Dev Builds » 20210621-2116

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 06:53:48 582113 4000 438 1597 1965 -103.64 ± 5.13 41 1172 693 93 1 -219.62 ± 12.94
ncm-dbt-02 06:55:54 585507 3994 465 1536 1993 -95.5 ± 5.11 34 1108 750 105 0 -199.9 ± 12.44
ncm-dbt-03 06:54:35 585150 3992 482 1586 1924 -98.65 ± 4.99 36 1113 767 79 1 -207.74 ± 12.27
ncm-dbt-04 06:55:25 565990 4014 460 1572 1982 -98.83 ± 5.08 45 1111 762 89 0 -205.83 ± 12.32
ncm-dbt-05 06:54:09 581498 4000 437 1600 1963 -104.02 ± 4.9 33 1168 728 71 0 -222.41 ± 12.6
20000 2282 7891 9827 -100.12 ± 2.26 189 5672 3700 437 2 -210.99 ± 5.59

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
438686 ncm-dbt-04 563939 14 3 6 5 -75.56 ± 67.89 0 3 4 0 0 -159.09 ± 177.81
438685 ncm-dbt-03 585801 492 50 193 249 -103.98 ± 14.81 7 140 88 11 0 -216.27 ± 36.54
438684 ncm-dbt-02 585759 494 49 193 252 -104.3 ± 13.82 3 147 88 9 0 -225.41 ± 36.5
438683 ncm-dbt-05 584874 500 59 216 225 -112.91 ± 14.38 7 152 82 9 0 -240.82 ± 37.9
438682 ncm-dbt-01 582235 500 54 201 245 -105.25 ± 14.66 5 150 82 13 0 -223.94 ± 37.87
438681 ncm-dbt-04 567165 500 60 198 242 -98.44 ± 13.98 4 140 96 10 0 -207.95 ± 34.86
438680 ncm-dbt-03 587494 500 65 207 228 -101.46 ± 13.85 4 143 94 9 0 -215.85 ± 35.23
438679 ncm-dbt-02 585632 500 54 190 256 -96.94 ± 13.96 2 144 92 12 0 -207.95 ± 35.71
438678 ncm-dbt-05 584874 500 65 192 243 -90.22 ± 14.01 4 130 105 11 0 -187.16 ± 33.21
438677 ncm-dbt-01 583154 500 55 197 248 -101.46 ± 15.07 4 151 78 17 0 -215.85 ± 38.65
438676 ncm-dbt-04 564683 500 55 205 240 -107.54 ± 14.52 5 152 81 12 0 -230.16 ± 38.13
438675 ncm-dbt-03 585548 500 59 201 240 -101.46 ± 14.01 5 141 95 9 0 -213.85 ± 35.03
438674 ncm-dbt-02 587494 500 67 183 250 -82.1 ± 13.98 1 129 105 15 0 -172.78 ± 33.32
438673 ncm-dbt-05 577561 500 49 203 248 -110.6 ± 13.39 1 161 79 9 0 -247.41 ± 38.64
438672 ncm-dbt-01 582235 500 57 209 234 -109.07 ± 14.99 9 145 85 11 0 -226.0 ± 37.21
438671 ncm-dbt-04 568077 500 56 189 255 -94.69 ± 14.54 4 139 93 14 0 -198.34 ± 35.52
438670 ncm-dbt-03 583154 500 65 195 240 -92.45 ± 13.74 2 137 100 11 0 -196.45 ± 34.12
438669 ncm-dbt-02 581028 500 57 204 239 -105.25 ± 14.2 6 144 91 9 0 -221.9 ± 35.86
438668 ncm-dbt-01 580406 500 53 198 249 -103.73 ± 14.19 7 139 96 8 0 -215.85 ± 34.78
438667 ncm-dbt-04 567126 500 65 198 237 -94.69 ± 14.09 4 136 99 11 0 -198.34 ± 34.31
438666 ncm-dbt-05 579702 500 51 198 251 -105.25 ± 13.39 3 148 92 7 0 -228.08 ± 35.56
438665 ncm-dbt-03 584453 500 67 191 242 -88.0 ± 13.97 4 127 108 11 0 -181.7 ± 32.68
438664 ncm-dbt-02 588302 500 66 184 250 -83.57 ± 15.43 4 132 92 22 0 -171.02 ± 35.61
438663 ncm-dbt-01 581111 500 40 194 266 -110.6 ± 14.22 4 157 78 11 0 -240.82 ± 38.87
438662 ncm-dbt-05 579702 500 50 195 255 -103.73 ± 14.19 6 142 93 9 0 -217.85 ± 35.44
438661 ncm-dbt-04 567601 500 48 196 256 -106.01 ± 13.56 3 150 89 8 0 -230.16 ± 36.26
438660 ncm-dbt-03 586604 500 58 207 235 -106.77 ± 14.05 6 145 91 8 0 -226.0 ± 35.82
438659 ncm-dbt-05 580738 500 56 197 247 -100.7 ± 13.03 2 143 99 6 0 -217.85 ± 34.06
438658 ncm-dbt-01 581402 500 60 195 245 -96.19 ± 13.95 2 143 93 12 0 -206.01 ± 35.51
438657 ncm-dbt-02 585042 500 49 199 252 -107.54 ± 14.37 9 139 95 7 0 -221.9 ± 34.93
438656 ncm-dbt-04 565823 500 49 198 253 -106.77 ± 14.82 11 135 96 8 0 -215.85 ± 34.78
438655 ncm-dbt-03 583992 500 62 187 251 -88.74 ± 14.28 3 133 100 14 0 -185.33 ± 34.19
438654 ncm-dbt-05 579744 500 58 201 241 -102.22 ± 14.33 4 147 87 12 0 -217.85 ± 36.77
438653 ncm-dbt-01 581111 500 63 205 232 -101.46 ± 14.32 6 139 97 7 1 -213.85 ± 34.58
438652 ncm-dbt-04 568593 500 63 194 243 -93.2 ± 14.81 7 130 100 13 0 -189.0 ± 34.18
438651 ncm-dbt-02 585464 500 62 187 251 -88.74 ± 14.86 5 131 98 16 0 -181.7 ± 34.58
438650 ncm-dbt-03 584159 500 56 205 239 -106.78 ± 14.05 5 147 91 6 1 -230.16 ± 35.78
438649 ncm-dbt-05 584790 500 49 198 253 -106.77 ± 14.05 6 145 91 8 0 -226.0 ± 35.82
438648 ncm-dbt-02 585337 500 61 196 243 -96.19 ± 14.71 4 142 89 15 0 -202.15 ± 36.32
438647 ncm-dbt-01 585253 500 56 198 246 -101.46 ± 14.63 4 148 84 14 0 -215.85 ± 37.4
438646 ncm-dbt-04 560904 500 61 188 251 -90.22 ± 14.75 7 126 104 13 0 -181.7 ± 33.45

Commit

Commit ID 9b82414b67aa7d1279e1cc99a6970ab766025bfa
Author MichaelB7
Date 2021-06-21 21:16:55 UTC
Make net nn-190f102a22c3.nnue the default net. Trained with the pytorch trainer: https://github.com/glinscott/nnue-pytorch python train.py i:/bin/all.binpack i:/bin/all.binpack --gpus 1 --threads 4 --num-workers 30 --batch-size 16384 --progress_bar_refresh_rate 300 --smart-fen-skipping --random-fen-skipping 3 --features=HalfKAv2^ --lambda=1.0 --max_epochs=440 --seed %random%%random% --default_root_dir exp/run_17 --resume-from-model ./pt/nn-75980ca503c6.pt This run is thus started from the previous master net. all.binpack equaled 4 parts Wrong_NNUE_2.binpack https://drive.google.com/file/d/1seGNOqcVdvK_vPNq98j-zV3XPE5zWAeq/view?usp=sharing plus two parts of Training_Data.binpack https://drive.google.com/file/d/1RFkQES3DpsiJqsOtUshENtzPfFgUmEff/view?usp=sharing Each set was concatenated together - making one large Wrong_NNUE 2 binpack and one large Training so the were approximately equal in size. They were then interleaved together. The idea was to give Wrong_NNUE.binpack closer to equal weighting with the Training_Data binpack passed LTC https://tests.stockfishchess.org/tests/view/60d09f52b4c17000d679517f LLR: 2.93 (-2.94,2.94) <0.50,3.50> Total: 32184 W: 1100 L: 970 D: 30114 Ptnml(0-2): 10, 878, 14193, 994, 17 passed STC https://tests.stockfishchess.org/tests/view/60d086c02114332881e7368e LLR: 2.93 (-2.94,2.94) <-0.50,2.50> Total: 11360 W: 1056 L: 906 D: 9398 Ptnml(0-2): 25, 735, 4026, 853, 41 closes https://github.com/official-stockfish/Stockfish/pull/3576 Bench: 4631244
Copyright 2011–2025 Next Chess Move LLC