Recent Releases of brainlearn
brainlearn - Santiago
Brainlearn Santiago Stockfish Patch Oct 12, 2024 Make low ply history size fixed Various improvements and some bugs fixed.
- C++
Published by amchess over 1 year ago
brainlearn - Brainlearn 29
Aligned with Stockfish 17. Fixed mcts. Fixed persisted learning and added the experience book option and an utility for reset (to be done for old experience files) and display content given a fen. Extended livebook to LiChess in addition to ChessDB. Removed opening variety and replaced with variety which is now a combo with three values: -off the default -standard without loss of elo -psychological more risky because with elo loss, but not as significant, useful for inviting the engine to a game with psychological/human sacrifices, i.e., strictly speaking, incorrect, but almost impossible to refute.
- C++
Published by amchess over 1 year ago
brainlearn - Brainlearn 28.1
Improved Opening Variety (Idea implemented in Hypnos) Extended LiveBook (also egtbs) Now, the ChessDB livebook is usable, again not in analysis or search for mate, even in endgame positions with at most 7 units on the board)
- C++
Published by amchess almost 2 years ago
brainlearn - Brainlearn 28
Brainlearn 28 Aligned to Stockfish 16.1 Improved Opening Variety based on Ibai Burgos idea in Hypnos
- C++
Published by amchess almost 2 years ago
brainlearn - Brainlearn 27
Feb 9, 2024 Fix the alignment of the transformer buffer
- C++
Published by amchess about 2 years ago
brainlearn - Brainlearn 26.5
Bug fix Updated the Makefile for latest mingw build
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 26.4
Fix mcts UCI_ShowWdl to true by default Stockfish patch Nov 20, 2023 Reformat some comments A great thanks for their help to -Omar Khalid in correcting -Peter Martan in professionally testing the mcts code.
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 26.3
Fix mcts (eliminated GoldDigger and fixed a bug) Stockfish patch Nov 20, 2023 Reformat some comments UCI_ShowWDL to true by default
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 26.2
Fixed mcts. Fixed native build (thanks to Marco Zerbinati) Stockfish patch Nov 7, 2023 Remove pawn history from ProbCut constructor
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 26.1
Fixed and improved MCTS Stockfish patch Oct 24, 2023 Update 5 search params for pruning at shallow depth
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 26.1
Brainlearn 26.1 Fixed and improved MCTS Stockfish patch Oct 24, 2023 Update 5 search params for pruning at shallow depth
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 26
BrainLearn 26 Fixed learning Stockfish patch Oct14, 2023 Use more continuation histories. This patch allows stats updates and movepicker bonuses for continuation history 3 plies deep - so counter counter move. Updates and movepicker usage are done with 1/4 multiplier compared to other histories.
- C++
Published by amchess over 2 years ago
brainlearn - BrainLearn 25.2
Fixed a bug in learning: no more entries with depth=0 Old experience files can be reduced by this tool from Ibai Burgos. A great thanks to him! The engine is always aligned with official Stockfish release. From now on, I'll be always only aligned with Stockfish official release.
- C++
Published by amchess over 2 years ago
brainlearn - BrainLearn 25.1
BrainLearn 25.1 Fix learning. Now, the experience file has a .exp extension and not .bin as before. So, simply, rename the file. Stockfish patch Update default net to nn-5af11540bbfe.nnue
- C++
Published by amchess over 2 years ago
brainlearn - BrainLearn 25
Remove optimism multiplier in nnue eval calculation The same formula had passed SPRT against an earlier version of master.
- C++
Published by amchess over 2 years ago
brainlearn - Brainlearn 24
Brainlearn 24 Added book management section (thanks to Khalid Omar PolyFish: see the Readme) Restored Minimum thinking time uci option Stockfish patch May 7,2023: Refine deeper post-lmr searches This patch improves logic conditions for performing deeper searches after passed LMR.
Instead of exceeding alpha by some margin now it requires to exceed the current best value - which may be lower than alpha (but never bigger since we update alpha with bestvalue if it exceeds alpha).
- C++
Published by amchess almost 3 years ago
brainlearn - BrainLearn 23.1
Bug fix Stockfish patch Depth Tweak and tuning tunes reduction related parameters, and introduces more reduction on found good moves.
credit for this patch goes also to candirufish Yoshie2000 dubslow peregrineshahin Vizvezdenec
- C++
Published by amchess almost 3 years ago
brainlearn - BrainLearn 23
BrainLearn 23 Stockfish patch Assign negative stat bonuses for quiet moves at Pv nodes This patch assigns negative stat bonuses for quiet moves at pv nodes which are searched at depth greater than this node assumes, so are extended.
Passed STC: https://tests.stockfishchess.org/tests/view/6426198bdb43ab2ba6f9cfa2 LLR: 2.94 (-2.94,2.94) <0.00,2.00> Total: 548944 W: 147287 L: 146254 D: 255403 Ptnml(0-2): 1662, 59772, 150671, 60605, 1762
Passed LTC: https://tests.stockfishchess.org/tests/view/642be4f177ff3301150d892d LLR: 2.95 (-2.94,2.94) <0.50,2.50> Total: 93352 W: 25400 L: 24994 D: 42958 Ptnml(0-2): 44, 8817, 28547, 9225, 43
closes https://github.com/official-stockfish/Stockfish/pull/4495
bench 5044536
- C++
Published by amchess almost 3 years ago
brainlearn - Brainlearn 22
Brainlearn 29 Aligned with Stockfish patch 02/27/23 Update default net to nn-52471d67216a.nnue
- C++
Published by amchess almost 3 years ago
brainlearn - Brainlearn 21.1
Stockfish patch Fix stack initialization This fixes a bug where on line 278 the Stack::staticEvals are initialized to 0. However VALUE_NONE is defined to be 32002 so this is a bug in master. It is probably due to the calculation of improvement, where staticEval prior to rootPos
- C++
Published by amchess about 3 years ago
brainlearn - BrainLearn 21
Improved optimsm Stockfish patch Update default net to nn-60fa44e376d9.nnue #4314
- C++
Published by amchess about 3 years ago
brainlearn - BrainLearn 20.2
Brainlearn 20.2 Aligned with Stockfish 15.1
- C++
Published by amchess about 3 years ago
brainlearn -
Brainlearn 20.1 Improved optimism based on Shashchess and alignement score/gui. Fixed Opening Variety
Stockfish patch Correctly output lowerbound/upperbound in threaded searches fixes the lowerbound/upperbound output by taking the alpha,beta bracket into account also if a bestThread is selected that is different from the master thread.
Instead of keeping track which bounds where used in the specific search, in this version we simply store the quality (exact, upperbound, lowerbound) of the score along with the actual score as information on rootMove.
closes #4239
No functional change
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 20
Brainlearn 20 Introduced contempt for quiescent positions
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 19.2.1
19.2.1 Normalize evaluation Normalizes the internal value as reported by evaluate or search to the UCI centipawn result used in output. This value is derived from the winratemodel() such that Stockfish outputs an advantage of "100 centipawns" for a position if the engine has a 50% probability to win from this position in selfplay at fishtest LTC time control.
The reason to introduce this normalization is that our evaluation is, since NNUE, no longer related to the classical parameter PawnValueEg (=208). This leads to the current evaluation changing quite a bit from release to release, for example, the eval needed to have 50% win probability at fishtest LTC (in cp and internal Value):
June 2020 : 113cp (237) June 2021 : 115cp (240) April 2022 : 134cp (279) July 2022 : 167cp (348)
With this patch, a 100cp advantage will have a fixed interpretation, i.e. a 50% win chance. To keep this value steady, it will be needed to update the winratemodel() from time to time, based on fishtest data. This analysis can be performed with a set of scripts currently available at https://github.com/vondele/WLD_model
fixes #4155 closes #4216
No functional change
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 19.2
Brainlearn 19.2 Improved montecarlo (thanks to Khalid Omar) Adjust reduction less at medium depths This patch dampens the reduction increase/decrease from statScore at mid-range depths. Inspired by patterns noticed in this tune: https://tests.stockfishchess.org/tests/view/635188930e5f47a8d0ffe8f5
Passed STC: https://tests.stockfishchess.org/tests/view/63599dfd6b27ef94d9ec04af LLR: 2.95 (-2.94,2.94) <0.00,2.00> Total: 87464 W: 23519 L: 23134 D: 40811 Ptnml(0-2): 319, 9599, 23524, 9958, 332
Passed LTC: https://tests.stockfishchess.org/tests/view/635a73046b27ef94d9ec2313 LLR: 2.94 (-2.94,2.94) <0.50,2.50> Total: 154792 W: 41746 L: 41214 D: 71832 Ptnml(0-2): 79, 15181, 46349, 15703, 84
closes #4213
Bench 4271738
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 19.1
BrainLearn 19.1 Change log: -Modified Makefile for Android build when curl library gives problems. (USELIVEBOOK flag) - Fixed debug assertion in MonteCarlo::printchildren() by using root instead of currentnode() which is supposed to be equal to root - Fixed MCTS TT updating. - As code cleanup, removed the bool mcts parameter to search() and qsearch() and replaced it with a Thread flag, so that each threads knows if it is running MCTS or not Thanks to Khalid Omar Stockfish patch Fix tablebase probe for dtz >1000 w/o 50 move rule For qn4N1/6R1/3K4/8/B2k4/8/8/8 w - - 0 1, white loses with DTZ 1034. See https://syzygy-tables.info/?fen=qn4N1/6R1/3K4/8/B2k4/8/8/8w--01
Prior to this fix, due to a too small hard-coded value, Stockfish interpreted this as winning. The new value picked (1<<18) is large enough to deal with the largest DTZ values that can be stored in the current syzygy format.
closes https://github.com/official-stockfish/Stockfish/pull/4187
No functional change.
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 19
BrainLearn 19 Improved mcts code (a great thanks to Khalid Omar for his help) Stockfish patch Tweak history initialization Simplify initialization of continuation history by using everywhere the same starting value.
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 18
BrainLearn 18 Cleaned code Aigned with Stockfish patch Jul 13 2022 Update default net to nn-ad9b42354671.nnue using trainer branch glinscott/nnue-pytorch#208 with a slightly tweaked loss function (power 2.5 instead of 2.6), otherwise same training as in the previous net update #4100
- C++
Published by amchess over 3 years ago
brainlearn - BrainLearn 17
BrainLearn 17 Aligned with stockfish patch Negative extension for ttMove that is less than alpha and value in the context of singular extensions
- C++
Published by amchess almost 4 years ago
brainlearn - BrainLearn 16
BrainLearn 16 Improved alignment score-gui. Fixed the experimental Montecarlo Introduced the option Syzygy50MoveRule, as in Stockfish, useful, effectively, for example, in correspondence chess. Stockfish patch Generalize the feature transform to use vect macros This commit generalizes the feature transform to use vect macros that are architecture defined instead of using a seperate code path for each one.
It should make some old architectures (MMX, including improvements by Fanael) faster and make further such improvements easier in the future.
- C++
Published by amchess almost 4 years ago
brainlearn - BrainLearn 15.2
BrainLearn 15.2 Refactored MCTS (thanks to Norman Schmidt https://github.com/FireFather/stockfish-plus/tree/master/src ) More refined MCTS Fix on learning Stockfish patch Author: Rui Coelho Date: Thu Jan 13 22:25:01 2022 +0100 Timestamp: 1642109101
Use complexity in search
This patch uses the complexity measure (from #3875) as a heuristic for null move pruning. Hopefully, there may be room to use it in other pruning techniques. I would like to thank vondele and locutus2 for the feedback and suggestions during testing.
- C++
Published by amchess about 4 years ago
brainlearn -
BrainLearn 15 Great improvement of learning system Stockfish patch Author: Michael Chaly Date: Mon Nov 15 18:20:10 2021 +0100 Timestamp: 1636996810
Simplify Pv nodes related logic in LMR
- C++
Published by amchess over 4 years ago
brainlearn - BrainLearn 14.2
Tuned eval Author: bmc4 Date: Mon Nov 8 13:14:18 2021 +0100 Timestamp: 1636373658
Simplify away the Reverse Move penalty
This simplifies the penalty for reverse move introduced in https://github.com/official-stockfish/Stockfish/pull/2294 .
- C++
Published by amchess over 4 years ago
brainlearn - BrainLearn 14.1
Aligned with Stockfish 14.1
- C++
Published by amchess over 4 years ago
brainlearn - BrainLearn 14
Stockfish patch: Author: FauziAkram Date: Sat Oct 23 12:27:32 2021 +0200 Timestamp: 1634984852
Tuning of a Null Move Parameter
Solved a random bug on some systems with very large experience files. A big thank you to all the users who took it over. Great general improvement on learning system.
- C++
Published by amchess over 4 years ago
brainlearn - BrainLearn 13.1
Restored the trend (dynamic contempt). Tuned the classical and nnue eval. Stockfish patch: Author: xoto10 Date: Sun Sep 12 18:03:56 2021 +0200 Timestamp: 1631462636
Update 2 search parameters after tune.
- C++
Published by amchess over 4 years ago
brainlearn - BrainLearn 13
BrainLearn 13
MCTS evolved Concurrent Experience BrainLearnDGV (see the related branch for infos) Stockfish patch Author: Tomasz Sobczyk Date: Tue May 18 18:06:23 2021 +0200 Timestamp: 1621353983 New NNUE architecture and net
- C++
Published by amchess almost 5 years ago
brainlearn - BrainLearn 12.1
BrainLearn 12.1 Removed montecarlo Stockfish patch Author: Unai Corzo Date: Wed Oct 14 19:32:12 2020 +0200 Timestamp: 1602696732
Scale factor tweak
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 12
BrainLearn 12
Introduced Montecarlo Stockfish patch Author: Stéphane Nicolet Date: Mon Sep 28 22:42:26 2020 +0200 Timestamp: 1601325746 Include pawns in NNUE scaling
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 11
Author: SFisGOD Date: Tue Sep 8 22:56:08 2020 +0200 Timestamp: 1599598568
Double probability of using classical eval
This patch doubles the moderate imbalance threshold and probability of using classical eval. So now if imbalance is greater than PawnValueMg / 4 then there is a 1/8 chance of using classical eval.
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 10
BrainLearn10 Corrected some bugs in learning and improved it. Aligned to Stockfish main dev. A great thanks to Omar Khalid for his help. Stockfish patch Author: syzygy1 Date: Tue Aug 18 20:15:07 2020 +0200 Timestamp: 1597774507
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 1.1
Improved NNUE All executables Windows and Linux for 20MB and 30MB nets
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn NNUE 1.0
BrainLearn
NNUE 1.0: search traditional or with Kelly Kiniama & Andrea Manzo learning mechanism eval with nn.bin
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 9.1.1
9.1.1 Some other bug fixing/improvements on learning part (thanks again to Omar Khalid: a very fruitful collaboration!) Stockfish patch Author: syzygy1 Date: Sat Jul 11 22:32:34 2020 +0200 Timestamp: 1594499554
Corrects a functional change in a cleanup patch.
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 9.1
Omar Khalid patch on learning Stockfish patch Author: Stefan Geschwentner Date: Mon Jul 6 17:56:18 2020 +0200 Timestamp: 1594050978
Add a rank based bonus for blocked pawns.
Fix for overevaluated blocked pawns on the 5th and 6th rank. This is a rewrite of the original idea that uses only two parameters. Thanks to rocky640 for pointing this out.
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 9.0
It's a pleasure to annonce the new BrainLearn 9 release with a lot of novelties: -new Self Q-learning, a learning variant optimized for self play -better contempt management to avoid the well know rollcoaster effect: the score is truly aligned with the gui -improved LiveBook management -latest Stockfish patch Author: Joost VandeVondele Date: Sat Jun 27 10:22:27 2020 +0200 Timestamp: 1593246147 Revert LTO for mingw on windows.
- C++
Published by amchess over 5 years ago
brainlearn - BrainLearn 8.0
BrainLearn 8.0 Improved large pages version Introduced depth in livebook and readonly in bin files management Simplified the learning algorithm and improved the performance management by the private application Improved score visualized by the gui
- C++
Published by amchess almost 6 years ago
brainlearn - BrainLearn 7.1
Windows support for large pages. Separated sources Stockfish patch: Author: Lolligerhans Date: Sat Jan 25 09:44:09 2020 +0100 Timestamp: 1579941849
Tweak RestrictedPiece bonus
Double the "RestrictedPiece" bonus for restricted moves targeting occupied squares.
- C++
Published by amchess about 6 years ago
brainlearn - BrainLearn 7
Aligned with Stockfish 11 Correcter and improved learning algorithm
- C++
Published by amchess about 6 years ago
brainlearn - BrainLearn 6.1
BrainLearn 6.1
Corrected default livebook default delay Stockfish patch Author: xoto10 Date: Wed Dec 25 00:12:07 2019 +0100 Timestamp: 1577229127
Tune search constants
- C++
Published by amchess about 6 years ago
brainlearn - BrainLearn 6.0
Stockfish patch Author: Guenther Demetz Date: Tue Dec 10 08:07:34 2019 +0100 Timestamp: 1575961654
Refine improving-logic
See the readme file: https://github.com/amchess/BrainLearn/blob/master/Readme.md
- C++
Published by amchess about 6 years ago
brainlearn - BrainLearn 5.0
5.0 version The idea is if someone has key moves in the pgn file, it captures the moves, especially played by opponents monsters like leela This version is able to convert pgn files to brainlearn format, especially when targeting those stronger engines like leela. Elo increment +8 after 1000 bullet games, against previous version
Stockfish patch Author: Stéphane Nicolet Date: Wed Sep 25 22:28:51 2019 +0200 Timestamp: 1569443331
Temporary patch to show the compiler for TCEC submission
This patch shows a description of the compiler used to compile Stockfish, when starting from the console.
Usage:
./stockfish
compiler
Example of output:
``` Stockfish 240919 64 POPCNT by T. Romstad, M. Costalba, J. Kiiski, G. Linscott
Compiled by clang++ 9.0.0 on Apple VERSION macro expands to: 4.2.1 Compatible Apple LLVM 9.0.0 (clang-900.0.38) ```
No functional change
- C++
Published by amchess over 6 years ago
brainlearn - BrainLearn 4.2.1
Builds for gcc 8. Bug corrected. Now, the file is written only when the gui sends the quit command. Important An infinite analysis must be stopped before any other operation. If not, the learning HashTable is not filled and the experience.bin file is incorrectly written.
- C++
Published by amchess over 6 years ago
brainlearn - BrainLearn 4.2
Corrected some bugs Code more robust Stockfish patch Author: noobpwnftw Date: Mon Sep 16 15:09:45 2019 +0200 Timestamp: 1568639385
Raise stack size to 8MB for pthreads
It seems there is no other way to specify stack size on std::thread than linker flags and the effective flags are named differently in many toolchains. On toolchains where pthread is always available, this patch changes the stack size change in our C++ code via pthread to ensure a minimum stack size of 8MB, instead of relying on linker defaults which may be platform-specific.
Also raises default stack size on OSX to current Linux default (8MB) just to be safe.
- C++
Published by amchess over 6 years ago
brainlearn - BrainLearn 4.1
Corrected bug on learning table update
- C++
Published by amchess over 6 years ago
brainlearn - BrainLearn 4.0
Stockfish patch Author: Stefan Geschwentner Date: Mon Aug 26 08:49:55 2019 +0200 Timestamp: 1566802195
Tweak Late Move Reduction at root
- C++
Published by amchess over 6 years ago