Dev Builds » 20210619-2124

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 06:47:26 584050 4000 477 1606 1917 -100.8 ± 5.13 44 1135 728 92 1 -211.38 ± 12.62
ncm-dbt-02 06:47:54 585607 4000 507 1575 1918 -95.07 ± 5.15 34 1110 747 108 1 -199.05 ± 12.46
ncm-dbt-03 06:47:21 584985 4000 526 1580 1894 -93.76 ± 5.09 38 1076 789 96 1 -194.81 ± 12.11
ncm-dbt-04 06:47:30 567078 4000 449 1598 1953 -102.69 ± 4.94 38 1145 745 72 0 -217.6 ± 12.45
ncm-dbt-05 06:47:03 581577 4000 462 1611 1927 -102.69 ± 5.0 39 1151 730 80 0 -217.35 ± 12.59
20000 2421 7970 9609 -98.99 ± 2.26 193 5617 3739 448 3 -207.9 ± 5.56

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
453391 ncm-dbt-04 567878 500 66 201 233 -96.19 ± 14.11 4 138 97 11 0 -202.15 ± 34.7
453390 ncm-dbt-01 583614 500 54 205 241 -108.3 ± 15.28 9 146 82 13 0 -223.94 ± 37.87
453389 ncm-dbt-02 585928 500 65 205 230 -99.95 ± 15.05 6 143 86 15 0 -207.95 ± 36.95
453388 ncm-dbt-03 585928 500 56 196 248 -99.95 ± 14.15 6 137 98 9 0 -207.95 ± 34.43
453387 ncm-dbt-05 584579 500 54 205 241 -108.3 ± 14.05 4 153 83 10 0 -234.38 ± 37.67
453386 ncm-dbt-04 566296 500 46 193 261 -105.25 ± 13.05 3 146 96 5 0 -228.08 ± 34.59
453385 ncm-dbt-01 582861 500 57 205 238 -106.01 ± 13.56 3 150 89 8 0 -230.16 ± 36.26
453384 ncm-dbt-02 585843 500 62 196 242 -95.44 ± 14.55 6 134 98 12 0 -196.45 ± 34.53
453383 ncm-dbt-03 586012 500 66 204 230 -98.44 ± 14.59 5 140 94 10 1 -207.95 ± 35.29
453382 ncm-dbt-05 580530 500 48 201 251 -109.83 ± 14.22 7 147 88 8 0 -232.26 ± 36.48
453381 ncm-dbt-04 566217 500 61 200 239 -99.2 ± 13.67 4 139 99 8 0 -209.91 ± 34.19
453380 ncm-dbt-01 584117 500 69 200 231 -93.2 ± 14.81 8 127 103 12 0 -187.16 ± 33.6
453379 ncm-dbt-02 584495 500 55 194 251 -99.2 ± 13.99 4 141 95 10 0 -209.91 ± 35.06
453378 ncm-dbt-05 584201 500 54 201 245 -105.25 ± 14.66 4 153 79 14 0 -226.0 ± 38.55
453377 ncm-dbt-03 583740 500 68 192 240 -88.0 ± 13.81 2 132 104 12 0 -185.33 ± 33.42
453376 ncm-dbt-04 564958 500 57 206 237 -106.77 ± 13.72 3 152 86 9 0 -232.26 ± 36.96
453375 ncm-dbt-01 583740 500 56 195 249 -99.2 ± 13.99 3 144 92 11 0 -211.87 ± 35.7
453374 ncm-dbt-02 585590 500 59 207 234 -106.01 ± 14.04 3 153 83 11 0 -230.16 ± 37.67
453373 ncm-dbt-05 575720 500 66 213 221 -105.25 ± 13.72 5 144 94 7 0 -223.94 ± 35.14
453372 ncm-dbt-03 585632 500 64 196 240 -93.95 ± 14.82 6 134 96 14 0 -192.71 ± 34.94
453371 ncm-dbt-04 568156 500 58 200 242 -101.46 ± 14.32 6 140 94 10 0 -211.87 ± 35.26
453370 ncm-dbt-01 585843 500 54 199 247 -103.73 ± 15.95 10 141 84 14 1 -211.87 ± 37.38
453369 ncm-dbt-03 583196 500 74 192 234 -83.57 ± 14.01 3 125 109 13 0 -172.78 ± 32.58
453368 ncm-dbt-05 580199 500 42 195 263 -109.83 ± 14.84 5 157 74 14 0 -236.51 ± 39.78
453367 ncm-dbt-02 586520 500 72 190 238 -83.57 ± 15.02 3 131 98 17 1 -174.55 ± 34.58
453366 ncm-dbt-01 584579 500 63 209 228 -104.49 ± 14.35 6 144 90 10 0 -219.87 ± 36.1
453365 ncm-dbt-04 567007 500 45 201 254 -112.14 ± 13.73 6 150 88 6 0 -240.82 ± 36.39
453364 ncm-dbt-03 585126 500 71 207 222 -96.94 ± 15.16 5 143 85 17 0 -202.15 ± 37.11
453363 ncm-dbt-05 582527 500 60 205 235 -103.73 ± 13.38 5 140 100 5 0 -219.87 ± 33.78
453362 ncm-dbt-02 584117 500 60 196 244 -96.94 ± 14.72 6 137 94 13 0 -200.24 ± 35.32
453361 ncm-dbt-04 567839 500 59 187 254 -90.96 ± 14.32 4 133 100 13 0 -189.0 ± 34.18
453360 ncm-dbt-01 585042 500 62 200 238 -98.44 ± 13.98 3 143 93 11 0 -209.91 ± 35.49
453359 ncm-dbt-02 585632 500 64 202 234 -98.44 ± 14.59 6 138 94 12 0 -204.07 ± 35.31
453358 ncm-dbt-03 585084 500 67 199 234 -93.95 ± 14.07 5 132 103 10 0 -194.57 ± 33.53
453357 ncm-dbt-05 579951 500 64 200 236 -96.94 ± 14.12 5 136 99 10 0 -202.15 ± 34.28
453356 ncm-dbt-01 582611 500 62 193 245 -93.2 ± 14.06 2 140 95 13 0 -198.34 ± 35.12
453355 ncm-dbt-04 568275 500 57 210 233 -109.83 ± 14.69 8 147 85 10 0 -230.16 ± 37.2
453354 ncm-dbt-02 586731 500 70 185 245 -81.37 ± 14.25 0 133 99 18 0 -172.78 ± 34.4
453353 ncm-dbt-05 584916 500 74 191 235 -82.83 ± 14.0 4 121 113 12 0 -169.27 ± 31.88
453352 ncm-dbt-03 585169 500 60 194 246 -95.44 ± 14.4 6 133 100 11 0 -196.45 ± 34.12

Commit

Commit ID ba01f4b95448bcb324755f4dd2a632a57c6e67bc
Author MichaelB7
Date 2021-06-19 21:24:35 UTC
Make net nn-75980ca503c6.nnue the default. trained with the Python command c:\nnue>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_10 --resume-from-model ./pt/nn-3b20abec10c1.pt ` 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 . Net nn-3b20abec10c1.nnue was chosen as the --resume-from-model with the idea that through learning, the manually hex edited values will be learned and will not need to be manually adjusted going forward. They would also be fine tuned by the learning process. passed STC: https://tests.stockfishchess.org/tests/view/60cdf91e457376eb8bcab66f LLR: 2.95 (-2.94,2.94) <-0.50,2.50> Total: 18256 W: 1639 L: 1479 D: 15138 Ptnml(0-2): 59, 1179, 6505, 1313, 72 passed LTC: https://tests.stockfishchess.org/tests/view/60ce2166457376eb8bcab6e1 LLR: 2.94 (-2.94,2.94) <0.50,3.50> Total: 18792 W: 654 L: 542 D: 17596 Ptnml(0-2): 9, 490, 8291, 592, 14 closes https://github.com/official-stockfish/Stockfish/pull/3570 Bench: 5020972
Copyright 2011–2026 Next Chess Move LLC