How do I measure the risk of a mutual fund, and which measures actually matter?

One word, three questions

"Risk" in a mutual fund is not one number, and the single biggest mistake in fund research is treating it as one. There are three questions, and a fund can look fine on one while failing another:

1. How much does it move? Standard deviation — the annualised spread of its monthly returns. A higher number means a bumpier ride, nothing more and nothing less.
2. How bad has it actually got? Maximum drawdown — the deepest peak-to-trough fall it has ever put an investor through. This is the one people feel, because it is the number at which they sell.
3. Was the return worth it? Sharpe and Sortino — return above the risk-free rate, divided by the risk taken to get it.

Four real funds

Direct plan, Growth option, computed on AMFI NAV history to 1 October 2026. These are the same rows the fund pages print, so nothing here can disagree with the rest of the site:

Return per unit of risk — Sharpe

The two flexi-cap funds earned roughly 0.58 and 0.57 of excess return per unit of volatility. The two large-cap funds earned 0.21 and 0.20 — less than half as much, for almost identical volatility. That gap is the entire argument for looking past the return column: these four funds do not differ much in how much they bounce, they differ in what they paid you for it.

A Sharpe ratio is only comparable between funds measured over the same window with the same risk-free rate. Ours uses 6.5% and 36 month-end returns for every fund, which is why these four can sit in one table at all.

The worst it has been — drawdown

HDFC Flexi Cap's deepest fall was −41.84%, from 4 July 2019 to 23 March 2020. It regained the old peak on 9 December 2020 — 8.6 months under water after the bottom. Parag Parikh fell least, −31.20%, and recovered in 3.4 months.

Drawdown is the honest risk number because it is the only one denominated in the thing that actually makes people sell. "Standard deviation 13%" is abstract. "This fund has been 42% down" is a question you can answer about yourself before you invest rather than during a crash.

The measure most people have never looked at

Up capture and down capture split performance into the two halves that matter. Parag Parikh captured 72.8% of its benchmark's rises and only 43.7% of its falls. HDFC Flexi Cap captured 100.2% of the upside and 73.1% of the downside — it kept pace going up while taking about three-quarters of the pain coming down.

Now compare the large-cap pair: roughly 99–102% up and 90% down. They follow the index closely in both directions. That is not a flaw — it is what a large-cap mandate does — but it tells you most of what you will get from them, and it is invisible in a star rating.

High risk, low risk — how to actually tell

A fund is riskier than another if it swings more (standard deviation), falls further (drawdown), and follows the market more completely on the way down (down capture, beta). Those three usually agree. When they disagree, the disagreement is the finding — a fund with moderate volatility and a very deep drawdown is one that is calm until it isn't.

And one caution worth more than any of the ratios: all of this is history. A fund's standard deviation is what it did, not a promise about what it will do. Measures describe a manager's past behaviour. They do not predict.

Where the numbers come from

Every figure above is arithmetic on AMFI's published NAV. One engine computes them for every scheme — 36 month-end returns annualised by √12, a 6.5% risk-free rate, drawdown on the full daily series, and beta and capture against the fund's own category TRI benchmark rather than an index fund. A window a fund is too young for shows as "—" rather than a number computed from four points.

Coverage: 8,564 schemes with full measures, 2,266 of them with a benchmark. Debt, gilt, liquid and arbitrage funds are deliberately given no equity benchmark, so they carry no beta or alpha — a bond fund's beta against an equity index is noise, and we would rather print nothing than print that.