Dev Builds » 20210923-2119

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 06:54:25 582771 4000 697 1340 1963 -56.34 ± 5.24 18 820 950 211 1 -112.52 ± 11.04
ncm-dbt-02 06:54:09 585786 4000 663 1328 2009 -58.3 ± 5.1 11 834 965 189 1 -118.14 ± 10.93
ncm-dbt-03 06:53:42 585201 4000 698 1317 1985 -54.2 ± 5.17 12 804 979 201 4 -109.64 ± 10.85
ncm-dbt-04 06:54:24 568149 4000 699 1332 1969 -55.45 ± 5.14 6 828 963 199 4 -113.49 ± 10.95
ncm-dbt-05 06:55:12 581724 4000 679 1330 1991 -57.05 ± 5.2 8 846 938 205 3 -116.39 ± 11.11
20000 3436 6647 9917 -56.27 ± 2.31 55 4132 4795 1005 13 -114.03 ± 4.91

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
436305 ncm-dbt-04 568872 500 81 167 252 -60.36 ± 13.65 0 105 126 19 0 -124.6 ± 30.12 ↓
436304 ncm-dbt-05 582235 500 82 170 248 -61.79 ± 14.55 1 110 115 24 0 -126.18 ± 31.84 ↓
436303 ncm-dbt-01 581194 500 86 167 247 -56.78 ± 14.46 4 95 129 22 0 -110.6 ± 29.78 ↓
436302 ncm-dbt-02 586477 500 83 166 251 -58.21 ± 14.93 1 109 112 28 0 -118.33 ± 32.29 ↓
436301 ncm-dbt-03 584034 500 90 160 250 -48.96 ± 14.74 1 98 121 30 0 -98.44 ± 31.02 ↓
436300 ncm-dbt-01 583656 500 93 180 227 -61.08 ± 15.31 4 105 116 24 1 -121.46 ± 31.7 ↓
436299 ncm-dbt-04 570188 500 87 165 248 -54.65 ± 15.02 1 105 116 27 1 -112.14 ± 31.71 ↓
436298 ncm-dbt-05 581028 500 90 162 248 -50.38 ± 14.41 2 94 128 26 0 -99.95 ± 30.0 ↓
436297 ncm-dbt-03 581527 500 94 152 254 -40.49 ± 14.12 0 87 135 27 1 -83.57 ± 29.08 ↓
436296 ncm-dbt-02 587028 500 85 157 258 -50.38 ± 14.55 0 101 120 29 0 -102.97 ± 31.15 ↓
436295 ncm-dbt-01 585126 500 85 171 244 -60.36 ± 14.76 1 110 113 26 0 -123.02 ± 32.15 ↓
436294 ncm-dbt-04 566375 500 88 174 238 -60.36 ± 14.49 1 106 123 18 2 -126.18 ± 30.59 ↓
436293 ncm-dbt-05 577602 500 75 167 258 -64.66 ± 14.53 1 113 113 23 0 -132.54 ± 32.13 ↓
436292 ncm-dbt-03 586520 500 82 176 242 -66.1 ± 14.59 0 118 108 24 0 -137.37 ± 32.9 ↓
436291 ncm-dbt-02 584748 500 83 161 256 -54.64 ± 14.22 1 100 125 24 0 -110.6 ± 30.39 ↓
436290 ncm-dbt-01 584622 500 74 154 272 -56.07 ± 14.83 2 103 118 27 0 -112.14 ± 31.42 ↓
436289 ncm-dbt-04 567561 500 85 177 238 -64.66 ± 14.67 3 108 117 22 0 -129.35 ± 31.52 ↓
436288 ncm-dbt-05 582819 500 87 166 247 -55.36 ± 15.57 1 109 110 28 2 -115.23 ± 32.58 ↓
436287 ncm-dbt-02 586223 500 80 172 248 -64.66 ± 14.12 0 113 116 21 0 -134.15 ± 31.65 ↓
436286 ncm-dbt-03 584537 500 96 151 253 -38.37 ± 14.66 2 84 131 33 0 -74.79 ± 29.69 ↓
436285 ncm-dbt-01 580365 500 93 169 238 -53.22 ± 14.95 1 104 115 30 0 -107.54 ± 31.86 ↓
436284 ncm-dbt-04 569071 500 87 168 245 -56.78 ± 14.6 0 108 115 27 0 -116.78 ± 31.85 ↓
436283 ncm-dbt-03 585042 500 86 171 243 -59.64 ± 14.99 2 107 116 24 1 -121.46 ± 31.7 ↓
436282 ncm-dbt-02 585421 500 96 170 234 -51.8 ± 14.08 1 96 129 24 0 -104.49 ± 29.82 ↓
436281 ncm-dbt-05 581777 500 81 171 248 -63.23 ± 14.34 2 107 120 21 0 -127.76 ± 31.05 ↓
436280 ncm-dbt-04 569190 500 95 175 230 -56.07 ± 14.29 1 102 123 24 0 -113.68 ± 30.67 ↓
436279 ncm-dbt-01 581569 500 89 166 245 -53.93 ± 14.73 4 94 127 25 0 -104.49 ± 30.12 ↓
436278 ncm-dbt-02 585421 500 84 164 252 -56.07 ± 14.96 2 103 119 25 1 -113.68 ± 31.27 ↓
436277 ncm-dbt-03 587919 500 70 172 258 -71.89 ± 14.1 1 117 116 15 1 -150.51 ± 31.54 ↓
436276 ncm-dbt-05 584958 500 79 161 260 -57.5 ± 14.63 1 106 117 26 0 -116.78 ± 31.56 ↓
436275 ncm-dbt-03 584285 500 83 168 249 -59.64 ± 15.12 5 100 120 25 0 -115.23 ± 31.11 ↓
436274 ncm-dbt-04 566809 500 82 156 262 -51.8 ± 14.35 0 100 125 24 1 -107.54 ± 30.4 ↓
436273 ncm-dbt-02 584495 500 79 169 252 -63.23 ± 14.47 4 102 124 20 0 -124.6 ± 30.44 ↓
436272 ncm-dbt-01 583740 500 91 181 228 -63.23 ± 14.88 2 111 112 25 0 -127.76 ± 32.29 ↓
436271 ncm-dbt-05 578588 500 92 169 239 -53.94 ± 15.25 0 109 110 30 1 -112.14 ± 32.58 ↓
436270 ncm-dbt-04 567126 500 94 150 256 -39.08 ± 15.08 0 94 118 38 0 -79.17 ± 31.44 ↓
436269 ncm-dbt-02 586477 500 73 169 258 -67.54 ± 14.09 2 110 120 18 0 -137.37 ± 30.98 ↓
436268 ncm-dbt-03 587749 500 97 167 236 -48.96 ± 14.2 1 93 132 23 1 -99.95 ± 29.4 ↓
436267 ncm-dbt-01 581902 500 86 152 262 -46.13 ± 14.72 0 98 120 32 0 -93.95 ± 31.17 ↓
436266 ncm-dbt-05 584790 500 93 164 243 -49.67 ± 14.24 0 98 125 27 0 -101.46 ± 30.43 ↓

Commit

Commit ID 73018a03375b4b72ee482eb5a4a2152d7e4f0aac
Author Stéphane Nicolet
Date 2021-09-23 21:19:06 UTC
Detect search explosions This patch detects some search explosions (due to double extensions in search.cpp) which can happen in some pathological positions, and takes measures to ensure progress in search even for these pathological situations. While a small number of double extensions can be useful during search (for example to resolve a tactical sequence), a sustained regime of double extensions leads to search explosion and a non-finishing search. See the discussion in https://github.com/official-stockfish/Stockfish/pull/3544 and the issue https://github.com/official-stockfish/Stockfish/issues/3532 . The implemented algorithm is the following: a) at each node during search, store the current depth in the stack. Double extensions are by definition levels of the stack where the depth at ply N is strictly higher than depth at ply N-1. b) during search, calculate for each thread a running average of the number of double extensions in the last 4096 visited nodes. c) if one thread has more than 2% of double extensions for a sustained period of time (6 millions consecutive nodes, or about 4 seconds on my iMac), we decide that this thread is in an explosion state and we calm down this thread by preventing it to do any double extension for the next 6 millions nodes. To calculate the running averages, we also introduced a auxiliary class generalizing the computations of ttHitAverage variable we already had in code. The implementation uses an exponential moving average of period 4096 and resolution 1/1024, and all computations are done with integers for efficiency. ----------- Example where the patch solves a search explosion: ``` ./stockfish ucinewgame position fen 8/Pk6/8/1p6/8/P1K5/8/6B1 w - - 37 130 go infinite ``` This algorithm does not affect search in normal, non-pathological positions. We verified, for instance, that the usual bench is unchanged up to depth 20 at least, and that the node numbers are unchanged for a search of the starting position at depth 32. ------------- See https://github.com/official-stockfish/Stockfish/pull/3714 Bench: 5575265
Copyright 2011–2025 Next Chess Move LLC
Feedback & Support Terms of Service Privacy Policy Desktop Version