3 [![Build Status](https://github.com/official-stockfish/Stockfish/actions/workflows/stockfish.yml/badge.svg)](https://github.com/official-stockfish/Stockfish/actions)
4 [![Build Status](https://ci.appveyor.com/api/projects/status/github/official-stockfish/Stockfish?branch=master&svg=true)](https://ci.appveyor.com/project/mcostalba/stockfish/branch/master)
6 [Stockfish](https://stockfishchess.org) is a free, powerful UCI chess engine
7 derived from Glaurung 2.1. Stockfish is not a complete chess program and requires a
8 UCI-compatible graphical user interface (GUI) (e.g. XBoard with PolyGlot, Scid,
9 Cute Chess, eboard, Arena, Sigma Chess, Shredder, Chess Partner or Fritz) in order
10 to be used comfortably. Read the documentation for your GUI of choice for information
11 about how to use Stockfish with it.
13 The Stockfish engine features two evaluation functions for chess.
14 The efficiently updatable neural network (NNUE) based evaluation is the default and by far the strongest.
15 The classical evaluation based on handcrafted terms remains available.
16 The strongest network is integrated in the binary and downloaded automatically during the build process.
17 The NNUE evaluation benefits from the vector intrinsics available on most CPUs (sse2, avx2, neon, or similar).
21 This distribution of Stockfish consists of the following files:
23 * [Readme.md](https://github.com/official-stockfish/Stockfish/blob/master/README.md), the file you are currently reading.
25 * [Copying.txt](https://github.com/official-stockfish/Stockfish/blob/master/Copying.txt), a text file containing the GNU General Public License version 3.
27 * [AUTHORS](https://github.com/official-stockfish/Stockfish/blob/master/AUTHORS), a text file with the list of authors for the project
29 * [src](https://github.com/official-stockfish/Stockfish/tree/master/src), a subdirectory containing the full source code, including a Makefile
30 that can be used to compile Stockfish on Unix-like systems.
32 * a file with the .nnue extension, storing the neural network for the NNUE
33 evaluation. Binary distributions will have this file embedded.
35 ## The UCI protocol and available options
37 The Universal Chess Interface (UCI) is a standard protocol used to communicate with
38 a chess engine, and is the recommended way to do so for typical graphical user interfaces
39 (GUI) or chess tools. Stockfish implements the majority of its options as described
40 in [the UCI protocol](https://www.shredderchess.com/download/div/uci.zip).
42 Developers can see the default values for UCI options available in Stockfish by typing
43 `./stockfish uci` in a terminal, but the majority of users will typically see them and
44 change them via a chess GUI. This is a list of available UCI options in Stockfish:
47 The number of CPU threads used for searching a position. For best performance, set
48 this equal to the number of CPU cores available.
51 The size of the hash table in MB. It is recommended to set Hash after setting Threads.
57 Let Stockfish ponder its next move while the opponent is thinking.
60 Output the N best lines (principal variations, PVs) when searching.
61 Leave at 1 for best performance.
64 Toggle between the NNUE and classical evaluation functions. If set to "true",
65 the network parameters must be available to load from file (see also EvalFile),
66 if they are not embedded in the binary.
69 The name of the file of the NNUE evaluation parameters. Depending on the GUI the
70 filename might have to include the full path to the folder/directory that contains the file.
71 Other locations, such as the directory that contains the binary and the working directory,
74 * #### UCI_AnalyseMode
75 An option handled by your GUI.
78 An option handled by your GUI. If true, Stockfish will play Chess960.
81 If enabled, show approximate WDL statistics as part of the engine output.
82 These WDL numbers model expected game outcomes for a given evaluation and
83 game ply for engine self-play at fishtest LTC conditions (60+0.6s per game).
85 * #### UCI_LimitStrength
86 Enable weaker play aiming for an Elo rating as set by UCI_Elo. This option overrides Skill Level.
89 If enabled by UCI_LimitStrength, aim for an engine strength of the given Elo.
90 This Elo rating has been calibrated at a time control of 60s+0.6s and anchored to CCRL 40/4.
93 Lower the Skill Level in order to make Stockfish play weaker (see also UCI_LimitStrength).
94 Internally, MultiPV is enabled, and with a certain probability depending on the Skill Level a
95 weaker move will be played.
98 Path to the folders/directories storing the Syzygy tablebase files. Multiple
99 directories are to be separated by ";" on Windows and by ":" on Unix-based
100 operating systems. Do not use spaces around the ";" or ":".
102 Example: `C:\tablebases\wdl345;C:\tablebases\wdl6;D:\tablebases\dtz345;D:\tablebases\dtz6`
104 It is recommended to store .rtbw files on an SSD. There is no loss in storing
105 the .rtbz files on a regular HDD. It is recommended to verify all md5 checksums
106 of the downloaded tablebase files (`md5sum -c checksum.md5`) as corruption will
107 lead to engine crashes.
109 * #### SyzygyProbeDepth
110 Minimum remaining search depth for which a position is probed. Set this option
111 to a higher value to probe less aggressively if you experience too much slowdown
112 (in terms of nps) due to tablebase probing.
114 * #### Syzygy50MoveRule
115 Disable to let fifty-move rule draws detected by Syzygy tablebase probes count
116 as wins or losses. This is useful for ICCF correspondence games.
118 * #### SyzygyProbeLimit
119 Limit Syzygy tablebase probing to positions with at most this many pieces left
120 (including kings and pawns).
123 Assume a time delay of x ms due to network and GUI overheads. This is useful to
124 avoid losses on time in those cases.
127 Lower values will make Stockfish take less time in games, higher values will
128 make it think longer.
131 Tells the engine to use nodes searched instead of wall time to account for
132 elapsed time. Useful for engine testing.
134 * #### Debug Log File
135 Write all communication to and from the engine into a text file.
137 For developers the following non-standard commands might be of interest, mainly useful for debugging:
139 * #### bench *ttSize threads limit fenFile limitType evalType*
140 Performs a standard benchmark using various options. The signature of a version (standard node
141 count) is obtained using all defaults. `bench` is currently `bench 16 1 13 default depth mixed`.
144 Give information about the compiler and environment used for building a binary.
147 Display the current position, with ascii art and fen.
150 Return the evaluation of the current position.
152 * #### export_net [filename]
153 Exports the currently loaded network to a file.
154 If the currently loaded network is the embedded network and the filename
155 is not specified then the network is saved to the file matching the name
156 of the embedded network, as defined in evaluate.h.
157 If the currently loaded network is not the embedded network (some net set
158 through the UCI setoption) then the filename parameter is required and the
159 network is saved into that file.
162 Flips the side to move.
165 ## A note on classical evaluation versus NNUE evaluation
167 Both approaches assign a value to a position that is used in alpha-beta (PVS) search
168 to find the best move. The classical evaluation computes this value as a function
169 of various chess concepts, handcrafted by experts, tested and tuned using fishtest.
170 The NNUE evaluation computes this value with a neural network based on basic
171 inputs (e.g. piece positions only). The network is optimized and trained
172 on the evaluations of millions of positions at moderate search depth.
174 The NNUE evaluation was first introduced in shogi, and ported to Stockfish afterward.
175 It can be evaluated efficiently on CPUs, and exploits the fact that only parts
176 of the neural network need to be updated after a typical chess move.
177 [The nodchip repository](https://github.com/nodchip/Stockfish) provided the first version of
178 the needed tools to train and develop the NNUE networks. Today, more advanced training tools are available
179 in [the nnue-pytorch repository](https://github.com/glinscott/nnue-pytorch/), while data generation tools
180 are available in [a dedicated branch](https://github.com/official-stockfish/Stockfish/tree/tools).
182 On CPUs supporting modern vector instructions
183 (avx2 and similar), the NNUE evaluation results in much stronger playing strength, even
184 if the nodes per second computed by the engine is somewhat lower (roughly 80% of nps
189 1) the NNUE evaluation depends on the Stockfish binary and the network parameter
190 file (see the EvalFile UCI option). Not every parameter file is compatible with a given
191 Stockfish binary, but the default value of the EvalFile UCI option is the name of a network
192 that is guaranteed to be compatible with that binary.
194 2) to use the NNUE evaluation, the additional data file with neural network parameters
195 needs to be available. Normally, this file is already embedded in the binary or it
196 can be downloaded. The filename for the default (recommended) net can be found as the default
197 value of the `EvalFile` UCI option, with the format `nn-[SHA256 first 12 digits].nnue`
198 (for instance, `nn-c157e0a5755b.nnue`). This file can be downloaded from
200 https://tests.stockfishchess.org/api/nn/[filename]
202 replacing `[filename]` as needed.
204 ## What to expect from the Syzygy tablebases?
206 If the engine is searching a position that is not in the tablebases (e.g.
207 a position with 8 pieces), it will access the tablebases during the search.
208 If the engine reports a very large score (typically 153.xx), this means
209 it has found a winning line into a tablebase position.
211 If the engine is given a position to search that is in the tablebases, it
212 will use the tablebases at the beginning of the search to preselect all
213 good moves, i.e. all moves that preserve the win or preserve the draw while
214 taking into account the 50-move rule.
215 It will then perform a search only on those moves. **The engine will not move
216 immediately**, unless there is only a single good move. **The engine likely
217 will not report a mate score, even if the position is known to be won.**
219 It is therefore clear that this behaviour is not identical to what one might
220 be used to with Nalimov tablebases. There are technical reasons for this
221 difference, the main technical reason being that Nalimov tablebases use the
222 DTM metric (distance-to-mate), while the Syzygy tablebases use a variation of the
223 DTZ metric (distance-to-zero, zero meaning any move that resets the 50-move
224 counter). This special metric is one of the reasons that the Syzygy tablebases are
225 more compact than Nalimov tablebases, while still storing all information
226 needed for optimal play and in addition being able to take into account
231 Stockfish supports large pages on Linux and Windows. Large pages make
232 the hash access more efficient, improving the engine speed, especially
233 on large hash sizes. Typical increases are 5..10% in terms of nodes per
234 second, but speed increases up to 30% have been measured. The support is
235 automatic. Stockfish attempts to use large pages when available and
236 will fall back to regular memory allocation when this is not the case.
240 Large page support on Linux is obtained by the Linux kernel
241 transparent huge pages functionality. Typically, transparent huge pages
242 are already enabled, and no configuration is needed.
244 ### Support on Windows
246 The use of large pages requires "Lock Pages in Memory" privilege. See
247 [Enable the Lock Pages in Memory Option (Windows)](https://docs.microsoft.com/en-us/sql/database-engine/configure-windows/enable-the-lock-pages-in-memory-option-windows)
248 on how to enable this privilege, then run [RAMMap](https://docs.microsoft.com/en-us/sysinternals/downloads/rammap)
249 to double-check that large pages are used. We suggest that you reboot
250 your computer after you have enabled large pages, because long Windows
251 sessions suffer from memory fragmentation, which may prevent Stockfish
252 from getting large pages: a fresh session is better in this regard.
254 ## Compiling Stockfish yourself from the sources
256 Stockfish has support for 32 or 64-bit CPUs, certain hardware
257 instructions, big-endian machines such as Power PC, and other platforms.
259 On Unix-like systems, it should be easy to compile Stockfish
260 directly from the source code with the included Makefile in the folder
261 `src`. In general it is recommended to run `make help` to see a list of make
262 targets with corresponding descriptions.
268 make build ARCH=x86-64-modern
271 When not using the Makefile to compile (for instance, with Microsoft MSVC) you
272 need to manually set/unset some switches in the compiler command line; see
273 file *types.h* for a quick reference.
275 When reporting an issue or a bug, please tell us which Stockfish version
276 and which compiler you used to create your executable. This information
277 can be found by typing the following command in a console:
283 ## Understanding the code base and participating in the project
285 Stockfish's improvement over the last decade has been a great community
286 effort. There are a few ways to help contribute to its growth.
288 ### Donating hardware
290 Improving Stockfish requires a massive amount of testing. You can donate
291 your hardware resources by installing the [Fishtest Worker](https://github.com/glinscott/fishtest/wiki/Running-the-worker:-overview)
292 and view the current tests on [Fishtest](https://tests.stockfishchess.org/tests).
294 ### Improving the code
296 If you want to help improve the code, there are several valuable resources:
298 * [In this wiki,](https://www.chessprogramming.org) many techniques used in
299 Stockfish are explained with a lot of background information.
301 * [The section on Stockfish](https://www.chessprogramming.org/Stockfish)
302 describes many features and techniques used by Stockfish. However, it is
303 generic rather than being focused on Stockfish's precise implementation.
304 Nevertheless, a helpful resource.
306 * The latest source can always be found on [GitHub](https://github.com/official-stockfish/Stockfish).
307 Discussions about Stockfish take place these days mainly in the [FishCooking](https://groups.google.com/forum/#!forum/fishcooking)
308 group and on the [Stockfish Discord channel](https://discord.gg/nv8gDtt).
309 The engine testing is done on [Fishtest](https://tests.stockfishchess.org/tests).
310 If you want to help improve Stockfish, please read this [guideline](https://github.com/glinscott/fishtest/wiki/Creating-my-first-test)
311 first, where the basics of Stockfish development are explained.
316 Stockfish is free, and distributed under the **GNU General Public License version 3**
317 (GPL v3). Essentially, this means you are free to do almost exactly
318 what you want with the program, including distributing it among your
319 friends, making it available for download from your website, selling
320 it (either by itself or as part of some bigger software package), or
321 using it as the starting point for a software project of your own.
323 The only real limitation is that whenever you distribute Stockfish in
324 some way, you MUST always include the license, the full source code, or a pointer
325 to where the source code can be found, to generate the exact binary
326 you are distributing. If you make any changes to the source code,
327 these changes must also be made available under the GPL v3.
329 For full details, read the copy of the GPL v3 found in the file named
330 [*Copying.txt*](https://github.com/official-stockfish/Stockfish/blob/master/Copying.txt).