Football Simulator

Football match data experiments for building versioned DS/ML models.

Project Layout

FutSIm/
├── data/
│   ├── raw/world_cup/          # Original football/FIFA datasets
│   └── processed/              # Generated training datasets
├── docs/                       # Notes and version writeups
├── models/                     # Saved model artifacts
├── notebooks/
│   ├── learning/               # Scratch/learning notebooks
│   └── v2/                     # V2 cleaning and model training notebooks
├── src/
│   ├── v1/                     # V1 rating scripts
│   └── v2/                     # V2 feature/training-data scripts
├── requirements.txt
└── .gitignore

All The Data used is stored here: Datasets

Typical Workflow

So This project consists of different version and what has been done in them has been explained briefly in docs. Looking forward, I don't see any way to improve accuracy or other metrics much, taking into account the randomness of football. So I guess I won't be moving forward with that, unless I learn some new concepts or get to know about things in ML or any other field. (One potential addition I might do in the near future is something with the player detail I have, maybe comparing players.)

So This was it for this project, it has V1, V2, and some random things I tried with AI in V3.