⚽ Premier League AI Predictor

A self-learning, full-stack machine learning application that predicts English Premier League match outcomes with ~80% accuracy on high-confidence bets.

🔗 Live Demo: Click here to visit the App (Note: The backend is hosted on a free tier, so the first request might take ~50 seconds to wake up the server. Please be patient!)

🚀 Key Features

🧠 The "Self-Learning" Engine

Unlike static projects, this application retrains itself every single day.

• Automated Pipeline: Every night, the system downloads the latest match results.

• Recalculation: It rebuilds the entire history (2015–2026), updating Elo ratings and Team Form from scratch.

• Auto-Deployment: A fresh XGBoost model is trained and instantly deployed to the API without manual intervention.

🎯 High-Precision Predictions

• 80.25% Accuracy on high-confidence predictions (>70% probability).

• Real-time Odds: Generates Home/Draw/Away probabilities instantly.

• Stats Breakdown: Explains why a team is favored by comparing Elo, Goals Scored, and Defensive Form.

Accuracy statistics on certain confidence levels

💻 Modern Dashboard (Frontend)

• Live Fixtures: Fetches real upcoming games via API and normalizes team names automatically.

• Clean UI: Dark Mode interface with low contrast elements.

• Dynamic Insights: Expandable "Head-to-Head" panels for deeper analysis.

⚙️ How It Works: The "Daily Job"

The heart of this project is the Backend Automation System.

• Data Ingestion:

  1. A Cron Job triggers daily_job.py.

  2. It fetches the latest results from football-data.co.uk.

  3. It merges new games with the 10-year historical database (PostgreSQL).

• Feature Engineering:

  The system replays history to calculate rolling metrics:

  1. Elo Ratings: Dynamic strength scores updated after every match.

  2. Form: Last 5 games (W/D/L), recent goal scoring, and defensive strength.

  3. Home Advantage: Weighted factors for home-ground performance.

• Model Retraining:

  1. An XGBoost Classifier is trained on the updated 4,000+ match dataset.

  2. The new model is serialized (pickle) and stored in the database, ready to serve predictions immediately.

🛠️ Tech Stack Frontend

• React (Vite): Blazing fast UI.

• Axios: API integration.

Backend

• Python (FastAPI): High-performance REST API.

• XGBoost & Scikit-Learn: Machine Learning engine.

• Pandas & NumPy: Advanced data manipulation.

• SQLAlchemy & PostgreSQL: Robust database management for historical data.

DevOps & Hosting

• Render: Backend & Database hosting + Cron Jobs.

• Vercel: Frontend distribution.

• GitHub Actions: CI/CD pipeline.

💻 How to Run Locally

Follow these steps to spin up the entire system on your machine.

1. Clone the Repository

git clone https://github.com/FAZ110/Premier-League-pred-model.git
cd Premier-League-pred-model

2. Backend Setup

Navigate to the backend and activate the environment.

cd backend
python -m venv venv

Windows:

venv\Scripts\activate

Mac/Linux:

source venv/bin/activate

Install dependencies: Bash

pip install -r requirements.txt

Database Configuration: You will need a local PostgreSQL database or a cloud URL. Create a .env file in the backend/ folder:

DATABASE_URL=postgresql://user:password@localhost/dbname
API_KEY=your_free_key_from_football-data.org

Initialize Data:

1. Clear and Backfill History (2015-Present)

python scripts/backfill.py

2. Calculate Stats & Train Model

python scripts/daily_job.py

Start the Server:

uvicorn app.main:app --reload

Backend running at: http://127.0.0.1:8000

3. Frontend Setup

Open a new terminal.

cd ../frontend
npm install
npm run dev

Frontend running at: http://localhost:5173

⚠️ Disclaimer

This tool is for informational and entertainment purposes only. While the model has achieved high accuracy in backtesting, sports outcomes are inherently unpredictable. Use at your own risk.

Sources:

• https://www.football-data.co.uk/englandm.php Match data from the database
• https://www.football-data.org Football API to fetch upcoming fixtures

Created by: Jan Dyląg