Lecture
What you hand in
- A notebook that runs top to bottom from a fresh session, with a fixed random seed
- A goal card: ticker, target, horizon, metric and baselines
- Walk-forward results on the most recent 12 months
- A one-page model card with an honest verdict
- A prompt log showing at least three AI-code fixes
- Five Forecast Week predictions
How it’s scored
| Criterion | Points |
|---|---|
| Runs and reproduces | 15 |
| Data done right | 20 |
| Fair comparison | 20 |
| Verification | 20 |
| Honest conclusion | 15 |
| Forecast Week | 10 |
A model that loses to the baseline but is tested and reported honestly still passes. Beating the baseline earns a bonus badge.