Free module · Foundations
Data: Availability, Revisions & Reproducible Research
Reconstruct what a trader could know, then preserve enough evidence to reproduce the decision.
timestamps · vintages · identifiers · survivorship · corporate actions · provenance
The building blocks
A historical value is not automatically a value that was available historically. Every observation needs an information timeline.
- Separate observation, publication, ingestion and revision times.
- Reconstruct a decision-time snapshot.
- Validate units, coverage and a reproducible manifest.
Lessons in this module
- Four clocks for one observation
- An as-of join that never selects the future
- First prints and revisions can reverse a signal
- Reconcile splits and cash distributions
- Universe membership and disappearing assets
- Coverage, missingness and stale values
- Purging labels that cross a test boundary
- Reproducible snapshots and discrepancy ledgers
Practice and apply
- Calculate usable information time — Release at epoch-second 1000, ingest 1003, processing delay 2 seconds.
- Reconcile split-adjusted wealth — 10 shares bought at $100 split into 20 shares, finish at $51; $10 total distributions.
- Retain the failed asset — Two assets each carry .5 original weight. Returns .1 and −1 over one period.
Work through the practice exercises · Quant development tools