Dev Builds » 20210923-2119

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 14. 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 11:53:18 1234376 3990 1229 766 1995 +40.5 +/- 5.35 3 271 1001 700 20 +79.01 +/- 10.75
ncm-dbt-02 11:48:52 1258372 3936 1275 720 1941 +49.32 +/- 5.45 3 253 914 782 16 +98.22 +/- 11.27
ncm-dbt-03 11:54:05 1260427 3972 1260 726 1986 +46.99 +/- 5.42 3 254 959 746 24 +91.83 +/- 10.99
ncm-dbt-04 00:31:12 1482618 172 58 33 81 +50.86 +/- 28.01 1 12 34 39 0 +108.43 +/- 58.6
ncm-dbt-05 11:49:57 1255543 3940 1225 692 2023 +47.29 +/- 5.39 1 244 974 723 28 +91.28 +/- 10.9
ncm-dbt-06 11:53:45 1257825 3990 1247 759 1984 +42.71 +/- 5.44 1 290 940 748 16 +83.97 +/- 11.11
20000 6294 3696 10010 +45.39 +/- 2.41 12 1324 4822 3738 104 +88.96 +/- 4.9

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
209339 ncm-dbt-02 1238115 272 88 51 133 +47.55 +/- 18.58 0 12 76 47 1 +94.21 +/- 38.44
209338 ncm-dbt-05 1232952 276 94 55 127 +49.42 +/- 20.47 0 15 73 46 4 +90.08 +/- 39.71
209337 ncm-dbt-03 1230665 304 98 51 155 +54.15 +/- 19.83 0 19 69 62 2 +106.03 +/- 41.23
209336 ncm-dbt-06 1233617 324 99 58 167 +44.2 +/- 18.96 0 24 73 65 0 +89.88 +/- 40.03
209335 ncm-dbt-01 1218659 324 107 57 160 +54.05 +/- 20.52 0 24 68 66 4 +101.44 +/- 41.37
209334 ncm-dbt-02 1245699 500 171 90 239 +56.78 +/- 15.25 0 32 105 113 0 +116.78 +/- 33.3
209333 ncm-dbt-05 1223450 500 163 100 237 +44.01 +/- 15.24 1 31 125 90 3 +86.52 +/- 30.48
209332 ncm-dbt-03 1239641 500 168 91 241 +53.93 +/- 15.12 0 30 115 103 2 +107.54 +/- 31.86
209331 ncm-dbt-06 1237124 500 159 94 247 +45.42 +/- 15.57 0 37 113 98 2 +89.48 +/- 32.12
209330 ncm-dbt-01 1194604 500 149 90 261 +41.19 +/- 14.95 0 33 128 86 3 +79.17 +/- 30.09
209329 ncm-dbt-02 1214603 500 162 97 241 +45.42 +/- 15.19 1 32 119 97 1 +92.46 +/- 31.31
209328 ncm-dbt-05 1207032 500 158 95 247 +44.01 +/- 15.24 0 34 122 91 3 +85.04 +/- 30.9
209327 ncm-dbt-03 1227893 500 157 83 260 +51.8 +/- 14.49 0 25 129 93 3 +101.46 +/- 29.84
209326 ncm-dbt-06 1238218 500 145 91 264 +37.67 +/- 15.86 1 40 115 92 2 +74.79 +/- 31.83
209325 ncm-dbt-01 1198639 500 169 88 243 +56.78 +/- 15.51 0 32 107 109 2 +113.68 +/- 33.0
209324 ncm-dbt-02 1214159 500 170 98 232 +50.38 +/- 14.81 0 29 122 97 2 +99.95 +/- 30.87
209323 ncm-dbt-05 1229014 500 147 83 270 +44.72 +/- 15.28 0 35 118 95 2 +88.0 +/- 31.44
209322 ncm-dbt-03 1222479 500 170 95 235 +52.51 +/- 15.31 0 30 119 97 4 +101.46 +/- 31.3
209321 ncm-dbt-01 1224458 500 145 96 259 +34.16 +/- 14.52 1 31 138 78 2 +67.55 +/- 28.74
209320 ncm-dbt-06 1220597 500 140 102 258 +26.45 +/- 15.52 0 45 125 77 3 +48.96 +/- 30.53
209319 ncm-dbt-02 1231267 500 160 86 254 +51.8 +/- 15.27 0 32 114 102 2 +102.97 +/- 32.0
209318 ncm-dbt-05 1235915 500 158 71 271 +61.07 +/- 15.44 0 26 117 101 6 +116.78 +/- 31.56
209317 ncm-dbt-03 1232675 500 154 87 259 +46.84 +/- 16.37 1 38 108 99 4 +90.97 +/- 32.79
209316 ncm-dbt-06 1230180 500 154 96 250 +40.48 +/- 14.91 0 34 126 88 2 +79.17 +/- 30.36
209315 ncm-dbt-01 1214606 500 153 100 247 +36.97 +/- 14.57 0 31 139 76 4 +68.99 +/- 28.6
209314 ncm-dbt-02 1227314 500 163 81 256 +57.5 +/- 16.4 0 35 104 105 6 +109.07 +/- 33.4
209313 ncm-dbt-05 1221245 500 157 91 252 +46.13 +/- 15.23 0 30 130 84 6 +85.04 +/- 29.79
209312 ncm-dbt-01 1200922 500 142 102 256 +27.85 +/- 15.37 1 42 124 82 1 +56.07 +/- 30.66
209311 ncm-dbt-03 1228190 500 149 105 246 +30.65 +/- 15.43 1 39 128 79 3 +58.93 +/- 30.13
209310 ncm-dbt-06 1219659 500 160 92 248 +47.54 +/- 14.79 0 31 121 97 1 +95.44 +/- 31.02
209309 ncm-dbt-02 1218490 500 154 88 258 +46.13 +/- 15.85 1 36 111 100 2 +92.46 +/- 32.39
209308 ncm-dbt-03 1229777 500 169 91 240 +54.65 +/- 15.02 1 25 122 99 3 +109.07 +/- 30.84
209307 ncm-dbt-01 1193497 500 150 110 240 +27.85 +/- 14.75 0 39 134 75 2 +53.22 +/- 29.35
209306 ncm-dbt-05 1233016 500 147 90 263 +39.78 +/- 15.0 0 35 125 88 2 +77.71 +/- 30.5
209305 ncm-dbt-06 1226239 500 174 98 228 +53.22 +/- 15.21 0 31 114 103 2 +106.01 +/- 32.0
209304 ncm-dbt-02 1244470 500 161 97 242 +44.72 +/- 15.03 1 31 122 95 1 +90.97 +/- 30.89
209303 ncm-dbt-05 1220565 500 153 78 269 +52.51 +/- 13.98 0 23 130 96 1 +106.01 +/- 29.66
209302 ncm-dbt-03 1231745 500 145 102 253 +29.95 +/- 15.39 0 43 123 82 2 +57.5 +/- 30.78
209301 ncm-dbt-06 1213242 500 159 89 252 +48.96 +/- 15.63 0 35 113 99 3 +95.44 +/- 32.13
209300 ncm-dbt-01 1184031 500 154 94 252 +41.89 +/- 14.87 1 30 129 88 2 +83.57 +/- 29.93
177958 ncm-dbt-01 1479974 166 60 29 77 +65.65 +/- 26.06 0 9 34 40 0 +136.36 +/- 59.13
177957 ncm-dbt-02 1491234 164 46 32 86 +29.73 +/- 27.04 0 14 41 26 1 +55.55 +/- 53.61
177956 ncm-dbt-06 1501553 166 57 39 70 +37.82 +/- 26.99 0 13 40 29 1 +72.18 +/- 54.33
177955 ncm-dbt-04 1482618 172 58 33 81 +50.86 +/- 28.01 1 12 34 39 0 +108.43 +/- 58.6
177954 ncm-dbt-03 1500785 168 50 21 97 +60.58 +/- 23.3 0 5 46 32 1 +120.41 +/- 49.27
177953 ncm-dbt-05 1496705 164 48 29 87 +40.43 +/- 28.8 0 15 34 32 1 +77.53 +/- 58.53

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 2011–2024 Next Chess Move LLC