Projects

An applied-machine-learning platform that reads six decades of national indicators to find countries tracing the same trajectories that preceded past crises, and surfaces the real historical precedents behind every signal.

Built twice: the retired Classic Cascade, a pure-statistics correlation pipeline, and the current Trajectory Engine, a fleet of per-event-type models. A separate statistical scoring layer, older than the fleet and unrelated to the retired recipes, still contributes country scores alongside it. The project page has the architecture of each generation, the pipeline steps, and how the fleet is cross-validated, including the result that was retracted and what was installed so it could not happen again.

A data-journalism study for TU Dortmund's KURT magazine, testing whether EA FIFA's potential ratings actually predict the careers players go on to have.

Players rated in the top potential bracket go on to become ten-year veterans just 8.8% of the time. The scrape, the cleaning and the analysis are in R; the extended study is published as a PDF and the article ran in KURT in May 2022.