Dev Builds » 20141208-0018

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 06:00:02 1988046 2385 180 682 1523 -74.24 +/- 8.22
ncm-et-4 06:00:19 1972086 2387 192 686 1509 -72.96 +/- 8.3
ncm-et-5 01:47:09 2020678 724 54 215 455 -78.57 +/- 15.12
ncm-et-6 01:47:25 2024836 702 51 186 465 -67.66 +/- 14.64
ncm-et-7 01:47:19 2010122 714 55 201 458 -72.06 +/- 14.98
ncm-et-8 01:47:05 2005136 711 62 207 442 -71.86 +/- 15.47
ncm-et-9 06:00:16 1990900 2393 175 728 1490 -81.77 +/- 8.37
ncm-et-10 06:00:16 1986692 2381 175 720 1486 -80.96 +/- 8.38
ncm-et-11 01:47:04 2016372 712 60 192 460 -65.17 +/- 14.95
ncm-et-12 01:47:28 2014872 716 40 182 494 -69.83 +/- 13.79
ncm-et-13 06:00:08 1990633 2382 183 667 1532 -71.59 +/- 8.18
ncm-et-14 01:46:50 2020673 719 52 219 448 -82.2 +/- 15.27
ncm-et-15 06:00:18 1987452 2351 166 680 1505 -77.21 +/- 8.24
ncm-et-16 01:47:21 2000542 723 52 196 475 -70.14 +/- 14.53
20000 1497 5761 12742 -75.23 +/- 2.84

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
83372 ncm-et-15 2019-05-30 12:03 00:22:28 1967853 143 12 37 94 -61.37 +/- 32.92 ↓
83371 ncm-et-4 2019-05-30 12:00 00:25:17 1968978 172 18 48 106 -61.23 +/- 31.93 ↓
83370 ncm-et-10 2019-05-30 11:59 00:25:57 1972591 164 18 43 103 -53.38 +/- 32.25 ↓
83369 ncm-et-3 2019-05-30 11:59 00:26:28 1962920 162 7 48 107 -89.88 +/- 30.09 ↓
83368 ncm-et-13 2019-05-30 11:58 00:26:51 1970323 176 12 50 114 -76.21 +/- 29.85 ↓
83367 ncm-et-9 2019-05-30 11:58 00:27:45 1967078 182 16 48 118 -61.73 +/- 29.58 ↓
83366 ncm-et-15 2019-05-30 10:46 01:16:10 1975603 500 34 151 315 -82.83 +/- 18.11 ↓
83365 ncm-et-4 2019-05-30 10:42 01:16:48 1967386 500 41 159 300 -83.57 +/- 18.95 ↓
83364 ncm-et-10 2019-05-30 10:42 01:16:16 1964153 500 45 139 316 -66.1 +/- 18.23 ↓
83363 ncm-et-13 2019-05-30 10:41 01:15:46 1965843 500 35 133 332 -68.99 +/- 17.29 ↓
83362 ncm-et-9 2019-05-30 10:41 01:15:31 1973897 500 38 152 310 -80.63 +/- 18.42 ↓
83361 ncm-et-3 2019-05-30 10:41 01:16:52 1965695 500 46 142 312 -67.55 +/- 18.43 ↓
83360 ncm-et-15 2019-05-30 09:28 01:16:30 1975294 500 37 146 317 -76.98 +/- 18.06 ↓
83359 ncm-et-4 2019-05-30 09:26 01:14:55 1962927 500 48 146 306 -68.99 +/- 18.75 ↓
83358 ncm-et-10 2019-05-30 09:25 01:15:33 1974009 500 35 155 310 -85.04 +/- 18.37 ↓
83357 ncm-et-13 2019-05-30 09:25 01:15:11 1982388 500 39 131 330 -64.66 +/- 17.46 ↓
83356 ncm-et-9 2019-05-30 09:25 01:14:54 1982992 500 33 145 322 -79.17 +/- 17.75 ↓
83355 ncm-et-3 2019-05-30 09:25 01:14:56 1965843 500 39 152 309 -79.9 +/- 18.48 ↓
83354 ncm-et-4 2019-05-30 08:09 01:15:51 1966001 500 31 155 314 -88.0 +/- 18.1 ↓
83353 ncm-et-10 2019-05-30 08:09 01:15:12 1972121 500 39 155 306 -82.1 +/- 18.63 ↓
83352 ncm-et-9 2019-05-30 08:09 01:14:51 1968320 500 44 151 305 -75.52 +/- 18.74 ↓
83351 ncm-et-3 2019-05-30 08:09 01:14:39 1983332 500 34 150 316 -82.1 +/- 18.06 ↓
83350 ncm-et-15 2019-05-30 08:09 01:17:54 1970958 500 38 139 323 -71.16 +/- 17.79 ↓
83349 ncm-et-13 2019-05-30 08:09 01:14:59 1973286 500 41 142 317 -71.16 +/- 18.12 ↓
18960 ncm-et-15 2018-04-14 18:10 00:30:53 2015313 208 15 59 134 -74.62 +/- 27.63 ↓
18959 ncm-et-8 2018-04-14 18:10 00:30:48 2025081 211 13 59 139 -76.98 +/- 26.73 ↓
18958 ncm-et-13 2018-04-14 18:10 00:31:22 2026390 206 14 60 132 -78.91 +/- 27.84 ↓
18957 ncm-et-6 2018-04-14 18:10 00:31:25 2025081 202 17 56 129 -67.93 +/- 28.4 ↓
18956 ncm-et-7 2018-04-14 18:09 00:31:43 1995001 214 18 54 142 -59.01 +/- 26.65 ↓
18955 ncm-et-12 2018-04-14 18:09 00:31:47 2010212 216 11 56 149 -73.46 +/- 25.07 ↓
18954 ncm-et-10 2018-04-14 18:09 00:32:07 2018070 217 12 67 138 -90.02 +/- 27.11 ↓
18953 ncm-et-11 2018-04-14 18:09 00:32:23 2013698 212 19 54 139 -57.89 +/- 27.13 ↓
18952 ncm-et-4 2018-04-14 18:09 00:32:26 1986246 215 20 56 139 -58.73 +/- 27.32 ↓
18951 ncm-et-9 2018-04-14 18:09 00:32:38 2028032 211 15 75 121 -101.6 +/- 30.08 ↓
18950 ncm-et-3 2018-04-14 18:08 00:32:41 2023936 223 22 62 139 -63.0 +/- 27.72 ↓
18949 ncm-et-5 2018-04-14 18:08 00:32:53 2019705 224 19 64 141 -70.76 +/- 27.31 ↓
18948 ncm-et-14 2018-04-14 18:08 00:32:47 2020510 219 13 72 134 -95.97 +/- 27.94 ↓
18947 ncm-et-16 2018-04-14 18:08 00:33:20 1982688 223 14 57 152 -67.84 +/- 25.16 ↓
18946 ncm-et-11 2018-04-14 16:53 01:14:41 2019046 500 41 138 321 -68.27 +/- 17.93 ↓
18945 ncm-et-14 2018-04-14 16:53 01:14:03 2020836 500 39 147 314 -76.25 +/- 18.24 ↓
18944 ncm-et-9 2018-04-14 16:53 01:14:37 2025082 500 29 157 314 -90.97 +/- 18.06 ↓
18943 ncm-et-10 2018-04-14 16:53 01:15:11 2019208 500 26 161 313 -96.19 +/- 18.04 ↓
18942 ncm-et-7 2018-04-14 16:53 01:15:36 2025244 500 37 147 316 -77.71 +/- 18.11 ↓
18941 ncm-et-3 2018-04-14 16:53 01:14:26 2026555 500 32 128 340 -67.55 +/- 16.84 ↓
18940 ncm-et-13 2018-04-14 16:53 01:15:59 2025572 500 42 151 307 -76.98 +/- 18.62 ↓
18939 ncm-et-15 2018-04-14 16:53 01:16:23 2019696 500 30 148 322 -83.57 +/- 17.69 ↓
18938 ncm-et-5 2018-04-14 16:53 01:14:16 2021651 500 35 151 314 -82.1 +/- 18.17 ↓
18937 ncm-et-16 2018-04-14 16:53 01:14:01 2018396 500 38 139 323 -71.16 +/- 17.79 ↓
18936 ncm-et-6 2018-04-14 16:53 01:16:00 2024591 500 34 130 336 -67.55 +/- 17.08 ↓
18935 ncm-et-8 2018-04-14 16:53 01:16:17 1985191 500 49 148 303 -69.71 +/- 18.9 ↓
18934 ncm-et-12 2018-04-14 16:53 01:15:41 2019533 500 29 126 345 -68.27 +/- 16.53 ↓
18933 ncm-et-4 2018-04-14 16:53 01:15:02 1980979 500 34 122 344 -61.79 +/- 16.67 ↓

Commit

Commit ID 158864270a055fe20dca4a87f4b7a8aa9cedfeb9
Author Ernesto Gatti
Date 2014-12-08 00:18:26 UTC
Simpler PRNG and faster magics search This patch replaces RKISS by a simpler and faster PRNG, xorshift64* proposed by S. Vigna (2014). It is extremely simple, has a large enough period for Stockfish's needs (2^64), requires no warming-up (allowing such code to be removed), and offers slightly better randomness than MT19937. Paper: http://xorshift.di.unimi.it/ Reference source code (public domain): http://xorshift.di.unimi.it/xorshift64star.c The patch also simplifies how init_magics() searches for magics: - Old logic: seed the PRNG always with the same seed, then use optimized bit rotations to tailor the RNG sequence per rank. - New logic: seed the PRNG with an optimized seed per rank. This has two advantages: 1. Less code and less computation to perform during magics search (not ROTL). 2. More choices for random sequence tuning. The old logic only let us choose from 4096 bit rotation pairs. With the new one, we can look for the best seeds among 2^64 values. Indeed, the set of seeds[][] provided in the patch reduces the effort needed to find the magics: 64-bit SF: Old logic -> 5,783,789 rand64() calls needed to find the magics New logic -> 4,420,086 calls 32-bit SF: Old logic -> 2,175,518 calls New logic -> 1,895,955 calls In the 64-bit case, init_magics() take 25 ms less to complete (Intel Core i5). Finally, when playing with strength handicap, non-determinism is achieved by setting the seed of the static RNG only once. Afterwards, there is no need to skip output values. The bench only changes because the Zobrist keys are now different (since they are random numbers straight out of the PRNG). The RNG seed has been carefully chosen so that the resulting Zobrist keys are particularly well-behaved: 1. All triplets of XORed keys are unique, implying that it would take at least 7 keys to find a 64-bit collision (test suggested by ceebo) 2. All pairs of XORed keys are unique modulo 2^32 3. The cardinality of { (key1 ^ key2) >> 48 } is as close as possible to the maximum (65536) Point 2 aims at ensuring a good distribution among the bits that determine an TT entry's cluster, likewise point 3 among the bits that form the TT entry's key16 inside a cluster. Details: Bitset card(key1^key2) ------ --------------- RKISS key16 64894 = 99.020% of theoretical maximum low18 180117 = 99.293% low32 305362 = 99.997% Xorshift64*, old seed key16 64918 = 99.057% low18 179994 = 99.225% low32 305350 = 99.993% Xorshift64*, new seed key16 65027 = 99.223% low18 181118 = 99.845% low32 305371 = 100.000% Bench: 9324905 Resolves #148
Copyright 2011–2024 Next Chess Move LLC
Feedback & Support Terms of Service Privacy Policy Desktop Version