I’ve built my own football match prediction model + website (currently in beta) – sharing details

Hi everyone,
over the last months I’ve been building my own football prediction engine as a side project, and I wanted to share what I’ve done so far — mainly from a technical/data perspective, not trying to promote anything.

The project consists of two parts:
1) the ML model

• mainly logistic regression + additional decision rules
• momentum and form weighting
• odds-as-features
• backtesting framework (rolling window)
• model calibration
• live-data layer planned for next iterations
• data source: Football-Data API + my own preprocessing pipeline

2) the website (beta)

• simple UI that displays win/draw/lose probabilities
• daily updated predictions
• still very early, so bugs are expected
• using a lightweight backend + caching to speed up requests

Right now I’m testing how stable the predictions are across leagues and how backtested accuracy compares to live performance. I’m also trying to avoid overfitting and looking at ways to improve calibration under different odds ranges.

If anyone is working on similar projects or has experience with sports ML models, I’d be happy to hear about your approaches, especially regarding feature engineering, probability calibration, or dataset cleaning.

Again — not promoting anything, just sharing that the model + website are in beta and still evolving.

AI-powered football prediction engine just nailed 3/3 matches yesterday:

• Arsenal 3-1 Bayern Munich ✔ Home win
• Manchester City 0-2 Bayer Leverkusen ✔Away win
• Frankfurt 0-3 Atalanta ✔Away win

Here is the tip for today

Processing img zv7zpct09p3g1...