Fantasy Football Optimization Suite

A comprehensive Fantasy Premier League (FPL) optimization tool that uses machine learning and linear programming to predict player performance and generate optimal team selections.

Overview

This project combines historical FPL data analysis, XGBoost machine learning models, and PuLP optimization to:

• Predict player performance for upcoming gameweeks
• Generate optimal team selections for single and multi-period strategies
• Scrape expert recommendations from Fantasy Football Scout
• Optimize transfers and captaincy decisions

Features

• Machine Learning Predictions: XGBoost model trained on historical player performance data
• Team Optimization: Single and multi-period FPL team optimization using linear programming
• Expert Integration: Fantasy Football Scout scraper for incorporating expert recommendations
• Historical Data: Comprehensive dataset covering 2023-2025 seasons
• Transfer Planning: Multi-gameweek transfer optimization

Project Structure

• model.py - XGBoost model training and prediction generation
• fpl_optimizer.py - Basic FPL optimization logic
• fpl_optimizer_single_period.py - Single gameweek optimization
• fpl_optimizer_multi_period.py - Multi-gameweek optimization with transfers
• ffs_scraper.py - Fantasy Football Scout recommendations scraper (standalone)
• data_download.py - Combined FPL API data collection and FFS scraping script
• data_2023/, data_2024/, data_2025/ - Historical and current season data

Installation

1. Clone the repository:

git clone [repository-url]
cd Fantasy_Football_v2

2. Install dependencies:

pip install -r requirements.txt

Or using uv:

uv pip install -r requirements.txt

Usage

Generate Predictions

python model.py

Optimize Team Selection

# Single gameweek optimization
python fpl_optimizer_single_period.py

# Multi-gameweek optimization with transfers
python fpl_optimizer_multi_period.py

Scrape Expert Recommendations (Optional - included in data_download.py)

python ffs_scraper.py

Download Latest Data and Expert Recommendations

python data_download.py

This command downloads both FPL API data and Fantasy Football Scout recommendations in a single run.

Data Sources

• FPL API: Official Fantasy Premier League API for player stats and fixtures
• Fantasy Football Scout: Expert recommendations and insights
• Historical Data: Multi-season player performance and team data

Model Performance

The XGBoost model uses features including:

• Player historical performance
• Fixture difficulty ratings
• Team form and statistics
• Previous season performance

Output Files

Optimized teams and predictions are saved to:

• data_2025/predictions/ - Model predictions and optimal teams
• data_2025/ffs_recommendations/ - Expert recommendations

Dependencies

Key libraries used:

• xgboost - Machine learning model
• pulp - Linear programming optimization
• pandas - Data manipulation
• numpy - Numerical computing
• scikit-learn - Model evaluation
• beautifulsoup4 - Web scraping
• requests - API calls

Future Thoughts

Model Export and Sharing

• Build XGBoost model based on previous years
• Think about pulp scoring: captain, bench and transfers

Considerations

• Dealing with changing team ids (relegations)
• Player transfers
• Free to change captain
• How many bench positions to actually use