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