Dev Builds » 20210614-0724

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 06:42:47 583088 4000 403 1658 1939 -112.81 ± 5.06 63 1191 684 62 0 -238.66 ± 13.01
ncm-dbt-02 06:40:55 586395 4000 418 1623 1959 -108.02 ± 5.15 64 1152 709 75 0 -225.23 ± 12.78
ncm-dbt-03 06:40:20 584973 4000 478 1648 1874 -104.68 ± 5.28 70 1120 720 90 0 -214.85 ± 12.69
ncm-dbt-04 06:39:48 567352 4000 428 1636 1936 -108.3 ± 5.17 55 1184 676 84 1 -228.6 ± 13.11
ncm-dbt-05 06:42:06 581383 4000 438 1620 1942 -105.82 ± 5.27 61 1155 690 93 1 -220.38 ± 12.97
20000 2165 8185 9650 -107.92 ± 2.32 313 5802 3479 404 2 -225.44 ± 5.77

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
441969 ncm-dbt-01 584117 500 49 211 240 -116.77 ± 14.37 9 151 83 7 0 -247.41 ± 37.61
441968 ncm-dbt-05 581943 500 58 198 244 -99.95 ± 13.84 3 144 93 10 0 -213.85 ± 35.47
441967 ncm-dbt-04 569071 500 51 202 247 -108.3 ± 14.21 6 148 87 9 0 -230.16 ± 36.73
441966 ncm-dbt-02 587962 500 60 195 245 -96.19 ± 14.71 9 127 104 10 0 -192.71 ± 33.34
441965 ncm-dbt-03 586562 500 68 207 225 -99.2 ± 15.04 7 139 90 14 0 -204.07 ± 36.13
441964 ncm-dbt-01 584832 500 41 200 259 -114.45 ± 13.21 4 156 85 5 0 -251.89 ± 37.03
441963 ncm-dbt-04 563587 500 62 200 238 -98.44 ± 14.29 4 142 92 12 0 -207.95 ± 35.71
441962 ncm-dbt-05 584285 500 54 205 241 -108.3 ± 16.69 12 147 71 20 0 -217.85 ± 40.1
441961 ncm-dbt-02 583950 500 51 204 245 -109.83 ± 14.22 6 150 85 9 0 -234.38 ± 37.19
441960 ncm-dbt-03 586858 500 69 196 235 -90.22 ± 15.86 10 125 97 18 0 -176.33 ± 34.76
441959 ncm-dbt-01 587367 500 44 203 253 -114.45 ± 14.53 9 149 84 8 0 -240.82 ± 37.4
441958 ncm-dbt-05 580282 500 61 212 227 -108.3 ± 14.98 8 147 83 12 0 -226.0 ± 37.66
441957 ncm-dbt-04 567997 500 46 190 264 -102.97 ± 13.54 4 143 96 7 0 -219.87 ± 34.73
441956 ncm-dbt-02 587325 500 63 200 237 -97.69 ± 15.17 6 141 87 16 0 -202.15 ± 36.72
441955 ncm-dbt-03 583992 500 61 208 231 -105.25 ± 14.81 5 151 80 14 0 -223.94 ± 38.31
441954 ncm-dbt-05 578095 500 48 205 247 -112.91 ± 13.89 7 149 88 6 0 -240.82 ± 36.39
441953 ncm-dbt-01 582778 500 48 216 236 -121.45 ± 14.5 8 161 72 9 0 -263.42 ± 40.52
441952 ncm-dbt-04 568037 500 53 201 246 -106.01 ± 15.26 6 152 76 16 0 -223.94 ± 39.17
441951 ncm-dbt-03 585885 500 58 213 229 -111.37 ± 13.89 6 150 87 7 0 -238.66 ± 36.67
441950 ncm-dbt-02 585126 500 42 207 251 -119.11 ± 14.84 12 148 83 7 0 -247.41 ± 37.61
441949 ncm-dbt-01 582485 500 58 216 226 -113.68 ± 14.54 7 154 79 10 0 -243.0 ± 38.64
441948 ncm-dbt-04 567442 500 54 211 235 -112.91 ± 14.54 7 153 80 10 0 -240.82 ± 38.39
441947 ncm-dbt-05 581610 500 56 191 253 -96.19 ± 15.43 6 141 85 18 0 -198.34 ± 37.08
441946 ncm-dbt-03 587792 500 54 202 244 -106.01 ± 14.82 9 140 91 10 0 -217.85 ± 35.89
441945 ncm-dbt-02 590353 500 49 207 244 -113.68 ± 14.85 9 150 81 10 0 -238.66 ± 38.15
441944 ncm-dbt-04 567640 500 52 220 228 -121.46 ± 14.98 8 163 69 9 1 -265.78 ± 41.36
441943 ncm-dbt-01 580779 500 52 210 238 -113.68 ± 14.69 10 146 86 8 0 -236.51 ± 36.93
441942 ncm-dbt-05 582402 500 55 206 239 -108.3 ± 14.37 8 143 91 8 0 -226.0 ± 35.82
441941 ncm-dbt-02 587707 500 52 199 249 -105.25 ± 14.04 8 137 99 6 0 -217.85 ± 34.06
441940 ncm-dbt-03 584243 500 55 205 240 -107.54 ± 14.68 11 135 97 7 0 -217.85 ± 34.53
441939 ncm-dbt-04 568196 500 46 196 258 -107.54 ± 14.98 9 143 87 11 0 -221.9 ± 36.76
441938 ncm-dbt-01 580530 500 57 205 238 -106.01 ± 14.36 9 137 97 7 0 -217.85 ± 34.53
441937 ncm-dbt-02 585590 500 56 205 239 -106.77 ± 14.21 6 146 89 9 0 -226.0 ± 36.29
441936 ncm-dbt-05 578918 500 53 198 249 -103.73 ± 15.67 10 139 88 12 1 -211.87 ± 36.55
441935 ncm-dbt-03 585000 500 63 202 235 -99.2 ± 15.04 9 133 96 12 0 -200.24 ± 34.91
441934 ncm-dbt-04 566849 500 64 216 220 -109.07 ± 15.14 11 140 89 10 0 -221.9 ± 36.31
441933 ncm-dbt-03 579455 500 50 215 235 -119.11 ± 15.15 13 147 82 8 0 -245.2 ± 37.88
441932 ncm-dbt-02 583154 500 45 206 249 -116.0 ± 14.37 8 153 81 8 0 -247.41 ± 38.13
441931 ncm-dbt-05 583531 500 53 205 242 -109.07 ± 14.05 7 145 91 7 0 -230.16 ± 35.78
441930 ncm-dbt-01 581818 500 54 197 249 -102.22 ± 14.18 7 137 98 8 0 -211.87 ± 34.38

Commit

Commit ID f8c779dbe538315aa6f65556d0acf11640558504
Author JWmer
Date 2021-06-14 07:24:07 UTC
Update default net to nn-8e47cf062333.nnue This net is the result of training on data used by the Leela project. More precisely, we shuffled T60 and T74 data kindly provided by borg (for different Tnn, the data is a result of Leela selfplay with differently sized Leela nets). The data is available at vondele's google drive: https://drive.google.com/drive/folders/1mftuzYdl9o6tBaceR3d_VBQIrgKJsFpl. The Leela data comes in small chunks of .binpack files. To shuffle them, we simply used a small python script to randomly rename the files, and then concatenated them using `cat`. As validation data we picked a file of T60 data. We will further investigate T74 data. The training for the NNUE architecture used 200 epochs with the Python trainer from the Stockfish project. Unlike the previous run we tried with this data, this run does not have adjusted scaling — not because we didn't want to, but because we forgot. However, this training randomly skips 40% more positions than previous run. The loss was very spiky and decreased slower than it does usually. Training loss: https://github.com/official-stockfish/images/blob/main/training-loss-8e47cf062333.png Validation loss: https://github.com/official-stockfish/images/blob/main/validation-loss-8e47cf062333.png This is the exact training command: python train.py --smart-fen-skipping --random-fen-skipping 14 --batch-size 16384 --threads 4 --num-workers 4 --gpus 1 trainingdata\training_data.binpack validationdata\val.binpack --- 10k STC result: ELO: 3.61 +-3.3 (95%) LOS: 98.4% Total: 10000 W: 1241 L: 1137 D: 7622 Ptnml(0-2): 68, 841, 3086, 929, 76 https://tests.stockfishchess.org/tests/view/60c67e50457376eb8bcaae70 10k LTC result: ELO: 2.71 +-2.4 (95%) LOS: 98.8% Total: 10000 W: 659 L: 581 D: 8760 Ptnml(0-2): 22, 485, 3900, 579, 14 https://tests.stockfishchess.org/tests/view/60c69deb457376eb8bcaae98 Passed LTC: LLR: 2.93 (-2.94,2.94) <0.50,3.50> Total: 9648 W: 685 L: 545 D: 8418 Ptnml(0-2): 22, 448, 3740, 596, 18 https://tests.stockfishchess.org/tests/view/60c6d41c457376eb8bcaaecf --- closes https://github.com/official-stockfish/Stockfish/pull/3550 Bench: 4877339
Copyright 2011–2025 Next Chess Move LLC