Dev Builds » 20210526-2321

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 06:46:57 580741 3996 365 1732 1899 -123.85 ± 5.13 85 1250 610 53 0 -264.38 ± 13.79
ncm-dbt-02 06:46:37 585133 4016 358 1763 1895 -126.91 ± 5.02 72 1315 567 54 0 -277.83 ± 14.32
ncm-dbt-03 06:44:05 585505 4000 363 1785 1852 -129.15 ± 4.93 82 1290 596 32 0 -281.68 ± 13.92
ncm-dbt-04 06:45:00 567111 4000 380 1792 1828 -128.16 ± 5.12 90 1281 580 49 0 -276.07 ± 14.15
ncm-dbt-05 06:46:48 583470 3988 344 1786 1858 -131.58 ± 5.13 92 1307 546 49 0 -286.15 ± 14.59
20000 1810 8858 9332 -127.92 ± 2.27 421 6443 2899 237 0 -277.12 ± 6.32

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
445481 ncm-dbt-02 587070 16 1 7 8 -136.89 ± 61.78 0 6 2 0 0 -337.87 ± 515.83
445480 ncm-dbt-05 584579 488 43 210 235 -123.89 ± 14.9 11 152 74 7 0 -263.03 ± 39.96
445479 ncm-dbt-01 581111 496 46 215 235 -123.31 ± 16.15 14 154 67 13 0 -254.73 ± 41.76
445478 ncm-dbt-03 585464 500 44 220 236 -127.76 ± 13.57 8 164 74 4 0 -282.94 ± 39.86
445477 ncm-dbt-04 568037 500 42 211 247 -122.24 ± 14.66 11 154 78 7 0 -258.75 ± 38.88
445476 ncm-dbt-02 583531 500 48 219 233 -123.81 ± 13.46 7 161 78 4 0 -273.0 ± 38.74
445475 ncm-dbt-01 579001 500 47 221 232 -126.17 ± 14.79 11 160 71 8 0 -270.57 ± 40.83
445474 ncm-dbt-05 585000 500 51 220 229 -122.24 ± 15.6 13 154 72 11 0 -254.16 ± 40.45
445473 ncm-dbt-03 588430 500 37 215 248 -129.35 ± 14.41 14 153 80 3 0 -273.0 ± 38.12
445472 ncm-dbt-04 564840 500 52 225 223 -125.38 ± 15.74 14 156 69 11 0 -261.07 ± 41.31
445471 ncm-dbt-02 586731 500 45 213 242 -121.45 ± 14.17 6 165 70 9 0 -268.17 ± 41.1
445470 ncm-dbt-05 584117 500 50 222 228 -124.6 ± 14.96 12 156 74 8 0 -263.42 ± 39.97
445469 ncm-dbt-01 579537 500 51 218 231 -120.67 ± 14.35 9 156 78 7 0 -258.75 ± 38.88
445468 ncm-dbt-03 584664 500 49 224 227 -126.97 ± 13.94 11 156 80 3 0 -273.0 ± 38.12
445467 ncm-dbt-04 566060 500 38 220 242 -132.54 ± 13.83 10 166 70 4 0 -293.29 ± 41.07
445466 ncm-dbt-02 579042 500 42 219 239 -128.55 ± 14.42 12 158 75 5 0 -275.45 ± 39.63
445465 ncm-dbt-05 582318 500 44 229 227 -134.95 ± 14.99 16 158 71 5 0 -285.49 ± 40.81
445464 ncm-dbt-01 579826 500 50 215 235 -119.11 ± 14.36 12 145 89 4 0 -247.41 ± 36.02
445463 ncm-dbt-04 566454 500 60 220 220 -115.22 ± 15.31 13 143 85 9 0 -234.38 ± 37.19
445462 ncm-dbt-02 587240 500 49 215 236 -119.89 ± 14.51 8 159 74 9 0 -258.75 ± 39.96
445461 ncm-dbt-03 586646 500 48 237 215 -138.18 ± 13.69 10 173 63 4 0 -312.48 ± 43.44
445460 ncm-dbt-05 579951 500 40 222 238 -132.54 ± 14.18 9 171 63 7 0 -295.94 ± 43.42
445459 ncm-dbt-01 579661 500 38 224 238 -135.76 ± 13.76 11 167 69 3 0 -301.33 ± 41.33
445458 ncm-dbt-02 586223 500 44 229 227 -134.95 ± 14.31 12 166 67 5 0 -295.94 ± 42.08
445457 ncm-dbt-04 570148 500 44 229 227 -134.95 ± 13.59 9 171 66 4 0 -304.07 ± 42.38
445456 ncm-dbt-03 584790 500 44 217 239 -125.38 ± 14.13 10 158 77 5 0 -270.57 ± 39.07
445455 ncm-dbt-05 581652 500 33 236 231 -149.68 ± 12.45 9 186 54 1 0 -359.68 ± 46.95
445454 ncm-dbt-01 583028 500 45 220 235 -126.97 ± 14.28 11 158 76 5 0 -273.0 ± 39.35
445453 ncm-dbt-03 586858 500 45 216 239 -123.81 ± 12.54 4 165 79 2 0 -280.42 ± 38.28
445452 ncm-dbt-02 583279 500 36 223 241 -136.56 ± 12.98 7 176 64 3 0 -315.35 ± 43.03
445451 ncm-dbt-04 565312 500 50 227 223 -128.55 ± 14.76 13 157 74 6 0 -273.0 ± 39.95
445450 ncm-dbt-05 587579 500 39 226 235 -136.56 ± 14.45 13 166 66 5 0 -298.62 ± 42.41
445449 ncm-dbt-03 582694 500 58 224 218 -119.89 ± 15.15 12 151 78 9 0 -249.64 ± 38.9
445448 ncm-dbt-01 580779 500 37 219 244 -132.54 ± 13.47 9 167 71 3 0 -295.94 ± 40.69
445447 ncm-dbt-02 586477 500 48 222 230 -126.17 ± 14.46 9 164 69 8 0 -275.45 ± 41.43
445446 ncm-dbt-04 567086 500 51 229 220 -129.35 ± 14.24 10 164 70 6 0 -282.94 ± 41.13
445445 ncm-dbt-03 584495 500 38 232 230 -142.25 ± 13.75 13 170 65 2 0 -318.25 ± 42.59
445444 ncm-dbt-01 582987 500 51 200 249 -106.77 ± 14.67 8 143 89 10 0 -221.9 ± 36.31
445443 ncm-dbt-02 586604 500 45 216 239 -123.81 ± 15.28 11 160 68 11 0 -263.42 ± 41.6
445442 ncm-dbt-04 568951 500 43 231 226 -137.37 ± 13.34 10 170 68 2 0 -309.64 ± 41.56
445441 ncm-dbt-05 582569 500 44 221 235 -128.55 ± 13.91 9 164 72 5 0 -282.94 ± 40.5

Commit

Commit ID f193778446acc6e60d7f0f99c6eb01489f89e962
Author Stéphane Nicolet
Date 2021-05-26 23:21:56 UTC
Do not use lazy evaluation inside NNUE This simplification patch implements two changes: 1. it simplifies away the so-called "lazy" path in the NNUE evaluation internals, where we trusted the psqt head alone to avoid the costly "positional" head in some cases; 2. it raises a little bit the NNUEThreshold1 in evaluate.cpp (from 682 to 800), which increases the limit where we switched from NNUE eval to Classical eval. Both effects increase the number of positional evaluations done by our new net architecture, but the results of our tests below seem to indicate that the loss of speed will be compensated by the gain of eval quality. STC: LLR: 2.95 (-2.94,2.94) <-2.50,0.50> Total: 26280 W: 2244 L: 2137 D: 21899 Ptnml(0-2): 72, 1755, 9405, 1810, 98 https://tests.stockfishchess.org/tests/view/60ae73f112066fd299795a51 LTC: LLR: 2.95 (-2.94,2.94) <-2.50,0.50> Total: 20592 W: 750 L: 677 D: 19165 Ptnml(0-2): 9, 614, 8980, 681, 12 https://tests.stockfishchess.org/tests/view/60ae88e812066fd299795a82 closes https://github.com/official-stockfish/Stockfish/pull/3503 Bench: 3817907
Copyright 2011–2025 Next Chess Move LLC