Dev Builds » 20221221-0614

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 06:59:07 582638 4000 1140 896 1964 +21.22 ± 5.23 1 354 1050 590 5 +41.89 ± 10.49
ncm-dbt-02 06:57:14 585761 4000 1180 850 1970 +28.73 ± 5.18 4 297 1073 617 9 +56.96 ± 10.34
ncm-dbt-03 06:58:24 586539 4000 1190 856 1954 +29.08 ± 5.26 5 312 1032 646 5 +58.57 ± 10.58
ncm-dbt-04 06:58:57 568974 4008 1181 892 1935 +25.1 ± 5.23 2 332 1052 611 7 +49.57 ± 10.47
ncm-dbt-05 06:58:07 582865 3992 1211 850 1931 +31.5 ± 5.1 3 276 1080 631 6 +63.0 ± 10.29
20000 5902 4344 9754 +27.12 ± 2.33 15 1571 5287 3095 32 +53.97 ± 4.67

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
412655 ncm-dbt-04 570029 8 2 2 4 -0.0 ± 125.78 0 1 2 1 0 -0.0 ± 296.58 ↓
412654 ncm-dbt-05 579248 492 154 115 223 +27.6 ± 14.93 1 37 131 76 1 +55.55 ± 29.7 ↓
412653 ncm-dbt-01 580862 500 142 112 246 +20.87 ± 14.76 0 45 130 75 0 +41.89 ± 29.89 ↓
412652 ncm-dbt-03 588515 500 148 101 251 +32.76 ± 14.95 0 40 123 87 0 +66.1 ± 30.78 ↓
412651 ncm-dbt-02 585632 500 145 111 244 +23.66 ± 14.58 1 39 135 75 0 +48.96 ± 29.23 ↓
412650 ncm-dbt-04 570268 500 148 109 243 +27.16 ± 15.08 1 40 129 79 1 +54.65 ± 30.01 ↓
412649 ncm-dbt-05 590396 500 152 113 235 +27.16 ± 14.95 1 40 128 81 0 +56.07 ± 30.14 ↓
412648 ncm-dbt-01 582569 500 135 114 251 +14.6 ± 15.4 0 53 124 72 1 +27.85 ± 30.66 ↓
412647 ncm-dbt-03 586350 500 139 106 255 +22.96 ± 14.66 0 42 134 73 1 +44.72 ± 29.37 ↓
412646 ncm-dbt-02 584832 500 137 102 261 +24.36 ± 14.24 1 36 140 73 0 +50.38 ± 28.55 ↓
412645 ncm-dbt-04 567839 500 138 115 247 +15.99 ± 14.51 0 46 135 69 0 +32.05 ± 29.26 ↓
412644 ncm-dbt-05 581735 500 155 97 248 +40.48 ± 13.71 0 27 138 85 0 +82.1 ± 28.65 ↓
412643 ncm-dbt-01 582861 500 146 116 238 +20.87 ± 14.76 0 44 133 72 1 +40.48 ± 29.51 ↓
412642 ncm-dbt-03 586223 500 145 114 241 +21.57 ± 15.91 2 48 117 83 0 +46.13 ± 31.53 ↓
412641 ncm-dbt-02 586181 500 146 95 259 +35.56 ± 14.08 0 32 135 83 0 +71.89 ± 29.14 ↓
412640 ncm-dbt-04 567839 500 159 105 236 +37.67 ± 15.38 1 38 117 94 0 +77.71 ± 31.57 ↓
412639 ncm-dbt-05 582903 500 137 98 265 +27.15 ± 14.04 0 35 142 72 1 +53.22 ± 28.26 ↓
412638 ncm-dbt-01 583489 500 141 120 239 +14.6 ± 14.27 0 45 139 66 0 +29.25 ± 28.74 ↓
412637 ncm-dbt-03 583698 500 152 110 238 +29.25 ± 15.46 1 42 122 84 1 +58.93 ± 30.91 ↓
412636 ncm-dbt-02 587325 500 150 126 224 +16.69 ± 15.19 1 48 127 74 0 +34.86 ± 30.28 ↓
412635 ncm-dbt-04 569111 500 152 121 227 +21.57 ± 14.42 0 40 141 67 2 +40.48 ± 28.44 ↓
412634 ncm-dbt-05 584706 500 154 117 229 +25.76 ± 13.81 0 34 146 69 1 +50.38 ± 27.71 ↓
412633 ncm-dbt-03 584832 500 145 115 240 +20.87 ± 14.11 2 33 149 65 1 +43.3 ± 27.33 ↓
412632 ncm-dbt-01 582027 500 153 108 239 +31.35 ± 14.47 0 36 134 79 1 +61.79 ± 29.32 ↓
412631 ncm-dbt-02 584622 500 155 99 246 +39.08 ± 15.34 0 38 120 90 2 +76.25 ± 31.18 ↓
412630 ncm-dbt-04 569390 500 139 109 252 +20.87 ± 14.37 0 41 139 69 1 +40.48 ± 28.71 ↓
412629 ncm-dbt-05 578588 500 147 109 244 +26.46 ± 14.9 1 40 129 80 0 +54.65 ± 30.01 ↓
412628 ncm-dbt-01 581902 500 148 105 247 +29.95 ± 14.89 0 40 128 81 1 +58.93 ± 30.13 ↓
412627 ncm-dbt-02 583992 500 151 103 246 +33.46 ± 13.94 0 32 138 80 0 +67.55 ± 28.74 ↓
412626 ncm-dbt-03 587834 500 147 91 262 +39.08 ± 14.44 0 33 128 89 0 +79.17 ± 30.09 ↓
412625 ncm-dbt-04 568116 500 144 111 245 +22.96 ± 14.91 0 44 130 75 1 +44.72 ± 29.89 ↓
412624 ncm-dbt-05 582152 500 166 99 235 +46.84 ± 14.88 0 31 123 94 2 +92.46 ± 30.75 ↓
412623 ncm-dbt-02 586054 500 147 112 241 +24.36 ± 14.24 0 38 140 71 1 +47.55 ± 28.56 ↓
412622 ncm-dbt-03 586308 500 163 109 228 +37.67 ± 14.74 0 34 130 84 2 +73.34 ± 29.83 ↓
412621 ncm-dbt-01 584034 500 142 111 247 +21.57 ± 15.31 1 45 127 76 1 +43.3 ± 30.28 ↓
412620 ncm-dbt-04 568951 500 149 106 245 +29.95 ± 15.02 0 42 123 85 0 +60.36 ± 30.78 ↓
412619 ncm-dbt-05 583196 500 146 102 252 +30.65 ± 13.89 0 32 143 74 1 +60.36 ± 28.08 ↓
412618 ncm-dbt-02 587452 500 149 102 249 +32.76 ± 15.32 1 34 138 71 6 +58.93 ± 28.79 ↓
412617 ncm-dbt-01 583363 500 133 110 257 +15.99 ± 14.51 0 46 135 69 0 +32.05 ± 29.26 ↓
412616 ncm-dbt-03 588558 500 151 110 239 +28.55 ± 14.67 0 40 129 81 0 +57.5 ± 30.0 ↓
412615 ncm-dbt-04 569230 500 150 114 236 +25.06 ± 14.68 0 40 136 72 2 +47.55 ± 29.1 ↓

Commit

Commit ID c620886181b5c0fb4a0c067b4173143a81819199
Author Linmiao Xu
Date 2022-12-21 06:14:58 UTC
Update default net to nn-335a9b2d8a80.nnue Created by retraining the master net with a combination of: the previous best dataset (Leela-dfrc_n5000.binpack), with about half the dataset filtered using depth6 multipv2 search to throw away positions where either of the 2 best moves are captures Leela T80 Oct and Nov training data rescored with best moves, adding ~9.5 billion positions Trained effectively the same way as the previous master net: python3 easy_train.py \ --experiment-name=leela-dfrc-filtered-T80-oct-nov \ --training-dataset=/data/leela-dfrc-filtered-T80-oct-nov.binpack \ --start-from-engine-test-net True \ --gpus="0," \ --start-lambda=1.0 \ --end-lambda=0.75 \ --gamma=0.995 \ --lr=4.375e-4 \ --tui=False \ --seed=$RANDOM \ --max_epoch=800 \ --auto-exit-timeout-on-training-finished=900 \ --network-testing-threads 20 \ --num-workers 6 Local testing at a fixed 25k nodes: experiments/experiment_leela-dfrc-filtered-T80-oct-nov/training/run_0/nn-epoch779.nnue localElo: run_0/nn-epoch779.nnue : 4.7 +/- 3.1 The new Leela T80 part of the dataset was prepared by downloading test80 training data from all of Oct 2022 and Nov 2022, rescoring with syzygy 6-piece tablebases and ~600 GB of 7-piece tablebases, saving best moves to exported .plain files, removing all positions with castling flags, then converting to binpacks and using interleave_binpacks.py to merge them together. Scripts used in this data conversion process are available at: https://github.com/linrock/lc0-data-converter Filtering binpack data using depth6 multipv2 search was done by modifying transform.cpp in the tools branch: https://github.com/linrock/Stockfish/tree/tools-filter-multipv2-no-rescore Links for downloading the training data (total size: 338 GB) are available at: https://robotmoon.com/nnue-training-data/ Passed STC: LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 30544 W: 8244 L: 7947 D: 14353 Ptnml(0-2): 93, 3243, 8302, 3542, 92 https://tests.stockfishchess.org/tests/view/63a0d377264a0cf18f86f82b Passed LTC: LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 32464 W: 8866 L: 8573 D: 15025 Ptnml(0-2): 19, 3054, 9794, 3345, 20 https://tests.stockfishchess.org/tests/view/63a10bc9fb452d3c44b1e016 closes https://github.com/official-stockfish/Stockfish/pull/4295 Bench 3554904
Copyright 2011–2025 Next Chess Move LLC
Feedback & Support Terms of Service Privacy Policy Desktop Version