Git Product home page Git Product logo

tenhou-python-bot's Introduction

Build Status

For now only Python 3.5+ is supported.

What do we have here?

Example of bot game

Mahjong hands calculation

We have a code which can calculate hand cost (han, fu, yaku and scores) based on the hand's tiles.

It supports features like:

  • Disable\enable aka dora in the hand
  • Disable\enable open tanyao yaku
  • By now it supports double yakumans (Dai Suushii, Daburu Kokushi musou, Suu ankou tanki, Daburu Chuuren Poutou). Later I plan to have a disabling option in settings for it.

The code was validated on tenhou.net phoenix replays in total on 8'527'296 hands, and results were the same in 100% cases.

So, we can say that our own hand calculator works the same way that tenhou.net hand calculation.

The example of usage you can find here: https://github.com/MahjongRepository/tenhou-python-bot/blob/master/project/validate_hand.py#L194

Mahjong bot

For research purposes we built a simple bot to play riichi mahjong on tenhou.net server.

0.0.5 version

It can reach a tempai and call a riichi. It doesn't know about dora, yaku, defence and etc. Only about tempai and riichi so far.

This version had played 335 games (hanchans) and achieved only first dan (初段) on the tenhou.net so far (and lost it later, and achieved it again...).

Rate was somewhere around R1350.

Stat:

Result
Average position 2.78
Win rate 20.73%
Feed rate 19.40%
Riichi rate 36.17%
Call rate 0%

So, even with the current simple logic it can play and win.

0.2.5 version

This version is much smarter than 0.0.x versions. It can open hand, go to defence and build hand effective (all supported features you can find in releases description).

This version had played 375 games (hanchans) and achieved second dan (二段).

Rate was somewhere around R1500.

Stat:

Result
Average position 2.65
Win rate 18.60%
Feed rate 10.59%
Riichi rate 15.64%
Call rate 34.89%

For the next version I have a plan to improve win rate, probably bot should push with good hands more often.

For developers

How to run it?

Run pythone main.py it will connect to the tenhou.net and will play a match. After the end of the match it will close connection to the server

Configuration instructions

  1. Put your own settings to the project/settings_local.py file. They will override settings from default settings.py file
  2. Also you can override some default settings with command argument. Use python main.py -h to check all available commands

Round reproducer

We built the way to reproduce already played round. This is really helpful when you want to reproduce table state and fix bot bad behaviour.

There are two options to do it.

Reproduce from tenhou log link

First you need to do dry run of the reproducer with command:

python reproducer.py -o "http://tenhou.net/0/?log=2017041516gm-0089-0000-23b4752d&tw=3&ts=2" -d

It will print all available tags in the round. For example we want to stop before discard tile to other player ron, in given example we had to chose <W59/> tag as a stop tag.

Next command will be:

python reproducer.py -o "http://tenhou.net/0/?log=2017041516gm-0089-0000-23b4752d&tw=3&ts=2" -t "<W59/>"

And output:

Hand: 268m28p23456677s + 6p
Discard: 2m

After this you can debug bot decisions.

Reproduce from our log

Sometimes we had to debug bot <-> server communication. For this purpose we built this reproducer.

Just use it with already played game:

python reproducer.py -l d6a5e_2017-04-13\ 09_54_01.log

It will send to the bot all commands that were send from tenhou in real game.

Code checking

This command will check the code style: flake8 --config=../.flake8

Contribution to the project

All PRs are welcomed anytime. Currently the project is in early stage and I'm working on the different parts of it in the same time, so before making any big changes it's better to check with me to avoid code duplication.

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.