Projections come from models trained on nflverse history, not from opinion or hand-tuned rankings.
About · a portfolio project
Built in the open, measured before published.
Win My League is a one-person applied data science project. It exists to show an honest, end-to-end forecasting system rather than to sell advice.
01
What this is
A weekly projection system for fantasy football built entirely on public nflverse data. Models are trained on a player's own history, evaluated on a forward-time holdout against a causal baseline, and published by a scheduled job that can refuse to publish. The site reads everything it shows from the artifacts that job commits.
02
Who builds it
Christopher Bratkovics builds and maintains the models, the API, and this site. He holds an M.S. in Applied Data Science and a B.S. in Computer Science and works in enterprise analytics and data engineering. The project started as a way to apply that work to a problem he cares about.
03
How it is built
Every projection ships with a floor and ceiling and names the model and feature version that produced it.
The modelling code, the API, and the committed artifacts are public in the repository.
Nothing is published until it has been scored against a baseline on a held-out season.
04
Where to start
Open the predictions for the latest published week, read the evaluation to see how the model compares with its baseline, or read how it works for the full pipeline.