Glossary
Rolling returns
Returns measured over every window of the same length within a history, which show the spread of results rather than one start and end date.
On this page
Rolling returns are the returns over every period of a fixed length inside a longer history: every three-year stretch, say, starting each month or each year. Instead of one number that depends on a chosen start and end date, they give a range of outcomes an investor could actually have had.
A single trailing figure, such as "10 years to December 31", can change a lot when the end date moves by a few months, because one crash or rally enters or leaves the window. Rolling returns show how much.
What they are used for
- Range of outcomes. The best, worst and median window say more about a fund than one average.
- Consistency. The share of windows in which a fund beat its benchmark is a common test.
- Horizon and risk. Longer windows usually show a narrower spread than shorter ones, though the worst long windows can still be poor, as sequence-of-returns risk shows.
Pitfalls
- Overlapping windows are not independent. Monthly three-year windows share most of their data, so 120 windows are not 120 separate observations.
- The history is finite. The worst window on record is not the worst possible.
- Same data, different look. Rolling returns do not add information, only a better view of the drawdowns and recoveries already in the record.
