Real-time analytics for fantasy
football managers
Fantasy Premier League has over 11 million active managers, but the official platform gives them almost nothing in terms of meaningful analysis. Basic stats don't tell you who's in form, who's about to rise in price, or which transfers will actually make a difference.
We built a data platform that solves that. A proprietary scoring algorithm processes multiple data sources daily, calculates performance ratings for every Premier League player, and delivers actionable insights through a fast, static frontend.
This is a standalone product we built from scratch — data pipeline, algorithm, frontend, infrastructure. It demonstrates what we can do with data engineering, not just web design.
A data engineering problem,
not a football problem
The official FPL platform provides basic stats, but nothing that helps managers make genuinely informed decisions. Points per game don't tell you who's in form. Historical data doesn't account for squad changes. Most third-party tools just rearrange the same surface-level numbers.
The challenge was designing an algorithm that combines multiple data sources into a single meaningful rating per player, position-aware and context-sensitive. Then building a pipeline that runs automatically, processes everything daily, and serves it through a frontend that loads instantly with zero runtime dependencies.
Every feature earned its place
Proprietary scoring algorithm
A multi-factor, position-aware rating system that weights metrics differently depending on whether a player is a goalkeeper, defender, midfielder, or forward. Not just stats rearranged — a genuine model.
Automated data pipeline
Daily automated ingestion from multiple data sources. Raw data is fetched, processed, scored, and stored without manual intervention. The entire pipeline runs on our own infrastructure.
Regression analysis
Research-backed detection of over and underperforming players. The system flags who is due a correction and who is genuinely elite, helping managers avoid traps and spot opportunities.
Player rating system
A confidence-tracked rating system inspired by competitive ranking models. Ratings adjust based on performance relative to peers in the same position and price bracket.
Static frontend architecture
The backend feeds a pre-rendered Astro frontend. Pages are built once daily with fresh data baked in. No runtime API calls. Instant loads, zero server dependency for end users.
Fixture and form analysis
Position-aware fixture difficulty ratings and rolling form windows. The system understands that a defender's upcoming fixtures matter differently to a forward's. Context-sensitive, not generic.
Data engineering that
runs itself
FPL Ratings runs autonomously. Every day the pipeline fetches fresh data, the algorithm processes and scores every player, and the frontend rebuilds with updated ratings. No manual intervention. The infrastructure handles everything.
This project proves we can build beyond websites. A proprietary algorithm, an automated data pipeline, a static frontend architecture, and infrastructure that runs unattended. It's the kind of engineering that most consultancies can't offer — and it's a product we built from the ground up.