Public work from Alt Data Forge: notes, worked examples and small tools about research data that only uses what was public at the time.
Most research data records what a value is. Far less of it records when anyone could have known it. A backtest or a model evaluation that sees a filing, a revised figure or a reassigned ticker before it was public looks better than it is. The work here is about dating, matching and checking public records so that does not happen.
| Folder | What it holds |
|---|---|
notes/ |
Short write-ups on point-in-time data problems, each with the check that shows the problem |
examples/ |
Small, runnable examples on public data |
| Project | What it does |
|---|---|
| split-adjustment-tool | Checks stock-split records against the as-traded prices before they are applied (Python) |
| us-gov-data-dictionary | Machine-readable data dictionaries for U.S. government datasets, with official definitions and source citations |
Everything published here is built from public records, such as SEC EDGAR filings and U.S. government open-data releases.
Code: Apache-2.0. Notes and text: CC BY 4.0.
Research and education only. Nothing here is investment advice.