A training trade ending on Friday and a test trade beginning on Thursday can share the same price shock. Putting their starting dates in different folds does not separate their outcome information. Purging removes such overlaps; embargoes provide an additional exclusion after a test interval when later training observations may remain too closely connected. This lesson translates the time-ordered design into interval rules and an auditable split ledger.
Learning objectives. By the end of this note you should be able to:
- identify all overlap configurations, including interval containment;
- implement purging with an explicit endpoint convention;
- explain when a post-test embargo has a role;
- justify an embargo from the information process rather than a universal percentage;
- distinguish leakage control from independent evidence and selection control.
In this lesson
- Purge information intervals
- A split that defeats a start-date check
- Add an embargo for post-test training
- Choose the embargo from the information process
- What these controls cannot repair
- Keep an exclusion ledger
Before you start
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Quantitative Research, Data and Backtesting syllabus.