Rolling Returns vs Trailing Returns: How Start Dates Change the Result

Rolling returns vs trailing returns compares a set of equal-length historical windows with one window ending on a stated date. A trailing three-year return gives you one result. Three-year rolling returns repeat the calculation as you move the start and end dates together. You can then see whether the latest result resembles earlier periods or sits near an extreme.

A strong trailing return can coexist with weak earlier rolling returns. Before calling an allocation consistent, check which dates you have examined and how much history those windows share.

Rolling returns vs trailing returns at a glance

“Trailing” describes the position of the window relative to an as-of date. “Rolling” describes repeating the calculation across dates. Each observation in a rolling-return series is itself a trailing return at that observation's end date.

The window length and the step between windows are separate choices. You could calculate three-year returns at each month-end or at each year-end. Those choices produce different numbers of observations and can produce different extremes.

Morningstar's explanation of rolling returns describes examining overlapping periods of equal length alongside a single trailing result. The calculation changes the historical question; it does not establish which investment will perform best next.

How to calculate rolling returns

For a positive value series with no external deposits or withdrawals, calculate each window's cumulative change as:

Cumulative return = ending value ÷ starting value − 1

To express a multi-year window as a compound annual rate:

Annualized return = (ending value ÷ starting value)^(1 ÷ years) − 1

Multiply the decimal result by 100 to express a percentage. For an exact three-year window, use three as the exponent's denominator. For a 36-month window defined by monthly observations, the exponent is 12 ÷ 36.

Rolling return and CAGR describe different parts of the calculation. Rolling tells you to repeat a window. CAGR tells you how to annualize the change inside each multi-year window. You can report rolling cumulative returns or rolling CAGRs, provided you label the figures.

FINRA explains why annualizing a return requires compounding. Dividing a three-year cumulative gain by three gives a different number. The CAGR calculator guide works through that distinction.

In a spreadsheet, put the dates in one column and the values in another. Divide each ending value by the value one full window earlier, apply the formula, and advance both endpoints by the chosen step. Keep the date pair beside each result so you can audit it.

A three-year rolling-return example

Suppose you have these hypothetical year-end index levels. They represent an invented price-only series with no cash flows, distributions, fees, or taxes. They are not stock prices or measured market returns.

Calculate every complete three-year window, advancing one year at a time:

As of the hypothetical 2025 year-end, the trailing three-year result is 21.64% a year: (162 ÷ 90)^(1 ÷ 3) − 1. That window begins at the series' lowest listed level. Starting at the 2021 year-end instead produces a three-year annualized result of 2.33%, despite ending one year earlier.

Across the four windows, the median annualized result is about 2.63%. The latest result sits at the top of this small sample. Reporting both observations gives the reader more context than presenting 21.64% alone.

The median is a summary of four hypothetical windows. It is not the compound return earned by an account over the full six years. You also cannot add the rolling returns together: doing so counts shared years more than once.

Overlapping periods share the same market history

The first two windows above cover year-end 2019 to year-end 2022 and year-end 2020 to year-end 2023. They share two years of price movement. A large decline inside those shared years affects both results.

Using monthly steps gives you more observations, but adjacent windows still share most of their history. Two five-year windows ending one month apart share 59 of their 60 monthly return intervals. A decade of complete monthly data contains 120 monthly return intervals and supports 61 five-year windows when you advance one month at a time. It does not provide 61 independent five-year experiences.

Three of the four illustrative windows above have positive returns. Calling that a 75% historical positive-window frequency describes the sample. Calling it a 75% chance of making money over the next three years would claim something the calculation cannot establish.

Record the full history examined, window length, step size, and number of complete windows. Include the worst result and the median alongside the latest result. A daily, monthly, or annual step can change the distribution, so compare studies that use the same settings.

Keep the return basis and benchmark consistent

Moving the dates does not fix mismatched data. A rolling price-return series excludes the distributions that a fund's total-return report may reinvest. Read price return vs total return before comparing them.

For each pair of investments, use the same window endpoints, currency, return definition, and annualization method. Use the same complete windows for both. A newer ETF cannot supply ten years of fund history before its launch; an older index history describes a different series.

Portfolio studies need another definition: what happens to the holdings and weights inside each window? Restarting a hypothetical allocation at each window's beginning differs from measuring rolling changes along one portfolio path whose weights have drifted since an earlier date. Label the method, including any cash flows or rebalancing assumptions, before comparing results.

You can also compare each window with a relevant benchmark over those exact dates. The benchmark guide explains how to choose a reference that resembles the exposure. A large-cap U.S. benchmark answers a narrower question than a claim about every kind of portfolio.

The SEC's performance-claims bulletin cautions against selecting only favorable periods and stresses the importance of methodology and benchmark choice. Including several windows helps expose date selection, but you still need to inspect the underlying price path. A positive three-year endpoint can conceal a severe decline during those three years.

Read an Equity Replay window with these limits in mind

Equity Replay lets you compare stocks and ETFs or compare portfolio allocations across its available historical ranges. Read the actual shared start date, end date, and data-through date. Missing history for one holding can shorten the usable window.

Selecting one month and then five years changes the window length. That is a multi-horizon comparison. A rolling-return study holds the length constant and moves both endpoints. Equity Replay's range views do not generate the rolling-return table described above; use a separate spreadsheet or analysis tool with a consistent dated series for that calculation.

For a replay, Equity Replay applies the entered weights once at the start of the selected historical window and combines normalized closing-price movements. It excludes dividends, fees, taxes, deposits, withdrawals, interim trades, and rebalancing. The replay methodology explains the calculation.

Treat the displayed historical change as one dated result under those assumptions. If a headline or factsheet calls a result “three-year performance,” check whether it means cumulative change or an annualized figure, and whether it describes only the latest window.

Equity Replay is an educational historical simulator, not investment advice, a broker, or a record of trades. Neither a single replay nor a rolling-return study predicts a future result or establishes a personal allocation or trading decision.