Data practices
The data problems that bias a backtest before any strategy is run, and how each is checked.
What this shelf covers
Data errors bias a backtest before a single rule runs. The notes treat survivorship and delisting bias, point-in-time values and revisions of macro series, bar timestamps across time zones and daylight saving changes, continuous futures built from rolled contracts, the bid-ask bounce that imitates mean reversion, and disagreement between crypto data sources. Each problem comes with a check that runs on the data itself.
8 notes
A Wide Panel Empties Itself Silently
Bar Timestamps, Time Zones and Daylight Saving: the Checklist Before Any Backtest
The Bid-Ask Bounce Manufactures Mean Reversion and Inflates Small-Cap Returns
Cleaning Tick Data, Step by Step
Point-in-Time Data, Revisions and Vintages
The Epps Effect: Short-Interval Correlation and Asynchronous Prices