https://github.com/kmock930/mahjong-strategy-simulation

Simulating agents in a Cantonese-style Mahjong game as a Multi-agent system.
https://github.com/kmock930/mahjong-strategy-simulation

agent-based-modeling agent-based-simulation game-theory jupyter-notebook markov-decision-processes modelling-research multi-agent-systems python3 shangting-distance unsupervised-learning

Last synced: about 2 months ago
JSON representation

Simulating agents in a Cantonese-style Mahjong game as a Multi-agent system.

• Host: GitHub
• URL: https://github.com/kmock930/mahjong-strategy-simulation
• Owner: kmock930
• Created: 2024-10-29T18:58:16.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-12-20T06:34:49.000Z (5 months ago)
• Last Synced: 2025-02-11T14:58:47.428Z (4 months ago)
• Topics: agent-based-modeling, agent-based-simulation, game-theory, jupyter-notebook, markov-decision-processes, modelling-research, multi-agent-systems, python3, shangting-distance, unsupervised-learning
• Language: Jupyter Notebook
• Homepage: https://github.com/kmock930/Mahjong-Strategy-Simulation/blob/main/COMP5900_Project_Report.pdf
• Size: 9.07 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Mahjong Simulation

## Paper

https://github.com/kmock930/Mahjong-Strategy-Simulation/blob/main/COMP5900_Project_Report.pdf

## Results

![Hu Rate vs Number of Games](HuRate.png)

![Hu Points vs Number of Games](HuPoints.png)

![DianPao Rate vs Number of Games](DianPaoRate.png)

![Average Score vs Number of Games](AvgScore.png)

## Problem Statement

Humans can make decision under high uncertainty. After learning the fundamentals of Game Theory, I am interested in understanding how humans make decisions in an environment full of uncertainties. Mahjong is a traditional board game from my home town which is an **imperfect information** game with high uncertainty where players do not know what tiles an opponent holds. As a result, I am going to proceed with my research via a simulation of a Cantonese-style Mahjong game.

## Methodology

1. Simulation 

2. Visualizing in a .txt file

3. Use of 3 types of Agents: 

    * Default strategy (baseline)

    * ShangTing distance

    * Markov Decision Process

## Acknowledgement

I would like to express gratitude to those open-source solutions which inspired my code solutions.

- MJai Simulator: https://github.com/gimite/mjai

- MJX Simulator: https://github.com/mjx-project/mjx

- RLCard: https://rlcard.org/