Factor Construction by Splitting Additive Market Variables
Summary
This factor research note proposes a general method for designing trading factors by dividing a target variable into informative parts and recombining or transforming them. It identifies three elements: an additive object to split, a discriminating variable that defines the split, and a transformation that produces the factor. Returns, turnover, volume, and average range are given as additive examples; market capitalization and valuation ratios are presented as unsuitable objects because their meanings do not add across time segments.
The split can use different sources of information, including traces of institutional participation in minute data, time of day, or whether a stock is near its high or low. The resulting factor can use one informative segment or combine segments through subtraction or division, which the authors argue can provide normalization and improve stability. Examples are described, but the document supplies no detailed formulas, validation results, or implementation guidance, so the framework is a research heuristic rather than demonstrated evidence of predictive performance.
Key ideas
- A split target should be additive across time periods while retaining the same meaning in each part.
- A useful splitting variable should distinguish parts with different information content.
- Splits may be based on trading behavior, intraday timing, or price position within a range.
- Selecting or comparing the resulting parts can create a new factor, with subtraction or division serving as normalization.
- The note outlines a construction framework but does not provide empirical validation in the supplied text.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.