This term
What it means
A single five-year return gives one starting date and one ending date, while rolling analysis repeats the calculation for other five-year intervals. It can show whether a favourable headline period was typical of the available history or unusually strong.
Window length and step size are separate settings, so a five-year window stepped monthly shares most of its observations with the previous window, whereas a window stepped every five years avoids overlap but supplies fewer observations. Returns can be cumulative over each window or annualised to a compounded yearly rate, so specify the convention instead of treating rolling and annualised as synonyms.
The window repeats; annualisation describes how each result is expressed. Total-return data include reinvested distributions while price-only data omit them, and mixing the two can create a performance gap caused by methodology rather than skill.
The 2024 CFA Institute Research Foundation brief on investment horizons notes that its multiyear historical calculations use overlapping periods and limits some analysis to annual frequency to reduce overlap. This shows why a large count of rolling observations is not the same as many independent market histories.
Overlapping windows can make poor returns cluster on a chart because the same weak year enters several consecutive calculations, which should not be described as several separate crises. A rolling chart can display a range, median or frequency of negative historical windows.
The median is not the worst result, and a count above zero does not give a guaranteed probability for a future investor. The first observation also requires a complete window, so a five-year series cannot be calculated from only three years of data without changing the definition.
Fees, currency and cash-flow methods still matter, because an investor adding or withdrawing money may experience a different personal result from the published fund series. Rolling a fund's time-weighted data does not automatically give the investor's money-weighted outcome.
A fund with a shorter history may also show fewer weak-market windows than an older fund, so a favourable comparison can reflect the sample rather than a superior strategy. Managers can use rolling results to test how sensitive a conclusion is to the starting date.
A long window can hide severe interim losses even when its final annualised return is positive, so pair the range of outcomes with a look at drawdowns.
In practice
Real-world examples
Example 1
A fictional report calculates each three-year window at annual steps. The windows use years 1-3,2-4 and 3-5. Years 2 and 3 appear more than once, so the three results are not independent samples.
Example 2
An analyst compares a price-only rolling index with a fund's reinvested total return. The team aligns income treatment before judging the difference. Repeated windows do not fix inconsistent inputs.
Example 3
A fund advertises its strongest ten-year result. A manager requests all available ten-year windows under one convention and checks the weakest periods too. The fuller history remains evidence, not a forecast.
Formula
For a total-return value series with consistent assumptions, window cumulative return = ending value / starting value -1
Note
An n-year annualised window return = (ending value / starting value)^(1/n) -1
A fictional three-year window rising from 100 to 133.1 has 33.1% cumulative return and 10% annualised return.
The next window's starting and ending values must be used separately; do not average those two values to construct its return.
With five annual returns and a three-year window, there are:
5 -3 +1 = 3
Complete windows at one-year steps.
With those annual data, a monthly step is not available unless monthly returns exist.
These formulas assume an appropriate return series and cash-flow treatment.
They do not turn overlapping historical outcomes into independent forecasts.
Case study
Seen in the real world
Fictional case study: Willow Fund appears strong over the latest five years. Its sales chart repeats five-year returns monthly and presents the large observation count as evidence of highly reliable future performance. The review keeps the repeated-window analysis but removes that claim.
It shows the range of historical outcomes, explains shared years and compares a benchmark using the same data convention. Management also examines interim drawdowns. The exercise broadens the historical view without pretending that a sliding window removes uncertainty or makes past returns predictive.
Watch out
Common mistakes
- Calling rolling returns an automatic average or smoothing guarantee. Each window is a separate calculation.
- Treating overlapping windows as independent trials. Shared observations affect interpretation.
- Comparing different window lengths, currencies or income treatment. Align the methods first.
Questions
People also ask
Must rolling returns be annualised?
No. Each window can be expressed cumulatively or annualised, provided the method is clear.
Do more windows mean more independent evidence?
Not necessarily. Closely stepped windows can share almost all their data.
Does a positive historical range guarantee future gains?
No. The sample describes past windows and cannot eliminate future investment risk.