Dev Builds » 20211121-2018

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 15. 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 06:56:07 583480 4000 740 1301 1959 -49.05 ± 5.07 9 756 1025 207 3 -99.01 ± 10.57
ncm-dbt-02 06:57:08 587050 4000 764 1343 1893 -50.65 ± 5.22 9 796 960 235 0 -101.84 ± 10.98
ncm-dbt-03 06:56:57 585115 4000 738 1294 1968 -48.61 ± 5.11 13 747 1026 211 3 -97.32 ± 10.57
ncm-dbt-04 06:57:43 567821 4012 721 1276 2015 -48.37 ± 5.25 12 776 975 241 2 -96.82 ± 10.9
ncm-dbt-05 06:57:44 579669 3988 756 1294 1938 -47.16 ± 5.1 5 755 1008 225 1 -95.37 ± 10.69
20000 3719 6508 9773 -48.77 ± 2.3 48 3830 4994 1119 9 -98.07 ± 4.8

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
435486 ncm-dbt-04 568156 12 2 5 5 -88.66 ± 75.75 0 3 3 0 0 -190.73 ± 238.47
435485 ncm-dbt-05 581735 488 91 152 245 -43.66 ± 14.31 2 83 133 26 0 -85.71 ± 29.23
435484 ncm-dbt-02 586858 500 109 174 217 -45.42 ± 14.94 0 99 117 34 0 -92.46 ± 31.58
435483 ncm-dbt-01 583824 500 89 164 247 -52.51 ± 14.92 1 101 122 24 2 -109.07 ± 30.84
435482 ncm-dbt-03 586562 500 93 152 255 -41.19 ± 14.17 1 86 134 29 0 -82.1 ± 29.23
435481 ncm-dbt-04 567205 500 84 167 249 -58.21 ± 14.11 1 103 124 22 0 -118.33 ± 30.49
435480 ncm-dbt-05 578753 500 105 161 234 -39.08 ± 14.04 0 85 137 27 1 -80.63 ± 28.81
435479 ncm-dbt-03 587113 500 98 157 245 -41.19 ± 13.89 0 87 135 28 0 -83.57 ± 29.08
435478 ncm-dbt-01 583070 500 82 159 259 -53.93 ± 13.91 1 97 130 22 0 -109.07 ± 29.63
435477 ncm-dbt-02 587197 500 97 172 231 -52.51 ± 14.25 1 98 126 25 0 -106.01 ± 30.26
435476 ncm-dbt-04 567403 500 96 171 233 -52.51 ± 14.52 0 102 122 25 1 -109.07 ± 30.84
435475 ncm-dbt-05 584034 500 99 157 244 -40.48 ± 13.85 0 86 136 28 0 -82.1 ± 28.94
435474 ncm-dbt-02 587028 500 98 176 226 -54.64 ± 14.49 2 99 124 25 0 -109.07 ± 30.54
435473 ncm-dbt-03 584916 500 91 166 243 -52.51 ± 14.11 1 97 128 24 0 -106.01 ± 29.96
435472 ncm-dbt-01 584285 500 94 164 242 -48.96 ± 14.87 3 93 125 29 0 -95.44 ± 30.46
435471 ncm-dbt-04 565430 500 85 160 255 -52.51 ± 15.31 5 94 122 29 0 -99.95 ± 30.87
435470 ncm-dbt-05 578465 500 92 162 246 -48.96 ± 14.87 0 102 116 32 0 -99.95 ± 31.72
435469 ncm-dbt-02 586308 500 101 180 219 -55.36 ± 13.98 0 102 125 23 0 -113.68 ± 30.37
435468 ncm-dbt-03 584579 500 85 160 255 -52.51 ± 14.79 0 104 118 27 1 -109.07 ± 31.43
435467 ncm-dbt-01 580987 500 89 162 249 -51.09 ± 14.04 1 95 130 24 0 -102.97 ± 29.68
435466 ncm-dbt-04 568713 500 90 144 266 -37.67 ± 15.25 2 88 122 38 0 -73.34 ± 30.91
435465 ncm-dbt-05 576210 500 86 160 254 -51.8 ± 14.35 0 101 122 27 0 -106.01 ± 30.85
435464 ncm-dbt-02 585464 500 81 161 258 -56.07 ± 14.96 2 104 116 28 0 -112.14 ± 31.71
435463 ncm-dbt-01 584034 500 104 158 238 -37.67 ± 14.22 0 85 135 29 1 -77.71 ± 29.11
435462 ncm-dbt-03 582068 500 99 157 244 -40.48 ± 15.42 4 86 124 36 0 -76.25 ± 30.64
435461 ncm-dbt-04 568156 500 95 152 253 -39.78 ± 14.61 1 88 128 33 0 -79.17 ± 30.09
435460 ncm-dbt-05 576537 500 78 150 272 -50.38 ± 14.55 1 98 123 28 0 -101.46 ± 30.72
435459 ncm-dbt-01 584117 500 83 162 255 -55.36 ± 14.93 2 103 117 28 0 -110.6 ± 31.57
435458 ncm-dbt-02 588643 500 86 155 259 -48.25 ± 14.16 2 90 133 25 0 -95.44 ± 29.29
435457 ncm-dbt-04 568673 500 95 160 245 -45.42 ± 15.44 2 97 115 36 0 -89.48 ± 31.85
435456 ncm-dbt-03 583531 500 76 160 264 -58.93 ± 14.15 2 100 129 18 1 -119.89 ± 29.69
435455 ncm-dbt-01 583112 500 99 162 239 -44.01 ± 14.47 1 91 128 30 0 -88.0 ± 30.06
435454 ncm-dbt-05 578465 500 95 182 223 -61.07 ± 14.38 2 105 121 22 0 -123.02 ± 30.92
435453 ncm-dbt-04 565666 500 81 155 264 -51.8 ± 14.35 1 98 125 26 0 -104.49 ± 30.42
435452 ncm-dbt-03 586223 500 103 171 226 -47.55 ± 15.05 0 101 117 31 1 -98.44 ± 31.58
435451 ncm-dbt-02 588004 500 84 167 249 -58.21 ± 15.32 1 112 106 31 0 -118.33 ± 33.16
435450 ncm-dbt-01 584411 500 100 170 230 -48.96 ± 13.36 0 91 138 21 0 -99.95 ± 28.49
435449 ncm-dbt-05 583154 500 110 170 220 -41.89 ± 14.87 0 95 120 35 0 -85.04 ± 31.17
435448 ncm-dbt-02 586901 500 108 158 234 -34.86 ± 15.7 1 92 113 44 0 -68.99 ± 32.07
435447 ncm-dbt-04 570989 500 93 162 245 -48.26 ± 15.22 0 103 114 32 1 -99.95 ± 32.0
435446 ncm-dbt-03 585928 500 93 171 236 -54.64 ± 13.94 5 86 141 18 0 -104.49 ± 27.96

Commit

Commit ID a5a89b27c8e3225fb453d603bc4515d32bb351c3
Author Stéphane Nicolet
Date 2021-11-21 20:18:08 UTC
Introduce Optimism Current master implements a scaling of the raw NNUE output value with a formula equivalent to 'eval = alpha * NNUE_output', where the scale factor alpha varies between 1.8 (for early middle game) and 0.9 (for pure endgames). This feature allows Stockfish to keep material on the board when she thinks she has the advantage, and to seek exchanges and simplifications when she thinks she has to defend. This patch slightly offsets the turning point between these two strategies, by adding to Stockfish's evaluation a small "optimism" value before actually doing the scaling. The effect is that SF will play a little bit more risky, trying to keep the tension a little bit longer when she is defending, and keeping even more material on the board when she has an advantage. We note that this patch is similar in spirit to the old "Contempt" idea we used to have in classical Stockfish, but this implementation differs in two key points: a) it has been tested as an Elo-gainer against master; b) the values output by the search are not changed on average by the implementation (in other words, the optimism value changes the tension/exchange strategy, but a displayed value of 1.0 pawn has the same signification before and after the patch). See the old comment https://github.com/official-stockfish/Stockfish/pull/1361#issuecomment-359165141 for some images illustrating the ideas. ------- finished yellow at STC: LLR: -2.94 (-2.94,2.94) <0.00,2.50> Total: 165048 W: 41705 L: 41611 D: 81732 Ptnml(0-2): 565, 18959, 43245, 19327, 428 https://tests.stockfishchess.org/tests/view/61942a3dcd645dc8291c876b passed LTC: LLR: 2.95 (-2.94,2.94) <0.50,3.00> Total: 121656 W: 30762 L: 30287 D: 60607 Ptnml(0-2): 87, 12558, 35032, 13095, 56 https://tests.stockfishchess.org/tests/view/61962c58cd645dc8291c8877 ------- How to continue from there? a) the shape (slope and amplitude) of the sigmoid used to compute the optimism value could be tweaked to try to gain more Elo, so the parameters of the sigmoid function in line 391 of search.cpp could be tuned with SPSA. Manual tweaking is also possible using this Desmos page: https://www.desmos.com/calculator/jhh83sqq92 b) in a similar vein, with two recents patches affecting the scaling of the NNUE evaluation in evaluate.cpp, now could be a good time to try a round of SPSA tuning of the NNUE network; c) this patch will tend to keep tension in middlegame a little bit longer, so any patch improving the defensive aspect of play via search extensions in risky, tactical positions would be welcome. ------- closes https://github.com/official-stockfish/Stockfish/pull/3797 Bench: 6184852
Copyright 2011–2025 Next Chess Move LLC