Raphael

KO | EN

gitlite — search

#javascript #ai #llm #android #deep-learning #linux #docker #awesome #cli #kubernetes #vue #pytorch #security #open-source

(★ 24)

Superhuman chess engine and GUI coded in C++ and SFML.

Raphael Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# <div align="center"><img src=src/assets/logo/Raphael_Logo_HL_NoBG_Text.png alt="Raphael Logo" width=400 /></div>

Raphael is a superhuman UCI Chess Engine built using C++.
It also comes with a GUI built using [SFML](https://www.sfml-dev.org/).

You can [scroll to the bottom](#raphael-engine) to download the prebuilt binaries and see the list of currently-implemented features.

Raphael is largely inspired by [Sebastian Lague's Coding Adventure series on implementing a Chess Engine](https://youtu.be/U4ogK0MIzqk).

Raphael uses its own original data for its NNUE and does not contain any LLM-generated code.

## Elo

The following are the historic elo for Raphael.

<table>
  <tr align="center">
    <th>Version</th>
    <th>Release Date</th>
    <th><a href="https://www.computerchess.org.uk/ccrl/404/cgi/compare_engines.cgi?family=Raphael">CCRL Blitz</a></th>
    <th><a href="https://www.computerchess.org.uk/ccrl/4040/cgi/compare_engines.cgi?family=Raphael">CCRL 40/15</a></th>
    <th><a href="http://www.cegt.net/40_40%20Rating%20List/40_40%20SingleVersion/rangliste.html">CEGT 40/20</a></th>
    <th><a href="https://ipmanchess.yolasite.com/r9-7945hx.php">Ipman R9</a></th>
    <th><a href="https://www.sp-cc.de">SPCC Top 15</a></th>
    <th><a href="https://e4e6.com/324/">Chess324 Top 15</th>
  </tr>
  <tr align="center">
    <td>4.2.0        </td> <td>Jun 27, 2026 </td>
    <td>3740*        </td> <td>3593 (#14)   </td>
    <td>             </td> <td>             </td>
    <td>3710 (#11)   </td> <td>3733 (#11)   </td>
  </tr>
  <tr align="center">
    <td>4.1.0        </td> <td>May 17, 2026 </td>
    <td>3716 (#26)   </td> <td>3614 (#20)   </td>
    <td>3555 (#22)   </td> <td>3513 (#18**) </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>4.0.0        </td> <td>Apr 26, 2026 </td>
    <td>3697*        </td> <td>3606 (#23)   </td>
    <td>3532 (#31)   </td> <td>3487 (#20**) </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>3.3.0        </td> <td>Apr 06, 2026 </td>
    <td>3670         </td> <td>3558         </td>
    <td>3521         </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>3.2.0        </td> <td>Mar 19, 2026 </td>
    <td>3612*        </td> <td>3484         </td>
    <td>3433         </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>3.1.0        </td> <td>Mar 01, 2026 </td>
    <td>3510         </td> <td>3416         </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>3.0.0        </td> <td>Feb 12, 2026 </td>
    <td>3252         </td> <td>3206         </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>2.3.0        </td> <td>Jan 25, 2026 </td>
    <td>3146*        </td> <td>3061         </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>2.2.0        </td> <td>Jan 08, 2026 </td>
    <td>3035*        </td> <td>2953         </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>2.1.0        </td> <td>Dec 31, 2025 </td>
    <td>2739*        </td> <td>2689         </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>2.0.0        </td> <td>Dec 23, 2025 </td>
    <td>2646*        </td> <td>             </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>1.8.0        </td> <td>Dec 27, 2024 </td>
    <td>2223*        </td> <td>             </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>1.7.6        </td> <td>Dec 16, 2024 </td>
    <td>1970         </td> <td>             </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>1.7.0        </td> <td>Aug 26, 2023 </td>
    <td>1853         </td> <td>             </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>1.6.0        </td> <td>Aug 20, 2023 </td>
    <td>1797*        </td> <td>             </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
  <tr align="center">
    <td>1.5.0        </td> <td>Aug 16, 2023 </td>
    <td>1764*        </td> <td>             </td>
    <td>             </td> <td>             </td>
    <td>             </td> <td>             </td>
  </tr>
</table>
*estimated<br>
**ranking based on best single version per engine

## Getting Started

Prebuilt binaries of the UCI engine for Windows and Linux/WSL are available on the [Releases](https://github.com/Orbital-Web/Raphael/releases) page.

Please refer to the [following section](#compiling-from-source) to compile the GUI and/or the engine yourself on Windows and Linux.
The [features section](#features) outline the supported commands and features of the GUI and UCI engine.

### Compiling From Source

Follow these steps to build Raphael yourself. Note that it is highly recommended you build on [WSL](https://learn.microsoft.com/en-us/windows/wsl/install) if you are on Windows.

1. Clone the repository with

```shell
    git clone https://github.com/Orbital-Web/Raphael.git
    ```

2. Ensure you have Make and g++ installed. If you are on Linux/WSL, you can do so by running:

```shell
    sudo apt-get install build-essential g++  # Linux/WSL
    ```

Otherwise,  if you are on Windows, follow [this guide](https://code.visualstudio.com/docs/cpp/config-mingw) to install MSYS2 and run the following command inside the MSYS2 UCRT64 terminal:

```shell
    pacman -S --needed mingw-w64-ucrt-x86_64-toolchain make  # Windows
    ```

3. Compile as follows:

```shell
    make -j uci       # build UCI engine
    make -j packages  # download SFML, required to build main
    make -j main      # build GUI
    ```

## Features

### Graphics User Interface (GUI)

The GUI is a quick and easy way to start engine battles or play against Raphael interactively.
You can start a quick GUI match against yourself and Raphael as follows:

```shell
main.exe
```

You can see other command-line arguments by running `main.exe -h`.
There are supports for board annotations (arrows and square highlights) using the right mouse button.

### Raphael (Engine)

Raphael is a UCI-compliant chess engine.
To use it in other UCI-compliant softwares, compile `uci.cpp` or download the prebuilt binaries from the [Releases](https://github.com/Orbital-Web/Raphael/releases) page.
To see all supported commands, run `uci.exe help`

<summary>Click to show the list of implemented features:</summary>

- [x] Search                                (`v1.0+`)
  - [x] Iterative deepening                 (`v1.1+`)
  - [x] Aspiration window                   (`v1.3+`)
  - [x] Aspiration widening                 (`v3.0+`)
  - [ ] Endgame table base
- [x] Alpha-beta search                     (`v1.0+`)
  - [x] Pruning                             (`v1.0+`)
    - [x] Alpha-beta pruning                (`v1.0+`)
    - [x] Transposition table cutoff        (`v1.1+`)
    - [x] Mate distance pruning             (`v1.6+`)
    - [x] Reverse futility pruning          (`v2.2+`)
    - [x] Null move pruning                 (`v2.2+`)
    - [x] Razoring                          (`v2.3+`)
    - [x] Probcut                           (`v4.1+`)
    - [x] Late move pruning                 (`v2.3+`)
    - [x] Futility pruning                  (`v2.3+`)
    - [x] SEE pruning                       (`v2.3+`)
    - [ ] Quiet history pruning
    - [ ] Noisy history pruning
    - [x] Multi-cut                         (`v3.3+`)
    - [x] Improving heuristics              (`v3.2+`)
    - [ ] Bad noisy futility pruning
  - [x] Transposition table                 (`v1.1+`)
    - [x] Prefetching                       (`v2.2+`)
    - [x] Aging                             (`v3.0+`)
    - [x] Clusters                          (`v3.3+`)
    - [x] Storing static evaluations        (`v3.3+`)
  - [x] Principle variation search          (`v2.1+`)
  - [x] Extensions                          (`v1.4+`)
    - ~~Check extensions~~                  (`v1.4+`)
    - ~~Pawn push extensions~~              (`v1.4+`)
    - ~~One reply extensions~~              (`v1.7+`)
    - [x] Singular extensions               (`v3.0+`)
    - [x] Double extensions                 (`v3.0+`)
    - [x] Triple extensions                 (`v4.0+`)
    - [x] Negative extensions               (`v3.0+`)
    - [x] Cutnode negative extensions       (`v3.1+`)
    - [x] Low depth singular extensions     (`v4.0+`)
    - [x] Hindsight extensions              (`v4.0+`)
  - [x] Reductions                          (`v1.5+`)
    - [x] Late move reductions              (`v1.5+`)
    - [x] Internal iterative reduction      (`v3.0+`)
  - [x] Move ordering & History             (`v1.0+`)
    - [x] MVV-LVA                           (`v1.0+`)
    - [x] Promotions                        (`v1.0+`)
    - [x] Hash move                         (`v1.6+`)
    - ~~[x] Killer heuristics~~             (`v1.3+`)
    - [x] Butterfly history                 (`v1.5+`)
    - [x] Capture history                   (`v2.3+`)
    - [x] Pawn history                      (`v4.3+`)
    - [x] Continuation history              (`v3.2+`)
    - [x] Threats in history                (`v3.2+`)
    - [x] SEE                               (`v1.7+`)
    - [x] Pawn correction history           (`v3.3+`)
    - [x] Major piece correction history    (`v3.3+`)
    - [x] Nonpawn correction history        (`v3.3+`)
    - [x] Continuation correction history   (`v3.3+`)
- [x] Quiescence search                     (`v1.0+`)
  - ~~Delta pruning~~                       (`v2.1+`)
  - [x] Futility pruning                    (`v2.3+`)
  - [x] SEE pruning                         (`v2.3+`)
  - [x] Late move pruning                   (`v3.2+`)
  - [x] Transposition table cutoff          (`v3.0+`)
  - [x] Storing into transposition table    (`v3.2+`)
- [x] Evaluation                            (`v1.0+`)
  - [x] Hand-crafted evaluation             (`v1.0+`)
    - [x] Materials                         (`v1.0+`)
    - [x] Piece-square tables               (`v1.0+`)
    - [ ] Midgame King safety
    - [ ] Endgame King opposition
    - [x] Endgame King proximity            (`v1.0+`)
    - [x] Evaluation tapering               (`v1.0+`)
    - [x] Passed Pawn                       (`v1.3+`)
    - [x] Isolated Pawn                     (`v1.3+`)
    - [x] Mobility                          (`v1.5+`)
    - [x] Bishop pair                       (`v1.8+`)
    - [x] Bishop-colored corner             (`v1.8+`)
    - [x] Draw evaluation                   (`v1.8+`)
    - [x] Evaluation texel tuning           (`v1.8+`)
  - [x] NNUE                                (`v2.0+`)
    - [x] Lizard SCReLU                     (`v2.0+`)
    - [x] Lazy updates                      (`v2.1+`)
    - [x] Horizontal mirroring              (`v3.1+`)
    - [x] Output buckets                    (`v3.1+`)
    - [x] King buckets                      (`v3.2+`)
    - [x] Finny tables                      (`v3.2+`)
    - [x] Finetuning                        (`v3.3+`)
    - [x] Pairwise multiplication           (`v4.0+`)
    - [x] Multilayer network                (`v4.1+`)
    - [x] Sparse affine                     (`v4.1+`)
    - [ ] Dual activations
    - [x] Threat inputs                     (`v4.2+`)
    - [ ] Pawn pair inputs
    - [ ] Relabeling/distillation
  - [x] Material Scaling                    (`v3.3+`)
  - [x] Halfmove clock scaling              (`v3.3+`)
- [x] Time management                       (`v1.0+`)
  - [x] Hard/soft limit                     (`v3.0+`)
  - [x] Node-based scaling                  (`v3.2+`)
  - [x] Bestmove stability                  (`v3.2+`)
  - [x] Score stability                     (`v3.2+`)
  - [ ] Complexity
- [x] Pondering                             (`v1.2+`)
- [x] Performance                           (`v1.8+`)
  - [x] Compiler optimizations              (`v1.8+`)
  - [x] Incremental selection sort          (`v2.2+`)
  - [x] Linux huge pages                    (`v2.2+`)
  - [x] Staged movegen                      (`v3.0+`)
  - [x] Upcoming cycle detection            (`v4.2+`)
- [x] Multithreading                        (`v4.0+`)
  - [x] Lazy SMP                            (`v4.0+`)
  - [x] Thread voting                       (`v4.2+`)
  - [x] NUMA thread pinning                 (`v4.2+`)
  - [x] Shared corrhist                     (`v4.2+`)
  - [ ] NUMA net replication
- [x] Tuning                                (`v3.3+`)
  - [x] SPSA                                (`v3.3+`)
  - [x] Fractional depth                    (`v4.0+`)

</details>

For a more in-depth documentation on the NNUE and how it was trained, refer to the [NNUE History](https://github.com/Orbital-Web/Raphael/blob/main/src/eval/net/history.txt).
All iterations of Raphael's NNUE were trained on self-generated training data.
The net files can be found on the [Raphael-Net](https://github.com/Orbital-Web/Raphael-Net) repository.

## UCI Options

<table>
  <tr>
    <th>Name</th>
    <th>Default</th>
    <th>Range</th>
    <th>Description</th>
  </tr>
  <tr>
    <td>Hash</td> <td>64</td> <td>[1, 524288]</td>
    <td>Memory allocated for transposition table (in MiB)</td>
  </tr>
  <tr>
    <td>Threads</td> <td>1</td> <td>[1, 2048]</td>
    <td>Number of search threads</td>
  </tr>
  <tr>
    <td>UCI_Chess960</td> <td>false</td> <td>true/false</td>
    <td>Whether to play Chess960 (frc/dfrc) games</td>
  </tr>
  <tr>
    <td>MoveOverhead</td> <td>10</td> <td>[0, 5000]</td>
    <td>Amount of time assumed to be lost to overhead per move (in ms)</td>
  </tr>
  <tr>
    <td>Datagen</td> <td>false</td> <td>true/false</td>
    <td>Whether to enable datagen mode (modifies search behavior)</td>
  </tr>
  <tr>
    <td>Softnodes</td> <td>false</td> <td>true/false</td>
    <td>Whether to use a soft node limit when sent go nodes</td>
  </tr>
  <tr>
    <td>SoftNodeHardLimitMultiplier</td> <td>1678</td> <td>[1, 5000]</td>
    <td>Scale factor of hard node limit when using softnodes</td>
  </tr>
</table>

## Acknowledgements

Raphael uses or has used the following tools throughout its development:

- [fastchess](https://github.com/Disservin/fastchess) for running SPRTs locally
- [C++ chess library](https://github.com/Disservin/chess-library) for movegen up until v2.3, and a strong source of inspiration for the custom movegen logic from v3.0 onwards
- [GediminasMasaitis's Texel Tuner](https://github.com/GediminasMasaitis/texel-tuner) for tuning the HCE parameters for v1.8
- [OpenBench](https://github.com/AndyGrant/OpenBench) for data generations and distributed SPRTs from v3.1 onwards
- [bullet](https://github.com/jw1912/bullet) for NNUE training from v3.1 onwards
- [Pawnocchio](https://github.com/JonathanHallstrom/pawnocchio) for data processing from v3.1 onwards
- [incbin](https://github.com/graphitemaster/incbin) for embedding network files from v3.1 onwards

## Special Thanks To

Furthermore, the following individuals have inspired me or have helped me tremendously throughout the development process of Raphael (in no particular order):

- [Sebastian Lague](https://www.youtube.com/c/SebastianLague) for inspiring me to start the development of Raphael through the Coding Adventures series
- [Jonathan Hallström](https://github.com/JonathanHallstrom), author of [Pawnocchio](https://github.com/JonathanHallstrom/pawnocchio) (and a contributor to Raphael!!)
- [Ciekce](https://github.com/Ciekce), author of [Stormphrax](https://github.com/Ciekce/Stormphrax)
- [Sp00ph](https://github.com/Sp00ph), author of [Icarus](https://github.com/Sp00ph/icarus)
- [Dan](https://github.com/kelseyde), author of [Hobbes](https://github.com/kelseyde/hobbes-chess-engine)
- [Eren](https://github.com/ProgramciDusunur), author of [Potential](https://github.com/ProgramciDusunur/Potential)
- [Tecci](https://github.com/Teccii), author of [Cherry](https://github.com/Teccii/cherry)
- [Mitsune Mei](https://x.com/mitsune_mei) for creating the Raphael logo
- [Zahrizhal Ali](https://github.com/ZahrizhalAli) for providing me with kaggle compute for running tests, datagen, etc.
- and many others on the Stockfish and AlphaBeta Discord servers

Raphael

Raphael Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Raphael

Raphael Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README