Dev Builds » 20260720-0906

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 06:44:13 583143 4000 1803 287 1910 +138.59 ± 4.31 0 22 470 1478 30 +332.55 ± 15.71
ncm-dbt-02 06:43:00 590484 4000 1812 264 1924 +141.85 ± 4.24 0 15 456 1495 34 +343.67 ± 15.94
ncm-dbt-03 06:43:53 577179 4000 1825 329 1846 +136.56 ± 4.48 1 21 503 1431 44 +320.08 ± 15.17
ncm-dbt-04 06:41:04 566417 4000 1818 298 1884 +138.99 ± 4.29 0 15 485 1465 35 +332.16 ± 15.44
ncm-dbt-05 06:45:44 601254 4000 1825 298 1877 +139.71 ± 4.41 0 19 480 1456 45 +331.0 ± 15.54
20000 9083 1476 9441 +139.14 ± 1.94 1 92 2394 7325 188 +331.77 ± 6.95

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
479488 ncm-dbt-03 577889 500 230 47 223 +133.34 ± 12.09 0 2 66 179 3 +315.35 ± 42.24
479487 ncm-dbt-05 604622 500 224 37 239 +136.56 ± 12.18 0 3 60 184 3 +327.18 ± 44.54
479486 ncm-dbt-01 581693 500 226 26 248 +147.19 ± 11.47 0 1 52 193 4 +366.78 ± 47.92
479485 ncm-dbt-02 591211 500 223 31 246 +140.62 ± 11.36 0 2 55 192 1 +349.43 ± 46.59
479484 ncm-dbt-04 567759 500 230 40 230 +138.99 ± 13.1 0 5 56 183 6 +327.18 ± 46.19
479483 ncm-dbt-03 580157 500 230 41 229 +138.18 ± 13.5 0 2 68 169 11 +309.64 ± 41.56
479482 ncm-dbt-05 604667 500 228 49 223 +130.14 ± 12.97 0 3 71 170 6 +295.94 ± 40.69
479481 ncm-dbt-02 590096 500 228 34 238 +142.25 ± 11.95 0 0 62 182 6 +339.63 ± 43.41
479480 ncm-dbt-04 562418 500 222 36 242 +135.76 ± 12.21 0 1 67 177 5 +318.25 ± 41.77
479479 ncm-dbt-01 581693 500 226 29 245 +144.71 ± 12.05 0 2 54 189 5 +352.8 ± 47.05
479478 ncm-dbt-03 577602 500 231 42 227 +138.18 ± 12.93 0 1 68 172 9 +315.35 ± 41.44
479477 ncm-dbt-02 589710 500 224 31 245 +141.44 ± 11.98 0 1 60 184 5 +339.63 ± 44.35
479476 ncm-dbt-01 582736 500 223 40 237 +133.34 ± 13.45 0 6 61 177 6 +306.84 ± 44.17
479475 ncm-dbt-05 600957 500 223 20 257 +149.68 ± 12.23 0 1 53 188 8 +363.2 ± 47.43
479474 ncm-dbt-04 564291 500 219 39 242 +130.94 ± 12.95 0 5 64 177 4 +304.07 ± 43.1
479473 ncm-dbt-03 575883 500 229 46 225 +133.34 ± 12.69 0 5 60 182 3 +315.35 ± 44.57
479472 ncm-dbt-01 584369 500 231 44 225 +136.56 ± 12.98 0 4 61 179 6 +318.25 ± 44.18
479471 ncm-dbt-05 595499 500 237 40 223 +144.71 ± 11.15 0 0 56 191 3 +359.67 ± 45.87
479470 ncm-dbt-04 567482 500 226 36 238 +138.99 ± 11.66 0 1 61 185 3 +336.46 ± 43.96
479469 ncm-dbt-02 590224 500 218 25 257 +141.44 ± 11.98 0 3 54 190 3 +346.12 ± 47.1
479468 ncm-dbt-03 575028 500 228 34 238 +142.25 ± 12.37 0 2 58 184 6 +339.63 ± 45.29
479467 ncm-dbt-04 567680 500 231 36 233 +143.07 ± 11.01 0 0 57 191 2 +356.21 ± 45.43
479466 ncm-dbt-01 581028 500 222 43 235 +130.14 ± 11.99 0 4 63 183 0 +312.48 ± 43.44
479465 ncm-dbt-02 590996 500 223 35 242 +137.37 ± 11.73 0 2 60 186 2 +333.32 ± 44.47
479464 ncm-dbt-05 598788 500 225 40 235 +134.95 ± 13.22 0 6 58 181 5 +315.35 ± 45.33
479463 ncm-dbt-03 577069 500 229 39 232 +138.99 ± 12.5 0 3 59 183 5 +330.23 ± 44.94
479462 ncm-dbt-04 566336 500 231 42 227 +138.18 ± 11.48 0 0 64 183 3 +333.32 ± 42.66
479461 ncm-dbt-01 583237 500 234 38 228 +143.89 ± 10.72 0 1 52 197 0 +366.78 ± 47.92
479460 ncm-dbt-02 587792 500 227 38 235 +138.18 ± 12.33 0 3 59 184 4 +330.23 ± 44.94
479459 ncm-dbt-05 600026 500 233 38 229 +143.07 ± 11.91 0 1 58 186 5 +346.12 ± 45.18
479458 ncm-dbt-04 565705 500 225 33 242 +140.62 ± 12.64 0 1 64 177 8 +327.18 ± 42.83
479457 ncm-dbt-03 576578 500 228 44 228 +134.15 ± 12.47 0 4 61 182 3 +318.25 ± 44.18
479456 ncm-dbt-02 590567 500 230 35 235 +143.07 ± 12.12 0 1 59 184 6 +342.85 ± 44.76
479455 ncm-dbt-01 583070 500 217 34 249 +133.34 ± 11.67 0 1 67 180 2 +318.25 ± 41.77
479454 ncm-dbt-05 602471 500 231 32 237 +146.36 ± 12.81 0 2 56 183 9 +346.12 ± 46.15
479453 ncm-dbt-04 569669 500 234 36 230 +145.54 ± 11.79 0 2 52 192 4 +359.68 ± 48.01
479452 ncm-dbt-03 577233 500 220 36 244 +134.15 ± 12.86 1 2 63 180 4 +318.25 ± 43.39
479451 ncm-dbt-02 593282 500 239 35 226 +150.51 ± 12.4 0 3 47 193 7 +370.41 ± 50.67
479450 ncm-dbt-01 587325 500 224 33 243 +139.81 ± 12.87 0 3 60 180 7 +327.18 ± 44.54
479449 ncm-dbt-05 603008 500 224 42 234 +132.54 ± 12.91 0 3 68 173 6 +304.07 ± 41.65

Commit

Commit ID f4bcd40409f94bd397a083c5d6243bac6dcc6d85
Author sscg13
Date 2026-07-20 09:06:41 UTC
Update nnue architecture to SFNNv16 and net nn-89cb98a217f7.nnue Failed STC (https://tests.stockfishchess.org/tests/view/6a4f36355529b8472df7ff7e): LLR: -2.94 (-2.94,2.94) <0.00,2.00> Total: 144768 W: 38003 L: 38051 D: 68714 Ptnml(0-2): 521, 17154, 37050, 17170, 489 Passed LTC (https://tests.stockfishchess.org/tests/view/6a501dad5529b8472df80103): LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 189888 W: 49613 L: 48984 D: 91291 Ptnml(0-2): 122, 20406, 53256, 21041, 119 Passed VLTC (https://tests.stockfishchess.org/tests/view/6a52892c5529b8472df80481): LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 46124 W: 12156 L: 11865 D: 22103 Ptnml(0-2): 12, 4523, 13696, 4824, 7 Passed VVLTC (https://tests.stockfishchess.org/tests/view/6a566a285529b8472df80b84): LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 44698 W: 11805 L: 11507 D: 21386 Ptnml(0-2): 9, 3819, 14395, 4117, 9 This PR adds pawn pair features, a new set of features that is indexed by pairs of pawns occupying the same or adjacent files (hence "3-wide"). Because these are a superset of pawn–pawn interactions in threat inputs (including the recently added pawn-pusher inputs), those are removed. HalfKA features remain unchanged. Pawn-pair features were invented by Jonathan for Pawnocchio, based on his observation that in a net trained on *all* pairs of pawns, the most important pawn pairs were those differing by at most one file. sscg13 made the necessary changes to the trainer (https://github.com/official-stockfish/nnue-pytorch/pull/502) and initial changes to inference. The slowdown was originally far too large (some 10%); following major optimizations from Jonathan and a bit from me, the slowdown is small enough for a comfortable LTC pass; on my machine I get a 3.5% slowdown overall. I expect there are further "easy" speedups in the new code, so this gap should narrow a bit. Nettest PR: https://github.com/vondele/nettest/pull/359 closes https://github.com/official-stockfish/Stockfish/pull/6982 Bench: 2718396 Co-Authored-By: anematode <timothy.herchen@gmail.com> Co-Authored-By: Jonathan Hallström <lmj.hallstrom@gmail.com>
Copyright 2011–2026 Next Chess Move LLC