Dev Builds » 20210519-1834

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 14. This yields an approximate Elo difference and establishes confidence in the strength of the dev builds.

Summary

Host Duration Avg Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo
ncm-dbt-01 11:44:51 1232136 3978 682 1303 1993 -54.68 +/- 5.48 31 799 922 234 3 -106.83 +/- 11.22
ncm-dbt-02 11:41:31 1258886 3958 686 1289 1983 -53.35 +/- 5.55 31 792 910 241 5 -104.33 +/- 11.29
ncm-dbt-03 11:44:57 1261803 3988 682 1298 2008 -54.1 +/- 5.51 41 774 942 234 3 -103.68 +/- 11.09
ncm-dbt-04 00:31:16 1475757 172 32 53 87 -42.63 +/- 26.67 2 28 46 9 1 -82.3 +/- 50.17
ncm-dbt-05 11:40:42 1264093 3932 682 1238 2012 -49.46 +/- 5.46 30 742 950 242 2 -95.66 +/- 11.04
ncm-dbt-06 11:46:15 1262525 3972 707 1272 1993 -49.76 +/- 5.52 33 757 943 248 5 -96.34 +/- 11.09
20000 3471 6453 10076 -52.19 +/- 2.45 168 3892 4713 1208 19 -101.2 +/- 4.96

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
210319 ncm-dbt-05 1228977 266 42 84 140 -55.32 +/- 19.45 1 52 68 12 0 -110.7 +/- 41.19
210318 ncm-dbt-02 1233456 294 53 104 137 -60.88 +/- 19.9 3 60 69 15 0 -117.76 +/- 41.19
210317 ncm-dbt-06 1224823 304 45 94 165 -56.49 +/- 18.87 1 63 72 16 0 -113.6 +/- 40.3
210316 ncm-dbt-01 1206385 314 47 100 167 -59.21 +/- 18.58 3 61 79 14 0 -114.63 +/- 38.23
210315 ncm-dbt-03 1255414 322 55 98 169 -46.67 +/- 17.44 0 60 84 17 0 -95.1 +/- 37.03
210314 ncm-dbt-05 1230514 500 87 151 262 -44.72 +/- 16.14 4 95 113 37 1 -86.52 +/- 32.11
210313 ncm-dbt-06 1215392 500 87 160 253 -51.09 +/- 14.31 3 91 132 24 0 -99.95 +/- 29.4
210312 ncm-dbt-02 1240085 500 87 149 264 -43.3 +/- 15.33 3 91 121 35 0 -83.57 +/- 31.04
210311 ncm-dbt-01 1187770 500 103 155 242 -36.26 +/- 16.02 2 91 116 39 2 -73.34 +/- 31.7
210310 ncm-dbt-03 1243658 500 90 164 246 -51.8 +/- 15.01 2 99 121 27 1 -104.49 +/- 31.0
210309 ncm-dbt-05 1231132 500 92 156 252 -44.72 +/- 15.41 3 93 119 35 0 -86.52 +/- 31.31
210308 ncm-dbt-06 1253002 500 95 157 248 -43.3 +/- 15.94 3 94 117 34 2 -86.52 +/- 31.58
210307 ncm-dbt-02 1226879 500 99 142 259 -29.95 +/- 16.22 3 84 118 43 2 -58.93 +/- 31.42
210306 ncm-dbt-01 1196315 500 95 153 252 -40.48 +/- 15.54 3 89 122 35 1 -79.17 +/- 30.91
210305 ncm-dbt-03 1226228 500 96 166 238 -48.96 +/- 15.63 6 90 122 32 0 -90.97 +/- 30.89
210304 ncm-dbt-05 1222609 500 97 153 250 -39.08 +/- 14.7 2 84 133 30 1 -77.71 +/- 29.4
210303 ncm-dbt-03 1212951 500 79 164 257 -59.64 +/- 15.38 6 98 122 23 1 -115.23 +/- 30.81
210302 ncm-dbt-06 1235641 500 92 151 257 -41.19 +/- 16.42 5 91 113 40 1 -77.71 +/- 32.09
210301 ncm-dbt-02 1240707 500 89 161 250 -50.38 +/- 15.58 2 102 113 32 1 -101.46 +/- 32.14
210300 ncm-dbt-01 1214333 500 87 167 246 -56.07 +/- 16.22 4 108 102 36 0 -109.07 +/- 33.67
210299 ncm-dbt-05 1228584 500 86 176 238 -63.23 +/- 15.01 6 100 122 22 0 -121.46 +/- 30.78
210298 ncm-dbt-03 1232103 500 85 164 251 -55.36 +/- 16.89 9 98 106 37 0 -99.95 +/- 33.09
210297 ncm-dbt-02 1223494 500 89 170 241 -56.78 +/- 15.63 4 104 111 31 0 -110.6 +/- 32.43
210296 ncm-dbt-06 1255853 500 100 164 236 -44.72 +/- 15.41 2 96 116 36 0 -88.0 +/- 31.71
210295 ncm-dbt-01 1193501 500 74 175 251 -71.16 +/- 14.92 7 105 120 18 0 -137.37 +/- 30.98
210294 ncm-dbt-05 1213505 500 79 146 275 -46.84 +/- 14.22 0 95 127 28 0 -95.44 +/- 30.17
210293 ncm-dbt-03 1224279 500 91 143 266 -36.26 +/- 16.14 8 75 129 37 1 -63.23 +/- 29.99
210292 ncm-dbt-01 1197308 500 95 174 231 -55.36 +/- 15.69 4 103 111 32 0 -107.54 +/- 32.43
210291 ncm-dbt-02 1206641 500 86 170 244 -58.93 +/- 15.6 2 110 110 26 2 -121.46 +/- 32.59
210290 ncm-dbt-06 1230990 500 83 165 252 -57.5 +/- 16.87 9 98 111 30 2 -107.54 +/- 32.43
210289 ncm-dbt-05 1252173 500 90 148 262 -40.48 +/- 15.66 5 85 123 37 0 -74.79 +/- 30.78
210288 ncm-dbt-03 1225328 500 78 175 247 -68.27 +/- 15.47 6 109 111 24 0 -132.54 +/- 32.44
210287 ncm-dbt-02 1232334 500 79 161 260 -57.5 +/- 14.89 4 99 122 25 0 -112.14 +/- 30.83
210286 ncm-dbt-06 1238250 500 92 158 250 -46.13 +/- 15.61 4 93 118 35 0 -88.0 +/- 31.44
210285 ncm-dbt-01 1190569 500 83 161 256 -54.64 +/- 15.15 3 101 117 29 0 -107.54 +/- 31.57
210284 ncm-dbt-05 1270772 500 79 166 255 -61.07 +/- 16.19 7 104 108 31 0 -115.23 +/- 32.87
210283 ncm-dbt-03 1237389 500 83 166 251 -58.21 +/- 15.32 3 106 112 29 0 -115.23 +/- 32.29
210282 ncm-dbt-02 1233753 500 79 170 251 -63.94 +/- 15.82 8 101 115 26 0 -119.89 +/- 31.85
210281 ncm-dbt-01 1212071 500 74 163 263 -62.51 +/- 15.37 5 105 114 26 0 -121.46 +/- 32.0
210280 ncm-dbt-06 1219203 500 84 171 245 -61.07 +/- 14.92 4 103 119 24 0 -119.89 +/- 31.25
178132 ncm-dbt-01 1490974 164 24 55 85 -66.47 +/- 23.34 0 36 41 5 0 -138.2 +/- 53.01
178131 ncm-dbt-06 1489579 168 29 52 87 -47.86 +/- 25.75 2 28 45 9 0 -88.74 +/- 50.66
178130 ncm-dbt-04 1475757 172 32 53 87 -42.63 +/- 26.67 2 28 46 9 1 -82.3 +/- 50.17
178129 ncm-dbt-02 1492631 164 25 62 77 -79.75 +/- 27.84 2 41 31 8 0 -158.44 +/- 61.98
178127 ncm-dbt-03 1498881 166 25 58 83 -70.0 +/- 26.34 1 39 35 8 0 -141.25 +/- 58.33
178126 ncm-dbt-05 1498571 166 30 58 78 -59.16 +/- 27.61 2 34 37 10 0 -112.62 +/- 56.61

Commit

Commit ID 754fc8a8b5ca7466926d54465eeb1df4d4a481ac
Author Stéphane Nicolet
Date 2021-05-19 18:34:37 UTC
Remove Tempo The Tempo variable was introduced 10 years ago in our search because the classical evaluation function was antisymmetrical in White and Black by design to gain speed: Eval(White to play) = -Eval(Black to play) Nowadays our neural networks know which side is to play in a position when they evaluate a position and are trained on real games, so the neural network encodes the advantage of moving as an output of search. This patch shows that the Tempo variable is not necessary anymore. STC: LLR: 2.94 (-2.94,2.94) <-2.50,0.50> Total: 33512 W: 2805 L: 2709 D: 27998 Ptnml(0-2): 80, 2209, 12095, 2279, 93 https://tests.stockfishchess.org/tests/view/60a44ceace8ea25a3ef03d30 LTC: LLR: 2.95 (-2.94,2.94) <-2.50,0.50> Total: 53920 W: 1807 L: 1760 D: 50353 Ptnml(0-2): 16, 1617, 23650, 1658, 19 https://tests.stockfishchess.org/tests/view/60a477f0ce8ea25a3ef03d49 We also tried a match (20000 games) at STC using purely classical, result was neutral: https://tests.stockfishchess.org/tests/view/60a4eebcce8ea25a3ef03db5 Note: there are two locations left in search.cpp where we assume antisymmetry of evaluation (in relation with a speed optimization for null moves in lines 770 and 1439), but as the values are just used for heuristic pruning this approximation should not hurt too much because the order of magnitude is still true most of the time. closes https://github.com/official-stockfish/Stockfish/pull/3481 Bench: 4015864
Copyright 2011–2024 Next Chess Move LLC