Dev Builds » 20230425-0617

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 09:29:44 1956688 3343 2856 5 482 +440.0 +/- 15.52
ncm-et-4 09:29:45 1955898 3347 2866 4 477 +442.91 +/- 15.59
ncm-et-9 09:29:41 1950060 3315 2852 1 462 +449.39 +/- 15.84
ncm-et-10 09:29:52 1948956 3328 2853 0 475 +445.75 +/- 15.61
ncm-et-13 09:29:49 1961281 3354 2895 5 454 +451.58 +/- 16.0
ncm-et-15 09:29:33 1954232 3313 2799 10 504 +426.46 +/- 15.18
20000 17121 25 2854 +442.53 +/- 6.36

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
185462 ncm-et-15 2023-04-27 08:13 00:53:51 1953594 313 255 0 58 +396.36 +/- 45.22
185461 ncm-et-9 2023-04-27 08:12 00:54:57 1947536 315 275 0 40 +467.51 +/- 55.19
185460 ncm-et-10 2023-04-27 08:11 00:55:59 1958439 328 280 0 48 +441.06 +/- 50.07
185459 ncm-et-3 2023-04-27 08:09 00:58:07 1960635 343 290 0 53 +430.84 +/- 47.52
185458 ncm-et-4 2023-04-27 08:08 00:59:13 1956037 347 295 0 52 +436.61 +/- 48.01
185457 ncm-et-13 2023-04-27 08:08 00:59:26 1964354 354 314 0 40 +489.08 +/- 55.26
185450 ncm-et-9 2023-04-27 06:47 01:25:05 1945403 500 439 0 61 +474.93 +/- 44.28
185449 ncm-et-15 2023-04-27 06:46 01:27:00 1950470 500 432 0 68 +454.76 +/- 41.82
185448 ncm-et-10 2023-04-27 06:45 01:25:52 1949955 500 433 0 67 +457.52 +/- 42.15
185447 ncm-et-3 2023-04-27 06:43 01:25:30 1955426 500 437 0 63 +468.95 +/- 43.54
185446 ncm-et-13 2023-04-27 06:42 01:25:25 1955568 500 433 2 65 +452.04 +/- 42.92
185445 ncm-et-4 2023-04-27 06:41 01:25:53 1949354 500 428 2 70 +438.95 +/- 41.29
185438 ncm-et-9 2023-04-27 05:19 01:26:49 1950246 500 423 0 77 +431.48 +/- 39.18
185437 ncm-et-15 2023-04-27 05:19 01:25:58 1954805 500 420 1 79 +421.93 +/- 38.73
185436 ncm-et-10 2023-04-27 05:19 01:25:35 1946765 500 433 0 67 +457.52 +/- 42.15
185435 ncm-et-3 2023-04-27 05:17 01:25:00 1958099 500 434 0 66 +460.32 +/- 42.48
185434 ncm-et-4 2023-04-27 05:16 01:24:29 1965851 500 437 1 62 +466.04 +/- 43.97
185433 ncm-et-13 2023-04-27 05:16 01:25:21 1954655 500 424 3 73 +426.65 +/- 40.41
185426 ncm-et-9 2023-04-27 03:53 01:25:47 1947316 500 427 1 72 +438.95 +/- 40.66
185425 ncm-et-3 2023-04-27 03:52 01:24:16 1962625 500 426 3 71 +431.48 +/- 41.0
185424 ncm-et-15 2023-04-27 03:52 01:26:40 1943351 500 422 1 77 +426.65 +/- 39.25
185423 ncm-et-4 2023-04-27 03:51 01:24:30 1956068 500 430 0 70 +449.35 +/- 41.19
185422 ncm-et-10 2023-04-27 03:51 01:26:49 1945558 500 433 0 67 +457.52 +/- 42.15
185421 ncm-et-13 2023-04-27 03:50 01:25:01 1974058 500 434 0 66 +460.32 +/- 42.48
185414 ncm-et-9 2023-04-27 02:28 01:24:28 1955002 500 417 0 83 +417.31 +/- 37.67
185413 ncm-et-15 2023-04-27 02:25 01:25:35 1956024 500 421 2 77 +421.93 +/- 39.29
185412 ncm-et-4 2023-04-27 02:25 01:25:10 1956185 500 430 0 70 +449.35 +/- 41.19
185411 ncm-et-3 2023-04-27 02:25 01:26:28 1957563 500 424 1 75 +431.48 +/- 39.8
185410 ncm-et-13 2023-04-27 02:24 01:25:10 1956489 500 422 0 78 +429.05 +/- 38.92
185409 ncm-et-10 2023-04-27 02:24 01:26:13 1949332 500 424 0 76 +433.94 +/- 39.45
185402 ncm-et-9 2023-04-27 01:01 01:25:47 1950867 500 427 0 73 +441.5 +/- 40.29
185401 ncm-et-10 2023-04-27 01:01 01:22:49 1947849 500 420 0 80 +424.27 +/- 38.4
185400 ncm-et-3 2023-04-27 01:00 01:24:30 1953135 500 423 0 77 +431.48 +/- 39.18
185399 ncm-et-4 2023-04-27 01:00 01:24:54 1951920 500 424 0 76 +433.94 +/- 39.45
185398 ncm-et-15 2023-04-27 00:59 01:26:01 1958776 500 415 4 81 +404.05 +/- 38.29
185397 ncm-et-13 2023-04-27 00:58 01:25:32 1956715 500 444 0 56 +490.71 +/- 46.32
185390 ncm-et-3 2023-04-26 23:34 01:25:53 1949335 500 422 1 77 +426.65 +/- 39.25
185389 ncm-et-9 2023-04-26 23:34 01:26:48 1954052 500 444 0 56 +490.71 +/- 46.32
185388 ncm-et-15 2023-04-26 23:34 01:24:28 1962609 500 434 2 64 +454.76 +/- 43.27
185387 ncm-et-4 2023-04-26 23:34 01:25:36 1955877 500 422 1 77 +426.65 +/- 39.25
185386 ncm-et-10 2023-04-26 23:34 01:26:35 1944798 500 430 0 70 +449.35 +/- 41.19
185385 ncm-et-13 2023-04-26 23:33 01:23:54 1967129 500 424 0 76 +433.94 +/- 39.45

Commit

Commit ID c3ce2204083400267592dc088b8ad9e88aed56b1
Author Linmiao Xu
Date 2023-04-25 06:17:22 UTC
Created by retraining the master net with these changes to the dataset: * Extending v6 filtering to data from T77 dec2021, T79 may2022, and T80 nov2022 * Reducing the number of duplicate positions, prioritizing position scores seen later in time * Using a binpack minimizer to reduce the overall data size Trained the same way as the previous master net, aside from the dataset changes: ``` python3 easy_train.py \ --experiment-name leela96-dfrc99-T60novdec-v2-T80augsep-v6-T80junjuloctnovT79aprmayT78jantosepT77dec-v6dd \ --training-dataset /data/leela96-dfrc99-T60novdec-v2-T80augsep-v6-T80junjuloctnovT79aprmayT78jantosepT77dec-v6dd.binpack \ --nnue-pytorch-branch linrock/nnue-pytorch/misc-fixes \ --start-from-engine-test-net True \ --early-fen-skipping 30 \ --start-lambda 1.0 \ --end-lambda 0.7 \ --max_epoch 900 \ --lr 4.375e-4 \ --gamma 0.995 \ --tui False \ --gpus "0," \ --seed $RANDOM ``` The new v6-dd filtering reduces duplicate positions by iterating over hourly data files within leela test runs, starting with the most recent, then keeping positions the first time they're seen and ignoring positions that are seen again. This ordering was done with the assumption that position scores seen later in time are generally more accurate than scores seen earlier in the test run. Positions are de-duplicated based on piece orientations, the first token in fen strings. The binpack minimizer was run with default settings after first merging monthly data into single binpacks. ``` python3 interleave_binpacks.py \ leela96-filt-v2.binpack \ dfrc99-filt-v2.binpack \ T60-nov2021-12tb7p-eval-filt-v2.binpack \ T60-dec2021-12tb7p-eval-filt-v2.binpack \ filt-v6/test80-aug2022-16tb7p-filter-v6.min-mar2023.binpack \ filt-v6/test80-sep2022-16tb7p-filter-v6.min-mar2023.binpack \ filt-v6-dd/test80-jun2022-16tb7p-filter-v6-dd.min-mar2023.binpack \ filt-v6-dd/test80-jul2022-16tb7p-filter-v6-dd.binpack \ filt-v6-dd/test80-oct2022-16tb7p-filter-v6-dd.binpack \ filt-v6-dd/test80-nov2022-16tb7p-filter-v6-dd.binpack \ filt-v6-dd/test79-apr2022-16tb7p-filter-v6-dd.binpack \ filt-v6-dd/test79-may2022-16tb7p-filter-v6-dd.binpack \ filt-v6-dd/test78-jantomay2022-16tb7p-filter-v6-dd.min-mar2023.binpack \ filt-v6-dd/test78-juntosep2022-16tb7p-filter-v6-dd.binpack \ filt-v6-dd/test77-dec2021-16tb7p-filter-v6-dd.binpack \ /data/leela96-dfrc99-T60novdec-v2-T80augsep-v6-T80junjuloctnovT79aprmayT78jantosepT77dec-v6dd.binpack ``` The code for v6-dd filtering is available along with training data preparation scripts at: https://github.com/linrock/nnue-data Links for downloading the training data components: https://robotmoon.com/nnue-training-data/ The binpack minimizer is from: #4447 Local elo at 25k nodes per move: nn-epoch859.nnue : 1.2 +/- 2.6 Passed STC: https://tests.stockfishchess.org/tests/view/643aad7db08900ff1bc5a832 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 565040 W: 150225 L: 149162 D: 265653 Ptnml(0-2): 1875, 62137, 153229, 63608, 1671 Passed LTC: https://tests.stockfishchess.org/tests/view/643ecf2fa43cf30e719d2042 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 1014840 W: 274645 L: 272456 D: 467739 Ptnml(0-2): 515, 98565, 306970, 100956, 414 closes https://github.com/official-stockfish/Stockfish/pull/4545 bench 3476305
Copyright 2011–2024 Next Chess Move LLC