Dev Builds » 20230612-1841

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 06:52:48 582520 4000 1333 729 1938 +52.87 ± 5.2 2 215 974 795 14 +106.01 ± 10.89
ncm-dbt-02 06:53:49 585223 3998 1292 731 1975 +49.08 ± 5.11 2 219 1000 772 6 +99.44 ± 10.73
ncm-dbt-03 06:55:00 585921 4002 1358 688 1956 +58.72 ± 5.07 1 187 961 845 7 +119.83 ± 10.95
ncm-dbt-04 06:53:39 570798 4000 1288 762 1950 +45.95 ± 5.17 3 237 997 757 6 +93.01 ± 10.76
ncm-dbt-05 06:54:30 585492 4000 1349 709 1942 +56.07 ± 5.02 0 186 997 808 9 +113.49 ± 10.73
20000 6620 3619 9761 +52.53 ± 2.29 8 1044 4929 3977 42 +106.28 ± 4.83

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
408594 ncm-dbt-03 587282 2 1 0 1 +189.7 ± 55.98 0 0 0 1 0 +1129.65 ± 376.02 ↓
408593 ncm-dbt-02 586139 498 162 95 241 +47.03 ± 13.44 0 22 138 89 0 +95.85 ± 28.51 ↓
408592 ncm-dbt-05 586266 500 165 87 248 +54.64 ± 14.36 0 25 123 101 1 +110.6 ± 30.69 ↓
408591 ncm-dbt-01 581777 500 163 101 236 +43.3 ± 15.2 1 32 123 92 2 +86.52 ± 30.76 ↓
408590 ncm-dbt-04 570348 500 162 94 244 +47.54 ± 14.26 0 27 129 93 1 +95.44 ± 29.88 ↓
408589 ncm-dbt-03 585084 500 166 85 249 +56.78 ± 13.62 0 20 129 101 0 +116.78 ± 29.72 ↓
408588 ncm-dbt-02 585211 500 171 100 229 +49.67 ± 13.97 1 21 135 92 1 +101.46 ± 28.94 ↓
408587 ncm-dbt-05 589155 500 170 93 237 +53.93 ± 13.63 0 21 131 98 0 +110.6 ± 29.47 ↓
408586 ncm-dbt-01 583028 500 165 83 252 +57.5 ± 14.76 0 27 115 107 1 +116.78 ± 31.85 ↓
408585 ncm-dbt-04 567958 500 154 93 253 +42.6 ± 13.98 0 26 139 83 2 +83.57 ± 28.49 ↓
408584 ncm-dbt-03 585421 500 156 98 246 +40.48 ± 14.26 0 31 130 89 0 +82.1 ± 29.8 ↓
408583 ncm-dbt-02 580821 500 169 88 243 +56.78 ± 14.73 0 28 113 109 0 +116.78 ± 32.15 ↓
408582 ncm-dbt-05 582152 500 173 89 238 +58.93 ± 14.96 0 27 114 107 2 +118.33 ± 32.0 ↓
408581 ncm-dbt-01 581402 500 168 85 247 +58.21 ± 14.79 0 27 114 108 1 +118.33 ± 32.0 ↓
408580 ncm-dbt-04 570829 500 160 90 250 +48.96 ± 15.13 1 30 118 100 1 +99.95 ± 31.44 ↓
408579 ncm-dbt-03 584832 500 166 82 252 +58.93 ± 15.09 0 28 112 108 2 +118.33 ± 32.29 ↓
408578 ncm-dbt-02 587749 500 164 88 248 +53.22 ± 14.95 0 30 115 104 1 +107.54 ± 31.86 ↓
408577 ncm-dbt-05 587749 500 169 97 234 +50.38 ± 13.44 0 20 139 90 1 +101.46 ± 28.32 ↓
408576 ncm-dbt-04 571471 500 160 89 251 +49.67 ± 14.51 0 29 121 100 0 +101.46 ± 31.01 ↓
408575 ncm-dbt-01 585885 500 164 99 237 +45.42 ± 15.44 1 32 121 93 3 +89.48 ± 31.03 ↓
408574 ncm-dbt-03 586689 500 173 82 245 +63.95 ± 14.23 0 22 115 113 0 +132.54 ± 31.82 ↓
408573 ncm-dbt-02 587494 500 156 87 257 +48.25 ± 14.57 1 26 127 95 1 +98.44 ± 30.15 ↓
408572 ncm-dbt-04 568713 500 169 106 225 +44.01 ± 14.86 0 33 122 94 1 +88.0 ± 30.9 ↓
408571 ncm-dbt-05 585337 500 157 87 256 +48.96 ± 14.74 0 29 124 95 2 +96.94 ± 30.59 ↓
408570 ncm-dbt-01 581319 500 177 85 238 +64.66 ± 14.8 0 24 112 112 2 +130.94 ± 32.29 ↓
408569 ncm-dbt-03 583405 500 179 92 229 +61.08 ± 15.05 1 24 114 109 2 +124.6 ± 31.99 ↓
408568 ncm-dbt-02 585464 500 152 88 260 +44.72 ± 14.24 0 28 131 90 1 +89.48 ± 29.62 ↓
408567 ncm-dbt-05 583908 500 165 88 247 +53.93 ± 13.77 0 22 129 99 0 +110.6 ± 29.78 ↓
408566 ncm-dbt-04 572074 500 148 99 253 +34.16 ± 15.16 1 38 122 89 0 +70.44 ± 30.91 ↓
408565 ncm-dbt-01 581652 500 163 88 249 +52.51 ± 14.25 0 25 126 98 1 +106.01 ± 30.26 ↓
408564 ncm-dbt-03 586520 500 175 71 254 +73.34 ± 14.15 0 19 108 123 0 +153.86 ± 32.88 ↓
408563 ncm-dbt-02 584076 500 158 87 255 +49.67 ± 14.64 0 29 122 98 1 +99.95 ± 30.87 ↓
408562 ncm-dbt-05 583489 500 187 91 222 +67.54 ± 14.23 0 19 118 111 2 +137.37 ± 31.31 ↓
408561 ncm-dbt-01 583572 500 163 96 241 +46.84 ± 14.36 0 27 131 90 2 +92.46 ± 29.6 ↓
408560 ncm-dbt-04 572679 500 167 98 235 +48.25 ± 15.09 1 30 119 99 1 +98.44 ± 31.3 ↓
408559 ncm-dbt-03 587155 500 169 82 249 +61.07 ± 13.96 0 19 127 102 2 +123.02 ± 29.98 ↓
408558 ncm-dbt-02 584832 500 160 98 242 +43.3 ± 15.08 0 35 119 95 1 +86.52 ± 31.31 ↓
408557 ncm-dbt-01 581527 500 170 92 238 +54.64 ± 13.94 0 21 132 95 2 +109.07 ± 29.32 ↓
408556 ncm-dbt-05 585885 500 163 77 260 +60.36 ± 14.35 0 23 119 107 1 +123.02 ± 31.23 ↓
408555 ncm-dbt-04 572316 500 168 93 239 +52.51 ± 13.98 0 24 127 99 0 +107.54 ± 30.1 ↓
408554 ncm-dbt-03 586901 500 173 96 231 +53.93 ± 14.19 0 24 126 99 1 +109.07 ± 30.24 ↓

Commit

Commit ID 38e61663d836e062af0bc002814ad5149c4b7729
Author AndrovT
Date 2023-06-12 18:41:27 UTC
Use block sparse input for the first layer. Use block sparse input for the first fully connected layer on architectures with at least SSSE3. Depending on the CPU architecture, this yields a speedup of up to 10%, e.g. ``` Result of 100 runs of 'bench 16 1 13 default depth NNUE' base (...ockfish-base) = 959345 +/- 7477 test (...ckfish-patch) = 1054340 +/- 9640 diff = +94995 +/- 3999 speedup = +0.0990 P(speedup > 0) = 1.0000 CPU: 8 x AMD Ryzen 7 5700U with Radeon Graphics Hyperthreading: on ``` Passed STC: https://tests.stockfishchess.org/tests/view/6485aa0965ffe077ca12409c LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 8864 W: 2479 L: 2223 D: 4162 Ptnml(0-2): 13, 829, 2504, 1061, 25 This commit includes a net with reordered weights, to increase the likelihood of block sparse inputs, but otherwise equivalent to the previous master net (nn-ea57bea57e32.nnue). Activation data collected with https://github.com/AndrovT/Stockfish/tree/log-activations, running bench 16 1 13 varied_1000.epd depth NNUE on this data. Net parameters permuted with https://gist.github.com/AndrovT/9e3fbaebb7082734dc84d27e02094cb3. closes https://github.com/official-stockfish/Stockfish/pull/4612 No functional change
Copyright 2011–2025 Next Chess Move LLC
Feedback & Support Terms of Service Privacy Policy Desktop Version