An event classifier for a long–short equity book is only as good as the clock that defines its observations. Sampling every calendar day produces a row whenever the market opens; sampling every fixed notional produces a row when information arrives. This note decides which bar type feeds the classifier, adapting the data-bars lesson without re-deriving bar statistics from scratch.
Learning objectives. By the end of this note you should be able to:
- state what each bar type triggers and what decision it encodes;
- argue when time bars are an acceptable default and when they distort the classifier's design matrix;
- document a sampling choice as a frozen research artefact for the running case.
In this lesson
- The decision: which observations are events?
- Method: freeze the clock before features
- Deliverable: sampling contract
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 Financial Machine Learning syllabus.