Make quantitative finance empirical again.

Most of finance education — and a lot of finance research — rewards the wrong thing: a backtest that looks brilliant on the data it was fit to. ConvexPi is built around the only question that matters, does it work on data you have never seen?

We teach the discipline of honest, out-of-sample evaluation, and we build the tools to practice it in the open: a simulation-first curriculum, a research library that is candid about which anomalies survive and which decayed after publication, an in-browser playground, a live market arena, and an open package of verified strategy replications. The grader, the data pipeline, the benchmark logic — all open, so students can inspect the machinery, instructors can adapt it, and researchers can extend it.