Dev Builds » 20260726-1044

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 06:44:03 584481 4000 1815 297 1888 +138.79 ± 4.42 1 16 492 1446 45 +327.94 ± 15.33
ncm-dbt-02 06:43:10 590546 4000 1840 293 1867 +141.74 ± 4.3 0 16 461 1483 40 +340.84 ± 15.85
ncm-dbt-03 06:43:23 576715 4000 1807 295 1898 +138.18 ± 4.29 0 14 495 1456 35 +329.08 ± 15.27
ncm-dbt-04 06:41:29 567354 4000 1822 301 1877 +139.1 ± 4.4 0 13 502 1436 49 +327.18 ± 15.15
ncm-dbt-05 06:43:59 602493 4000 1814 291 1895 +139.3 ± 4.39 1 20 472 1469 38 +332.55 ± 15.68
20000 9098 1477 9425 +139.42 ± 1.95 2 79 2422 7290 207 +331.46 ± 6.9

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
480813 ncm-dbt-05 597556 500 231 42 227 +138.18 ± 12.33 0 1 65 178 6 +324.17 ± 42.47
480812 ncm-dbt-02 590653 500 221 47 232 +126.17 ± 12.68 0 3 74 169 4 +288.06 ± 39.79
480811 ncm-dbt-01 583237 500 231 31 238 +147.19 ± 11.92 0 1 54 189 6 +359.68 ± 46.95
480810 ncm-dbt-03 576414 500 219 31 250 +137.37 ± 12.56 0 1 67 175 7 +318.25 ± 41.77
480809 ncm-dbt-04 567799 500 233 41 226 +140.62 ± 12.02 0 1 61 183 5 +336.46 ± 43.96
480808 ncm-dbt-05 616450 500 225 35 240 +138.99 ± 11.88 0 2 59 186 3 +336.46 ± 44.87
480807 ncm-dbt-01 584664 500 221 40 239 +131.74 ± 13.12 1 3 64 178 4 +309.64 ± 43.08
480806 ncm-dbt-03 576210 500 228 47 225 +131.74 ± 12.14 0 2 68 177 3 +309.64 ± 41.56
480805 ncm-dbt-02 589710 500 227 37 236 +138.99 ± 12.7 0 3 60 181 6 +327.18 ± 44.54
480804 ncm-dbt-04 567522 500 224 33 243 +139.81 ± 12.67 0 1 65 176 8 +324.17 ± 42.47
480803 ncm-dbt-05 599937 500 236 36 228 +147.19 ± 12.36 0 3 50 191 6 +359.68 ± 49.05
480802 ncm-dbt-04 568355 500 238 42 220 +143.89 ± 13.12 0 2 60 178 10 +333.32 ± 44.47
480801 ncm-dbt-02 592461 500 230 39 231 +139.8 ± 12.05 0 0 65 179 6 +330.23 ± 42.3
480800 ncm-dbt-03 577602 500 235 24 241 +156.38 ± 11.11 0 0 45 199 6 +401.91 ± 51.63
480799 ncm-dbt-01 585211 500 223 36 241 +136.56 ± 11.33 0 0 65 183 2 +330.23 ± 42.3
480798 ncm-dbt-05 598788 500 227 34 239 +141.44 ± 12.41 0 3 56 186 5 +339.63 ± 46.2
480797 ncm-dbt-04 567244 500 223 33 244 +138.99 ± 11.66 0 0 64 182 4 +333.32 ± 42.66
480796 ncm-dbt-02 592590 500 242 44 214 +145.54 ± 11.79 0 1 55 189 5 +356.21 ± 46.49
480795 ncm-dbt-03 578712 500 217 35 248 +132.54 ± 12.12 0 1 70 175 4 +309.64 ± 40.79
480794 ncm-dbt-01 583782 500 232 38 230 +142.25 ± 12.16 0 3 54 189 4 +346.12 ± 47.1
480793 ncm-dbt-04 566020 500 217 35 248 +132.54 ± 12.12 0 2 67 178 3 +312.48 ± 41.9
480792 ncm-dbt-05 602248 500 222 30 248 +140.62 ± 11.81 0 2 57 188 3 +342.85 ± 45.71
480791 ncm-dbt-02 589112 500 231 33 236 +145.54 ± 12.64 0 3 53 187 7 +349.43 ± 47.56
480790 ncm-dbt-03 577684 500 225 48 227 +128.55 ± 12.82 0 3 72 170 5 +293.29 ± 40.39
480789 ncm-dbt-01 586054 500 226 47 227 +130.14 ± 12.59 0 2 72 171 5 +298.62 ± 40.29
480788 ncm-dbt-04 567126 500 229 29 242 +147.19 ± 12.14 0 1 55 187 7 +356.21 ± 46.49
480787 ncm-dbt-05 601758 500 234 44 222 +138.99 ± 11.88 0 0 65 180 5 +330.23 ± 42.3
480786 ncm-dbt-01 583196 500 228 40 232 +137.37 ± 12.56 0 4 58 184 4 +327.18 ± 45.37
480785 ncm-dbt-03 570709 500 229 40 231 +138.18 ± 12.53 0 3 60 182 5 +327.18 ± 44.54
480784 ncm-dbt-02 589197 500 235 35 230 +147.19 ± 11.24 0 1 51 195 3 +370.41 ± 48.42
480783 ncm-dbt-04 566888 500 221 48 231 +125.38 ± 12.89 0 5 70 172 3 +288.06 ± 41.11
480782 ncm-dbt-02 587325 500 222 30 248 +140.62 ± 12.02 0 2 58 186 4 +339.63 ± 45.29
480781 ncm-dbt-01 584369 500 225 29 246 +143.89 ± 13.31 0 2 61 176 11 +330.23 ± 44.08
480780 ncm-dbt-05 603008 500 222 36 242 +135.76 ± 12.62 1 3 57 187 2 +330.23 ± 45.79
480779 ncm-dbt-03 578177 500 230 32 238 +145.54 ± 11.56 0 3 48 197 2 +366.78 ± 50.11
480778 ncm-dbt-04 567878 500 237 40 223 +144.71 ± 12.68 0 1 60 180 9 +339.63 ± 44.35
480777 ncm-dbt-02 593326 500 232 28 240 +150.51 ± 11.96 0 3 45 197 5 +377.87 ± 51.84
480776 ncm-dbt-05 600203 500 217 34 249 +133.34 ± 13.81 0 6 63 173 8 +301.33 ± 43.45
480775 ncm-dbt-03 578218 500 224 38 238 +135.76 ± 11.8 0 1 65 181 3 +324.17 ± 42.47
480774 ncm-dbt-01 585337 500 229 36 235 +141.44 ± 12.81 0 1 64 176 9 +327.18 ± 42.83

Commit

Commit ID db98633b1f8bbcad850bee892ec143ff8723ba80
Author Carlos Esparza
Date 2026-07-26 10:44:45 UTC
Avoid recomputing threat/pp accumulation for some king moves. Since the separate threat and psq accumulators were merged in 7c7fe322eae6fa3a594c036eb4fc2553fd4e6a11, any king move requires a full recomputation of the threat (and now also pawn-pair) accumulation, even though these only change when the king switches between the left and right half of the board. This patch adds a method `update_accumulator_hybrid` which incrementally computes the accumulator change of a king move as ``` <new acc> = <previous acc> - <previous psq acc> + <new psq acc> + <threat/pp changes> ``` thus avoiding re-adding all threat/pp features. `<previous psq acc>` and `<new psq acc>` can be computing efficiently using the accumulator cache ("Finny table"). When there are few pieces on the board there typically are only few threat/pp features so computing their sum from scratch is cheaper than computing `<previous psq acc>` from the cache entry. Therefore we only do "hybrid" updates for positions with at least `MIN_PC_COUNT_HYBRID = 15` pieces. For simplicity this also doesn't handle castling moves. passed STC: https://tests.stockfishchess.org/tests/view/6a5b8a7d5529b8472df8158b LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 37792 W: 9855 L: 9557 D: 18380 Ptnml(0-2): 65, 3955, 10578, 4213, 85 local benchmark: ``` Result of 100 runs (counting nodes/1B cycles) ============================================= base (...tockfish.f4bcd4) = 264725 +/- 3700 test (...tockfish.4e77c9) = 266318 +/- 3701 speedup % = +0.60 +/- 0.08 [95% CI; t-test] likelihood(speedup > 0) = 1.0000 CPU: 16 x Intel(R) Core(TM) Ultra 9 185H Hyperthreading: on ``` I don't know if `MIN_PC_COUNT_HYBRID = 15` is optimal. On my computer (avx-vnni) 12 was the optimum, but that did not perform well [on fishtest](https://tests.stockfishchess.org/tests/view/6a5769345529b8472df80d7e). It would be good if other devs could try benchmarking different thresholds and report. The first version of this patch was coded up by gpt-5.5-high. I made many changes, but probably most of the lines of code are LLM-written. closes https://github.com/official-stockfish/Stockfish/pull/6990 No functional change
Copyright 2011–2026 Next Chess Move LLC