| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Linmiao Xu
Date: Sun Jan 7 21:15:52 2024 +0100 Timestamp: 1704658552 Dual NNUE with L1-128 smallnet Credit goes to @mstembera for: - writing the code enabling dual NNUE: https://github.com/official-stockfish/Stockfish/pull/4898 - the idea of trying L1-128 trained exclusively on high simple eval positions The L1-128 smallnet is: - epoch 399 of a single-stage training from scratch - trained only on positions from filtered data with high material difference - defined by abs(simple_eval) > 1000 ```yaml experiment-name: 128--S1-only-hse-v2 training-dataset: - /data/hse/S3/dfrc99-16tb7p-eval-filt-v2.min.high-simple-eval-1k.binpack - /data/hse/S3/leela96-filt-v2.min.high-simple-eval-1k.binpack - /data/hse/S3/test80-apr2022-16tb7p.min.high-simple-eval-1k.binpack - /data/hse/S7/test60-2020-2tb7p.v6-3072.high-simple-eval-1k.binpack - /data/hse/S7/test60-novdec2021-12tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test77-nov2021-2tb7p.v6-3072.min.high-simple-eval-1k.binpack - /data/hse/S7/test77-dec2021-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test77-jan2022-2tb7p.high-simple-eval-1k.binpack - /data/hse/S7/test78-jantomay2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test78-juntosep2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test79-apr2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test79-may2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack # T80 2022 - /data/hse/S7/test80-may2022-16tb7p.high-simple-eval-1k.binpack - /data/hse/S7/test80-jun2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test80-jul2022-16tb7p.v6-dd.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-aug2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test80-sep2022-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test80-oct2022-16tb7p.v6-dd.high-simple-eval-1k.binpack - /data/hse/S7/test80-nov2022-16tb7p-v6-dd.min.high-simple-eval-1k.binpack # T80 2023 - /data/hse/S7/test80-jan2023-3of3-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test80-feb2023-16tb7p-filter-v6-dd.min-mar2023.unmin.high-simple-eval-1k.binpack - /data/hse/S7/test80-mar2023-2tb7p.v6-sk16.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-apr2023-2tb7p-filter-v6-sk16.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-may2023-2tb7p.v6.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-jun2023-2tb7p.v6-3072.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-jul2023-2tb7p.v6-3072.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-aug2023-2tb7p.v6.min.high-simple-eval-1k.binpack - /data/hse/S7/test80-sep2023-2tb7p.high-simple-eval-1k.binpack - /data/hse/S7/test80-oct2023-2tb7p.high-simple-eval-1k.binpack start-from-engine-test-net: False nnue-pytorch-branch: linrock/nnue-pytorch/L1-128 engine-test-branch: linrock/Stockfish/L1-128-nolazy engine-base-branch: linrock/Stockfish/L1-128 num-epochs: 500 lambda: 1.0 ``` Experiment yaml configs converted to easy_train.sh commands with: https://github.com/linrock/nnue-tools/blob/4339954/yaml_easy_train.py Binpacks interleaved at training time with: https://github.com/official-stockfish/nnue-pytorch/pull/259 Data filtered for high simple eval positions with: https://github.com/linrock/nnue-data/blob/32d6a68/filter_high_simple_eval_plain.py https://github.com/linrock/Stockfish/blob/61dbfe/src/tools/transform.cpp#L626-L655 Training data can be found at: https://robotmoon.com/nnue-training-data/ Local elo at 25k nodes per move of L1-128 smallnet (nnue-only eval) vs. L1-128 trained on standard S1 data: nn-epoch399.nnue : -318.1 +/- 2.1 Passed STC: https://tests.stockfishchess.org/tests/view/6574cb9d95ea6ba1fcd49e3b LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 62432 W: 15875 L: 15521 D: 31036 Elo +1.97 Ptnml(0-2): 177, 7331, 15872, 7633, 203 Passed LTC: https://tests.stockfishchess.org/tests/view/6575da2d4d789acf40aaac6e LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 64830 W: 16118 L: 15738 D: 32974 Elo +2.04 Ptnml(0-2): 43, 7129, 17697, 7497, 49 closes https://github.com/official-stockfish/Stockfish/pulls Bench: 1330050 Co-Authored-By: mstembera <5421953+> see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: mstembera
Date: Sun Jan 7 13:41:50 2024 +0100 Timestamp: 1704631310 Introduce BAD_QUIET movepicker stage Split quiets into good and bad as we do with captures. When we find the first quiet move below a certain threshold that has been sorted we consider all subsequent quiets bad. Inspired by @locutus2 idea to skip bad captures. Passed STC: https://tests.stockfishchess.org/tests/view/6597759f79aa8af82b95fa17 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 138688 W: 35566 L: 35096 D: 68026 Elo +1.18 Ptnml(0-2): 476, 16367, 35183, 16847, 471 Passed LTC: https://tests.stockfishchess.org/tests/view/6598583c79aa8af82b960ad0 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 84108 W: 21468 L: 21048 D: 41592 Elo +1.73 Ptnml(0-2): 38, 9355, 22858, 9755, 48 closes https://github.com/official-stockfish/Stockfish/pull/4970 Bench: 1336907 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Disservin
Date: Sun Jan 7 13:38:55 2024 +0100 Timestamp: 1704631135 Add .git-blame-ignore-revs Add a `.git-blame-ignore-revs` file which can be used to skip specified commits when blaming, this is useful to ignore formatting commits, like clang-format #4790. Github blame automatically supports this file format, as well as other third party tools. Git itself needs to be told about the file name to work, the following command will add it to the current git repo. `git config blame.ignoreRevsFile .git-blame-ignore-revs`, alternatively one has to specify it with every blame. `git blame --ignore-revs-file .git-blame-ignore-revs search.cpp` Supported since git 2.23. closes https://github.com/official-stockfish/Stockfish/pull/4969 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Michael Chaly
Date: Sun Jan 7 13:37:28 2024 +0100 Timestamp: 1704631048 Tweak usage of correction history Instead of using linear formula use quadratic one. Maximum impact of correction history is doubled this way, it breaks even with previous formula on half of maximum value. Passed STC: https://tests.stockfishchess.org/tests/view/659591e579aa8af82b95d7e8 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 225216 W: 57616 L: 57019 D: 110581 Elo +0.92 Ptnml(0-2): 747, 26677, 57201, 27198, 785 Passed LTC: https://tests.stockfishchess.org/tests/view/6596ee0b79aa8af82b95f08a LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 73314 W: 18524 L: 18125 D: 36665 Elo +1.89 Ptnml(0-2): 41, 8159, 19875, 8524, 58 closes https://github.com/official-stockfish/Stockfish/pull/4967 Bench: 1464785 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Miguel Lahoz
Date: Sun Jan 7 13:37:12 2024 +0100 Timestamp: 1704631032 Remove unneeded operator overload macros Only Direction type is using two of the enable overload macros. Aside from this, only two of the overloads are even being used. Therefore, we can just define the needed overloads and remove the macros. closes https://github.com/official-stockfish/Stockfish/pull/4966 No functional change. see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Thu Jan 4 15:56:53 2024 +0100 Timestamp: 1704380213 Remove redundant int cast Remove a redundant int cast in the calculation of fwdOut. The variable OutputType is already defined as std::int32_t, which is an integer type, making the cast unnecessary. closes https://github.com/official-stockfish/Stockfish/pull/4961 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Disservin
Date: Thu Jan 4 15:54:23 2024 +0100 Timestamp: 1704380063 Use type aliases instead of enums for Value types The primary rationale behind this lies in the fact that enums were not originally designed to be employed in the manner we currently utilize them. The Value enum was used like a type alias throughout the code and was often misused. Furthermore, changing the underlying size of the enum to int16_t broke everything, mostly because of the operator overloads for the Value enum, were causing data to be truncated. Since Value is now a type alias, the operator overloads are no longer required. Passed Non-Regression STC: https://tests.stockfishchess.org/tests/view/6593b8bb79aa8af82b95b401 LLR: 2.95 (-2.94,2.94) <-1.75,0.25> Total: 235296 W: 59919 L: 59917 D: 115460 Elo +0.00 Ptnml(0-2): 743, 27085, 62054, 26959, 807 closes https://github.com/official-stockfish/Stockfish/pull/4960 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: RainRat
Date: Thu Jan 4 15:51:56 2024 +0100 Timestamp: 1704379916 Fix typo in tbprobe.cpp closes https://github.com/official-stockfish/Stockfish/pull/4959 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Disservin
Date: Thu Jan 4 15:51:04 2024 +0100 Timestamp: 1704379864 Change the Move enum to a class This changes the Move enum to a class, this way all move related functions can be moved into the class and be more self contained. closes https://github.com/official-stockfish/Stockfish/pull/4958 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Viren6
Date: Thu Jan 4 15:49:33 2024 +0100 Timestamp: 1704379773 Modify ttPV reduction This patch modifies ttPV reduction by reducing 1 more unless ttValue is above alpha. Inspired from @pb00068 https://tests.stockfishchess.org/tests/view/658060796a3b4f6202215f1f Passed STC: https://tests.stockfishchess.org/tests/view/6591867679aa8af82b958328 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 37856 W: 9727 L: 9407 D: 18722 Elo +2.94 Ptnml(0-2): 99, 4444, 9568, 4672, 145 Passed LTC: https://tests.stockfishchess.org/tests/view/6591d9b679aa8af82b958a6c LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 128256 W: 32152 L: 31639 D: 64465 Elo +1.39 Ptnml(0-2): 64, 14364, 34772, 14851, 77 closes https://github.com/official-stockfish/Stockfish/pull/4957 Bench: 1176235 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Thu Jan 4 15:47:37 2024 +0100 Timestamp: 1704379657 Simplification of partial_insertion_sort formula. Passed STC: https://tests.stockfishchess.org/tests/view/6590110879aa8af82b9562e9 LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 134880 W: 34468 L: 34355 D: 66057 Elo +0.29 Ptnml(0-2): 476, 16060, 34220, 16243, 441 Passed LTC: https://tests.stockfishchess.org/tests/view/659156ca79aa8af82b957f07 LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 60780 W: 15179 L: 14996 D: 30605 Elo +1.05 Ptnml(0-2): 27, 6847, 16464, 7020, 32 closes https://github.com/official-stockfish/Stockfish/pull/4955 Bench: 1338331 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Disservin
Date: Thu Jan 4 15:47:10 2024 +0100 Timestamp: 1704379630 Update copyright year closes https://github.com/official-stockfish/Stockfish/pull/4954 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Disservin
Date: Thu Jan 4 15:45:33 2024 +0100 Timestamp: 1704379533 Silence security alert warning about possible infinite loop As some have noticed, a security alert has been complaining about a for loop in our TB code for quite some now. Though it was never a real issue, so not of high importance. A few lines earlier the symlen vector is resized `d->symlen.resize(number<uint16_t, LittleEndian>(data));` while this code seems odd at first, it resizes the array to at most (2 << 16) - 1 elements, basically making the infinite loop issue impossible to occur. closes https://github.com/official-stockfish/Stockfish/pull/4953 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Joseph Huang
Date: Thu Jan 4 15:45:03 2024 +0100 Timestamp: 1704379503 Lower MultiPV max to MAX_MOVES Link max value of MultiPV to that of MAX_MOVES which is 256 closes https://github.com/official-stockfish/Stockfish/pull/4951 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Disservin
Date: Mon Jan 1 02:31:25 2024 +0100 Timestamp: 1704072685 Fix formatting in search.cpp fixes the formatting for 1fe562fdf32c153f82929660197f8b97469f76b4 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Stefan Geschwentner
Date: Sun Dec 31 20:09:09 2023 +0100 Timestamp: 1704049749 Tweak static eval history update Modify the applied static eval bonus for main and pawn history with different factors for positive and negative values. Passed STC: https://tests.stockfishchess.org/tests/view/659132e179aa8af82b957bb0 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 12512 W: 3308 L: 3027 D: 6177 Elo +7.80 Ptnml(0-2): 32, 1372, 3189, 1609, 54 Passed LTC: https://tests.stockfishchess.org/tests/view/65913e3d79aa8af82b957cd2 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 35946 W: 9128 L: 8809 D: 18009 Elo +3.08 Ptnml(0-2): 19, 3879, 9862, 4190, 23 closes https://github.com/official-stockfish/Stockfish/pull/4952 Bench: 1392883 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Michael Chaly
Date: Sun Dec 31 20:00:06 2023 +0100 Timestamp: 1704049206 Introduce static evaluation correction history Idea from Caissa (https://github.com/Witek902/Caissa) chess engine. With given pawn structure collect data with how often search result and by how much it was better / worse than static evalution of position and use it to adjust static evaluation of positions with given pawn structure. Details: 1. excludes positions with fail highs and moves producing it being a capture; 2. update value is function of not only difference between best value and static evaluation but also is multiplied by linear function of depth; 3. maximum update value is maximum value of correction history divided by 2; 4. correction history itself is divided by 32 when applied so maximum value of static evaluation adjustment is 32 internal units. Passed STC: https://tests.stockfishchess.org/tests/view/658fc7b679aa8af82b955cac LLR: 2.96 (-2.94,2.94) <0.00,2.00> Total: 128672 W: 32757 L: 32299 D: 63616 Elo +1.24 Ptnml(0-2): 441, 15241, 32543, 15641, 470 Passed LTC: https://tests.stockfishchess.org/tests/view/65903f6979aa8af82b9566f1 LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 97422 W: 24626 L: 24178 D: 48618 Elo +1.60 Ptnml(0-2): 41, 10837, 26527, 11245, 61 closes https://github.com/official-stockfish/Stockfish/pull/4950 Bench: 1157852 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Sun Dec 31 19:58:10 2023 +0100 Timestamp: 1704049090 Mark square_bb() as constexpr closes https://github.com/official-stockfish/Stockfish/pull/4949 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Sun Dec 31 19:57:34 2023 +0100 Timestamp: 1704049054 Simplify the improving flag calculation Passed STC: https://tests.stockfishchess.org/tests/view/658ec29979aa8af82b9547f6 LLR: 2.93 (-2.94,2.94) <-1.75,0.25> Total: 93408 W: 23747 L: 23587 D: 46074 Elo +0.60 Ptnml(0-2): 340, 11178, 23527, 11300, 359 Passed LTC: https://tests.stockfishchess.org/tests/view/658f73e479aa8af82b9555b6 LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 64026 W: 15984 L: 15806 D: 32236 Elo +0.97 Ptnml(0-2): 31, 7113, 17552, 7281, 36 closes https://github.com/official-stockfish/Stockfish/pull/4948 Bench: 1143749 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Sun Dec 31 19:54:27 2023 +0100 Timestamp: 1704048867 Cleanup comments Tests used to derive some Elo worth comments: https://tests.stockfishchess.org/tests/view/656a7f4e136acbc573555a31 https://tests.stockfishchess.org/tests/view/6585fb455457644dc984620f closes https://github.com/official-stockfish/Stockfish/pull/4945 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Tobias Steinmann
Date: Sun Dec 31 19:51:04 2023 +0100 Timestamp: 1704048664 Update Makefile for android x86-64 builds For developing an Android GUI it can be helpful to use the Emulator on Windows. Therefor an android_x86-64 library of Stockfish is needed. It would be nice to compile it "out-of-the-box". This change is originally suggested by Craftyawesome closes https://github.com/official-stockfish/Stockfish/pull/4927 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Shahin M. Shahin
Date: Sat Dec 30 12:19:48 2023 +0100 Timestamp: 1703935188 Fix scores from reverse futility pruning This fixes futility pruning return values after recent tweaks, `eval` is guaranteed to be less than the mate-in range but it can be as low value such that the average between eval and beta can still fall in the mated-in range when beta is as low in mated range. i.e. (eval + beta) / 2 being at mated-range which can break mates. Passed non-regression STC: https://tests.stockfishchess.org/tests/view/658f3eed79aa8af82b955139 LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 117408 W: 29891 L: 29761 D: 57756 Elo +0.38 Ptnml(0-2): 386, 13355, 31120, 13429, 414 Passed non-regression LTC: https://tests.stockfishchess.org/tests/view/658f8b7a79aa8af82b9557bd LLR: 2.94 (-2.94,2.94) <-1.75,0.25> Total: 60240 W: 14962 L: 14786 D: 30492 Elo +1.02 Ptnml(0-2): 22, 6257, 17390, 6425, 26 changes signature at higher depth e.g. `128 1 15` closes https://github.com/official-stockfish/Stockfish/pull/4944 Bench: 1304666 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Linmiao Xu
Date: Sat Dec 30 11:08:03 2023 +0100 Timestamp: 1703930883 Update default net to nn-b1e55edbea57.nnue Created by retraining the master big net `nn-0000000000a0.nnue` on the same dataset with the ranger21 optimizer and more WDL skipping at training time. More WDL skipping is meant to increase lambda accuracy and train on fewer misevaluated positions where position scores are unlikely to correlate with game outcomes. Inspired by: - repeated reports in discord #events-discuss about SF misplaying due to wrong endgame evals, possibly due to Leela's endgame weaknesses reflected in training data - an attempt to reduce the skewed dataset piece count distribution where there are much more positions with less than 16 pieces, since the target piece count distribution in the trainer is symmetric around 16 The faster convergence seen with ranger21 is meant to: - prune experiment ideas more quickly since fewer epochs are needed to reach elo maxima - research faster potential trainings by shortening each run ```yaml experiment-name: 2560-S7-Re-514G-ranger21-more-wdl-skip training-dataset: /data/S6-514G.binpack early-fen-skipping: 28 start-from-engine-test-net: True nnue-pytorch-branch: linrock/nnue-pytorch/r21-more-wdl-skip num-epochs: 1200 lr: 4.375e-4 gamma: 0.995 start-lambda: 1.0 end-lambda: 0.7 ``` Experiment yaml configs converted to easy_train.sh commands with: https://github.com/linrock/nnue-tools/blob/4339954/yaml_easy_train.py Implementations based off of Sopel's NNUE training & experimentation log: https://docs.google.com/document/d/1gTlrr02qSNKiXNZ_SuO4-RjK4MXBiFlLE6jvNqqMkAY - Experiment 336 - ranger21 https://github.com/Sopel97/nnue-pytorch/tree/experiment_336 - Experiment 351 - more WDL skipping The version of the ranger21 optimizer used is: https://github.com/lessw2020/Ranger21/blob/b507df6/ranger21/ranger21.py The dataset is the exact same as in: https://github.com/official-stockfish/Stockfish/pull/4782 Local elo at 25k nodes per move: nn-epoch619.nnue : 6.2 +/- 4.2 Passed STC: https://tests.stockfishchess.org/tests/view/658a029779aa8af82b94fbe6 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 46528 W: 11985 L: 11650 D: 22893 Elo +2.50 Ptnml(0-2): 154, 5489, 11688, 5734, 199 Passed LTC: https://tests.stockfishchess.org/tests/view/658a448979aa8af82b95010f LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 265326 W: 66378 L: 65574 D: 133374 Elo +1.05 Ptnml(0-2): 153, 30175, 71254, 30877, 204 This was additionally tested with the latest DualNNUE and passed SPRTs: Passed STC vs. https://github.com/official-stockfish/Stockfish/pull/4919 https://tests.stockfishchess.org/tests/view/658bcd5c79aa8af82b951846 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 296128 W: 76273 L: 75554 D: 144301 Elo +0.84 Ptnml(0-2): 1223, 35768, 73617, 35979, 1477 Passed LTC vs. https://github.com/official-stockfish/Stockfish/pull/4919 https://tests.stockfishchess.org/tests/view/658c988d79aa8af82b95240f LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 75618 W: 19085 L: 18680 D: 37853 Elo +1.86 Ptnml(0-2): 45, 8420, 20497, 8779, 68 closes https://github.com/official-stockfish/Stockfish/pull/4942 Bench: 1304666 see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: FauziAkram
Date: Sat Dec 30 11:05:19 2023 +0100 Timestamp: 1703930719 Refactor bestvalue adjustment in qsearch closes https://github.com/official-stockfish/Stockfish/pull/4935 No functional change see source |
| Windows x64 for Haswell CPUs Windows x64 for modern computers + AVX2 Windows x64 for modern computers Windows x64 + SSSE3 Windows x64 Linux x64 for Haswell CPUs Linux x64 for modern computers + AVX2 Linux x64 for modern computers Linux x64 + SSSE3 Linux x64 | Author: Michael Chaly
Date: Sat Dec 30 11:01:32 2023 +0100 Timestamp: 1703930492 Adjust value returned after TT cutoff Instead of returning value from TT in case of a fail high return mix between it and beta. Passed STC: https://tests.stockfishchess.org/tests/view/658465395457644dc98446c7 LLR: 2.93 (-2.94,2.94) <0.00,2.00> Total: 220704 W: 56404 L: 55811 D: 108489 Elo +0.93 Ptnml(0-2): 750, 26214, 55921, 26627, 840 Passed LTC: https://tests.stockfishchess.org/tests/view/6585c3f55457644dc9845db9 LLR: 2.97 (-2.94,2.94) <0.50,2.50> Total: 124980 W: 31169 L: 30658 D: 63153 Elo +1.42 Ptnml(0-2): 57, 14147, 33603, 14594, 89 closes https://github.com/official-stockfish/Stockfish/pull/4934 Bench: 1191093 see source |