ConvexPi

For instructors

Teach with ConvexPi

Run a simulation-first quantitative-finance course with zero setup and zero manual grading. Students work in Colab, submit strategies, and are scored automatically on a hidden out-of-sample market — so there are no answer keys to leak and no curve-fitting to reward. Everything is free and open-source.

1

Create your classroom

A classroom is a private cohort with its own roster, join code, and leaderboard. Name it for your course (e.g. “FINA 6090 — Fall 2026”) and set the dates.

Create a classroom
2

Share the join code

Each classroom gets a 6-character join code. Share it (or the join link) with your students; they create a free account and enroll themselves at /classroom/join. Private classrooms are visible only to enrolled members.

3

Assign the curriculum

The 9-mission curriculum is your ready-made syllabus — from the overfitting trap through alpha discovery, real data, and market microstructure. Missions 1–3 work well as a first half; 4–6 as a research-project phase; 7–9 are advanced electives. Each runs in Colab with no install.

For reading, point students at the topic surveys (momentum, value, quality, low-vol, the factor zoo) and the paper wikis they synthesize.

4

Students submit — graded automatically

Students build a strategy (each competition has a Colab starter) and submit code. The grader runs it on a hidden out-of-sample market and returns an OOS Sharpe, overfitting ratio, and alpha-discovery report — no manual grading, and in-sample curve-fitting earns nothing. Run the class against your classroom cohort or the always-open public leaderboard.

5

Monitor on your dashboard

Your instructor dashboard shows the roster, each student’s submissions and grades, the class’s average overfitting ratio, and who hasn’t submitted yet — a live view of who’s actually learning the out-of-sample lesson.

Go to your dashboard

Why it works for a course

  • No setup: everything runs in Colab; the only install is pip install convexpi-lab.
  • Cheat-resistant: grading is on a hidden holdout with a private seed — no answer key, and overfitting is visible, not rewarded.
  • Open-source: the curriculum, grader, and replications are all public; adapt them freely.