Dev Builds » 20240320-1529

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 09:46:04 1195759 3338 1459 242 1637 +132.78 +/- 5.28 0 37 428 1154 50 +300.8 +/- 16.49
ncm-dbt-02 09:42:17 1233261 3324 1479 239 1606 +136.18 +/- 5.25 1 27 421 1157 56 +310.35 +/- 16.62
ncm-dbt-03 09:45:33 1231050 3336 1505 234 1597 +139.4 +/- 5.07 0 23 402 1192 51 +323.73 +/- 17.01
ncm-dbt-04 09:46:42 1226294 3368 1510 237 1621 +138.17 +/- 5.23 1 24 423 1173 63 +314.71 +/- 16.57
ncm-dbt-05 09:43:08 1225526 3304 1469 261 1574 +133.19 +/- 5.15 0 26 438 1142 46 +303.65 +/- 16.28
ncm-dbt-06 09:46:32 1232287 3330 1484 234 1612 +137.12 +/- 5.01 1 17 424 1177 46 +318.05 +/- 16.53
20000 8906 1447 9647 +136.14 +/- 2.11 3 154 2536 6995 312 +311.77 +/- 6.76

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
338813 ncm-dbt-05 1228253 304 129 21 154 +129.05 +/- 16.88 0 4 38 108 2 +299.54 +/- 56.41
338812 ncm-dbt-02 1198434 324 145 13 166 +150.26 +/- 17.56 0 4 30 120 8 +350.63 +/- 63.73
338811 ncm-dbt-06 1213347 330 152 25 153 +140.96 +/- 16.67 0 2 41 115 7 +320.65 +/- 54.21
338810 ncm-dbt-03 1219010 336 155 24 157 +143.02 +/- 17.06 0 6 30 127 5 +338.04 +/- 62.9
338809 ncm-dbt-01 1209688 338 140 23 175 +125.44 +/- 16.52 0 6 42 119 2 +288.37 +/- 53.38
338808 ncm-dbt-04 1222593 368 160 24 184 +134.77 +/- 14.78 0 2 48 130 4 +313.47 +/- 49.86
338807 ncm-dbt-05 1218082 500 233 46 221 +136.56 +/- 14.45 0 4 69 163 14 +295.94 +/- 41.39
338806 ncm-dbt-02 1246888 500 210 44 246 +119.89 +/- 13.16 0 4 81 160 5 +265.78 +/- 37.95
338805 ncm-dbt-06 1227539 500 227 37 236 +138.99 +/- 12.5 0 3 59 183 5 +330.23 +/- 44.94
338804 ncm-dbt-03 1221049 500 234 33 233 +148.01 +/- 12.74 0 0 60 179 11 +346.11 +/- 44.19
338803 ncm-dbt-01 1206878 500 214 27 259 +136.56 +/- 14.27 0 6 62 171 11 +304.07 +/- 43.81
338802 ncm-dbt-04 1216727 500 219 39 242 +130.94 +/- 14.89 1 8 60 172 9 +293.29 +/- 44.36
338801 ncm-dbt-05 1228846 500 222 39 239 +133.34 +/- 13.26 0 3 69 170 8 +301.33 +/- 41.33
338800 ncm-dbt-06 1245132 500 227 25 248 +148.84 +/- 12.27 0 0 57 184 9 +356.21 +/- 45.43
338799 ncm-dbt-02 1229567 500 223 41 236 +132.54 +/- 14.01 1 5 62 175 7 +304.07 +/- 43.81
338798 ncm-dbt-03 1216831 500 221 33 246 +137.37 +/- 12.76 0 3 62 179 6 +321.19 +/- 43.77
338797 ncm-dbt-01 1181765 500 215 40 245 +126.97 +/- 12.85 0 3 74 168 5 +288.06 +/- 39.79
338796 ncm-dbt-04 1231049 500 230 34 236 +143.89 +/- 11.87 0 0 60 184 6 +346.11 +/- 44.19
338795 ncm-dbt-05 1212615 500 215 43 242 +124.6 +/- 13.09 0 7 66 175 2 +288.06 +/- 42.4
338794 ncm-dbt-03 1224724 500 224 35 241 +138.18 +/- 12.93 0 2 65 175 8 +318.25 +/- 42.59
338793 ncm-dbt-02 1254483 500 221 36 243 +134.95 +/- 13.22 0 5 61 178 6 +312.48 +/- 44.19
338792 ncm-dbt-06 1242754 500 232 32 236 +147.19 +/- 13.38 0 1 61 175 13 +336.46 +/- 43.96
338791 ncm-dbt-01 1181753 500 229 36 235 +141.44 +/- 14.68 0 8 53 177 12 +318.25 +/- 47.23
338790 ncm-dbt-04 1216581 500 227 36 237 +139.81 +/- 14.55 0 5 63 168 14 +306.84 +/- 43.45
338789 ncm-dbt-05 1233659 500 222 40 238 +132.54 +/- 13.65 0 5 66 171 8 +298.62 +/- 42.41
338788 ncm-dbt-03 1248798 500 225 38 237 +136.56 +/- 13.55 0 4 64 173 9 +309.64 +/- 43.08
338787 ncm-dbt-06 1241789 500 212 38 250 +126.17 +/- 12.1 0 3 71 175 1 +295.94 +/- 40.69
338786 ncm-dbt-02 1250507 500 222 20 258 +148.85 +/- 12.7 0 2 53 186 9 +356.21 +/- 47.52
338785 ncm-dbt-01 1183886 500 217 41 242 +127.76 +/- 13.57 0 4 74 164 8 +282.94 +/- 39.86
338784 ncm-dbt-04 1235113 500 226 36 238 +138.99 +/- 13.67 0 3 65 171 11 +312.48 +/- 42.67
338783 ncm-dbt-03 1244850 500 222 27 251 +143.07 +/- 12.34 0 2 57 185 6 +342.85 +/- 45.71
338782 ncm-dbt-05 1230290 500 226 39 235 +136.56 +/- 11.98 0 0 68 177 5 +321.18 +/- 41.27
338781 ncm-dbt-06 1246572 500 216 34 250 +132.54 +/- 13.28 1 3 64 177 5 +309.64 +/- 43.08
338780 ncm-dbt-02 1218440 500 233 41 226 +140.62 +/- 14.35 0 3 67 165 15 +306.84 +/- 41.99
338779 ncm-dbt-01 1222204 500 214 35 251 +130.14 +/- 12.39 0 4 65 179 2 +306.84 +/- 42.73
338778 ncm-dbt-04 1218527 500 227 36 237 +139.81 +/- 13.26 0 1 68 170 11 +315.35 +/- 41.44
338777 ncm-dbt-03 1242089 500 224 44 232 +130.94 +/- 13.51 0 6 64 174 6 +298.62 +/- 43.1
338776 ncm-dbt-05 1226942 500 222 33 245 +138.18 +/- 12.93 0 3 62 178 7 +321.19 +/- 43.77
338775 ncm-dbt-02 1234513 500 225 44 231 +131.74 +/- 13.12 0 4 67 173 6 +301.33 +/- 42.04
338774 ncm-dbt-01 1184144 500 230 40 230 +138.99 +/- 14.03 0 6 58 176 10 +315.35 +/- 45.33
338773 ncm-dbt-06 1208879 500 218 43 239 +126.97 +/- 13.41 0 5 71 168 6 +285.49 +/- 40.81
338772 ncm-dbt-04 1243473 500 221 32 247 +138.18 +/- 13.5 0 5 59 178 8 +318.25 +/- 44.96

Commit

Commit ID 9b92ada935ddf920491156be22f609afaca4d840
Author Robert Nurnberg
Date 2024-03-20 15:29:35 UTC
Base WDL model on material count and normalize evals dynamically This PR proposes to change the parameter dependence of Stockfish's internal WDL model from full move counter to material count. In addition it ensures that an evaluation of 100 centipawns always corresponds to a 50% win probability at fishtest LTC, whereas for master this holds only at move number 32. See also https://github.com/official-stockfish/Stockfish/pull/4920 and the discussion therein. The new model was fitted based on about 340M positions extracted from 5.6M fishtest LTC games from the last three weeks, involving SF versions from e67cc979fd2c0e66dfc2b2f2daa0117458cfc462 (SF 16.1) to current master. The involved commands are for [WDL_model](https://github.com/official-stockfish/WDL_model) are: ``` ./updateWDL.sh --firstrev e67cc979fd2c0e66dfc2b2f2daa0117458cfc462 python scoreWDL.py updateWDL.json --plot save --pgnName update_material.png --momType "material" --momTarget 58 --materialMin 10 --modelFitting optimizeProbability ``` The anchor `58` for the material count value was chosen to be as close as possible to the observed average material count of fishtest LTC games at move 32 (`43`), while not changing the value of `NormalizeToPawnValue` compared to the move-based WDL model by more than 1. The patch only affects the displayed cp and wdl values. closes https://github.com/official-stockfish/Stockfish/pull/5121 No functional change
Copyright 2011–2024 Next Chess Move LLC