The whole platform, one page
How ConvexPi fits together
An open platform for empirical finance, organized around three things you do: learn the method, experiment against hidden holdouts and live markets, and contribute to a shared research commons.
New here? The path
- 1. Run Mission 1 — feel overfitting first-hand.
- 2. Enter a competition — get scored on hidden data.
- 3. Publish a project — share it; earn reputation.
Learn
Learn the method
Why out-of-sample is the only honest test — and the tools to find real signal.
- Get started
The 30-minute on-ramp.
- Curriculum (6 missions)
The core course: overfitting → alpha discovery.
- Lectures ↗
p-hacking, the Information Coefficient.
- Market-making lesson
Earn the spread; manage inventory & adverse selection.
- How the exchange works
The matching engine, in plain English.
- The realistic exchange (L3)
Queue position, latency, the cancel race.
- Glossary
OOS Sharpe, overfitting ratio, maker/taker, L2/L3…
Experiment
Experiment against reality
Test ideas on hidden holdouts and live markets. Reality is the final test set.
- Playground
Run code in the browser, no setup.
- Competitions
Submit strategies, connect agents, forecast — ranked out of sample.
- S&P next-day
Predict tomorrow’s index move; scored live on real prices.
- Realistic exchange (L3)
Trade a real order-by-order book with true queue position.
- Agent arena
Write a trading agent for the live limit-order book.
Contribute
Contribute to the commons
Everything is open. Replicate the canon, publish your work, earn reputation.
- Replication library
Canonical strategies, recomputed and scored OOS.
- Projects showcase
Publish a notebook from GitHub; we run, render, and rank it.
- Papers & wikis
The finance-ML literature, summarized.
- Anomaly graveyard
Which published edges actually survived.
- Contributors
The reputation leaderboard.