Ori Knows Ball ⚽️

Predicting the 2025–26 UEFA Champions League — from the group stage to the final — using machine learning (Random Forest classification).

---

✨ Features

• Match-level pre-game predictions (H/D/A)
• Round-by-round simulation (Monte Carlo from per-match probabilities)
• Incremental upgrades: Elo → rolling form → team stats (FBref)
• Optional inputs (future): news, injuries, sentiment

---

🧰 Tech Stack

• Core: Python (pandas, NumPy), scikit-learn
• Data: FBref (CSV) + ClubElo-style ratings (CSV with from/to)
• Viz: Matplotlib (static reports)
• IO: CSV / Parquet (for fast cached datasets)
• App (TBD): Streamlit or Flask UI

---

📦 Repository Layout

• fbref_data//.csv # schedule, shooting, passing, ...
• data/elo_filtered/.csv # time-ranged Elo per team
• scripts/preview_clean_v1.py # schema check + cleaned schedule generation

🚦 Project Status

• Scope: 2025–26 UEFA Champions League (group → final)
• Current focus: data hygiene (schedule parsing), per-match Elo join, baseline RandomForestClassifier
• Risks/Next: team-name normalization, header flattening for FBref tables, leakage-proof rolling windows, calibration

---

🔖 Versioning

• Current: 0.1.0-dev (data engineering/research/training)

Credit

• Data Scraping: https://github.com/probberechts/soccerdata