⚽ FPL AI-Powered Team Optimizer 🤖

A sophisticated analytics tool that leverages machine learning and AI-driven insights to build your optimal Fantasy Premier League (FPL) squad.
This project goes beyond basic statistics — combining a predictive ML model, a mathematical optimizer, and qualitative AI scouting reports to give you a decisive edge in FPL team selection.

---

---

✨ Key Features

• 📊 Live Data Integration – Fetches real-time player data from the official FPL API for the most current stats.
• 🔮 Predictive Analytics – Uses a pre-trained XGBoost model to forecast expected points (xP) for every player.
• 🤖 AI Scouting Reports – Integrates with Google's Gemini API to provide insights on player form, injuries, and tactical roles.
• ⚙️ Optimal Squad Generation – Uses the PuLP optimization library to build the best 15-player squad within budget and rules.
• 📈 Fixture Difficulty Rating (FDR) – Calculates attacking & defensive strength to assess match difficulty.
• 🔧 Manual Overrides – Adjust player scores with your own scouting bias in main.py.

---

🚀 System Workflow

graph LR
A[Live FPL API Data] --> B[Data Processing & FDR Calculation]
B --> C[XGBoost Point Prediction]
C --> D[Gemini AI Scouting]
D --> E[PuLP Squad Optimization]
E --> F[Final Team Output]

Loading

---

🛠 Technology Stack

Category	Technology
Core Language	Python 3
Data Science	Pandas, NumPy, Scikit-learn
Machine Learning	XGBoost
Optimization	PuLP
API & Web	Requests, Google Generative AI, python-dotenv

---

📦 Setup & Usage

1️⃣ Prerequisites

• Python 3.8+
• Git
• Google AI Studio API Key (for Gemini AI scouting)

---

2️⃣ Installation

# Clone the repository
git clone https://github.com/your-username/your-repo-name.git
cd your-repo-name

# Create & activate a virtual environment
python -m venv venv
source venv/bin/activate  # Windows: .\venv\Scripts\activate

# Install all dependencies
pip install -r requirements.txt

---

3️⃣ Configuration

Create a .env file in the project root:

GEMINI_API_KEY="YOUR_API_KEY_HERE"

(Optional) Customize scout_overrides and player_team_overrides in main.py to add your own bias.

---

4️⃣ Run the Optimizer

python main.py

The script outputs:

• Starting XI
• Bench
• Captain

---

📂 Project Structure

.
├── data/
│   ├── epl_player_stats_24_25.csv
│   ├── fpl_playerstats_2023-24.csv
│   ├── fpl_playerstats_2024-25.csv
│   └── fpl_xgb_model.json
│
├── scripts/
│   ├── data_processing.py
│   ├── fixtures.py
│   ├── fpl_playerstats_2024-25.csv
│   ├── model_trainer.py
│   ├── optimized_main.py
│   ├── scout.py
│   └── scraping_script.py
│
├── .env
├── .gitattributes
├── .gitignore
├── README.md
└── requirements.txt

---

🎯 Limitations

• Static Prediction Model – The XGBoost model does not retrain automatically during the season.
• Single Gameweek Focus – Optimizes only for the upcoming gameweek.
• Outdated Fixtures Risk – fixtures.py must be updated manually for each season.

---

🔮 Future Roadmap

• Automated weekly model retraining.
• Multi-gameweek lookahead for strategic planning.
• Weekly transfer & captaincy suggestion module.

---