The Three-Year Number That Changes Every Day

Last Tuesday a friend forwarded me two factsheets of the same fund. Same scheme name. Same fund manager. Same ISIN. One was the December 2021 edition. The other was December 2024. The three-year CAGR on the first sheet read 18.3 percent. The three-year CAGR on the second read 11.4 percent. He wanted to know what went wrong.

Nothing went wrong. The fund did not change. The window changed. A three-year return is not a property of the fund. It is a property of the three years you happened to measure.

The factsheet picked one window. There were dozens. Most of them told a different story.

Same index, eleven different answers

I pulled the Nifty 50 TRI calendar-year returns from 2013 to 2025 and ran every possible three-year window. Same index, same rule, same market. Eleven windows. The CAGR ranged from 8.8 percent to 18.3 percent. A spread of 9.5 percentage points on the same asset.

The best window, 2019 to 2021, caught the post-COVID rally. The worst, 2015 to 2017, opened on a flat year and closed before the real move. Both are real. Both are correct. Neither is the whole picture. The factsheet printed one of these bars and called it the three-year return. The other ten did not make the page.

The three months that rewrote the number

Here is where it gets uncomfortable. I took the Nifty 50 TRI level at the end of December 2019 and again at the end of March 2020. Three months apart. Same index.

The three-year CAGR ending December 2019 was approximately 18.1 percent. The three-year CAGR ending March 2020, after COVID pulled the index down roughly 28 percent in one quarter, was approximately 3.8 percent. A gap of over 14 percentage points. Same fund. Same three-year horizon. Three months between the two measurement dates.

Source . NSE . Nifty 50 TRI . illustrative fund matching index exactly

The fund that "delivered 18.3 percent" in the December 2021 factsheet and "only 11.4 percent" in the December 2024 factsheet is the same portfolio. The market it operated in changed, but the bigger shift was just the starting line. The 2021 factsheet's three-year window opened in January 2019, before a pandemic and a rally. The 2024 factsheet's window opened in January 2022, after the rally was done. One window caught the up-leg. The other missed it.

What rolling returns actually show you

A rolling return takes a fixed horizon, say three years, and slides it forward one day (or one month) at a time. Instead of a single number you get a line. Instead of one answer you get every answer. The range tells you what was possible. The distribution tells you what was typical.

When someone tells you a fund delivered 14 percent over three years, a rolling return chart asks: 14 percent was one window. What were the other two hundred? Was 14 the middle, the top, or the bottom of the range? A number near the top of its own rolling range is a fund that happened to catch a good window. A number near the bottom might be a perfectly fine fund in a bad stretch.

Most fund comparison apps today offer rolling return charts. Advisorkhoj, PrimeInvestor, Value Research, and the more recent AMFI updates all let you toggle from trailing returns to rolling windows. The rolling chart is not hidden. It is just not the first thing the comparison table shows, and most investors never scroll past the first number.

The rupee version of the gap

Put it in money because percentages are slippery. Ten lakh invested at the best three-year window on this index, the 18.3 percent CAGR from 2019 to 2021, grows to roughly Rs 16.55 lakh. The same ten lakh at the worst three-year window, the 8.8 percent from 2015 to 2017, grows to roughly Rs 12.88 lakh. A gap of Rs 3.67 lakh on the same market, the same horizon, the same starting capital. The only variable is which three years the clock happened to cover.

Nobody chose the window. The factsheet printed the one that ended on the date it was published. The investor read a single line and formed a belief about the fund. That belief might have been completely different sixty days earlier or sixty days later.

What the factsheet does not lie about

This is not a scandal. The factsheet is not falsified. SEBI mandates performance disclosure over one-year, three-year, five-year, and since-inception periods, compared against the benchmark TRI. The number is real, audited, and correctly computed. Starting April 2025 SEBI also requires daily disclosure of the Information Ratio, a risk-adjusted return measure, across multiple horizons.

The limitation is structural, not dishonest. A point-to-point number collapses an entire experience into one figure. It tells you how the journey ended but not how rough or smooth the road was along the way. Two funds with the same three-year return could have arrived there on entirely different paths, one grinding up steadily and the other crashing and recovering.

The five-year rolling returns on the same Nifty 50 TRI data ranged from 8.6 percent to 17.2 percent. A longer horizon narrowed the spread from 9.5 percentage points to 8.7, which is better but not small. Even at ten years, the range does not collapse to a single point. Markets are not that kind.

Where this breaks, and the honest half

Rolling returns are better but not perfect. They still depend on the length of history available. A fund that has existed for four years cannot show a meaningful five-year rolling chart. And a fund with a short history operating in a benign market has no rolling window that includes a real drawdown, so the chart looks smoother than it should.

Point-to-point returns are also exactly what you need when comparing two funds over the same specific period. If you want to know which fund did better from January 2022 to December 2024, the trailing return is the right tool. The problem only arises when you treat that one window as the fund's identity rather than one frame from a longer film.

The factsheet number is fine for what it is. It is a timestamp, not a verdict. The mistake is reading a single frame as the entire film and then picking a fund because you liked that frame. The film has many frames, and most of them are playing at a different speed.

Three things I keep on the checklist

Ask for the range before the number. When someone tells you a fund did 14 percent over three years, the first question is what was the best and worst three-year window in its history. If 14 is the ceiling, you are looking at a fund near its peak reading. If 14 is the floor, you are looking at one that held up even in bad stretches. The range is the character. The number is just one sample.

Compare on the same window. Two funds measured on different factsheet dates are not comparable, even if both say three-year. A December 2021 factsheet and a December 2024 factsheet are showing two completely different markets. Pull both funds' returns over the identical calendar period before you rank them. Every fund analysis site lets you do this. The comparison table at RupeeCase lines them up side by side.

Check the worst window, not the best. A fund's worst rolling three-year return tells you what it felt like to hold through its roughest stretch. If that number is a loss, you now know a loss was possible on a three-year hold. If you cannot stomach that floor, the fund is wrong for you regardless of how high the ceiling reads today.

A factsheet is a photograph. The rolling return is the time-lapse. The photograph is not lying, but anyone who buys a house from a single photograph deserves the surprise that comes with the walkthrough.