How should I select a mutual fund, and are star ratings any use?

Start with the role, not the fund

"Which fund should I buy" is unanswerable until you have decided what it is for. A fund is not good or bad in the abstract — it is suited or unsuited to a job. Settle the category first: how long the money is for, how much of a fall you can hold through, and whether this is the stable part of your portfolio or the aggressive part. The category determines most of your return and all of your volatility. The specific fund within it matters far less than people think.

Then shortlist, then judge

Within one category, cut the list on things that are facts rather than opinions — a long enough track record to measure, a mandate that matches the job, and cost. Then judge what is left on:

• Rolling returns — the median, the worst window, and how often a window ended negative. Consistency is visible here and nowhere else.
• Risk-adjusted return — Sharpe and Sortino over the same window, so you are not paying for return with volatility you did not notice.
• Maximum drawdown and recovery — the fall you would have had to sit through, with the dates.
• Benchmark-relative behaviour — alpha, and up/down capture against the fund's own TRI. A fund that captures 100% of the upside and 73% of the downside is doing something specific, and you can see what.
• Cost — from the factsheet. The one input that is certain.

Category by category, the thing that actually differs

• Large cap — the honest question is whether the fund beats its index after costs at all. Check alpha against the NIFTY 100 TRI over several windows, and consider an index fund as the benchmark for the decision, not just for the chart.
• Mid and small cap — drawdown is the whole decision. These funds have fallen far further than large caps and taken longer to recover. Look at the worst rolling window before the best one.
• Flexi cap — the mandate is wide, so the fund's actual market-cap split from its disclosed portfolio matters more than its label. Two flexi-cap funds can be running completely different books.
• ELSS — a three-year lock-in means a bad entry cannot be undone, so consistency outranks peak return.
• Index funds — do not judge on return, judge on tracking error and tracking difference. The fund's job is to match the index, so the question is how closely and at what cost.
• Debt — credit quality and maturity, not past return. A debt fund's past return mostly records the interest-rate cycle it lived through.

Are five-star funds better?

A star rating is a backward-looking summary of risk-adjusted past returns within a category. It is not wrong, but two things make it less useful than it looks: it compresses several different measures into one number so you cannot see which one is driving it, and ratings change — today's five-star fund was often a three-star fund a few years ago and will be again.

You are better off reading the inputs directly. That is also why there are no ratings here: MF Analyser is not a SEBI-registered Investment Adviser or Research Analyst, and a star is a recommendation with the working hidden.

NFOs

A new fund offer has no track record, by definition. Everything in this page is unmeasurable for an NFO — no rolling returns, no drawdown, no Sharpe, no disclosed portfolio. You are buying a strategy document and a fund house's reputation. That can be a reasonable thing to do; it is not research, and the "₹10 NAV is cheap" argument is simply false — a ₹10 NAV buys the same assets per rupee as a ₹2,000 one.

When to replace a fund you already own

Three things that are real signals, and one that is not. Real: the mandate changed, so the fund no longer does the job you bought it for; it has lagged its own benchmark across most windows rather than most recently; or you now hold something that overlaps it heavily, so it is adding cost without adding exposure.

Not a signal: one bad year. If a fund's history shows one year in six ending negative, a negative year is the fund behaving normally. Selling on it converts ordinary volatility into a permanent loss — and the switch itself may cost you exit load and capital gains tax.