Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Test Positions / Stockfish 280517 FB dropping half points
- - By Peter Grayson (****) [gb] Date 2017-06-13 21:43
The following position arose in the first game of a recent Gauntlet match when Stockfish 280517 Faster Build did not make the most of an advantageous position. Following Stockfish 8's 50..h7, Stockfish 280517 FB played 51.b5 missing a key winning move that should have been found within the context of the game conditions given the indicated ply depths and move times. 51.Qe6+ is a forced win.

[Event "Stockfish 170528 FB test"]
[Site "Newport, South Wales"]
[Date "2017.06.09"]
[Round "1.1"]
[White "Stockfish 280517 64 POPCNT_FB"]
[Black "Stockfish 8 64 POPCNT"]
[Result "1/2-1/2"]
[ECO "B90"]
[Annotator "0.24;0.09"]
[SetUp "1"]
[FEN "1rk4r/8/3Q4/3P4/1P3p2/8/1P6/1K6 b - - 0 50"]
[PlyCount "160"]
[EventDate "2017.06.09"]
[EventType "simul"]
[Source "Grayson"]

{Intel(R) Xeon(R) CPU E5-2687W 0 @ 3.10GHz 3092 MHz  W=41.8 plies; 14,538kN/s;
86,093,177 TBAs  B=40.8 plies; 14,519kN/s; 70,939,125 TBAs} 50... Rh7 {[%eval
168,40] [%emt 0:00:11]} 51. b5 {[%eval 181,42] [%emt 0:00:00] (Qxf4)} (51. Qe6+
Rd7 52. b5 f3 53. b6 Rbb7 54. Ka2 f2 55. Qc6+ Kd8 (55... Kb8 56. Qf6 $18) 56.
Qf6+ Ke8 57. Qxf2 Ke7 (57... Rxd5 58. Qf3 Rbd7 59. b7 Ra5+ 60. Kb1 Rb5 61. Qe2+
Kf7 62. Qxb5 $18) 58. Qc5+ Rd6 {Stockfish 8 64:} 59. Qe3+ Kf7 60. Qf3+ Ke7 61.
Qe4+ Kd8 62. Qh4+ Kc8 63. Qh8+ Kd7 64. Qh7+ Kc8 65. Qg8+ Kd7 66. Qf7+ Kc8 67.
Qe8+ Rd8 68. Qc6+ Kb8 69. d6 Rc8 70. Qb5 Rd8 71. Kb3 Rg7 72. Qd5 Rb7 73. Qc6
Rc8 74. Qb5 Rd8 75. Kb4 Rg7 76. Qc6 Rc8 77. d7 Rxc6 78. d8=Q+ Rc8 79. Qd6+ Ka8
80. Qd5+ Rb7 81. Kb5 Kb8 82. Ka6 Rh7 83. Qe5+ Ka8 84. Qe4+ Kb8 85. Qxh7 Rc7 86.
bxc7+ {[%eval 12829,45]}) 51... Rbb7 {[%eval 158,30] [%emt 0:00:07]} 52. Qf8+ {
[%eval 177,42] [%emt 0:00:00]} Kc7 {[%eval 158,32] [%emt 0:00:03]} 53. Qxf4+ {
[%eval 184,41] [%emt 0:00:13]} Kb6 {[%eval 158,37] [%emt 0:00:01]} 54. Qf6+ {
[%eval 177,40] [%emt 0:00:06]} Kxb5 {[%eval 158,38] [%emt 0:00:02]} 55. Qc6+ {
[%eval 173,45] [%emt 0:00:08]} Kb4 {[%eval 158,40] [%emt 0:00:03]} 56. d6 {
[%eval 173,46] [%emt 0:00:00]} Rbg7 {[%eval 158,38] [%emt 0:00:02] (Ra7)} 57.
Kc1 {[%eval 173,43] [%emt 0:00:09]} Rb7 {[%eval 158,39] [%emt 0:00:03] (Ra7)}
58. Kc2 {[%eval 173,44] [%emt 0:00:09] (Kd2)} Rh2+ {[%eval 151,38] [%emt 0:00:
14]} 59. Kd3 {[%eval 173,46] [%emt 0:00:00]} Rh3+ {[%eval 149,34] [%emt 0:00:
05]} 60. Ke4 {[%eval 173,47] [%emt 0:00:04] (Kd4)} Rh4+ {[%eval 138,38] [%emt
0:00:10]} 61. Ke5 {[%eval 173,45] [%emt 0:00:04]} Rh5+ {[%eval 137,37] [%emt 0:
00:00]} 62. Kf4 {[%eval 173,48] [%emt 0:00:13] (Kf6)} Rh4+ {[%eval 130,31]
[%emt 0:00:08]} 63. Ke3 {[%eval 173,49] [%emt 0:00:04] (Kg3)} Rh3+ {[%eval 130,
38] [%emt 0:00:14]} 64. Kd4 {[%eval 173,52] [%emt 0:00:00] (Ke2)} Rh4+ {[%eval
0,38] [%emt 0:00:07]} 65. Kd3 {[%eval 173,50] [%emt 0:00:00] (Kd5)} Rh3+ {
[%eval 0,78] [%emt 0:00:11]} 66. Kc2 {[%eval 173,51] [%emt 0:00:02] (Ke4)} Rh2+
{[%eval 0,79] [%emt 0:00:12]} 67. Kc1 {[%eval 173,52] [%emt 0:00:00] (Kd3)}
Rhh7 {[%eval 0,81] [%emt 0:00:11]} 68. Kd2 {[%eval 173,46] [%emt 0:00:10]} Rh2+
{[%eval 123,40] [%emt 0:00:49] (Ra7)} 69. Ke3 {[%eval 173,54] [%emt 0:00:25]
(Kd1)} Rh3+ {[%eval 0,84] [%emt 0:00:08]} 70. Ke2 {[%eval 173,49] [%emt 0:00:
04]} Rh2+ {[%eval 116,38] [%emt 0:00:17] (Rhh7)} 71. Kf3 {[%eval 173,49] [%emt
0:00:09]} Rhh7 {[%eval 116,40] [%emt 0:00:00]} 72. Ke4 {[%eval 173,49] [%emt 0:
00:02] (Qd5)} Rh4+ {[%eval 0,91] [%emt 0:00:04]} 73. Ke5 {[%eval 173,52] [%emt
0:00:20] (Kf5)} Rh5+ {[%eval 0,74] [%emt 0:00:05]} 74. Kf6 {[%eval 173,49]
[%emt 0:00:00] (Ke4)} Rh6+ {[%eval 35,34] [%emt 0:00:13]} 75. Kf5 {[%eval 173,
51] [%emt 0:00:00] (Kg5)} Rh5+ {[%eval 130,43] [%emt 0:00:17] (Rhh7)} 76. Kg6 {
[%eval 173,52] [%emt 0:00:15]} Rhh7 {[%eval 130,42] [%emt 0:00:00]} 77. Qd5 {
[%eval 173,51] [%emt 0:00:03]} Rhd7 {[%eval 130,42] [%emt 0:00:06]} 78. b3 {
[%eval 166,47] [%emt 0:00:29]} Ka3 {[%eval 130,43] [%emt 0:00:00] (Rb5)} 79.
Kf5 {[%eval 173,45] [%emt 0:00:11]} Rxb3 {[%eval 130,37] [%emt 0:00:00]} 80.
Qc5+ {[%eval 173,46] [%emt 0:00:18] (Qa5+)} Ka2 {[%eval 130,34] [%emt 0:00:19]}
81. Qc2+ {[%eval 173,52] [%emt 0:00:00] (Qa5+)} Ka3 {[%eval 116,32] [%emt 0:00:
09]} 82. Qc1+ {[%eval 173,48] [%emt 0:00:00] (Ke6)} Ka2 {[%eval 130,42] [%emt
0:00:18]} 83. Qc5 {[%eval 173,52] [%emt 0:00:00] (Qd2+)} Rbb7 {[%eval 116,35]
[%emt 0:00:11]} 84. Qc4+ {[%eval 173,53] [%emt 0:00:00] (Ke4)} Ka3 {[%eval 116,
44] [%emt 0:00:10]} 85. Ke5 {[%eval 173,53] [%emt 0:00:02] (Ke6)} Rh7 {[%eval
116,43] [%emt 0:00:11]} 86. Qd4 {[%eval 173,52] [%emt 0:00:03] (Kd5)} Kb3 {
[%eval 108,34] [%emt 0:00:11] (Rhd7)} 87. Kd5 {[%eval 173,54] [%emt 0:00:14]
(Ke6)} Ra7 {[%eval 108,42] [%emt 0:00:18] (Ka3)} 88. Qe4 {[%eval 173,48] [%emt
0:00:14] (Qd2)} Rh5+ {[%eval 108,43] [%emt 0:00:09]} 89. Kc6 {[%eval 173,52]
[%emt 0:00:00]} Rhh7 {[%eval 108,48] [%emt 0:00:08]} 90. Kb5 {[%eval 173,51]
[%emt 0:00:07] (Qb1+)} Rh5+ {[%eval 108,48] [%emt 0:00:09] (Rab7+)} 91. Kb6 {
[%eval 173,54] [%emt 0:00:14]} Rhh7 {[%eval 108,51] [%emt 0:00:00]} 92. Kc5 {
[%eval 173,53] [%emt 0:00:02] (Qd4)} Rh5+ {[%eval 0,81] [%emt 0:00:05]} 93. Kc6
{[%eval 173,53] [%emt 0:00:06] (Kb6)} Rhh7 {[%eval 0,85] [%emt 0:00:08]} 94.
Qd4 {[%eval 173,54] [%emt 0:00:00] (Qb1+)} Rhf7 {[%eval 108,47] [%emt 0:00:11]
(Ka3)} 95. Qc5 {[%eval 133,46] [%emt 0:01:15] (Qb6+)} Rg7 {[%eval 108,46]
[%emt 0:00:09]} 96. Qd4 {[%eval 107,42] [%emt 0:00:07] (Qf5)} Rgf7 {[%eval 0,
94] [%emt 0:00:04] (Rh7)} 97. Qd3+ {[%eval 133,42] [%emt 0:00:05] (Qb6+)} Kb4 {
[%eval 108,45] [%emt 0:00:14]} 98. Qe4+ {[%eval 133,52] [%emt 0:00:37] (Qd2+)}
Kc3 {[%eval 108,44] [%emt 0:00:45]} 99. Kb5 {[%eval 133,53] [%emt 0:00:00]
(Kb6)} Rfb7+ {[%eval 108,41] [%emt 0:00:13] (Kd2)} 100. Kc5 {[%eval 132,41]
[%emt 0:00:19]} Ra5+ {[%eval 108,43] [%emt 0:00:00]} 101. Kc6 {[%eval 125,23]
[%emt 0:00:00]} Raa7 {[%eval 108,46] [%emt 0:00:28]} 102. Qe3+ {[%eval 93,36]
[%emt 0:00:05]} Kc4 {[%eval 108,46] [%emt 0:00:00]} 103. Qc5+ {[%eval 97,34]
[%emt 0:00:06]} Kb3 {[%eval 108,43] [%emt 0:00:11] (Kd3)} 104. Kd5 {[%eval 173,
47] [%emt 0:00:04] (Qd5+)} Rh7 {[%eval 108,37] [%emt 0:00:03]} 105. Qb5+ {
[%eval 173,41] [%emt 0:00:04] (Qc8)} Ka3 {[%eval 108,41] [%emt 0:00:03]} 106.
Kc6 {[%eval 173,43] [%emt 0:00:29] (Qc4)} Rhg7 {[%eval 108,36] [%emt 0:00:05]
(Rhf7)} 107. Qb1 {[%eval 173,43] [%emt 0:00:21] (Qd5)} Ra6+ {[%eval 108,26]
[%emt 0:00:03] (Rgf7)} 108. Kd5 {[%eval 173,41] [%emt 0:00:13]} Ra5+ {[%eval
108,40] [%emt 0:00:00]} 109. Ke6 {[%eval 173,40] [%emt 0:00:08] (Ke4)} Raa7 {
[%eval 108,22] [%emt 0:00:02]} 110. Qe4 {[%eval 173,39] [%emt 0:00:00]} Rh7 {
[%eval 108,31] [%emt 0:00:01] (Rgb7)} 111. Qc4 {[%eval 118,22] [%emt 0:00:05]
(Kd5)} Rhg7 {[%eval 38,21] [%emt 0:00:01] (Rhb7)} 112. Qe2 {[%eval 49,25]
[%emt 0:00:02] (Qb5)} Kb3 {[%eval 0,23] [%emt 0:00:02] (Kb4)} 113. Kd5 {[%eval
23,25] [%emt 0:00:01] (Qb5+)} Kb4 {[%eval 0,42] [%emt 0:00:02]} 114. Qe4+ {
[%eval 0,35] [%emt 0:00:00] (Qf1)} Kb5 {[%eval 0,46] [%emt 0:00:03]} 115. Qe8+
{[%eval 0,49] [%emt 0:00:00] (Qb1+)} Ka5 {[%eval 0,48] [%emt 0:00:04]} 116. Qe4
{[%eval 0,53] [%emt 0:00:00] (Qb8)} Rh7 {[%eval 0,44] [%emt 0:00:02] (Kb5)}
117. Kc5 {[%eval 0,36] [%emt 0:00:01] (Kc4)} Rh5+ {[%eval 0,43] [%emt 0:00:02]}
118. Kd4 {[%eval 0,45] [%emt 0:00:00] (Kc4)} Rah7 {[%eval 0,44] [%emt 0:00:03]
(Kb6)} 119. Qc6 {[%eval 0,48] [%emt 0:00:01] (Qd3)} Rh4+ {[%eval 0,42] [%emt 0:
00:02]} 120. Kc3 {[%eval 0,57] [%emt 0:00:00] (Kd3)} Rh3+ {[%eval 0,46] [%emt
0:00:03] (Rh2)} 121. Kb2 {[%eval 0,69] [%emt 0:00:11] (Kd2)} R3h4 {[%eval 0,58]
[%emt 0:00:13] (Rh2+)} 122. Qc3+ {[%eval 0,64] [%emt 0:00:11] (d7)} Kb5 {
[%eval 0,73] [%emt 0:00:11]} 123. d7 {[%eval 0,93] [%emt 0:00:02] (Qd3+)} Rxd7
{[%eval 0,114] [%emt 0:00:13]} 124. Qe5+ {[%eval 0,0] [%emt 0:00:00]} Ka4 {
[%eval 0,1] [%emt 0:00:00]} 125. Kc3 {[%eval 0,0] [%emt 0:00:00]} Ra7 {[%eval
0,0] [%emt 0:00:00]} 126. Qd6 {[%eval 0,0] [%emt 0:00:00]} Rh3+ {[%eval 0,0]
[%emt 0:00:00]} 127. Kc4 {[%eval 0,0] [%emt 0:00:00]} Rh4+ {[%eval 0,1] [%emt
0:00:00]} 128. Kc3 {[%eval 0,0] [%emt 0:00:00]} Rh3+ {[%eval 0,0] [%emt 0:00:
00]} 129. Kc4 {[%eval 0,0] [%emt 0:00:00]} Rh4+ {[%eval 0,1] [%emt 0:00:00]}
130. Kc3 {[%eval 0,0] [%emt 0:00:00]} 1/2-1/2
Parent - - By Eelco de Groot (***) Date 2017-06-15 18:26
Hello Peter,

You are the first it seems to mention the existence of a Stockfish Faster Build in this forum as far as I can see so could you share some details? Is it just supposed to be faster than Abrok.eu compiles it seems hardly worth to give it a new name, and why not use asmFish then. If not is there sourcecode, who are the authors do they have real names, in case there is sources too. What did they change.

If you can't reproduce the suboptimal move in any way, there is not a lot you can do I think. It can be number of things for instance a real bug, effect of new code or old code but not a real bug or effect of some interaction with hashcontents.
Parent - By Peter Grayson (****) [gb] Date 2017-06-16 01:03
Hello Eelco,

The faster build compiles have been around for a while and can be downloaded from here ...

http://chess.ultimaiq.net/stockfish.html

The engines should be the same code as the Abrok.eu compile but measure around 8.5% faster running speed for the equivalent POPCOUNT compile I used in this test. They are compiled by I. Ivec and L. Ljubičić. I added the "FB" suffix to facilitate easy identification in my test games database to distinguish it from others. The engine download does not show a datestamp but will show the date when the engine is run in the command line mode or from within a GUI. The engines do not use LP hash. Perhaps surprisingly, the asmFish conversion has not always produced the best results using my fixed opening lines although the single thread output indicates asmFish is the fastest version I have seen. What happens when additional CPU cores and hash are allocated is unclear although the indicated asmFish kN/s is always higher.

> If you can't reproduce the suboptimal move in any way, there is not a lot you can do I think. It can be number of things for instance a real bug, effect of new code or old code but not a real bug or effect of some interaction with hashcontents.


The number of cores and amount of hash allocation also affects engine behaviour and of course it is difficult to faithfully reproduce the state of hash at the point of the critical move. This can provide the unexpected situation when running with 16 cores and 16 Gb hash for example can fail to produce the best move but 4 cores and 4 Gb hash can find it relatively quickly. This is more likely to be the exception but does occur frequently enough as a reminder of MP variability and probably also influenced by move pruning and time control when the influence of the latter cannot be reproduced in analysis.

The Brainfish compile looks to be similar speed and has more often than not produced marginally better results than asmFish. I did recently identify an endgame issue with the asmFish compile and the the author kindly replied and confirmed it had been addressed so I will probably check the next asmFish compile when it may prove to give the better result.
Parent - - By Eelco de Groot (***) Date 2017-06-15 20:39
My Stockfish sees it at depth 47, after about 3 minutes+, if I start with the diagram (I could not load the pgn, maybe because of all the Chessbase &^& :mad: unreadable IMO, not your fault.) 51.Qc6+ wins also?

1rk4r/8/3Q4/3P4/1P3p2/8/1P6/1K6 b - -


Engine: Sf20170524_020 HT MOD (i7 6700, 4 threads + 3 hyperthreads, no rablebases, 512 MB)
by T. Romstad, M. Costalba, J. Kiiski, G. Linscott

30/39  0:01   +1.49    1...Rh7 2.Qxf4 Rb6 3.Kc2 Rhb7 4.Kc3 Kd8
                       5.Qf8+ Kd7 6.Kb3 Rb8 7.Qf7+ Kd8
                       8.Qg8+ Ke7 9.Qg4 R8b7 10.Qd4 Kd6
                       11.Qc5+ Kd7 12.Qc4 Kd6 13.Qe4 Kd7
                       14.Qf5+ Kd8 (18.379.782) 15040

31/45  0:03   +2.11--  1...Rh7 2.Qxf4 (48.513.222) 15274

31/46  0:03   +2.11    1...Rh7 2.Qxf4 Rb6 3.Kc2 Rc7+ 4.Kb3 Kb7
                       5.Qe4 Rd7 6.Kc3 Rbd6 7.Kc4 Rc7+
                       8.Kd4 Ka6 9.Qe8 Kb7 10.Qb5+ Rb6
                       11.Qa4 Ra6 12.Qb3 Rb6 13.b5 Ka7
                       14.Qb4 Kb7 (51.975.776) 15286
.
.
.
45/73  1:58   +1.77    1...Rh7 2.b5 Rbb7 3.Qf8+ Kc7 4.Qxf4+ Kb6
                       5.Qd6+ Kxb5 6.Qc6+ Kb4 7.d6 Rbg7
                       8.Kc1 Rb7 9.Kc2 Rh2+ 10.Kd3 Rh3+
                       11.Kd4 Rh4+ 12.Ke5 Rh5+ 13.Ke4 Rh4+
                       14.Ke3 Rh3+ (1.764.499.578) 14947

46/73  2:24   +1.77    1...Rh7 2.b5 Rbb7 3.Qf8+ Kc7 4.Qxf4+ Kb6
                       5.Qd6+ Kxb5 6.Qc6+ Kb4 7.d6 Rbg7
                       8.Kc1 Rb7 9.Kc2 Rh2+ 10.Kd3 Rh3+
                       11.Kd4 Rh4+ 12.Ke5 Rh5+ 13.Ke4 Rh4+
                       14.Ke3 Rh3+ (2.151.058.580) 14890

47/73  3:05   +1.85--  1...Rh7 2.Qc6+ (2.771.457.334) 14950

47/73  3:14   +1.92--  1...Rh7 2.Qe6+ (2.905.503.687) 14942

47/73  3:26   +2.03--  1...Rh7 2.Qe6+ (3.072.695.227) 14907

47/73  3:40   +2.18--  1...Rh7 2.Qc6+ (3.288.856.746) 14888

47/73  3:54   +2.39--  1...Rh7 2.Qc6+ (3.485.810.982) 14862

47/73  4:09   +2.67--  1...Rh7 2.Qc6+ (3.694.914.989) 14826

47/73  4:59   +3.04--  1...Rh7 2.Qc6+ (4.414.391.368) 14759

47/73  5:42   +3.53--  1...Rh7 2.Qc6+ (5.037.358.098) 14713

47/87  7:27   +4.15--  1...Rh7 2.Qc6+ (6.584.562.631) 14716

47/87  7:31   +4.15    1...Rh7 2.Qc6+ Kd8 3.Qf6+ Kc8 4.Qe6+ Rd7
                       5.b5 Kc7 6.Qc6+ Kd8 7.d6 Rbb7 8.Ka2 f3
                       9.Qxf3 Ra7+ 10.Kb3 Rxd6 11.Qf8+ Kc7
                       12.Qf4 Rb7 13.Kb4 Rb8 14.Qc4+ Kd8 (6.655.469.556) 14725

48/87  7:32   +4.15    1...Rh7 2.Qc6+ Kd8 3.Qf6+ Kc8 4.Qe6+ Rd7
                       5.b5 Kc7 6.Qc6+ Kd8 7.d6 Rbb7 8.Ka2 f3
                       9.Qxf3 Ra7+ 10.Kb3 Rxd6 11.Qf8+ Kc7
                       12.Qf4 Rb7 13.Kb4 Rb8 14.Qc4+ Kd8 (6.657.167.081) 14725

49/87  7:34   +4.15    1...Rh7 2.Qc6+ Kd8 3.Qf6+ Kc8 4.Qe6+ Rd7
                       5.b5 Kc7 6.Qc6+ Kd8 7.d6 Rbb7 8.Ka2 f3
                       9.Qxf3 Ra7+ 10.Kb3 Rxd6 11.Qf8+ Kc7
                       12.Qf4 Rb7 13.Kb4 Rb8 14.Qc4+ Kd8 (6.695.060.736) 14730

50/87  7:44   +4.15    1...Rh7 2.Qc6+ Kd8 3.Qf6+ Kc8 4.Qe6+ Rd7
                       5.b5 Kc7 6.Qc6+ Kd8 7.d6 Rbb7 8.Ka2 f3
                       9.Qxf3 Ra7+ 10.Kb3 Rxd6 11.Qf8+ Kc7
                       12.Qf4 Rb7 13.Kb4 Rb8 14.Qc4+ Kd8 (6.843.200.500) 14724

51/88  9:40   +4.22--  1...Rh7 2.Qc6+ (8.526.417.862) 14691

51/88  11:34  +4.30--  1...Rh7 2.Qc6+ (10.231.283.394) 14731

51/88  13:15  +4.41--  1...Rh7 2.Qc6+ (11.735.820.619) 14745

51/88  13:29  +4.41    1...Rh7 2.Qc6+ Rc7 3.Qe6+ Rd7 4.b5 f3
                       5.b6 Rbb7 6.Ka2 f2 7.Qc6+ Kd8 8.Qf6+ Ke8
                       9.Qxf2 Rd6 10.Qe3+ Kf7 11.Qf3+ Ke7
                       12.Qe4+ Kd8 13.Qh4+ Ke8 14.Qh5+ Kd8 (11.954.289.662) 14762

52/88  16:42  +4.48--  1...Rh7 2.Qc6+ (14.789.924.416) 14758

52/88  16:51  +4.55--  1...Rh7 2.Qc6+ (14.935.340.313) 14761
Parent - By Peter Grayson (****) [gb] Date 2017-06-16 01:21
Thanks for the analysis. Maybe Qc6 is winning too.
I've attached the game pgn with analysis at move 51

Best regards,
Peter
Attachment: Stockfishmissedwin.zip - Stockfish 280517 missed win pgn (3k)
Up Topic The Rybka Lounge / Test Positions / Stockfish 280517 FB dropping half points

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill