Build from
Real Data

Four case studies on real, public data. Each starts from a question, works through it in code you can read — and run in your browser — and reports what the data actually showed, including what went wrong on the way.

Real-data cases

Four cases, real datasets

Every number on a case page is printed by the code above it, run on the data named at the top. Each case is also a notebook, for JupyterLite or your own machine.

4 Cases Real Data Metrics Runs in the browser
Evidence note: Model-evaluation cases use established statistical methods. Market backtests are educational and heuristic until validated with point-in-time data, costs, slippage, and out-of-sample testing.