What is the difference between rolling returns and point-to-point returns?

A point-to-point return measures one specific start date to one specific end date. A rolling return computes the same holding period starting on every date in the sample, producing a distribution instead of a single observation — so you can see the best, worst and median experience rather than the one the calendar happened to select.

Why does a fund's trailing return change without the fund doing anything?

Because the window moves. A trailing figure improving sharply often reflects a bad period dropping out of the back of the window rather than a good period being added at the front.

What is a good rolling-return spread?

There is no universal threshold, but the spread tells you how much of the headline was timing. Two funds with the same median can differ entirely: tightly clustered outcomes were delivered whenever you started, while a wide spread means the median required a fortunate entry date.

Lens · 16 September 2026

Rolling returns: the number that survives a different start date

A point-to-point return is one draw from a distribution, and the draw is chosen by the start date. What rolling returns measure instead, how to read the spread, and why the median matters more than the headline.

3 min read

Every point-to-point return is a single observation: what happened between one specific day and another specific day. Change either date and the number changes, sometimes by a great deal. That is not a flaw in the calculation. It is a flaw in reporting one number as though it were the answer.

The start date is doing the work

A five-year return quoted today begins five years ago, which is an arbitrary point chosen by the calendar rather than by anything about the investment. If that date happened to sit near a market low, the figure flatters. Near a high, it punishes. The fund did not change; the window did.

What rolling returns do instead

A rolling return computes the same holding period over and over, starting on every date in the sample, and looks at the whole set of results. Instead of one three-year number you get every three-year number that was available, which turns a single observation into a distribution.

for each date d: return(d, d + period) — then describe the set

The output is not a number but a spread: a minimum, a median, a maximum, and how often the result fell below whatever threshold you care about.

Illustrative shape, not a real fund. The point is the variation, not the values — a single point-to-point figure reports one of these columns and says nothing about the rest.

How to read the spread

Statistic	What it answers
Median	The middle outcome. Closer to a typical experience than any single trailing figure.
Minimum	The worst entry date in the sample — the case an investor actually has to sit through.
Spread (max − min)	How much of the headline number was timing rather than the strategy.
Share below a threshold	How often the holding period failed to clear whatever return you needed. Frequency, not just magnitude.

What each statistic tells you, and what it does not

The limits

Rolling windows overlap heavily, so the observations are not independent — a single bad year appears in many windows and the count of "periods" overstates how much evidence you have.
The sample still only contains the market regimes that actually occurred. A fund with no history through a drawdown has no rolling window that shows one.
A longer holding period produces fewer distinct windows, so a ten-year rolling return over a twelve-year history is describing almost nothing.

None of that makes the exercise less useful than the alternative. A trailing return has every one of these limitations too, plus the additional one of being a single draw — and it is the only one of the two that presents itself as definitive.

Common questions

What is the difference between rolling returns and point-to-point returns?: A point-to-point return measures one specific start date to one specific end date. A rolling return computes the same holding period starting on every date in the sample, producing a distribution instead of a single observation — so you can see the best, worst and median experience rather than the one the calendar happened to select.
Why does a fund's trailing return change without the fund doing anything?: Because the window moves. A trailing figure improving sharply often reflects a bad period dropping out of the back of the window rather than a good period being added at the front.
What is a good rolling-return spread?: There is no universal threshold, but the spread tells you how much of the headline was timing. Two funds with the same median can differ entirely: tightly clustered outcomes were delivered whenever you started, while a wide spread means the median required a fortunate entry date.

XIRR vs CAGR: which return number is telling you the truth?

CAGR and XIRR answer different questions, and using the wrong one can overstate a SIP return by several percentage points. The formulas, a worked example, and when each applies.

3 min read

Ten funds, one bet: the concentration behind a diversified portfolio

Holding more funds spreads the fee, not necessarily the risk. How to compute what you actually own at the stock level, why the top holdings dominate, and what a genuine reduction in concentration looks like.

2 min read

All insights.

Browse insights Talk to us

Or write to us at [email protected]