Where does MF Analyser get its data, how are the figures calculated, and how far should I trust them?

Two sources, both public

• NAV — AMFI's daily publication, for every scheme, back to each scheme's inception where AMFI carries it.
• Portfolios — the monthly disclosures AMCs are required to file: every security, its ISIN, its quantity, its value, its share of net assets.
• Benchmarks — 32 Total Return indices, daily values from 2 January 2012, from the index providers.

Nothing is estimated, modelled, crowd-sourced or licensed from a data vendor whose method we cannot see. If a figure is here, it is arithmetic on one of those three, and the arithmetic is published.

How the measures are computed

Stated precisely enough to be checked:

• Volatility, Sharpe, Sortino — from 36 month-end returns, annualised by √12, with a risk-free rate of 6.5%. Sortino divides by downside deviation only.
• Drawdown — on the full daily NAV series, reported with the peak date, the trough date, and the date the old peak was regained.
• Rolling returns — every overlapping window of the stated length, stepped daily, reported as worst, median, best, and the share of windows that ended negative.
• Beta, alpha, capture — against the fund's own category TRI, never against an index fund, over a common 36-month window.
• CAGR — point-to-point between two NAV dates. XIRR — for cash flows on actual dates.

A measure a fund has too little history for is shown as "—", never as zero and never computed from too few points. Null is not nought, and treating it as nought is how a league table ends up topped by a fund that launched last month.

Every formula, so you can reproduce it

Risk measures use a 36-month window of month-end returns with a 6.5% risk-free rate:

Beta, Alpha and Treynor are blank where no valid benchmark history exists — never estimated. Yes, the arithmetic can be reproduced in a spreadsheet from the same AMFI NAV; that is the point of stating it.

One engine, so nothing disagrees

The fund pages, the screener, the rankings, the compare page and the portfolio analysis all read the same computed row. This is deliberate, and it is the lesson from getting it wrong: the site once had the same return computed in three places from three caches, and they drifted apart, so a fund showed one 5-year return on its own page and a different one in the rankings. There is now a single source, and any figure that disagrees with another figure on this site is a bug worth reporting.

What MF Analyser does not have

A research tool that hides its gaps is less useful than one that states them:

• No expense ratio, AUM or portfolio turnover. Not in our data. Take them from the AMC's factsheet.
• Holdings are not universal. 49 fund houses have filed portfolios — most of the industry by assets, not all of it.
• Holdings are a month-end photograph, published in arrears. Not what the fund holds today.
• No equity benchmark on debt funds, so no beta or alpha for them. That is a decision, not an omission.
• No forecasts, no ratings, no recommendations. MF Analyser is not a SEBI-registered Investment Adviser or Research Analyst.

How to check us rather than trust us

Every figure carries its as-of date and its window, so it can be verified against the same AMFI NAV we used. If a number here disagrees with a number elsewhere, the usual causes are a different as-of date, the Regular plan against the Direct plan, or a benchmark that is a price index rather than a TRI — and because our date, plan and benchmark are all stated, you can tell which.

That is the only durable answer to "is it accurate": not a promise, but enough stated method that you do not need one.