Practical Part of the Master's Thesis: Data Science in Football

This repository contains the practical part of a master's thesis focused on the use of Data Science methods in football. The project is divided into separate analytical blocks that are methodologically related, but each of them can also be read on its own.

Repository Overview

• xG/
  Exploratory analysis of expected-value metrics in football. This section works with xG, xGA, and xPts, compares leagues, identifies long-term overperformance and underperformance, studies Leicester City's 2015/16 Premier League season, and explores team styles through clustering.

• match_prediction/
  Match-result prediction and season simulation for the Premier League, La Liga, Bundesliga, Serie A, and Ligue 1. Models are developed on 2021/22–2023/24 and evaluated chronologically on 2024/25. The analysis compares structural and market-informed classifiers with reference approaches and uses an independent Poisson score model to simulate final league positions and their uncertainty.

Recommended Reading Path

Depending on the reader's goal, the project can be approached in several ways:

1. For a quick overview, start with this file and then move to the README of the selected section.
2. For the methodological interpretation of the xG analysis, open xG/xG_analysis.ipynb and use xG/src/ as supporting documentation.
3. For the predictive pipeline, start with match_prediction/README.md, then continue with match_prediction/notebooks/README.md and follow the notebook order.
4. For results and visual outputs, use xG/outputs/ and match_prediction/outputs/. The Czech presentation notebooks prepare the tables and figures used in the thesis.

Project Structure

DP/
|-- README.md
|-- xG/
|   |-- README.md
|   |-- xG_analysis.ipynb
|   |-- xG_thesis_outputs_cs.ipynb
|   |-- data/
|   |-- outputs/
|   `-- src/
`-- match_prediction/
    |-- README.md
    |-- notebooks/
    |-- src/
    |-- data/
    `-- outputs/

How to Approach the Repository

• The repository is primarily notebook-driven: the main analytical narrative is developed in .ipynb files.
• Shared logic is moved into src/ modules to keep the work reproducible and reusable.
• Data folders are separated by pipeline stage into raw, interim, and processed.
• Outputs intended for interpretation and presentation are stored separately in outputs.
• README files inside subfolders act as local guides that explain what the folder contains, why it exists, and when it matters in the workflow.