Dev Builds » 20180304-1555

You are viewing an old NCM Stockfish dev build test. You may find the most recent dev build tests using Stockfish 15 as the baseline here.

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 7. This yields an approximate Elo difference and establishes confidence in the strength of the dev builds.

Summary

Host Duration Avg Base NPS Games Wins Losses Draws Elo
ncm-et-3 06:22:17 2001434 2496 1125 72 1299 +156.34 ± 9.11
ncm-et-4 06:20:56 2000837 2520 1150 77 1293 +158.0 ± 9.16
ncm-et-5 02:05:03 2012724 834 371 23 440 +154.39 ± 15.61
ncm-et-6 02:08:03 2025656 850 387 30 433 +155.54 ± 15.91
ncm-et-7 02:06:17 2021143 824 353 20 451 +148.9 ± 15.27
ncm-et-8 02:07:13 2009128 830 370 25 435 +153.72 ± 15.75
ncm-et-9 06:21:45 2000678 2511 1087 93 1331 +145.48 ± 9.02
ncm-et-10 06:22:38 1982207 2468 1079 82 1307 +148.84 ± 9.08
ncm-et-11 02:06:21 2018072 836 391 20 425 +165.7 ± 15.94
ncm-et-12 02:05:51 1953265 811 369 24 418 +157.83 ± 16.1
ncm-et-13 06:22:26 2001023 2519 1125 103 1291 +149.56 ± 9.23
ncm-et-14 02:05:52 2019696 831 399 31 401 +165.3 ± 16.68
ncm-et-15 04:14:55 1983890 1670 706 55 909 +143.0 ± 10.83
20000 8912 655 10433 +152.54 ± 3.22

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
52986 ncm-et-10 2018-09-08 22:21 00:20:28 1981381 130 54 6 70 +134.64 ± 39.68
52985 ncm-et-3 2018-09-08 22:17 00:25:05 1989085 162 73 4 85 +158.05 ± 35.62
52984 ncm-et-4 2018-09-08 22:16 00:25:56 1987983 175 77 7 91 +147.19 ± 34.84
52983 ncm-et-15 2018-09-08 22:15 00:26:23 1983733 170 71 4 95 +144.76 ± 33.21
52982 ncm-et-13 2018-09-08 22:15 00:27:14 1989558 178 84 9 85 +156.11 ± 36.65
52981 ncm-et-9 2018-09-08 22:14 00:28:01 1988929 185 82 7 96 +149.43 ± 33.88
52980 ncm-et-10 2018-09-08 21:04 01:16:25 1981383 500 218 13 269 +151.35 ± 19.88
52979 ncm-et-4 2018-09-08 21:00 01:14:48 1989243 500 236 15 249 +164.93 ± 21.0
52978 ncm-et-3 2018-09-08 20:59 01:16:41 1993356 500 245 15 240 +172.78 ± 21.5
52977 ncm-et-9 2018-09-08 20:58 01:15:00 1990032 500 238 22 240 +160.64 ± 21.66
52976 ncm-et-15 2018-09-08 20:57 01:16:58 1983419 500 223 19 258 +150.51 ± 20.62
52975 ncm-et-13 2018-09-08 20:56 01:17:14 1987665 500 221 20 259 +148.02 ± 20.6
52974 ncm-et-10 2018-09-08 19:46 01:16:47 1984047 500 207 17 276 +138.99 ± 19.64
52973 ncm-et-4 2018-09-08 19:43 01:16:30 1989558 500 241 13 246 +171.02 ± 21.1
52972 ncm-et-3 2018-09-08 19:42 01:16:38 1986097 500 206 13 281 +141.44 ± 19.27
52971 ncm-et-15 2018-09-08 19:41 01:15:28 1984833 500 211 18 271 +141.44 ± 19.92
52970 ncm-et-9 2018-09-08 19:40 01:16:48 1989718 500 221 21 258 +147.19 ± 20.67
52969 ncm-et-13 2018-09-08 19:39 01:16:06 1987353 500 214 20 266 +142.26 ± 20.23
52968 ncm-et-4 2018-09-08 18:24 01:17:53 1989875 500 212 18 270 +142.26 ± 19.98
52967 ncm-et-3 2018-09-08 18:24 01:16:50 1989086 500 229 9 262 +164.07 ± 20.1
52966 ncm-et-13 2018-09-08 18:24 01:14:38 1988458 500 228 19 253 +154.7 ± 20.89
52965 ncm-et-9 2018-09-08 18:23 01:15:31 1987352 500 202 15 283 +136.56 ± 19.23
52964 ncm-et-10 2018-09-08 18:23 01:21:32 1908339 500 222 12 266 +155.54 ± 20.0
52963 ncm-et-15 2018-09-08 18:23 01:16:06 1983576 500 201 14 285 +136.56 ± 19.1
4962 ncm-et-12 2018-03-05 03:01 00:47:38 1993937 311 151 11 149 +168.47 ± 27.41
4961 ncm-et-7 2018-03-05 03:00 00:49:10 2027046 324 140 11 173 +146.43 ± 25.0
4960 ncm-et-14 2018-03-05 02:59 00:49:42 2019208 331 137 14 180 +135.6 ± 24.55
4959 ncm-et-5 2018-03-05 02:59 00:49:30 2021325 334 161 12 161 +166.71 ± 26.35
4958 ncm-et-3 2018-03-05 02:58 00:50:32 2026555 334 162 13 159 +166.71 ± 26.6
4957 ncm-et-11 2018-03-05 02:58 00:50:47 2017587 336 164 6 166 +177.32 ± 25.53
4956 ncm-et-10 2018-03-05 02:58 00:50:32 2017420 338 156 13 169 +156.84 ± 25.63
4955 ncm-et-8 2018-03-05 02:58 00:50:52 2021651 330 156 11 163 +163.81 ± 26.07
4954 ncm-et-4 2018-03-05 02:57 00:51:33 2023446 345 155 11 179 +154.45 ± 24.66
4953 ncm-et-13 2018-03-05 02:57 00:51:37 2027046 341 159 12 170 +160.25 ± 25.52
4952 ncm-et-9 2018-03-05 02:57 00:51:41 2022630 326 134 11 181 +137.9 ± 24.24
4951 ncm-et-6 2018-03-05 02:56 00:52:35 2023938 350 155 16 179 +146.01 ± 24.92
4950 ncm-et-5 2018-03-05 01:42 01:15:33 2004123 500 210 11 279 +146.36 ± 19.3
4949 ncm-et-14 2018-03-05 01:42 01:16:10 2020184 500 262 17 221 +186.25 ± 22.65
4948 ncm-et-4 2018-03-05 01:42 01:14:16 2024917 500 229 13 258 +160.64 ± 20.45
4947 ncm-et-7 2018-03-05 01:41 01:17:07 2015241 500 213 9 278 +150.51 ± 19.28
4946 ncm-et-12 2018-03-05 01:41 01:18:13 1912594 500 218 13 269 +151.35 ± 19.88
4945 ncm-et-11 2018-03-05 01:41 01:15:34 2018558 500 227 14 259 +158.08 ± 20.43
4944 ncm-et-9 2018-03-05 01:41 01:14:44 2025408 500 210 17 273 +141.44 ± 19.79
4943 ncm-et-13 2018-03-05 01:40 01:15:37 2026063 500 219 23 258 +143.89 ± 20.72
4942 ncm-et-3 2018-03-05 01:40 01:16:31 2024428 500 210 18 272 +140.62 ± 19.87
4941 ncm-et-8 2018-03-05 01:40 01:16:21 1996606 500 214 14 272 +147.19 ± 19.76
4940 ncm-et-10 2018-03-05 01:40 01:16:54 2020673 500 222 21 257 +148.02 ± 20.72
4939 ncm-et-6 2018-03-05 01:40 01:15:28 2027374 500 232 14 254 +162.35 ± 20.7

Commit

Commit ID 450f04969c5699fb9a4b39b883c2f37d122de290
Author Stefano Cardanobile
Date 2018-03-04 15:55:58 UTC
Using a S-curve for the optimism measure Add a logarithmic term in the optimism computation, increase the maximal optimism and lower the contempt offset. This increases the dynamics of the optimism aspects, giving a boost for balanced positions without skewing too much on unbalanced positions (but this version will enter panic mode faster than previous master when behind, trying to draw faster when slightly behind). This helps, since optimism is in general a good thing, for instance at LTC, but too high optimism rapidly contaminates play. passed STC: LLR: 2.96 (-2.94,2.94) [0.00,5.00] Total: 159343 W: 34489 L: 33588 D: 91266 http://tests.stockfishchess.org/tests/view/5a8db9340ebc590297cc85b6 passed LTC: LLR: 2.97 (-2.94,2.94) [0.00,5.00] Total: 47491 W: 7825 L: 7517 D: 32149 http://tests.stockfishchess.org/tests/view/5a9456a80ebc590297cc8a89 It must be mentioned that a version of the PR with contempt 0 did not pass STC [0,5]. The version in the patch, which uses default contempt 12, was found to be as strong as current master on different matches against SF7 and SF8, both at STC and LTC. One drawback maybe is that it raises the draw rate in self-play from 56% to 59%, giving a little bit less sensitivity for SF developpers to find evaluation improvements by selfplay tests in fishtest. Possible further work: • tune the values accurately, while keeping in mind the drawrate issue • check whether it is possible to remove linear and offset term • try to simplify the S-shape curve Bench: 5934644
Copyright 2011–2024 Next Chess Move LLC