# SmartPlay

> AI-powered Fantasy Premier League analytics. Free, open-source expected points predictions, fixture ratings, squad optimiser, and transfer assistant.

SmartPlay is built by Qazybek Beken at UC Berkeley. It predicts FPL points using a multi-bucket XGBoost ensemble trained on 251 features from the FPL and Understat APIs, then optimises squad decisions with a mixed-integer linear program powered by HiGHS.

## Key Pages

- [Home](https://smartplayfpl.com): Player table with xPts, fixture ratings, and filters
- [Squad Optimiser](https://smartplayfpl.com/build): Enter your FPL ID to get optimal transfers, captain, and lineup
- [Player Comparison](https://smartplayfpl.com/players/compare): Side-by-side player stat comparison
- [How It Works](https://smartplayfpl.com/under-the-hood): Model architecture, solver details, and accuracy benchmarks
- [Ethics](https://smartplayfpl.com/ethics): Transparency principles and strict no-gambling policy
- [Credits](https://smartplayfpl.com/credits): Open-source projects and researchers we build on

## Technical Details

- Prediction model: multi-bucket mixture of position-specific XGBoost regressors (28 models)
- Feature engineering: built on OpenFPL (Daniel Groos, Groos Analytics)
- Solver: mixed-integer linear program via HiGHS, multi-gameweek lookahead
- Stack: Next.js 15, PostgreSQL, Python ML pipeline, TypeScript data pipeline
- All predictions based on publicly available FPL API and Understat data only

## Links

- [Full LLM context](https://smartplayfpl.com/llms-full.txt)
- [Contact](mailto:hello@smartplayfpl.com)