An embargo period is an additional block of observations excluded from training immediately after a validation information interval.
Its purpose is to create more temporal separation after direct overlap has already been removed.
Embargo is not the same as purging
Label overlap creates direct information contamination when a training label uses observations that intersect the validation label window. Purging removes those overlapping training rows.
Embargo goes one step further. It removes a nearby post-validation buffer even when direct label overlap has ended.
That makes embargo a research-design choice rather than a universal percentage.
Use the Purged Cross-Validation calculator to see the distinction mechanically, and the CPCV calculator to see how purge and embargo requirements sit inside a broader multi-path validation design.
A larger embargo does not manufacture independent data
More separation can reduce one kind of temporal contamination, but it also removes training observations. Residual dependence is related to effective sample size, but ESS and embargo answer different questions. It does not repair bad point-in-time features, survivorship bias, search multiplicity, or a reused holdout set.
The correct buffer depends on the information horizon and the research process, not on one fixed rule copied across strategies.
Sources: RiskLab AI, Cross-Validation in Finance and Grizzly Bulls Purged Cross-Validation.
Continue Research
Continue from the concept into the Grizzly Bulls research surface that best matches the next question. These links are research continuations, not recommendations or required steps.
Size purge and embargo geometry
See which training rows overlap a validation information window and how an additional embargo changes the retained sample.
Place embargo inside CPCV
See how purge and embargo mechanics fit within a broader combinatorial train/test path design.
Explore more topics in the Financial Research Encyclopedia.