Learn Lesson 2 of 6 7 minutes
Can a fund beat its category most of the time and still be behind?
Counting how many times a fund won tells you nothing about what it won. Two funds in this database make the point better than any argument: one beat its category in most of its three-year stretches and is behind across all of them, and one did the opposite.
Built on P4 — rolling three-year windows, counted and averaged.
Here is a sentence you will see on fund pages, in advertisements and in newspaper tables: this fund has beaten its category in most rolling three-year periods. It sounds like a strong claim. It is a claim about frequency, and frequency is half of an answer.
The other half is magnitude. A fund can win nine windows by a point each and lose one by twenty, and the count will call it a winner while the arithmetic calls it a loser. Both statements are true. Only one of them describes what happened to the money.
The identity, which is the whole lesson
Take every three-year stretch a fund has lived through, one starting each month, and for each one write down how far ahead of or behind its category median it finished. Then:
average margin = (share of windows won × average win) + (share lost × average loss)
That is not an approximation. It is exactly true, because a window that ties contributes nothing to either side. It means the two halves multiply, and it means a high share of windows won can be entirely paid back by what the losing windows cost.
A fund that won most of its windows and is still behind
UTI - Transportation and Logistics Fund (Equity Scheme - Sectoral/ Thematic) has 109 rolling three-year windows in this build, the first ending Sep 2017.
How often
59%of its three-year windows beat the category median — a clear majority
How much
−1.14percentage points a year, averaged across every one of those windows — behind
The decomposition says why. It was ahead by 4.61 percentage points a year in the 64 windows it won, and behind by 9.30 percentage points a year in the 45 it lost. More wins than losses, and the losses were the bigger events — so the average across all 109 of them lands at −1.14% points a year.
The single worst window ended Mar 2020, 14.06 percentage points a year behind its category; the best ended Aug 2024, 11.76 percentage points a year ahead.
And one more reading, which a single headline would hide: the typical window was +2.14% points a year, a long way from the average of −1.14% points. When the median sits that far from the mean, a small number of windows are carrying the whole figure — and a reader who is shown only one of the two is being shown the fund backwards.
Both numbers, on the fund's own page: UTI - Transportation and Logistics Fund
And a fund the counting makes look worse than it was
The error runs both ways, which is why this lesson is about arithmetic rather than about any particular fund. Axis Small Cap Fund lost more of its windows than it won — and came out ahead across all of them.
How often
42%of its windows beat the category median — a minority
How much
+1.36percentage points a year, averaged across every window — ahead
Ahead by 5.96 percentage points a year in the 46 windows it won, behind by only 2.00 percentage points in the 63 it lost. Fewer wins, bigger wins. Count the windows and you rank this fund below the one above; do the arithmetic and the order reverses.
Check it: Axis Small Cap Fund · every fund, ordered by average margin · the same funds, ordered by windows won
Those last two orderings are worth putting side by side. They are built from the same windows of the same funds and they do not agree, which is the clearest possible demonstration that "consistent" is not one measurement.
How often does this happen?
In this build
36 of 304equity funds with a full three-year record — 11.8% — have their count and their average pointing in different directions
12 won most of their windows and are behind across all of them; 24 lost most and are ahead. It is not a corner case invented to justify a metric — it changes the reading of roughly one fund in ten.
Counted across every equity fund this build will answer for, at the moment this page was rendered.
What this measure still cannot do
It cannot forecast. We tested the share of windows won directly, over ten years and 646 equity funds: as a predictor of the next twelve months it scored a top-minus-bottom quintile spread of −0.68% points a year (t = -0.59, p = 0.557), with an error interval running from −2.63 to +1.81 — indistinguishable from nothing, and pointing the wrong way. There is no reason at all to assume that weighting a statistic that does not predict by its payoff makes it predict.
So read both halves as what they are: a better description of a record that has already happened, on a named fund, over a stated sample.
The argument against this lesson
Rolling windows overlap almost completely — consecutive three-year windows share thirty-five of their thirty-six months — so a hundred of them are nowhere near a hundred independent observations. The fund above has 109 rolling windows and only 4 non-overlapping ones, and that second number is the sample size to keep in mind. The comparison is also only as good as the category behind it: where an AMFI category is a grab bag rather than a peer group, a large average shortfall is a statement about the comparator, not about the fund. And a fund that changed manager or mandate carries windows in its average that belong to somebody else's tenure. None of this makes the average wrong; it makes it a description with a stated sample, which is all it ever claimed to be.
Everything above is a description of what has already happened, measured from what fund houses, AMFI and the exchanges publish. It is not advice, not a recommendation to buy, sell or hold anything, and not a forecast — neither Anveshan nor Lineage Money is a SEBI-registered investment adviser or research analyst.