Home Original page

GitHub - UtkarshRahim/fpl-intelligence

An AI-powered Fantasy Premier League advisor that recommends transfers, evaluates hit strategies, and signals wildcard timing — using live FPL data and Claude AI.

What it does

  • My Squad — loads your 15 players via your FPL Manager ID, shows prices, points, form, and injury status
  • Transfers — formula-ranked upgrade suggestions with an AI analysis button per player (Claude reasons about fixtures, form, and injury risk)
  • Strategy — AI recommends whether to use your free transfer, take a -4 hit, or bank for next week, with ranked priority transfers
  • Wildcard — compares your squad score vs an optimal squad and signals when to pull the trigger
  • Settings — on-demand data refresh (pulls latest FPL data without any cron job)

Stack

data/          Python — fetches FPL API, writes to SQLite
backend/       Node.js + Express — reads DB, serves REST API, calls Claude
frontend/      React + Vite — dashboard UI

Prerequisites

Setup

1. Clone and install

git clone <your-repo-url>
cd fpl-intelligence

Python layer:

cd data
python3 -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install requests pandas scikit-learn aiohttp fpl

Node backend:

React frontend:

2. Add your API key

# backend/.env
ANTHROPIC_API_KEY=sk-ant-your-key-here
PORT=3001

3. Run the data pipeline (once per gameweek)

cd data
source venv/bin/activate
python pipeline.py
# takes ~2 minutes, populates data/db/fpl.db

4. Start the servers

Open three terminals:

# Terminal 1 — backend
cd backend && npm run dev

# Terminal 2 — frontend
cd frontend && npm run dev

Open http://localhost:5173, enter your FPL Manager ID and hit Load.

Finding your Manager ID

Go to fantasy.premierleague.com → click Points → the number in the URL is your ID:

fantasy.premierleague.com/entry/1234567/event/...
                                 ^^^^^^^

Mobile (same WiFi)

# Find your machine's local IP
ipconfig getifaddr en0      # Mac
# e.g. 192.168.1.42

Create frontend/.env.local:

VITE_API_URL=http://192.168.1.42:3001/api

Make sure vite.config.js has host: true and backend/index.js listens on 0.0.0.0. Then open http://192.168.1.42:5173 on your phone.

API endpoints

Method Endpoint Description
GET /api/my-team/:managerId Your 15 players for the current GW
GET /api/free-transfers/:managerId Free transfers banked + made this GW
POST /api/transfers Formula-ranked transfer suggestions
POST /api/ai-transfers Claude analysis for a specific player swap
POST /api/strategy Full transfer strategy (hits, banking, priority order)
POST /api/wildcard Wildcard signal vs optimal squad
GET /api/fixtures All fixtures with team names and FDR
GET /api/value Top 50 players by points per million
POST /api/refresh Triggers Python pipeline to refresh DB data

How the AI works

The formula (form, points-per-game, ICT index, availability) narrows candidates first. Claude then receives the player's stats, the top alternatives, and both teams' next 5 fixtures — and reasons about whether the transfer is worth making, flagging rotation risk, injury news, and fixture swings. The strategy endpoint additionally factors in your exact free transfer count (calculated from your full transfer history) to decide whether a -4 hit is worth taking.

Project structure

fpl-intelligence/
├── data/
│   ├── venv/                  # Python virtual environment (gitignored)
│   ├── db/
│   │   ├── setup.py           # Creates SQLite schema
│   │   └── fpl.db             # SQLite database (gitignored)
│   ├── fetchers/
│   │   └── fpl_api.py         # FPL API calls
│   └── pipeline.py            # Orchestrates fetch + store
├── backend/
│   ├── index.js               # Express API + Claude integration
│   ├── .env                   # API keys (gitignored)
│   └── package.json
├── frontend/
│   ├── src/
│   │   ├── App.jsx            # Main app, routing, squad/wildcard tabs
│   │   ├── TransferCard.jsx   # Per-player transfer card with AI button
│   │   └── StrategyPanel.jsx  # Full strategy tab
│   ├── .env.local             # Local overrides e.g. VITE_API_URL (gitignored)
│   └── vite.config.js
└── README.md

Refreshing data

No cron needed. Go to the Settings tab in the UI and hit Refresh FPL data. It triggers the Python pipeline on demand — run it at the start of each gameweek after the deadline.