PitchLink Euro 2024 — Tactical Network Engine

Elite Command Center for football tactical analysis featuring Pitch Overlay, Edge Intelligence, Community Clusters, and Scouting AI.

---

✨ Overview

PitchLink Euro 2024 is a specialized tactical engine that tracks passing networks, identifies tactical micro-structures (clusters), and dynamically scouts key playmakers using graph theory. Built for analysts and scouts to dissect match and tournament data using interactive network visualizations.

🌐 Live Demo: https://pitchlink-euro-2024-production.up.railway.app/app

---

🧠 Core Features

• 🕸️ Network Intelligence — Models passing behavior via PageRank and Betweenness Centrality
• 🧩 Tactical Units — Uses the Louvain algorithm to automatically detect player communities and positional clusters
• 🕵️ Scouting AI — Generates real-time text briefings highlighting progressive passing hubs and focal points
• 🔍 Advanced Filtering — Isolate by team, scale by centrality, or filter out low-value connections
• 📈 Leaderboard Roster — Interactive datatables dynamically updating based on on-pitch selections
• 🎨 Elite Dark UI — Custom 100vh Bokeh dark mode with neon green/cyan terminal aesthetics

---

📁 Project Structure

tactical-engine/
│
├── app/
│   ├── main.py               # Core Bokeh application & graph logic
│   └── templates/
│       └── index.html        # Custom Jinja2 dark-mode wrapper
├── data/
│   └── raw_passes.csv        # Passing event dataset
├── src/                      
│   ├── build_graph.py        # Graph processing utilities
│   ├── compute_metrics.py    # Math/algorithmic models
│   └── fetch_statsbomb.py    # Data ingestion scripts
├── Dockerfile                # Production server specs
├── requirements.txt          # Python dependencies
├── LICENSE
└── README.md

---

⚙️ Local Development Setup

1️⃣ Install Dependencies

# 1. Create and activate a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows use: venv\Scripts\activate

# 2. Install dependencies
pip install -r requirements.txt

2️⃣ Data Availability

data/raw_passes.csv is already included in this repository, so you can run the app immediately.

Optional: regenerate the dataset from StatsBomb source data:

pip install statsbombpy
python src/fetch_statsbomb.py

3️⃣ Start the Application Server

Run the Bokeh directory app locally:

bokeh serve app --show

4️⃣ Access the App

Your browser will automatically open, or you can navigate to:

http://localhost:5006/app

---

🔬 Tactical Metrics Explained

Centrality (PageRank)

Calculates a player's importance not just by how many passes they receive, but by the quality of the players passing to them. Identifies the true focal point of a team's transition system.

Louvain Communities

Groups players who interact with each other significantly more than with the rest of the team, successfully uncovering tactical "pods" (e.g., a left-sided triangle of LB-LW-LCM).

Progressive Passes

Highlights vertical line-breaking actions, helping to distinguish between safe lateral recycling and high-value infiltration passing.

---

🎨 UI Features

• Glow Selection — Clicking a node dims non-associated players and brightens exact passing routes
• Auto-Search — Real-time JS-based autocomplete that immediately centers the pitch on the queried player
• Granular Edges — Pass lines scale in opacity and thickness based on the volume and progression-value
• Crosshair Tracking — Tactical crosshair cursors and hover-tooltips for precise spatial analysis

---

🚀 Deployment

This application is containerized and optimized for cloud platforms.

Deploying to Railway / Render (Docker)

1. Connect your GitHub repository to Railway.app.
2. Railway will automatically detect the Dockerfile.
3. The Dockerfile natively binds to the $PORT variable for seamless web port mapping.
4. Set memory allocation. (Graph generation is memory-intensive; 1GB+ recommended).
5. Open your generated domain to view the application!

---

📊 App Data Flow

Data Ingestion (CSV) 
        ↓
NetworkX Directed/Undirected Graphs Created
        ↓
Louvain & PageRank Algorithms Computed
        ↓
ColumnDataSources synced to Bokeh Interface
        ↓
Jinja2 Custom HTML Matrix applies Dark CSS
        ↓
JS Callbacks trigger real-time UI/Scouting changes

---

🐛 Troubleshooting

Blank White Screen on Load

• Ensure you are running bokeh serve app (the directory) rather than bokeh serve app/main.py. The directory setup is required to utilize index.html.

Internal Server Error (500) during Deployment

• Ensure scipy is in your requirements.txt (required under the hood by NetworkX PageRank).
• Ensure the platform is injecting a valid $PORT environment variable that the Docker container binds to.

---

📝 Technical Stack

Layer	Technology
Backend / Logic	Python 3.10+, NetworkX, SciPy, Pandas
Visual Framework	Bokeh
Community Detection	Python-Louvain
Infrastructure	Docker, Bash
UI/UX	HTML5, CSS3, Bokeh CustomJS

---

🪪 License

MIT License — Copyright © 2026 Navodhya Fernando

See LICENSE for full details.

---

👨‍💻 Developer

Navodhya Fernando
Data & Web System Engineer @DreamShift INC
Data Science Undergraduate at National Innovation Centre (NIBM), Colombo 05, Sri Lanka
In partnership with Coventry University