A Python project to train and predict soccer match outcomes using historical data and a Random Forest classifier.

• Features
• Prerequisites
• Installation
• Usage
• Project Structure
• Configuration
• Contributing
• License

• Calculates team-level features (goals scored/conceded averages, form, points) over the last 5 matches
• Supports training on historical CSV datasets
• Uses RandomForestClassifier for prediction
• Simple API:
  • train(...) to build the model
  • predict_match(home, away) to get win/draw/lose probabilities

• Python 3.7+
• pandas
• numpy
• scikit-learn

Install required packages:

pip install pandas numpy scikit-learn

1. Clone this repository:

git clone https://github.com/your-username/FootballMatchPredictor.git
cd FootballMatchPredictor

2. (Optional) Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate   # on macOS/Linux
venv\Scripts\activate      # on Windows

3. Install dependencies:

pip install -r requirements.txt

1. Prepare a dataset CSV with columns:
   • Date, HomeTeam, AwayTeam, FTHG, FTAG
2. Update path in code or pass DataFrame/CSV path to train().

from football_predictor import FootballMatchPredictor

# Initialize for Premier League
predictor = FootballMatchPredictor('Premier League')
# Train on 2024-25 season data
predictor.train("Datasets/season-2425.csv")
# Predict a fixture
result = predictor.predict_match("Arsenal", "Liverpool")
print(result)
# Output: {'home_win': 0.65, 'draw': 0.20, 'away_win': 0.15}

.
├── src/
│   └── football_predictor.py  # FootballMatchPredictor class
├── Datasets/
│   └── season-2425.csv        # Sample training data
├── requirements.txt           # Python dependencies
└── README.md                  # This file

• Feature window: currently uses last 5 matches for form/goals averages
• Model: Random Forest with 100 trees; adjust hyperparameters in __init__

Contributions are welcome! Please open issues or pull requests to:

1. Improve feature engineering
2. Add support for other leagues
3. Experiment with different models

This project is licensed under the MIT License – see LICENSE for details.