I build forecasting tools and interactive applications around sports and music. These projects bring together data analysis, statistical modeling, and web development.
These are ongoing personal projects. Each links to a public demo and source code, with documentation explaining its methods and limitations.
Race forecasts and dashboards for several motorsport series. I brought the models and dashboards for each sport together around shared data, calibration and evaluation tools.
The public catalog distinguishes working prediction projects from series still in development. Project catalog
NBA game forecasts, season simulations and an interactive history of the league. I built a Python modeling pipeline and a dashboard for exploring the published predictions.
In the documented historical benchmark, the margin model improves on a constant baseline but trails forecasts from closing market odds. Read the benchmark
Men’s NCAA tournament history, seed probabilities and interactive brackets. I built a trained prediction pipeline and a browser app for comparing matchups and exploring the road through a tournament. The 2027 field awaits announcement.
The documented rolling evaluation covers 377 games from 2021 through 2026. The seed model improves probability scores over a fixed seed heuristic; the team form challenger does not improve the primary model. Read the Evaluation
NFL game probabilities, season projections, and playoff scenarios. I built a model that accounts for football scoring patterns and an interactive Forecast Lab for exploring matchups and possible outcomes.
The documentation compares forecasts with Elo and historical market prices, and explains the limits of its evaluation and playoff simulations. Read the Evaluation
Match probabilities, season projections and knockout brackets. I connected football data and forecasting models to an interface for following teams and exploring possible outcomes.
Published evaluations separate match, season and tournament performance. The historical match benchmark trails the market. Read the evaluation
A web application for exploring songs through lyrics and sound. I combined language analysis, audio features processed in the browser and a music discovery interface in one application.
The source documents its transformer and keyword engines, audio processing, and fallbacks. Mood labels are exploratory interpretations. How the analysis works