac8 {0.00/16 13s} 16. Nd5
{-0.03/16 28s} Rxc4 {-0.02/15 8s}
{[d]5rk1/p2b2bp/1p1p1np1/n2Np1B1/2r1P3/5NP1/PP3PBP/R4RK1 w - - }

17. b3 { ! }{The present version plays this also but does not see that this is a really good move

5rk1/p2b2bp/1p1p1np1/n2Np1B1/2r1P3/5NP1/PP3PBP/R4RK1 w - -

Engine: Glaurung 080422 Mjolnir 004 Build 6 (64 MB, Athlon 2009 MHz)
by Tord Romstad

11.01 0:02 -0.43 17.Nxf6+ Bxf6 18.Rad1 Bxg5 19.Nxg5 Rc2
20.f3 Rxb2 21.Rxd6 Bb5 22.Rf2 Nc4
23.Rxb2 Nxd6 (694.598) 284

11.03 0:04 +0.41 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bc6
20.Rxd6 Kg7 21.Rc1 Re2 22.Be3 e4
23.Nd4 (1.331.106) 291

12.01 0:06 +0.50 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bc6
20.Rxd6 Kg7 21.Rc1 Re2 22.Be3 Bxf3
23.Bxf3 Rxa2 24.Rd7+ Rf7 (1.992.184) 287

13.01 0:10 +0.05 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bxg5
20.Rxd6 Bh3 21.Nxg5 Bxg2 22.Kxg2 Re2
23.Nh3 Rf7 24.Rd5 Nc6 (3.087.621) 286

14.01 0:20 +0.03 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bxg5
20.Rxd6 Bh3 21.Nxg5 Bxg2 22.Kxg2 Re2
23.Nh3 Rf7 24.Rd5 Kg7 25.Rd8 (5.830.034) 284

15.01 0:43 -0.05 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bxg5
20.Rxd6 Bh3 21.Nxg5 Bxg2 22.Kxg2 Re2
23.Nh3 Rf7 24.Rd5 Kg7 25.Rd8 Rc2 (12.511.636) 285

16.01 1:32 -0.07 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bxg5
20.Rxd6 Bh3 21.Nxg5 Bxg2 22.Kxg2 Re2
23.Nh3 Rf7 24.Rd5 Kg7 25.Rd3 e4
26.Rd8 (26.687.298) 288

17.01 3:09 +0.07 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bxg5
20.Rxd6 Bh3 21.Nxg5 Bxg2 22.Kxg2 Re2
23.Nh3 Rf7 24.Re6 Kg7 25.Kf1 Rc2
26.Rxe5 Nc6 (55.193.464) 291

18.01 6:28 +0.72 17.b3 Rxe4 18.Nxf6+ Bxf6 19.Rfd1 Bxg5
20.Rxd6 Bg4 21.Nxg5 Rd4 22.Bd5+ Kg7
23.Rc1 Rc8 24.Rxc8 Bxc8 25.Rd8 Bf5
26.Rg8+ Kf6 27.Nxh7+ Ke7 28.Rg7+ Kd6
29.Nf6 Rxd5 30.Nxd5 (113.399.127) 291

best move: b2-b3 time: 8:43.516 min n/s: 291.363 nodes: 152.520.000 }
{+0.50/15 22s} 17... Rxe4 {-0.47/15 11s} 18. Nxf6+ {+0.62/15 14s} Bxf6 {-0.50/17 18s} 19. Rfd1 {+0.82/16 20s} Bxg5 {-0.53/16 13s} 20. Rxd6 {+0.68/16 16s} Bf5 {-0.76/16 36s} 21. Nxg5 {+0.76/16 19s}
Rd4 {-0.70/17 25s} 22. Rxd4 {+0.78/16 46s} exd4 {-0.77/17
5s} 23. b4 {+0.74/16 20s} h6 {-0.54/17 4s} 24. Nf3
{+0.45/17 18s} Nc6 {-0.70/19 17s} 25. b5 {+0.78/17 12s} Bg4
{-1.18/18 21s} 26. bxc6 {+1.43/18 11s} Bxf3 {-1.31/18 4s}
27. Bxf3 {+1.54/20 14s} Rxf3 {-1.19/19 4s} 28. Rc1
{+1.64/20 12s} Rf8 {-1.05/20 16s} 29. Kf1 {+1.64/19 10s}
Rc8 {-1.25/20 21s} 30. c7 {+1.98/20 15s} Kf8 {-1.32/20 18s}
31. Ke2 {+1.54/18 12s} Ke8 {-1.86/21 2:04m} 32. Rc6
{+2.23/17 11s} h5 {-1.89/20 27s} 33. f4 {+2.90/17 12s} Kd7
{-1.80/19 9s} 34. Rxg6 {+2.86/18 14s} Kxc7 {-1.90/18 16s}
35. Kd3 {+3.15/16 14s} a5 {-1.80/17 10s} 36. Rh6 {+3.49/16
12s} Re8 {-1.90/17 15s} 37. Rxh5 {+3.45/16 15s} Re3+
{-1.74/16 6s} 38. Kxd4 {+3.47/17 13s} Ra3 {-1.72/16 7s}
39. g4 {+3.47/15 25s} Rxa2 {-1.49/14 5s} 40. Rh7+ {+3.47/14
12s} (40. Rb5) 40... Kd6 {-1.48/15 8s} 41. Rh6+ {+3.43/14
14s} Kc7 {-1.39/14 2s} 42. h4 {+3.37/14 18s} a4 {-1.47/15
9s} 43. Rh7+ {+3.41/13 7s} Kb8 {-1.41/16 10s} 44. Re7
{+3.31/14 26s} Rf2 {-1.30/16 8s} 45. f5 {+3.11/14 9s} a3
{-1.60/15 8s} 46. Re1 {+3.19/14 24s} Kc7 {-1.92/15 12s}
47. Ke5 {+3.37/14 10s} a2 {-1.36/14 6s} 48. Rc1+ {+2.29/14
17s} Kd7 {-1.64/16 7s} 49. h5 {+2.09/14 11s} b5 {0.00/15
6s} 50. g5 {+0.84/13 14s} b4 {+3.49/15 15s} 51. Rd1+
{-0.56/12 7s} Kc7 {+5.04/16 11s} 52. Rc1+ {-1.54/13 7s} Kb6
{+5.84/15 14s} 53. h6 {-2.31/13 7s} b3 {+5.85/14 6s} 54. h7
{-6.19/12 12s} b2 {+7.37/14 17s} 55. h8=Q {-5.60/12 5s}
Re2+ {+7.20/14 17s} 56. Kd5 {-10.15/13 20s} bxc1=Q
{+9.19/14 7s} 57. Qb8+ {-9.09/13 8s} Ka5 {+10.76/14 7s}
58. Qa7+ {-10.50/13 7s} Kb4 {+12.46/14 13s} 59. Qb6+
{-10.50/13 8s} Kc3 {+13.20/14 15s} 60. Qc5+ {-12.41/12 6s}
Kd2 {+15.46/14 44s} 61. Qd4+ {-15.47/12 7s} Kc2 {+16.19/14
37s} 62. Qa4+ {-17.82/12 14s} Kb1 {+16.19/14 12s} 63. Qb3+
{-17.82/12 7s} Ka1 {+16.32/14 7s} 64. g6 {-80.26/12 12s}
Rd2+ {+M12/12 1s} 65. Ke5 {-80.26/12 7s} Qb2+ {+M11/11 0s}
66. Qxb2+ {-80.31/11 0s} Kxb2 {+M10/12 0s} 67. g7
{-86.66/13 7s} a1=Q {+M9/11 0s} 68. g8=Q {-93.11/12 4s}
Kc2+ {+M8/9 0s} 69. Ke6 {-M7/13 2s} Qa2+ {+M7/8 0s} 70. Kf6
{-M6/9 0s} Qxg8 {+M5/6 0s} 71. Ke7 {-M5/7 0s} Qg7+ {+M3/3
0s} 72. Ke6 {-M3/8 0s} Rd4 {+M2/2 0s} 73. f6 {-M2/6 0s}
Qd7+ {+M2/1 0s} 74. Ke5 {-M1/6 0s} Qd5# {+M1/1 0s} 0-1

Strelka output:


8/2k5/1p5R/p7/3K1PP1/8/r6P/8 w - -

Engine: Strelka 2.0 B (64 MB)
by Vasik Rajlich, Jury Osipov

8.00 0:00 +1.56 42.g5 a4 43.g6 a3 44.Rh7+ Kd6 45.g7 Rg2
46.h4 a2 47.Rh6+ Ke7 (58.321) 1240

9.00 0:00 +1.57 42.h4 a4 43.Rh7+ Kd6 44.h5 Rb2
45.Rh6+ Ke7 46.g5 a3 47.g6 Rd2+
48.Kc4 (147.767) 859

10.00 0:00 +1.51 42.g5 a4 43.g6 a3 44.g7 Rd2+ 45.Kc4 a2
46.g8Q a1Q 47.Rh7+ Rd7 48.Rxd7+ Kxd7
49.Qf7+ Kc6 50.Qe6+ Kc7 (290.424) 1029

11.22 0:00 +1.44 42.g5 a4 43.g6 a3 44.Rh7+ Kd6 45.Ra7 Rg2
46.Rxa3 Rxg6 47.h4 Rg8 (565.336) 881

12.22 0:01 +1.43 42.g5 a4 43.g6 a3 44.g7 Rd2+ 45.Kc4 a2
46.g8Q a1Q 47.Rh7+ Rd7 48.Rxd7+ Kxd7
49.Qf7+ Kd6 50.Qg6+ Kd7 51.Qh7+ Kd6
52.Qd3+ Ke6 53.Qd5+ Kf6 (1.162.963) 791

13.22 0:04 +1.45 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.g6 Rg2
46.Rxa3 Rxg6 47.Rb3 Ke7 48.Re3+ Kf7
49.f5 Rd6+ 50.Ke5 (3.006.584) 737

14.22 0:06 +1.50 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.g6 Rg2
46.Rxa3 Rxg6 47.Rb3 Kd7 48.f5 Rd6+
49.Ke5 Ke7 (5.028.194) 729

15.22 0:12 +1.48 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.g6 Rg2
46.Rxa3 Rxg6 47.Rb3 Ke7 48.f5 Rh6
49.Ke5 b5 50.Rxb5 Rxh2 (8.925.281) 726

16.22 0:23 +1.48 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.g6 Rg2
46.Rxa3 Rxg6 47.Rb3 Ke7 48.f5 Rh6
49.Ke5 b5 50.Rxb5 Rxh2 (17.146.209) 735

17.22 0:50 +1.73 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.h4 Rf2
46.Ke4 a2 47.g6 Rh2 48.g7 Rg2
49.Rxa2 Rxg7 50.Rd2+ Kc5 51.h5 Re7+
52.Kf5 b5 (37.451.075) 743

18.22 2:19 +1.98 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 Rf2
46.Ke3 Rc2 47.Rb7 Kc6 48.Ra7 Kd5
49.g6 Rc3+ 50.Kd2 Rc8 51.g7 Ke6
52.h6 (103.968.721) 746

19.22 4:16 +2.09 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 Rf2
46.Ke3 Rc2 47.Rb7 Kc6 48.Ra7 Kd5
49.h6 Rc3+ 50.Kd2 Rc8 51.h7 Ke6
52.Rc7 Rf8 (192.132.525) 748

20.22 12:03 +2.31 42.g5 a4 43.Rh7+ Kd6 44.h4 Ke6 45.g6 Rf2
46.Ke4 Re2+ 47.Kf3 Re1 48.Rb7 Rf1+
49.Ke3 a3 50.Rxb6+ Kf5 51.Ra6 Re1+
52.Kd4 a2 53.Rxa2 Rd1+ 54.Ke3 Kxg6
55.Ke4 (535.094.436) 739

21.22 22:58 +2.39 42.g5 a4 43.Rh7+ Kd6 44.h4 Ke6 45.g6 Rf2
46.Ke4 Re2+ 47.Kf3 Re1 48.Rb7 Rf1+
49.Ke3 a3 50.Rxb6+ Kf5 51.Ra6 Re1+
52.Kd4 Re4+ 53.Kd5 Re7 54.h5 Kxf4
55.h6 (1.007.199.581) 730

best move: g4-g5 time: 25:12.203 min n/s: 727.996 nodes: 1.100.833.568

This is my best output, both engines were calculating in parallel:


8/2k5/1p5R/p7/3K1PP1/8/r6P/8 w - -

Engine: Glaurung 080422 Mjolnir 004 Build 6 (64 MB)
by Tord Romstad

2.00 0:00 +3.50 42.h4 b5 (301) 1

3.00 0:00 +3.70 42.h4 Ra4+ 43.Ke5 b5 (835) 3

4.00 0:00 +3.88 42.h4 b5 43.Rh7+ Kd6 44.h5 Rd2+
45.Ke4 (1.961) 7

5.00 0:00 +3.92 42.h4 Rd2+ 43.Kc3 Rg2 44.g5 b5 (5.350) 20

5.00 0:00 +4.09 42.g5 Rd2+ 43.Ke4 Rg2 44.h4 b5 (8.305) 29

6.00 0:00 +4.33 42.g5 Rg2 43.h4 Rg4 44.Ke5 b5 (17.581) 53

7.00 0:00 +4.03 42.g5 Rd2+ 43.Ke3 Rd7 44.f5 b5 45.h4 a4 (29.341) 75

8.00 0:00 +4.11 42.g5 Rd2+ 43.Ke3 Rd8 44.f5 a4 45.g6 Re8+
46.Kd4 a3 (50.729) 111

9.00 0:00 +3.82 42.g5 a4 43.f5 a3 44.h4 Rf2 45.Rh7+ Kd6
46.Ra7 Rd2+ 47.Kc4 (148.015) 178

9.00 0:01 +4.07 42.h4 Rd2+ 43.Kc3 Rf2 44.f5 b5
45.Ra6 b4+ 46.Kd3 Rf3+ 47.Ke4 Ra3 (220.985) 202

10.01 0:01 +3.92 42.h4 a4 43.Rh7+ Kd6 44.Rh6+ Ke7
45.Rxb6 a3 46.Rb7+ Kd6 47.Rb6+ Ke7
48.f5 Rf2 (431.103) 226

11.01 0:04 +3.86 42.h4 a4 43.Rh7+ Kd6 44.h5 b5 45.h6 Rd2+
46.Ke3 Rg2 47.Kf3 Rh2 48.g5 (1.064.626) 249

12.01 0:08 +3.76 42.h4 a4 43.Rh7+ Kd6 44.Ra7 Rh2
45.h5 Rh4 46.Rxa4 Rxg4 47.Ke4 Rh4
48.Rd4+ Ke6 (2.070.264) 255

13.01 0:21 +3.80 42.h4 a4 43.Rh7+ Kd6 44.h5 Rd2+
45.Ke3 Rh2 46.Ra7 b5 47.h6 Rxh6
48.Ra6+ Kd5 49.Rxh6 b4 (5.482.385) 261

14.01 0:39 +3.70 42.h4 a4 43.Rh7+ Kd6 44.Ra7 Rh2
45.h5 Rh4 46.Rxa4 Rxg4 47.Ke4 Rh4
48.Rd4+ Ke6 49.Rd5 Kf6 50.Rd6+ Kg7 (10.295.685) 261

14.07 1:07 +3.94 42.g5 Rd2+ 43.Ke3 Rd7 44.h4 a4
45.Rh8 b5 46.h5 a3 47.f5 Rg7 48.Kf4 a2
49.Ra8 (17.806.421) 262

15.01 1:27 +3.90 42.g5 Rd2+ 43.Ke3 Rd7 44.h4 a4
45.Rh8 b5 46.h5 b4 47.Ra8 a3 48.f5 Kd6
49.h6 Rf7 (22.977.449) 262

16.01 3:28 +3.98 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 b4
46.g6 Rd2+ 47.Ke3 Rh2 48.g7 Rg2
49.h6 a3 50.Rh8 Ke6 (54.579.302) 262

17.01 5:56 +4.11 42.g5 a4 43.Rh7+ Kd6 44.h4 a3 45.h5 Rf2
46.Ke3 Rh2 47.h6 a2 48.Ra7 Ke6
49.Ke4 Re2+ 50.Kd3 Rg2 51.Kc3 (93.630.478) 262

18.01 14:08 +4.41 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 b4
46.Rb7 b3 47.h6 Rf2 48.Ke4 b2 49.h7 Re2+
50.Kd4 Re7 51.Rxb2 Rxh7 52.f5 (222.623.564) 262

best move: g4-g5 time: 25:12.172 min n/s: 261.417 nodes: 395.280.000

The eval for 42. g5 seems right but was too high for 42. h4. I believe at least Glaurung 2.0.1 also plays 42. h4

Eelco

[Eelco de Groot]

Build No.8:

[d]8/2k5/1p5R/p7/3K1PP1/8/r6P/8 w - -

Engine: Glaurung 080422 Mjolnir 004 Build 8 (64 MB)
by Tord Romstad

9.00 0:00 +3.94 42.g5 a4 43.f5 a3 44.f6 Rf2 45.Ke3 a2
46.Kxf2 a1Q (97.278) 151

10.00 0:01 +3.86 42.g5 a4 43.f5 Rd2+ 44.Kc3 Rd5
45.Rf6 a3 46.h4 a2 47.Kb2 Rd2+
48.Ka1 (230.756) 202

11.01 0:03 +3.84 42.g5 a4 43.f5 Rf2 44.f6 a3 45.Rh7+ Kd6
46.Rb7 Ke6 47.Rxb6+ Kf5 48.h4 Rf4+
49.Kd5 Rxh4 (830.635) 242

11.03 0:04 +3.92 42.h4 a4 43.Rh7+ Kd6 44.h5 a3 45.Ra7 Rf2
46.Rxa3 Rxf4+ 47.Kc3 Rxg4 48.h6 Rg3+
49.Kb4 (1.057.644) 248

12.01 0:06 +3.96 42.h4 a4 43.Rh7+ Kd6 44.Ra7 a3 45.f5 b5
46.g5 b4 47.Kc4 Rb2 48.h5 (1.686.110) 253

13.01 0:12 +3.98 42.h4 a4 43.Rh7+ Kd6 44.Ra7 Rh2
45.h5 Rh4 46.Rxa4 Rxg4 47.Ke4 Rh4
48.Rd4+ Ke6 49.Rd5 (3.153.399) 259

13.03 0:18 +4.09 42.g5 a4 43.f5 a3 44.Rh7+ Kd6 45.f6 Rg2
46.h4 Rg4+ 47.Kd3 b5 48.Ra7 Rg3+
49.Kd4 b4 50.Ra6+ Kd7 (4.759.046) 260

14.01 0:27 +4.17 42.g5 a4 43.f5 a3 44.Rh7+ Kd6 45.f6 Rg2
46.h4 Rg4+ 47.Kd3 b5 48.Ra7 Rg3+
49.Ke4 Rg4+ 50.Kf5 Ra4 (7.268.739) 260

15.01 1:00 +4.25 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.f5 Rxh2
46.f6 Ke6 47.f7 Rd2+ 48.Ke4 Rd8
49.g6 Kf6 50.Rxa3 Kxg6 (16.042.942) 263

16.01 2:02 +4.25 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.f5 Rxh2
46.f6 Ke6 47.f7 Rd2+ 48.Ke4 Rd8
49.g6 Kf6 50.Rxa3 Kxg6 (32.261.503) 262

17.01 4:09 +4.33 42.g5 a4 43.Rh7+ Kd6 44.Ra7 b5
45.Ra6+ Ke7 46.f5 Rxh2 47.f6+ Kf7
48.Ra7+ Ke6 49.f7 Rh8 50.Kc5 a3
51.g6 Kf6 52.Ra6+ Kg7 53.Kxb5 (65.218.853) 261

18.01 11:07 +4.66 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 b4
46.h6 b3 47.Rb7 Rf2 48.g6 Rxf4+
49.Kc3 Ke6 50.g7 Rf3+ 51.Kb2 Rf2+
52.Ka3 Rg2 53.Kxa4 (173.472.822) 259

19.01 32:02 +5.23 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 b4
46.Rb7 b3 47.h6 Rf2 48.g6 Rxf4+
49.Kc3 Kc6 50.Rb8 Rf3+ 51.Kb4 Kc7
52.Rb5 Rf6 (500.537.812) 260

best move: g4-g5 time: 35:43.718 min n/s: 259.769 nodes: 556.860.000

h4 is played later on move 44 but I am gambling that a transposition with h4 as first move is not the same. Anyway for the moment I think I will leave it like this. The PV is in good agreement with that of Strelka and the eval seems good

This is now my version of the first part of Glaurung's passed pawn evaluation:

Code: Select all

  // evaluate_passed_pawns() evaluates the passed pawns for both sides.

  void evaluate_passed_pawns(const Position& pos, EvalInfo& ei) {
    bool hasUnstoppable[2] = {false, false};
    int movesToGo[2] = {100, 100};
    int blockerCount, attackedsquareCount, defendedsquareCount;

    for(Color c = WHITE; c <= BLACK; c++) {
      Square ourKingSq = pos.king_square(c);
      Square theirKingSq = pos.king_square(opposite_color(c));
      Bitboard b = ei.pi->passed_pawns() & pos.pawns_of_color(c), b2;

      while(b) {
        Square s = pop_1st_bit(&b);
        assert(pos.piece_on(s) == pawn_of_color(c));
        assert(pos.pawn_is_passed(c, s));

        int r = int(pawn_rank(c, s) - RANK_2);
        int tr = Max(0, r * (r-1));
        Square blockSq = s + ((c == WHITE)? DELTA_N : DELTA_S);

        // Base bonus based on rank:
        Value mbonus = Value(20 * tr);
        Value ebonus = Value(r * r * 25);
        // [EdG: Edge pawns are less likely to promote:]
		if (square_file(s) == FILE_A || square_file(s) == FILE_H) {
			mbonus -= Value(20);
			ebonus -= Value(20);
		}
        // Adjust bonus based on king proximity:
        ebonus -= Value(square_distance(ourKingSq, blockSq) * 3 * tr);
        ebonus += Value(square_distance(theirKingSq, blockSq) * 6 * tr);

        // If the pawn is free to advance, increase bonus.  Note that the
        // use of pos.see() below causes occasional symmetry problems,
        // because the static exchange evaluator is not always symmetrical
        // in positions with complicated X-ray attacks.
        b2 = squares_in_front_of(c, s);
        blockerCount =
          count_1s_max_15( b2 & pos.occupied_squares());
        attackedsquareCount =
          count_1s_max_15( b2 & ei.attackedBy[opposite_color(c)][0]);
        defendedsquareCount =
          count_1s_max_15( b2 & ei.attackedBy[c][0]);
        mbonus += Value(40 + tr - 10 * blockerCount - 8 * attackedsquareCount + 4 * defendedsquareCount);
        ebonus += Value(40 + tr - 10 * blockerCount - 8 * attackedsquareCount + 4 * defendedsquareCount);
        if(pos.square_is_empty(blockSq) && pos.see(s, blockSq) == 0) {
          
          if((pos.pieces_of_color(opposite_color(c)) & b2) == EmptyBoardBB)
            ebonus += Value(8 * tr);
          if((pos.pieces_of_color(c) & b2) == EmptyBoardBB)
            ebonus += Value(tr);
          if((~ei.attackedBy[c][0] & ei.attackedBy[opposite_color(c)][0] & b2)
             == EmptyBoardBB)
            ebonus += Value(6 * tr);
        }

Edit: I noticed a serious bug, I put in twice FILE_A instead of FILE_H
I changed the code now. Still, it seemed to work as it is, but I now have to test of course what happens if I penalize pawn h which is what the change was meant for...

[Eelco de Groot]

Edit: I noticed a serious bug, I put in twice FILE_A instead of FILE_H
I changed the code now. Still, it seemed to work as it is, but I now have to test of course what happens if I penalize pawn h which is what the change was meant for...

With the corrected version I get a comparable result. This version still does not play the endgame correctly later on I think but the score for 42. g5 does keep going up, if it is a draw this version would have seen it by now I think. I checked a very little bit with ChessTiger that also thinks White is still good here, but on move 47 it was already a draw only.

Build No. 9 after more than twelve hours:


[d]8/2k5/1p5R/p7/3K1PP1/8/r6P/8 w - -

Engine: Glaurung 080422 Mjolnir 004 (64 MB)
by Tord Romstad


10.00 0:00 +3.80 42.g5 a4 43.f5 a3 44.f6 Rg2 45.Rh7+ Kd6
46.Ra7 Rg4+ 47.Kc3 Ke6 (286.645) 436

11.00 0:01 +3.78 42.g5 a4 43.f5 Rd2+ 44.Ke3 Rg2 45.g6 a3
46.Rh7+ Kd6 47.Ra7 a2 48.h4 (642.270) 478

11.02 0:01 +3.96 42.h4 a4 43.f5 a3 44.Rh7+ Kd6 45.Ra7 Ra1
46.f6 a2 47.f7 Rf1 48.Rxa2 Rf4+
49.Ke3 Rxf7 (967.017) 487

12.01 0:03 +3.86 42.h4 a4 43.f5 a3 44.Rh7+ Kd6 45.Ra7 Rg2
46.g5 Rg4+ 47.Kc3 b5 48.Kb3 b4 (1.608.019) 497

13.01 0:07 +3.74 42.h4 a4 43.Rh7+ Kd6 44.Ra7 Rh2
45.h5 Rh4 46.g5 Rxf4+ 47.Ke3 Ke5
48.g6 Re4+ 49.Kd3 Rh4 (3.769.146) 501

13.03 0:13 +4.05 42.g5 a4 43.f5 a3 44.g6 Rd2+ 45.Kc3 a2
46.Rh7+ Kd6 47.Ra7 Rxh2 48.f6 Ke6
49.f7 Rf2 (6.668.697) 505

14.01 0:21 +4.23 42.g5 Rd2+ 43.Ke5 a4 44.Rh3 Rg2
45.Rc3+ Kd7 46.h4 Re2+ 47.Kf6 b5
48.h5 Rh2 49.Rd3+ Kc6 50.h6 b4 (10.654.361) 506

15.01 0:46 +4.25 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.f5 Rxh2
46.f6 Ke6 47.f7 Rd2+ 48.Ke4 Rd8
49.g6 Kf6 50.Rxa3 Kxg6 (23.962.171) 512

16.01 1:09 +4.25 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.f5 Rxh2
46.f6 Ke6 47.f7 Rd2+ 48.Ke4 Rd8
49.g6 Kf6 50.Rxa3 Kxg6 (35.540.903) 512

17.01 2:01 +4.11 42.g5 a4 43.Rh7+ Kd6 44.Ra7 a3 45.f5 Rxh2
46.f6 Ke6 47.f7 Rh4+ 48.Kc3 Rh8
49.Kb4 a2 50.g6 Kf6 51.Rxa2 Kxg6 (61.880.457) 510

18.01 6:03 +4.29 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.g6 b4
46.Rb7 b3 47.h5 Rf2 48.h6 Rxf4+
49.Kc3 Rf3+ 50.Kb4 Rf4+ 51.Ka3 Rh4
52.g7 Rg4 (184.991.194) 508

19.01 17:11 +4.76 42.g5 a4 43.Rh7+ Kd6 44.h4 b5 45.h5 b4
46.Rb7 Rd2+ 47.Ke3 Rb2 48.Ra7 a3
49.h6 Ke6 50.Ra6+ Kf5 51.Rf6+ Kg4
52.g6 a2 53.g7 a1Q 54.g8Q+ Kh3
55.Qe6+ (520.989.554) 505

20.01 32:23 +4.84 42.g5 a4 43.Rh7+ Kd6 44.h4 Rf2
45.Ke4 b5 46.Rh6+ Ke7 47.f5 Re2+
48.Kd4 Rf2 49.Ke5 Re2+ 50.Kd5 Rd2+
51.Kc5 Rf2 52.Rh7+ Kf8 53.f6 a3
54.Ra7 a2 55.Kxb5 (980.961.592) 504

21.01 101:50 +5.33 42.g5 Rd2+ 43.Ke5 a4 44.Rh3 b5 45.g6 Rd8
46.f5 b4 47.Rh4 Re8+ 48.Kd5 Rd8+
49.Ke6 Re8+ 50.Kf7 Rb8 51.g7 a3
52.Rh8 Rb7 53.g8Q Kc6+ 54.Kf6 b3
55.Qc4+ (3.110.325.407) 509

22.01 195:43 +5.33 42.g5 Rd2+ 43.Ke5 a4 44.Rh3 b5 45.g6 Rd8
46.f5 b4 47.Rh4 Re8+ 48.Kd5 Rd8+
49.Ke6 Re8+ 50.Kf7 Rb8 51.g7 a3
52.Rh8 Rb7 53.Kf6 b3 54.g8Q Kc6
55.Qc4+ (5.975.159.528) 508

23.01 824:37 +5.78 42.g5 Rd2+ 43.Ke5 a4 44.Rh3 b5 45.g6 Rd8
46.f5 b4 47.Rh4 Re8+ 48.Kd5 Rd8+
49.Ke6 Re8+ 50.Kf7 Rb8 51.g7 a3
52.Rxb4 Rxb4 53.g8Q Ra4 54.Qg3+ Kc6
55.Qc3+ (25.090.032.436) 507

best move: g4-g5 time: 837:50.594 min n/s: 506.978 nodes: 25.486.110.000

Chess Tiger 2007.1 in Gambit Style after 47... a2 in the game continuation and after some forward analysis thinks now it should clearly be a draw:

[d]8/2k5/1p6/4KP2/6PP/8/p4r2/4R3 w - - 0 1

00:00:04.9 0,49 8 44410 Ra1 b5 g5 b4 g6 b3 g7 b2 Rxa2 Rxf5+ Kxf5 b1Q+ Kf6
00:00:09.9 0,74 8 62938 Rc1+ Kd7 g5 Re2+ Kd4 Rc2 Ra1 Ke7 f6+ Kf7 Kd5 Kg6 Ke5
00:00:10.0 1,20 9 121474 Rc1+ Kd7 g5 b5 g6 Re2+ Kd4 Rc2 Ra1 Ke7 h5 Kf6 h6 Kxf5 h7
00:00:15.0 0,92 10 286790 Rc1+ Kd7 h5 Re2+ Kf6 Rh2 Kg6 Rb2 Ra1 Rg2 g5
00:00:34.9 0,20 11 855348 Rc1+ Kd7 g5 b5 g6 Re2+ Kf6 b4 Rg1 Rg2 Rd1+ Kc6 Ra1 b3 h5
00:00:35.0 0,40 11 968780 Ra1 b5 Kd5 Kd7 f6 b4 Kc4 Rxf6 Rxa2 Rf4+ Kb3 Rxg4 h5
00:00:45.0 0,40 12 1244796 Ra1 b5 Kd5 Kd7 f6 b4 Kc4 Rxf6 Rxa2 Rf4+ Kb3 Rxg4 h5
00:01:10.0 0,56 13 2134322 Ra1 b5 Kd4 Kd6 h5 Ke7 g5 Rxf5 Rxa2 Rxg5 h6 Kf7 Ke4
00:01:45.0 0,73 14 3375090 Ra1 b5 Kd4 Kd6 h5 Ke7 h6 Kf6 Kc3 Kg5 h7 Rh2 Rxa2 Rxh7 Ra5 b4+ Kb3 Kxg4
00:02:40.0 0,76 15 5395117 Ra1 b5 Kd4 Kd6 h5 Ke7 Kc3 Kf6 Kb3 Kg5 Rxa2 Rf3+ Kb4 Kxg4
00:06:44.9 0,53 16 9585195 Ra1 b5 Kd4 Kd6 h5 Rg2 g5 Rxg5 Rxa2 Rxh5 Ra6+ Kd7 f6 Ke8 Ra7 Rf5 Ra6
00:18:09.9 0,66 17 21325878 Ra1 b5 Kd4 Kd6 h5 Rg2 g5 Rxg5 Rxa2 Rg4+ Kc3 Ke5 Ra5 Rg3+ Kb4 Rh3 Rxb5+ Kf6 Rb6+ Kg5 f6
00:38:24.9 0,65 18 36050081 Ra1 b5 Kd4 Kd6 h5 Rg2 g5 Rxg5 Rxa2 Rg4+ Kc3 Ke5 h6 Rc4+ Kb3 Rh4 Ra6 Kxf5 Ka3
02:06:24.9 0,24 19 103466495 Ra1 b5 Kd4 b4 Kc4 Rf4+ Kb3 Rxg4 h5 Rg5 h6 Rxf5 Rxa2 Rf3+ Kc4 Rc3+ Kb5 b3 Rb2 Rh3 Kc4 Rxh6 Rxb3
02:38:15.8 0,18 20 266949924 Ra1 b5 Kd4 b4 Kc4 Rf4+ Kb3 Rxg4 h5 Rh4 Rxa2 Rxh5 f6 Rh3+ Kc4 b3 Ra7+ Kd6
02:43:56.6 0,06 21 614114481 Ra1 b5 Kd4 b4 Kc4 Rf4+ Kb3 Rxg4 h5 Rh4 f6 Rxh5 f7 Rf5 Rxa2 Rf3+ Kxb4
02:53:18.8 0,10 21 1195577929 Rc1+ Kd7 g5 b5 g6 Re2+ Kd5 Rd2+ Ke5 Re2+ Kd5 Rd2+ Ke5 Re2+ Kd5 Rd2+ Ke5 Re2+