Rolling Returns

Rolling returns are returns computed on a moving window.

Instead of "What was the 1-year return from Jan 1 to Dec 31?", rolling returns ask: "What was the 1-year return as of each day?"

Why it matters

They show which asset is winning at any point. One asset can dominate one year and lag the next -- rolling returns make that obvious.

How we compute them

For each day $t$, we compute the return over the last $N$ days:

$$R_t = \frac{P_t}{P_{t-N}} - 1$$

We use calendar-day windows (30/90/180/365). For trading-day assets, that means fewer observations inside the window, and we do not forward-fill missing days.

Caveat

Rolling windows overlap, so the points are not independent. Treat the chart as a trend view, not a statistical test.