Mahjong

KO | EN

gitlite — search

#javascript #python #hacktoberfest #typescript #llm #go #machine-learning #docker #nodejs #java #cli #vue #security #ios

(★ 24)

Python 图形化麻将游戏 (带蒙特卡洛AI)

Mahjong 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Mahjong

A four-player Mahjong game written in Python/pygame, with a Monte-Carlo Tree Search (MCTS) AI running in a multi-process C++ engine.

## Features

- **Playable GUI**: you play the South seat against three computer opponents, with click-to-select tiles for discarding, and support for chow (吃), pung (碰), and win (和) responses.
- **MCTS AI**: `MCPlayer` picks actions by simulating random playouts with `mc_search`, implemented in C++ (`c_src/`) and exposed to Python through [pybind11](https://github.com/pybind/pybind11) with multi-process rollouts. A pure-Python MCTS reference lives in `montecarlo.py`.
- **Fast win detection**: winning hands are detected with a precomputed lookup table (`complete.pickle`, 9,362 hand keys) derived from the classic `AgariIndex` algorithm (`AgariIndex.java` is kept for reference), instead of enumerating melds at runtime.
- **Headless environment**: `pyenv.py` runs games entirely in the terminal (MCTS vs. random players), useful for benchmarking AI strength without the GUI.

## Code Layout

| Path | Description |
|---|---|
| `controller.py` | Game loop and rules for the pygame GUI version (entry point) |
| `pyenv.py` | Terminal-only game environment with `RandomPlayer` / `MCPlayer` (entry point) |
| `players.py` | Player logic: hand management, chow/pung/win checks via the lookup table; the GUI `AIPlayer` uses a simple discard policy |
| `graphics.py`, `sprites.py`, `resource.py` | pygame rendering, tile sprites, asset loading |
| `montecarlo.py` | Pure-Python MCTS reference implementation (superseded by the C++ engine) |
| `c_src/` | C++ MCTS engine (`mctree.cpp`, `player.cpp`) and pybind11 bindings (`export.cpp`) |
| `complete.pickle` | Precomputed winning-hand key table loaded at startup |
| `AgariIndex.java` | Original Java source of the winning-hand table generator (reference only) |
| `tiles/` | Tile face images |

## Setup

Requires Python 3 with `pygame` and `numpy`, plus CMake, pybind11, and a C++14 compiler for the AI engine.

1. Install Python dependencies:

```bash
pip install pygame numpy pybind11
```

2. Build the C++ MCTS module into `build/`:

```bash
mkdir build && cd build
cmake ..
cmake --build . --config Release
```

The Python code imports the compiled module as `build.Release.clib` (MSVC layout); on Linux/macOS, place the built `clib` extension under `build/` and adjust the import accordingly.

## Play

```bash
python controller.py   # GUI game: you (South) vs. 3 computer players
python pyenv.py        # terminal simulation: MCTS vs. random players
```

In the GUI, click tiles to raise/select them and discard; prompts appear when you can chow, pung, or declare a win. MCTS strength can be tuned via `NUM_PROCS` and `NUM_SEARCH` at the top of `pyenv.py`.

## Notes

- The game implements a simplified Chinese ruleset over the standard 34 tile types (136 tiles), with chow/pung/win interactions.
- UI text is in Chinese.

Mahjong

Mahjong 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Mahjong

Mahjong 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README