search

tryingsomestuff / Minic

A simple chess engine to learn and play with
GNU General Public License v3.0
96 stars 12 forks source link
ai chess engine game-development

readme


Minic

Minic is a chess engine I'm developing to learn about chess programming and modern C++ (please see this lovely wiki for more details on chess programming).

Minic has no graphic interface (GUI) but is compatible with both CECP (xboard) and UCI protocol so you can use it in your favorite software (for instance Cutechess, Arena, Banksia, Winboard/Xboard, c-chess-cli, ...).
Minic is currently one of the 15 best engines in major rating lists and the strongest french one.

Here are some shortcuts to navigate in this document :

Support Minic development

Generating data, learning process, tuning, optimization and testing of a chess engine is quite hardware intensive ! I have some good hardware at home but this is far from enough. This is why I opened a Patreon account for Minic ; if you want to support Minic development, it is the place to be ;-)

History & the NNUE Minic story

Code style

For a year and a half Minic was (mainly) a one-file-code with very dense lines. This is of course very wrong in terms of software design... So why is it so? First reason is that Minic was first developped as a week-end project (in mid-october 2018), the quick-and-dirty way, and since then I was having fun going on the way of minimal lines of code direction ; being a "small" code size engine was part of the stakes in developing it the first few months.

Until version 2 of Minic, some optional features such as evaluation and searchs tuning, perft, tests, uci support, book generation ... were available in the Add-Ons directory ; they are now fused with the source code for a while.

Nowadays, in fact since the release of version "2", the engine is written in a more classic C++ style, although some very dense lines may still be present and recall Minic past compacity...

More details about Minic history in the next paragraphs...

Week-end project (October 2018)

Initially, the code size of Minic was supposed not to go above 2000sloc. It started as a week-end project in october 2018 (http://talkchess.com/forum3/viewtopic.php?f=2&t=68701). But, as soon as more features (especially SMP, material tables and bitboard) came up, I tried to keep it under 4000sloc and then 5000sloc, ... This is why this engine was named Minic, this stands for "Minimal Chess" (and is not related to the GM Dragoljub Minić) but it has not much to do with minimalism anymore nowadays... For the record, here is a link to the very first published version of Minic (https://github.com/tryingsomestuff/Minic/blob/dbb2fc7f026d5cacbd35bc379d8a1cdc1cad5674/minic.cc).

Version "1" (October 2019)

Version "1" was published as a one year anniversary release in october 2019. At this point Minic has already gone from a 1800 Elo 2-days-of-work engine, to a 2800 Elo engine being invited at TCEC qualification league !

Version "2" (April 2020)

Version "2" was released for April 1st 2020 (during covid-19 confinement in France). For this version, the one file Minic was splitted into many header and source files, and commented a lot more, without negative impact on speed and strength.

NNUE from release 2.47 to release 2.53 (from Stockfish implementation)

Minic2, since release 2.47 of August 8th 2020 (http://talkchess.com/forum3/viewtopic.php?f=2&t=73521&hilit=minic2&start=50#p855313), has the possibility to be build using a shameless copy of the NNUE framework of Stockfish. Integration of NNUE was done easily and I hoped this can be done for any engines, especially if NNUE is released as a standalone library (see https://github.com/dshawul/nncpu-probe for instance). First tests shown that "MinicNNUE" is around 200Elo stronger than Minic, around the level of Xiphos or Ethereal at this date at short TC and maybe something like 50Elo higher at longer TC (so around Komodo11).

When using a NNUE network with this Stockfish implementation, it is important that Minic is called "MinicNNUE". Indeed, MinicNNUE (with copy/pasted SF NNUE implementation), won't be the official Minic, as this is not my own work at all.

Later on, starting from version 2.50, the NNUE learner from NodChip repository has also been ported to Minic so that networks can be built using Minic data and search can be done. The genFen part of Nodchip was not ported and instead replaced by an internal process to produce training data. This included both extracting position from fixed depth game and from random positions.

Nets I built are still available at https://github.com/tryingsomestuff/NNUE-Nets.

Version "3" (November 2020)

Version "3" of Minic is released in november 2020 (during second covid-19 confinement) as a 2 years anniversary release and is not using, nor compatible with, SF NNUE implementation anymore. More about this just below...

NNUE from release 3.00 (initially from Seer implementation)

Starting from release 3.00, Minic is not using Stockfish NNUE implementation anymore and is no more compatible with SF nets. It was too much foreign code inside Minic to be fair, to be maintained, to be fun. Seer chess engine author is offering a very well written implementation of NNUE that I borrowed and adapt to Minic (https://github.com/connormcmonigle/seer-nnue). The code was more or less 400 lines. I choose to keep some Stockfish code just for binary sfens format conversion as everyone (or at least many many) is using this data format for now. Training code is an external tool written in Python without any dependency to engine, also first adapted from Seer repository and then taking ideas from Gary Linscott pytorch trainer (https://github.com/glinscott/nnue-pytorch). A new story was written in Minic 3 and the code has diverged quite a lot from the initial one.

Nets I built are available at https://github.com/tryingsomestuff/NNUE-Nets. Beware there is no retrocompatibility of the net from version to version.

Some original stuff that were maybe introduced in Minic

Minic NNUE "originality" status

In brief, Minic NNUE world is vastly inspired from what others are doing and is using pure Minic data (Minic generated or otherwise rescored).

Testing and strength

Minic is currently in the 20 best engines with a Elo rating around 3400 at CCRL scale.

Various NNUE nets strenght

This table shows the evolution of nets strength of various Minic nets on an AVX2 hardware at short TC (10s+0.1). Results will be a lot different on older hardware where NNUE evaluation is much slower. I thus encourage users to only use Minic with NNUE nets on recent hardware. As we see, nets strenght is increasing version after version, this is due to better data, new net topologies but also of course to the fact that I rescore (or regenerate) data with previous version of Minic before training a new net. Moreover, the decision to revert SF implementation and start my own work based on Seer initial implementation led to a 2 years net training journey to fill the performance gap ...


Rank Name                          Elo     +/-   Games   Score    Draw 
   1 minic_2.53_nn-97f742aaefcd     165      24     458   72.1%   42.8% 
   2 minic_3.17_NiNe3               112      24     457   65.5%   44.0% 
   3 minic_2.53_napping_nexus        97      24     458   63.6%   42.1% 
   4 minic_3.18                      97      25     457   63.6%   40.5% 
   5 minic_3.17                      -5      24     457   49.3%   41.4% 
   6 minic_2.53_nascent_nutrient     -8      25     457   48.9%   40.5% 
   7 minic_3.14                     -55      23     458   42.1%   46.7% 
   8 minic_3.08                     -57      25     457   41.9%   40.9% 
   9 minic_3.02_nettling_nemesis    -72      25     458   39.7%   39.3% 
  10 minic_3.06_nocturnal_nadir    -114      25     457   34.1%   38.9% 
  11 minic_3.04_noisy_notch        -155      27     458   29.0%   33.2%

More details about those nets I built are available at https://github.com/tryingsomestuff/NNUE-Nets.

A word on NNUE and vectorization

In this table, Minic 3.19 is used to compare NNUE performances on various CPU architecture (effect of vectorisation).

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 minic_3.19_slylake             68      23     422   59.7%   50.7% 
   2 minic_3.19_sandybridge         14      23     423   52.0%   53.4% 
   3 minic_3.19_nehalem            -27      23     421   46.1%   52.3% 
   4 minic_3.19_core2              -55      24     422   42.2%   49.3%

What does this say ? Well ... for NNUE, using AVX2 is very important. This can explain some strange results during some testing process and in rating list where I sometimes see my nets underperforming a lot. So please, use AVX2 hardware (and the corresponding Minic binary, i.e. the "skylake" one for Intel or at least the "znver1" for AMD) for NNUE testing if possible.

Threading performances

I'd love to own a big enough hardware to test with more than 8 threads ... Here are 3s+0.1 TC results to illustrate threading capabilities. Minic is thus scaling very well.

Rank Name                          Elo     +/-   Games   Score    Draw 
   1 minic_3.19_8                   123      34     156   67.0%   58.3% 
   2 minic_3.19_6                    94      36     156   63.1%   54.5% 
   3 minic_3.19_4                    20      33     156   52.9%   63.5% 
   4 minic_3.19_2                   -45      35     156   43.6%   59.0% 
   5 minic_3.19_1                  -206      41     156   23.4%   42.9%

Moreover, speed test on CCC hardware, 2x AMD EPYC 7H12 (128 physical cores) : 

250 threads : 101,008,415 NPS
125 threads :  73,460,779 NPS

and speed test on a 2x Intel(R) Xeon(R) Platinum 8269CY CPU @ 2.50GHz (52 physical cores) :

104 threads :  46,291,408 NPS
52 threads  :  25,175,817 NPS

suggest Minic is reacting quite well to hyperthreading.

Home testing

Here are some fast TC results of a gauntlet tournament (STC 10s+0.1) for Minic 3.36.

Rank Name                          Elo     +/-   Games   Score    Draw 
   0 minic_3.36                     22       5    9852   53.1%   50.6% 
   1 seer                          103      11    1642   64.4%   57.9% 
   2 rofChade3                      52      11    1642   57.4%   55.8% 
   3 Uralochka3.39e-avx2           -47      11    1642   43.3%   57.3% 
   4 komodo-13.02                  -78      13    1642   39.0%   41.7% 
   5 Wasp600-linux-avx             -80      12    1642   38.7%   49.2% 
   6 stockfish.8                   -81      13    1642   38.6%   41.9%

Random mover

Minic random-mover (level = 0) stats are the following :

   7.73%  0-1 {Black mates}
   7.50%  1-0 {White mates}
   2.45%  1/2-1/2 {Draw by 3-fold repetition}
  21.99%  1/2-1/2 {Draw by fifty moves rule}
  54.16%  1/2-1/2 {Draw by insufficient mating material}
   6.13%  1/2-1/2 {Draw by stalemate}

Rating Lists & competitions

Progress

Here is 4 years of CCRL progress (single thread) image

CCRL

CEGT

FGRL

SP-CC

GRL

BRUCE

Ipman Chess

Test Suite (quite old results here)

TCEC stats

TCEC hardware: Minic is at 3440 (https://tcec-chess.com/bayeselo.txt)

TCEC

Here are Minic results at TCEC main event (https://tcec-chess.com/)

TCEC15: 8th/10 in Division 4a (https://www.chessprogramming.org/TCEC_Season_15)
TCEC16: 13th/18 in Qualification League (https://www.chessprogramming.org/TCEC_Season_16)
TCEC17: 7th/16 in Q League, 13th/16 in League 2 (https://www.chessprogramming.org/TCEC_Season_17)
TCEC18: 4th/10 in League 3 (https://www.chessprogramming.org/TCEC_Season_18)
TCEC19: 3rd/10 in League 3 (https://www.chessprogramming.org/TCEC_Season_19)
TCEC20: 2nd/10 in League 3, 9th/10 in League 2 (https://www.chessprogramming.org/TCEC_Season_20)
TCEC21: 1st/12 in League 3, 6th/10 in League 2 (https://www.chessprogramming.org/TCEC_Season_21)
TCEC22: 2nd/8 in League 2, 8th/8 in League 1 (well tried ;) ) (https://www.chessprogramming.org/TCEC_Season_22)
TCEC23: 3rd/12 in League 1 ! (https://tcec-chess.com/#div=l1&game=1&season=23)
TCEC24: 9rd/12 in League 1 ! (https://tcec-chess.com/#div=l1&game=1&season=24)
TCEC25: 8rd/12 in League 1 ! (https://tcec-chess.com/#div=l1&game=1&season=25)

Release process

WARNING : the former Dist directory as been REMOVED from the repository because it was starting to be too big. Unofficial releases are not available anymore here. All (including unofficial) releases are available in a new repo, here : https://github.com/tryingsomestuff/Minic-Dist, also available as a git submodule.

Some stable/official ones will still be made available as github release (https://github.com/tryingsomestuff/Minic/releases). I "officially release" (create a github version) as soon as I have some validated elo (at least +10) or an important bug fix.

In a github release, a tester shall only use the given (attached) binaries. The full "source" package always contains everything (source code, test suites, opening suite, books, ...) but using git "submodule" so that the main repository remains small.

Binaries are named following this convention :

Linux 64:
-- Intel --
* minic_X.YY_linux_x64_skylake         : fully optimized Linux64 (popcnt+avx2+bmi2)  
* minic_X.YY_linux_x64_sandybridge     : optimized Linux64 (popcnt+avx)  
* minic_X.YY_linux_x64_nehalem         : optimized Linux64 (popcnt+sse4.2)  
* minic_X.YY_linux_x64_core2           : basic Linux64 (nopopcnt+sse3)  

-- AMD --
* minic_X.YY_linux_x64_znver3          : fully optimized Linux64 (popcnt+avx2+bmi2)  
* minic_X.YY_linux_x64_znver1          : almost optimized Linux64 (popcnt+avx2)  
* minic_X.YY_linux_x64_bdver1          : optimized Linux64 (nopopcnt+avx)  
* minic_X.YY_linux_x64_barcelona       : optimized Linux64 (nopopcnt+sse4A)  
* minic_X.YY_linux_x64_athlon64-sse3   : basic Linux64 (nopopcnt+sse3)  

Windows 64:
Some as for Linux with naming convention like this minic_X.YY_mingw_x64_skylake.exe

Windows 32:
* minic_X.YY_mingw_x32_pentium2.exe    : very basic Windows32  

Others:
* minic_X.YY_android                   : android armv7  
* minic_X.YY_linux_x32_armv7           : RPi armv7
* minic_X.YY_linux_x64_armv8           : RPi armv8

Please note that for Linux binaries to work you will need a recent libc installed on your system.
Please note that Win32 binaries are very slow so please use Win64 one if possible.
Please note that Minic has always been a little weaker under Windows OS (probably due to cross-compilation lacking PGO).

How to compile

A minimal working example on Linux for dev version would be

git clone https://github.com/tryingsomestuff/Minic.git
cd Minic
git submodule update --init Fathom
make
# done exe in Dist/Minic3

Syzygy EGT

To compile with SYZYGY support you'll need to clone https://github.com/jdart1/Fathom as Fathom directory and activate WITH_SYZYGY definition at compile time (this is default behaviour). This can be done using the given git submodule or by hand. To use EGT just specify syzygyPath in the command line or using the GUI option.

How to run

add the command line option "-xboard" to go to xboard/winboard mode or -uci for UCI. If not option is given Minic will default to use UCI protocol. Please note that if you want to force specific option from command line instead of using protocol option, you have to first specific protocol as the first command line argument. For instance minic -uci -minOutputLevel 0 will give a very verbose Minic using uci.

Other available options (depending on compilation option, see config.hpp) are mainly for development or debug purpose. They do not start the protocol loop. Here is an incompleted list :

Options

GUI/protocol (Xboard or UCI)

Starting from release 1.00 Minic support setting options using protocol (both XBoard and UCI). Option priority are as follow : command line option can be override by protocol option. This way, Minic supports strength limitation, FRC, pondering, can use contempt, ...

If compiled with the WITH_SEARCH_TUNING definition, Minic can expose all search algorithm parameters so that they can be tweaked. Also, when compiled with WITH_PIECE_TUNING, Minic can expose all middle- and end-game pieces values.

Command line

Minic comes with some command line options :

Debug option

"Classic options"

NNUE net

Remark for Windows users : it may be quite difficult to get the path format for the NNUE file ok under Windows. Here is a working example (thanks to Stefan Pohl) for cutechess-cli as a guide:

cutechess-cli.exe -engine name="Minic3.06NoNa" cmd="C:/Engines/Minic/minic_3.06_mingw_x64_nehalem.exe" dir="C:/Engines/Minic" option.NNUEFile=C:/Engines/Minic/nocturnal_nadir.bin option.Hash=256 option.Threads=1 proto=uci

Starting from Minic 3.07, no need to worry about this, the official corresponding net is embeded inside the binary using the INCBIN technology.

Adjust strength

Minic strength can be ajdusted using the level option (from command line or using protocol option support, using value from 0 to 100). Level 0 is a random mover, 1 to 30 very weak, ..., level 100 is full strength. For now it uses multipv, maximum depth adjustment and randomness to make Minic play weaker moves (also a fixed node option is available to avoid randomness if needed).

Current level Elo are more or less so that even a beginner can beat low levels Minic. From level 50 or 60, you will start to struggle more! You can also use the UCI_Elo parameter if UCI_LimitStrenght is activated but the Elo fit is not good especially at low level. Level functionnaly will be enhanced in a near future.

Please also note that is nodesBasedLevel is activated, then no randomness is used to decrease Elo level of Minic, only number of nodes searched is changing.

Training data generation

There are multiple ways of generating sfen data with Minic.

Other options

Style

Moreover, Minic implements some "style" parameters when using HCE (hand crafted evalaution in opposition to NNUE) that allow the user to boost or minimize effects of :

Default values are 50 and a range from 0 to 100 can be used. A value of 100 will double the effect, while a value of 0 will disable the feature (it is probably not a good idea to put material awareness to 0 for instance ...).

Videos

Minic on youtube, most often loosing to stronger engines ;-) :

GM Matthew Sadler (Silicon Road youtube channel) game analysis

Thanks

Of course many/most idea in Minic are taken from the beautifull chess developer community. Here's a very incomplete list of open-source engines that were inspiring for me:

Arasan by Jon Dart
Berserk by Jay Honnold
CPW by Pawel Koziol and Edmund Moshammer
Deepov by Romain Goussault
Defenchess by Can Cetin and Dogac Eldenk
Demolito by Lucas Braesch
Dorpsgek by Dan Ravensloft
Ethereal by Andrew Grant
Galjoen by Werner Taelemans
Koivisto by Kim Kåhre and Finn Eggers
Madchess by Erik Madsen
Rodent by Pawel Koziol
RubiChess by Andreas Matthies
Seer by Connor McMonigle
Stockfish by the stockfish contributors (https://github.com/official-stockfish/Stockfish/blob/master/AUTHORS)
Texel by Peter Österlund
Topple by Vincent Tang
TSCP by Tom Kerrigan
Vajolet by Marco Belli
Vice by BlueFeverSoft
Weiss by Terje Kirstihagen
Winter by Jonathan Rosenthal
Xiphos by Milos Tatarevic
Zurichess by Alexandru Moșoi

Many thanks also to all testers for all those long time control tests, they really are valuable inputs in the chess engine development process.

Also thanks to TCEC and CCC for letting Minic participate to many event, it is fun to see Minic on such a great hardware.

Thanks to Karlson Pfannschmidt for the bayesian optimization chess-tuning-tools

And of course thanks to all the members of the talkchess forum and CPW, and to H.G. Muller and Joost Buijs for hosting the well-known friendly monthly tourney.

Info