Dev Builds » 20210923-2119

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 08:22:19 1961141 3346 2862 2 482 +442.47 +/- 15.5
ncm-et-4 08:22:21 1960166 3324 2774 7 543 +415.53 +/- 14.61
ncm-et-9 08:22:34 1959710 3348 2876 7 465 +445.3 +/- 15.81
ncm-et-10 08:22:44 1956946 3308 2829 7 472 +440.33 +/- 15.69
ncm-et-13 08:22:29 1959233 3337 2889 8 440 +453.87 +/- 16.26
ncm-et-15 08:22:25 1960562 3337 2853 1 483 +442.35 +/- 15.48
20000 17083 32 2885 +439.65 +/- 6.33

Test Detail

ID Host Started (UTC) Duration Base NPS Games Wins Losses Draws Elo CLI PGN
151784 ncm-et-10 2021-09-25 04:45 00:46:49 1955871 308 267 0 41 +458.75 +/- 54.45
151783 ncm-et-4 2021-09-25 04:43 00:49:07 1962834 324 269 1 54 +409.65 +/- 47.13
151782 ncm-et-13 2021-09-25 04:42 00:49:51 1958318 337 301 0 36 +499.4 +/- 58.47
151781 ncm-et-15 2021-09-25 04:41 00:51:19 1963392 337 290 0 47 +450.06 +/- 50.65
151780 ncm-et-9 2021-09-25 04:40 00:52:13 1959235 348 307 1 40 +476.93 +/- 55.34
151779 ncm-et-3 2021-09-25 04:39 00:52:28 1947695 346 310 0 36 +504.23 +/- 58.48
151778 ncm-et-10 2021-09-25 03:29 01:15:01 1956942 500 432 0 68 +454.76 +/- 41.82
151777 ncm-et-13 2021-09-25 03:26 01:14:50 1964772 500 428 0 72 +444.08 +/- 40.59
151776 ncm-et-15 2021-09-25 03:26 01:14:19 1959565 500 422 0 78 +429.05 +/- 38.92
151775 ncm-et-4 2021-09-25 03:25 01:16:52 1958473 500 413 1 86 +406.2 +/- 37.04
151774 ncm-et-3 2021-09-25 03:24 01:14:24 1969401 500 432 0 68 +454.76 +/- 41.82
151773 ncm-et-9 2021-09-25 03:23 01:15:51 1958012 500 425 1 74 +433.94 +/- 40.08
151772 ncm-et-10 2021-09-25 02:13 01:15:20 1959703 500 425 0 75 +436.43 +/- 39.73
151771 ncm-et-15 2021-09-25 02:11 01:14:04 1960770 500 445 0 55 +494.02 +/- 46.77
151770 ncm-et-4 2021-09-25 02:11 01:13:57 1977492 500 417 2 81 +412.8 +/- 38.27
151769 ncm-et-13 2021-09-25 02:09 01:16:37 1949431 500 421 3 76 +419.61 +/- 39.58
151768 ncm-et-3 2021-09-25 02:08 01:15:34 1962783 500 417 0 83 +417.31 +/- 37.67
151767 ncm-et-9 2021-09-25 02:07 01:15:39 1955125 500 435 0 65 +463.15 +/- 42.83
151766 ncm-et-10 2021-09-25 00:56 01:16:58 1954657 500 420 1 79 +421.93 +/- 38.73
151765 ncm-et-15 2021-09-25 00:54 01:16:42 1958318 500 435 0 65 +463.15 +/- 42.83
151764 ncm-et-4 2021-09-25 00:53 01:16:35 1955761 500 416 1 83 +412.8 +/- 37.74
151763 ncm-et-3 2021-09-25 00:53 01:14:47 1959542 500 424 1 75 +431.48 +/- 39.8
151762 ncm-et-13 2021-09-25 00:52 01:16:08 1950874 500 420 2 78 +419.61 +/- 39.03
151761 ncm-et-9 2021-09-25 00:51 01:14:44 1954805 500 429 2 69 +441.5 +/- 41.61
151760 ncm-et-15 2021-09-24 23:39 01:14:12 1959545 500 416 0 84 +415.04 +/- 37.43
151759 ncm-et-10 2021-09-24 23:38 01:17:05 1957098 500 423 2 75 +426.65 +/- 39.83
151758 ncm-et-3 2021-09-24 23:38 01:14:21 1961782 500 435 0 65 +463.15 +/- 42.83
151757 ncm-et-4 2021-09-24 23:37 01:16:09 1950826 500 414 0 86 +410.58 +/- 36.97
151756 ncm-et-13 2021-09-24 23:36 01:15:18 1959848 500 442 1 57 +481.1 +/- 45.96
151755 ncm-et-9 2021-09-24 23:36 01:14:44 1961240 500 427 1 72 +438.95 +/- 40.66
151754 ncm-et-10 2021-09-24 22:22 01:15:16 1957414 500 429 3 68 +438.95 +/- 41.92
151753 ncm-et-15 2021-09-24 22:22 01:15:59 1958350 500 423 1 76 +429.05 +/- 39.52
151752 ncm-et-4 2021-09-24 22:22 01:14:15 1956235 500 426 0 74 +438.95 +/- 40.01
151751 ncm-et-3 2021-09-24 22:21 01:15:29 1958472 500 421 0 79 +426.65 +/- 38.66
151750 ncm-et-9 2021-09-24 22:21 01:14:22 1963717 500 420 1 79 +421.93 +/- 38.73
151749 ncm-et-13 2021-09-24 22:21 01:15:06 1966305 500 439 1 60 +471.92 +/- 44.73
151748 ncm-et-15 2021-09-24 21:06 01:15:50 1963994 500 422 0 78 +429.05 +/- 38.92
151747 ncm-et-4 2021-09-24 21:06 01:15:26 1959543 500 419 2 79 +417.32 +/- 38.77
151746 ncm-et-3 2021-09-24 21:06 01:15:16 1968312 500 423 1 76 +429.05 +/- 39.52
151745 ncm-et-13 2021-09-24 21:05 01:14:39 1965085 500 438 1 61 +468.96 +/- 44.35
151744 ncm-et-9 2021-09-24 21:05 01:15:01 1965842 500 433 1 66 +454.76 +/- 42.55
151743 ncm-et-10 2021-09-24 21:05 01:16:15 1956941 500 433 1 66 +454.76 +/- 42.55

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