All files are generated with NumPy seed 20260813.
| File | Grain | Purpose |
|---|---|---|
| asset_returns.csv | 520 business days by six asset columns | Axes, broadcasting, portfolio matrix multiplication, rolling features |
| asset_metadata.csv | One row per ticker | Costs, asset classes and shape-safe joins |
| factor_panel.csv | One date-ticker row | Signals, realised and forward returns, vectorised backtesting |
| ml_observations.csv | One observation row | Six features, group, label, score and sample weight |
| market_prices.csv | One date-ticker row | Daily prices and volume |
| model_predictions.csv | One date-ticker-model row | Join and evaluation exercises |
| instruments.csv | One row per ticker | Multipliers and reference data |
| quotes.csv | One timestamp-ticker quote row | Asynchronous alignment and resampling |
| trades.csv | One irregular execution row | Cash flows, joins, grouping and slippage |
| macro_releases.csv | One publication-series row | Point-in-time feature alignment |
Important: these data are pedagogical simulations, not investment data. The forward_return column is intentionally provided for evaluation; it must never be used as a feature at the same date.