Scaling a Signal by Volatility

A 5% price move means something completely different for a sleepy utility stock than for a small biotech that swings 5% most days. If a signal is built from raw price moves — say, recent return as a momentum measure — and fed straight into position weights, the book ends up dominated by whichever stocks happen to be the most volatile, not by whichever stocks the signal is most confident about. Scaling by volatility is the step that separates "this stock moved a lot" from "this stock moved a lot for a stock like this."

The mechanics, stated in words

Divide the raw signal for each stock by that stock's own volatility (typically a trailing realised volatility, sometimes a forecast one). A momentum score for the calm utility gets divided by a small number and grows relatively larger; the same raw score for the volatile biotech gets divided by a large number and shrinks. What's left measures how unusual the move was relative to that stock's normal behaviour, which is closer to what a signal is actually trying to capture — genuine information, not just noisy volatility.

There is a second, separate reason to do this: risk. Even a perfectly informative signal, sized without regard to volatility, will produce a book where the volatile names dominate the portfolio's risk contribution, drowning out the calmer names' contribution regardless of how good the calm names' signal scores are. Volatility scaling at the signal-construction stage does some of the work that position-sizing would otherwise have to do downstream, and it does it before the ranking or blending step, so the ranks themselves reflect risk-adjusted conviction rather than raw magnitude.

Stock	Raw signal (recent return)	Trailing volatility	Vol-scaled signal
Utility co.	+3%	12% annualised	0.25
Biotech co.	+18%	70% annualised	0.26

A worked example

Two stocks both show up as "top movers" on a raw-return momentum screen: the utility up 3% and the biotech up 18%. Read raw, the biotech looks like the far stronger signal. Divide each by a trailing annualised volatility of 12% and 70% respectively, and the two scores come out to about 0.25 and 0.26 — nearly identical. The 18% biotech move was unremarkable for that stock; the 3% utility move was, relative to its own history, just as unusual. Without scaling, the book would have been built almost entirely around the biotech; scaled, the two stocks compete on close to equal footing, which is the more defensible reading of "which move was actually informative."