Dev Builds » 20230701-1134

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 06:53:39 584084 4000 1316 769 1915 +47.81 ± 5.17 3 226 1002 759 10 +96.19 ± 10.73
ncm-dbt-02 06:54:19 585165 4000 1380 736 1884 +56.43 ± 5.13 2 197 966 825 10 +114.45 ± 10.93
ncm-dbt-03 06:55:12 586406 4000 1348 735 1917 +53.67 ± 5.1 3 197 993 798 9 +108.88 ± 10.76
ncm-dbt-04 06:53:15 570574 4000 1320 741 1939 +50.65 ± 5.04 4 196 1024 769 7 +102.97 ± 10.57
ncm-dbt-05 06:54:25 584732 4000 1350 708 1942 +56.25 ± 5.12 3 195 967 827 8 +114.65 ± 10.92
20000 6714 3689 9597 +52.96 ± 2.29 15 1011 4952 3978 44 +107.39 ± 4.82

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
393508 ncm-dbt-01 582611 500 150 101 249 +34.16 ± 14.26 0 32 139 77 2 +66.1 ± 28.61
393507 ncm-dbt-03 586774 500 174 100 226 +51.8 ± 14.49 0 26 126 96 2 +102.97 ± 30.28
393506 ncm-dbt-05 582235 500 164 86 250 +54.65 ± 14.49 1 24 121 104 0 +113.68 ± 30.97
393505 ncm-dbt-04 571431 500 175 104 221 +49.67 ± 13.69 1 20 136 93 0 +102.97 ± 28.77
393504 ncm-dbt-02 585379 500 165 97 238 +47.55 ± 14.66 1 25 132 89 3 +93.95 ± 29.45
393503 ncm-dbt-03 587919 500 170 93 237 +53.93 ± 14.73 0 28 118 103 1 +109.07 ± 31.43
393502 ncm-dbt-05 579413 500 165 92 243 +51.09 ± 14.31 0 27 123 100 0 +104.49 ± 30.71
393501 ncm-dbt-01 581985 500 159 99 242 +41.89 ± 14.61 1 29 130 89 1 +85.04 ± 29.79
393500 ncm-dbt-04 572195 500 154 82 264 +50.38 ± 13.72 0 23 132 95 0 +102.97 ± 29.38
393499 ncm-dbt-02 587197 500 171 91 238 +56.07 ± 13.87 0 22 126 102 0 +115.23 ± 30.2
393498 ncm-dbt-03 583908 500 162 91 247 +49.67 ± 15.16 0 33 114 102 1 +99.95 ± 32.0
393497 ncm-dbt-05 582611 500 170 87 243 +58.21 ± 15.06 1 25 116 106 2 +118.33 ± 31.7
393496 ncm-dbt-02 585211 500 177 90 233 +61.07 ± 14.92 0 26 113 109 2 +123.02 ± 32.15
393495 ncm-dbt-04 570469 500 164 84 252 +56.07 ± 14.83 1 25 118 105 1 +115.23 ± 31.41
393494 ncm-dbt-01 586901 500 164 90 246 +51.8 ± 14.08 0 23 132 93 2 +102.97 ± 29.38
393493 ncm-dbt-03 585717 500 175 85 240 +63.23 ± 14.06 0 20 121 108 1 +129.35 ± 30.88
393492 ncm-dbt-04 569949 500 168 88 244 +56.07 ± 14.43 0 25 121 103 1 +113.68 ± 30.97
393491 ncm-dbt-05 583656 500 166 100 234 +46.13 ± 14.58 1 27 128 93 1 +93.95 ± 30.03
393490 ncm-dbt-02 585759 500 173 103 224 +48.96 ± 14.07 1 23 131 95 0 +101.46 ± 29.54
393489 ncm-dbt-01 584285 500 171 89 240 +57.5 ± 14.76 1 25 115 109 0 +119.89 ± 31.85
393488 ncm-dbt-04 570068 500 172 101 227 +49.67 ± 13.97 0 24 132 93 1 +99.95 ± 29.4
393487 ncm-dbt-01 581652 500 175 88 237 +61.07 ± 14.52 0 24 116 109 1 +124.6 ± 31.69
393486 ncm-dbt-05 585928 500 164 83 253 +56.78 ± 14.05 0 20 132 95 3 +112.14 ± 29.29
393485 ncm-dbt-03 586520 500 163 84 253 +55.36 ± 14.39 2 19 128 100 1 +115.23 ± 29.89
393484 ncm-dbt-02 581693 500 175 91 234 +58.93 ± 14.83 0 27 113 109 1 +119.89 ± 32.15
393483 ncm-dbt-04 569509 500 158 93 249 +45.42 ± 14.81 1 30 122 97 0 +93.95 ± 30.89
393482 ncm-dbt-02 583950 500 179 80 241 +69.71 ± 14.31 0 20 112 117 1 +143.89 ± 32.25
393481 ncm-dbt-01 582903 500 165 102 233 +44.01 ± 14.99 0 34 120 95 1 +88.0 ± 31.17
393480 ncm-dbt-05 584201 500 171 85 244 +60.36 ± 13.79 0 19 127 103 1 +123.02 ± 29.98
393479 ncm-dbt-03 586943 500 167 105 228 +43.3 ± 13.17 0 22 144 84 0 +88.0 ± 27.7
393478 ncm-dbt-03 585590 500 166 84 250 +57.5 ± 14.36 0 23 124 101 2 +115.23 ± 30.51
393477 ncm-dbt-02 583405 500 170 85 245 +59.64 ± 14.18 0 22 122 105 1 +121.46 ± 30.78
393476 ncm-dbt-04 570388 500 161 93 246 +47.55 ± 13.99 1 21 139 87 2 +95.44 ± 28.38
393475 ncm-dbt-05 588387 500 172 82 246 +63.23 ± 14.88 0 27 106 117 0 +130.94 ± 33.2
393474 ncm-dbt-01 583992 500 167 94 239 +51.09 ± 14.45 1 23 130 94 2 +102.97 ± 29.68
393473 ncm-dbt-04 570589 500 168 96 236 +50.38 ± 14.68 0 28 124 96 2 +99.95 ± 30.58
393472 ncm-dbt-01 588345 500 165 106 229 +41.19 ± 15.08 0 36 120 93 1 +82.1 ± 31.18
393471 ncm-dbt-05 591426 500 178 93 229 +59.64 ± 14.73 0 26 114 109 1 +121.46 ± 32.0
393470 ncm-dbt-03 587877 500 171 93 236 +54.65 ± 14.89 1 26 118 104 1 +112.14 ± 31.42
393469 ncm-dbt-02 588728 500 170 99 231 +49.67 ± 15.16 0 32 117 99 2 +98.44 ± 31.58

Commit

Commit ID 915532181f11812c80ef0b57bc018de4ea2155ec
Author Linmiao Xu
Date 2023-07-01 11:34:30 UTC
Update NNUE architecture to SFNNv7 with larger L1 size of 2048 Creating this net involved: - a 5-step training process from scratch - greedy permuting L1 weights with https://github.com/official-stockfish/Stockfish/pull/4620 - leb128 compression with https://github.com/glinscott/nnue-pytorch/pull/251 - greedy 2- and 3- cycle permuting with https://github.com/official-stockfish/Stockfish/pull/4640 The 5 training steps were: 1. 400 epochs, lambda 1.0, lr 9.75e-4 UHOx2-wIsRight-multinet-dfrc-n5000-largeGensfen-d9.binpack (178G) nodes5000pv2_UHO.binpack data_pv-2_diff-100_nodes-5000.binpack wrongIsRight_nodes5000pv2.binpack multinet_pv-2_diff-100_nodes-5000.binpack dfrc_n5000.binpack large_gensfen_multipvdiff_100_d9.binpack ep399 chosen as start model for step2 2. 800 epochs, end-lambda 0.75, skip 16 LeelaFarseer-T78juntoaugT79marT80dec.binpack (141G) T60T70wIsRightFarseerT60T74T75T76.binpack test78-junjulaug2022-16tb7p.no-db.min.binpack test79-mar2022-16tb7p.no-db.min.binpack test80-dec2022-16tb7p.no-db.min.binpack ep559 chosen as start model for step3 3. 800 epochs, end-lambda 0.725, skip 20 leela96-dfrc99-v2-T80dectofeb-sk20-mar-v6-T77decT78janfebT79apr.binpack (223G) leela96-filt-v2.min.binpack dfrc99-16tb7p-eval-filt-v2.min.binpack test80-dec2022-16tb7p-filter-v6-sk20.min-mar2023.binpack test80-jan2023-16tb7p-filter-v6-sk20.min-mar2023.binpack test80-feb2023-16tb7p-filter-v6-sk20.min-mar2023.binpack test80-mar2023-2tb7p-filter-v6.min.binpack test77-dec2021-16tb7p.no-db.min.binpack test78-janfeb2022-16tb7p.no-db.min.binpack test79-apr2022-16tb7p.no-db.min.binpack ep499 chosen as start model for step4 4. 800 epochs, end-lambda 0.7, skip 24 0dd1cebea57 dataset https://github.com/official-stockfish/Stockfish/pull/4606 ep599 chosen as start model for step5 5. 800 epochs, end-lambda 0.7, skip 28 same dataset as step4 ep619 became nn-1b951f8b449d.nnue For the final step5 training: python3 easy_train.py \ --experiment-name L1-2048-S5-sameData-sk28-S4-0dd1cebea57-shuffled-S3-leela96-dfrc99-v2-T80dectofeb-sk20-mar-v6-T77decT78janfebT79apr-sk20-S2-LeelaFarseerT78T79T80-ep399-S1-UHOx2-wIsRight-multinet-dfrc-n5000-largeGensfen-d9 \ --training-dataset /data/leela96-dfrc99-T60novdec-v2-T80juntonovjanfebT79aprmayT78jantosepT77dec-v6dd-T80apr.binpack \ --early-fen-skipping 28 \ --nnue-pytorch-branch linrock/nnue-pytorch/misc-fixes-L1-2048 \ --engine-test-branch linrock/Stockfish/L1-2048 \ --start-from-engine-test-net False \ --start-from-model /data/experiments/experiment_L1-2048-S4-0dd1cebea57-shuffled-S3-leela96-dfrc99-v2-T80dectofeb-sk20-mar-v6-T77decT78janfebT79apr-sk20-S2-LeelaFarseerT78T79T80-ep399-S1-UHOx2-wIsRight-multinet-dfrc-n5000-largeGensfen-d9/training/run_0/nn-epoch599.nnue --max_epoch 800 \ --lr 4.375e-4 \ --gamma 0.995 \ --start-lambda 1.0 \ --end-lambda 0.7 \ --tui False \ --seed $RANDOM \ --gpus 0 SF training data components for the step1 dataset: https://drive.google.com/drive/folders/1yLCEmioC3Xx9KQr4T7uB6GnLm5icAYGU Leela training data for steps 2-5 can be found at: https://robotmoon.com/nnue-training-data/ Due to larger L1 size and slower inference, the speed penalty loses elo at STC. Measurements from 100 bench runs at depth 13 with x86-64-modern on Intel Core i5-1038NG7 2.00GHz: sf_base = 1240730 +/- 3443 (95%) sf_test = 1153341 +/- 2832 (95%) diff = -87388 +/- 1616 (95%) speedup = -7.04330% +/- 0.130% (95%) Local elo at 25k nodes per move (vs. L1-1536 nn-fdc1d0fe6455.nnue): nn-epoch619.nnue : 21.1 +/- 3.2 Failed STC: https://tests.stockfishchess.org/tests/view/6498ee93dc7002ce609cf979 LLR: -2.95 (-2.94,2.94) <0.00,2.00> Total: 11680 W: 3058 L: 3299 D: 5323 Ptnml(0-2): 44, 1422, 3149, 1181, 44 LTC: https://tests.stockfishchess.org/tests/view/649b32f5dc7002ce609d20cf Elo: 0.68 ± 1.5 (95%) LOS: 80.5% Total: 40000 W: 10887 L: 10809 D: 18304 Ptnml(0-2): 36, 3938, 11958, 4048, 20 nElo: 1.50 ± 3.4 (95%) PairsRatio: 1.02 Passed VLTC 180+1.8: https://tests.stockfishchess.org/tests/view/64992b43dc7002ce609cfd20 LLR: 3.06 (-2.94,2.94) <0.00,2.00> Total: 38086 W: 10612 L: 10338 D: 17136 Ptnml(0-2): 9, 3316, 12115, 3598, 5 Passed VLTC SMP 60+0.6 th 8: https://tests.stockfishchess.org/tests/view/649a21fedc7002ce609d0c7d LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 38936 W: 11091 L: 10820 D: 17025 Ptnml(0-2): 1, 2948, 13305, 3207, 7 closes https://github.com/official-stockfish/Stockfish/pull/4646 Bench: 2505168
Copyright 2011–2025 Next Chess Move LLC