Repairing same bug over and over again

[Henk]

Thought I already repaired it three times. Some bugs never go away. See endgame.

[pgn]
[Event "Computer Chess Game"]
[Site "HP"]
[Date "2016.05.08"]
[Round "-"]
[White "Fairy-Max 4.8S"]
[Black "SkipperWinb"]
[Result "1/2-1/2"]
[TimeControl "120"]
[Annotator "1. +0.13 1... +0.00"]

1. c4 {+0.13/8} e5 {+0.00/20 2.5} 2. Nc3 {+0.13/8 4} Nf6 {+0.00/17 2.4} 3.
h4 {+0.10/8 4} Nc6 {+0.08/17 2.3} 4. Nf3 {+0.04/9 3} Be7 {+0.03/16 2.3} 5.
d4 {+0.04/9 1.9} exd4 {+0.15/15 2.3} 6. Nxd4 {-0.24/9 2.1} O-O
{+0.16/14 2.3} 7. Nf5 {-0.08/8 1.4} Bb4 {+0.09/14 2.2} 8. a3 {+0.04/9 2.2}
Bxc3+ {+0.80/15 2.2} 9. bxc3 {+0.12/9 2.9} d5 {+0.11/14 2.2} 10. Ne3
{+0.29/10 2.3} dxc4 {+0.21/15 2.0} 11. Qxd8 {+0.20/11 2.0} Rxd8
{+0.23/17 2.1} 12. Nxc4 {+0.19/10 3} Ne4 {+0.23/18 2.0} 13. Bf4
{+0.31/10 2.3} Nxc3 {+0.04/15 1.9} 14. Bxc7 {+0.28/9 1.1} Rd7
{+0.10/17 1.9} 15. Be5 {-0.07/9 1.4} Nxe5 {+0.18/16 1.8} 16. Nxe5
{-0.12/11 3} Rc7 {+0.21/15 1.8} 17. g3 {-0.12/9 1.2} Re7 {+0.21/15 1.8} 18.
Nc4 {-0.22/9 1.7} Be6 {+0.09/14 1.7} 19. Ne3 {+0.00/9 1.3} Nd5
{+0.03/15 1.7} 20. Nc2 {-0.01/10 4} Bg4 {+0.00/14 1.7} 21. Nd4
{+0.07/9 1.0} Rd8 {+0.06/15 2.1} 22. Bh3 {-0.04/8 1.0} Bxe2 {+0.54/15 2.0}
23. Nxe2 {-0.25/9 0.9} Nc3 {+0.45/16 2.1} 24. Bg4 {-0.36/11 3} Rde8
{+0.57/16 1.9} 25. O-O {-0.44/11 2.2} Nxe2+ {+0.67/17 1.8} 26. Kg2
{-0.61/10 8} g6 {+0.65/16 1.8} 27. h5 {-0.49/10 0.9} gxh5 {+0.75/17 1.8}
28. Bxh5 {-0.53/11 1.3} Nd4 {+0.63/15 1.8} 29. Rac1 {-0.48/10 2.6} a5
{+0.53/13 1.6} 30. Rc4 {-0.28/9 0.9} Rd8 {+0.54/16 1.6} 31. Rc5
{-0.25/10 0.9} b6 {+0.56/15 1.6} 32. Rg5+ {-0.25/10 0.8} Kh8 {+0.56/18 1.5}
33. Rd1 {-0.45/10 0.8} f6 {+0.53/15 1.5} 34. Rb5 {-0.35/10 0.7} Rd6
{+0.53/16 1.5} 35. Rb2 {-0.32/11 1.1} f5 {+0.52/15 1.5} 36. Rc1
{-0.30/10 0.6} Kg7 {+0.56/14 1.4} 37. Rc8 {-0.34/9 0.6} b5 {+0.58/14 1.4}
38. Ra8 {-0.33/9 0.9} Kh6 {+0.52/15 1.4} 39. Be8 {-0.19/9 0.9} Re5
{+0.45/16 1.3} 40. Rb8 {+0.12/9 0.9} Rdd5 {+0.47/16 1.3} 41. Rb6+
{+0.25/9 0.8} Kg7 {+0.49/14 0.9} 42. Rb7+ {+0.23/10 1.1} Kf8 {+0.49/16 0.9}
43. Bh5 {+0.14/10 1.1} f4 {+0.54/14 0.9} 44. Rb8+ {+0.01/10 1.5} Kg7
{+0.58/13 0.9} 45. Rb7+ {-0.09/10 1.1} Kh6 {+0.53/14 0.9} 46. Bg4
{-0.27/10 1.3} fxg3 {+0.58/11 0.9} 47. fxg3 {-0.40/10 1.3} Re3
{+0.63/13 1.0} 48. Rb1 {-0.37/9 1.4} Rxa3 {+0.83/9 0.8} 49. Rh1+
{-0.32/9 0.7} Kg6 {+0.58/12 0.7} 50. Rhxh7 {-0.03/9 1.0} Rd8 {+0.63/10 0.8}
51. Rbg7+ {+0.25/8 0.5} Kf6 {+0.65/13 0.7} 52. Rf7+ {+0.22/9 0.5} Ke5
{+0.71/9 0.7} 53. Re7+ {+0.03/9 1.9} Kd6 {+0.70/11 0.7} 54. Rd7+
{-0.12/10 0.6} Rxd7 {+1.06/12 0.6} 55. Rxd7+ {-0.09/10 0.5} Kc5
{+1.20/13 0.7} 56. Rc7+ {-0.37/12 1.0} Kd5 {+1.20/12 0.6} 57. Rd7+
{-0.38/12 1.3} Kc4 {+1.24/14 0.6} 58. Rc7+ {-0.39/12 0.4} Kd3
{+1.22/15 0.6} 59. Bh5 {-0.42/11 0.4} Ra2+ {+1.04/11 0.6} 60. Kh3
{-0.34/12 0.4} b4 {+1.06/13 0.6} 61. Bg6+ {-0.35/13 0.4} Kd2 {+0.63/14 0.6}
62. Rd7 {-0.38/13 0.4} Kc3 {+0.61/13 0.6} 63. Rc7+ {-0.34/13 0.3} Kb2
{+0.74/16 0.6} 64. Rc4 {-0.76/13 0.8} Ne2 {+1.15/14 0.6} 65. Bd3
{-0.80/12 0.5} Ka3 {+1.36/14 0.5} 66. Re4 {-0.92/11 0.5} Nc3 {+1.35/15 0.5}
67. Re5 {-0.97/10 0.3} a4 {+2.02/13 0.5} 68. Rc5 {-1.07/11 0.4} Kb2
{+1.90/15 0.5} 69. Rc6 {-1.07/10 0.5} b3 {+1.35/14 0.5} 70. Rb6
{-1.47/10 0.3} Kc1 {+1.35/15 0.5} 71. Rc6 {-1.50/12 0.4} Rc2 {+1.28/12 0.5}
72. Rf6 {-2.36/10 0.3} a3 {+1.82/11 0.5} 73. Bxc2 {-2.99/11 0.7} bxc2
{+4.64/12 0.5} 74. Rb6 {-5.80/10 0.3} a2 {+6.66/15 0.4} 75. Ra6
{-7.88/12 0.5} Kd1 {+8.89/14 0.4} 76. Rd6+ {-9.97/12 0.2} Ke2
{+9.07/15 0.4} 77. Re6+ {-10.09/11 0.2} Kf2 {+9.09/13 0.4} 78. Rf6+
{-10.22/12 0.5} Kg1 {+9.07/15 0.4} 79. Ra6 {-11.04/11 0.5} c1=Q
{+9.07/10 0.4} 80. Rxa2 {-11.11/11 0.3} Qf1+ {+10.22/9 0.4} 81. Kh4
{-11.10/11 0.3} Nxa2 {+10.12/9 0.4} 82. Kh5 {-11.33/11 0.3} Qh3+
{+11.08/9 0.4} 83. Kg5 {-11.40/12 0.3} Qxg3+ {+11.08/9 0.4} 84. Kf6
{-11.46/11 0.2} Qf2+ {+11.08/8 0.4} 85. Ke5 {-11.49/11 0.2} Kf1
{+11.05/7 0.4} 86. Kd5 {-11.48/11 0.4} Nc3+ {+11.08/7 0.3} 87. Kc4
{-11.47/10 0.2} Ne2 {+11.08/12 0.3} 88. Kd5 {-11.43/11 0.3} Nd4
{+11.07/8 0.3} 89. Ke5 {-11.36/11 0.4} Qb2 {+11.00/8 0.3} 90. Kd5
{-11.30/11 0.3} Qd2 {+11.00/12 0.4} 91. Ke4 {-11.34/11 0.2} Ke1
{+11.00/12 0.3} 92. Ke5 {-11.36/11 0.3} Kf2 {+11.00/8 0.3} 93. Kd5
{-11.36/11 0.5} Kg1 {+11.00/7 0.3} 94. Ke5 {-11.33/11 0.3} Kg2
{+11.00/7 0.3} 95. Ke4 {-11.33/10 0.2} Kh3 {+11.00/12 0.3} 96. Kd5
{-11.34/10 0.2} Kh4 {+11.00/7 0.3} 97. Kd6 {-11.33/10 0.2} Qe2
{+11.00/6 0.3} 98. Kc5 {-11.30/11 0.3} Qf2 {+10.98/9 0.3} 99. Kd5
{-11.30/11 0.3} Qb2 {+11.00/12 0.3} 100. Kc4 {-11.29/10 0.2} Kh3
{+11.00/12 0.3} 101. Kd5 {-11.29/11 0.2} Kg2 {+11.00/12 0.3} 102. Ke4
{-11.31/11 0.5} Kg3 {+11.00/12 0.3} 103. Ke5 {-11.30/11 0.5} Kf3
{+11.00/6 0.3} 104. Kf6 {-11.32/10 0.2} Qc2 {+11.01/7 0.3} 105. Ke5
{-11.32/11 0.3} Ke3 {+11.00/8 0.3} 106. Kf6 {-11.32/10 0.1} Kd2
{+11.00/7 0.3} 107. Ke5 {-11.26/10 0.2} Kd3 {+11.01/7 0.3} 108. Kf6
{-11.30/10 0.2} Qf2+ {+11.01/7 0.3} 109. Kg5 {-11.31/9 0.2} Qa2
{+11.01/7 0.3} 110. Kg6 {-11.29/10 0.2} Kc4 {+11.01/9 0.3} 111. Kf6
{-11.26/9 0.1} Qg2 {+11.01/8 0.3} 112. Kf7 {-11.28/9 0.2} Kd3
{+11.00/7 0.2} 113. Kf6 {-11.30/10 0.1} Ke2 {+11.00/8 0.3} 114. Ke7
{-11.31/10 0.1} Kd2 {+11.00/7 0.2} 115. Kf7 {-11.31/10 0.1} Ke3
{+11.00/9 0.2} 116. Kf6 {-11.32/10 0.1} Kd3 {+11.00/8 0.2} 117. Kf7
{-79.93/11 0.4} Kc3 {+11.00/7 0.3} 118. Kf6 {-11.33/10 0.2} Kb2
{+11.00/7 0.2} 119. Ke5 {-11.36/10 0.2} Qf2 {+10.98/7 0.2} 120. Kd6
{-11.45/10 0.1} Qc2 {+11.01/7 0.2} 121. Ke5 {-11.32/10 0.1} Kc3
{+11.01/8 0.2} 122. Kf6 {-11.42/11 0.3} Qh2 {+11.01/7 0.2} 123. Kg5
{-11.41/10 0.2} Kb2 {+11.00/7 0.2} 124. Kf6 {-11.42/8 0.1} Ka3
{+11.00/6 0.2} 125. Kg5 {-11.36/9 0.1} Kb4 {+11.00/7 0.2} 126. Kf6
{-11.38/10 0.1} Qe2 {+11.00/6 0.2} 127. Kf7 {-79.86/10 0.2} Kc3
{+11.00/8 0.2} 128. Kf6 {-79.91/10 0.1} Kb2 {+11.00/8 0.2} 129. Kf7
{-79.93/10 0.3} Ka3 {+11.00/8 0.2} 130. Kf6 {-79.92/10 0.1} Ka4
{+11.00/7 0.2} 131. Kg6 {-79.89/8 0.1} Qh2 {+11.00/8 0.2} 132. Kf6
{-79.80/10 0.1} Kb5 {+11.00/6 0.2} 133. Ke7 {-11.41/9 0.1} Kc6
{+11.00/6 0.2}
{Xboard adjudication: 50-move rule} 1/2-1/2
[/pgn]

[hgm]

So what is the bug here? Perhaps you are 'repairing' the wrong thing.

[Henk]

Simpel KQNk endgame. Maybe I have to use table bases to end this loop. Or put all pieces except king and queen in a corner and not allowing them to move. Maybe first make KQk endgame handling perfect.

[hgm]

Well, Fairy-Max would have no trouble at all winning this, and it doesn't use tablebases. It could even win this before I made the weights of the centralization table (Fairy-Max' substitute for piece-square tables) tunable on a per-piece basis, and then multiplied that of a bare King (detected in the root) by a factor 5 to 10.

All that is needed is that the eval of a King in a corner is a little worse than that of one in the center, plus the delayed loss bonus. The game you show doesn't seem to have that; there is no attempt at all to drive the bare King towards the corner, although with a Queen this should be totally trivial. There also is no attempt to centralize the own King.

[Daniel Anulliero]

Simpel KQNk endgame. Maybe I have to use table bases to end this loop. Or put all pieces except king and queen in a corner and not allowing them to move. Maybe first make KQk endgame handling perfect.

[Henk]

No. First step: Make sure your code is executed.

[flok]

No. First step: Make sure your code is executed.

I think creating unittests and running them with code coverage checking enabled may help for this.

[Henk]

Not to forget: in unittests at least fill hash table with 'normal' values. But maybe that's not possible for what would be 'normal' in that case.

[Henk]

Test position should include enough history. Probably only moves of the game where position is from. So test position without game is no test position !? That would mean WAC (Win at Chess) test positions can not be used.

[flok]

Test position should include enough history. Probably only moves of the game where position is from. So test position without game is no test position !? That would mean WAC (Win at Chess) test positions can not be used.

I wonder if WAC may be wrong in some cases.