Agentic FPL Manager — Technical Case Study
Next-Generation Fantasy Premier League Tactical Intelligence
An enterprise-grade, agentic intelligence platform engineered to replace subjective intuition with game-theory mini-league optimization, multi-gameweek transfer trajectories, and a real-time conversational AI tactical copilot.
Executive Problem Statement
Fantasy Premier League (FPL) is played by over 11 million global managers. Traditional companion apps are static number calculators that lack contextual understanding of mini-league game theory, rotation risk patterns ("Pep Roulette"), and long-term transfer horizon planning.
Agentic FPL Manager was architected from the ground up to solve this through autonomous agentic evaluation, strict system prompt constraints, and structured interaction protocols powered by Google Gemini.
System Architecture & Key Features
AI Tactical Copilot
Powered by Google Gemini with structured tool calling. Ingests live player metrics, fixture difficulty ratings, and returns deterministic transfer feasibility advice.
Mini-League "Rival Spy"
Computes Effective Ownership (EO) across competitive mini-leagues. Formulates differential captaincy recommendations to mathematically swing point deficits against league leaders.
5-Gameweek Horizon Engine
Multi-week roadmap evaluating rolling free transfers vs point hits (-4), and model-driven chip deployment windows (Wildcard, Free Hit, Bench Boost, Triple Captain).
Interactive Visual Pitch
Dynamic formation detector (3-5-2, 4-3-3, 3-4-3) with real-time captain/vice-captain badges, injury alert flags, and detailed player performance popups.
System Architecture Diagram
Engineering Stack & Deployment
The frontend is built with Next.js 15 (App Router), React 19, and TypeScript for end-to-end type safety. Styling strictly adheres to official Premier League and FPL design token guidelines via Tailwind CSS. Relational data schemas and player history are persisted in PostgreSQL, deployed continuously with automated GitHub Actions and Vercel Edge networks.
Project Metadata
- Role / Author: Asman Makero Musa
- Category: Agentic AI & Web Platform
- Deployment: Live on Vercel
- Database: PostgreSQL (Supabase)
- AI Engine: Google Gemini (Tool Calling)
- Frontend: Next.js 15 • React 19 • TS
- Live Domain: agentic-fpl.vercel.app