Chapter 3 · Introduction

Chapter 3 — The Statistics of Returns

Returns are random: how to measure how much they swing, how badly they can fall, how they move together, and how much return each unit of risk really bought — with simulated market data you can regenerate.

Chapters 1 and 2 treated returns as if they were known in advance: an FD at 7%, a SIP at an assumed 12%. A share or an equity fund does not work like that. Its NAV moves every trading day, some days by a fraction of a percent, a few days by several percent, and over a few months it can fall a long way before it recovers. Two funds that both "returned 12% a year on average" can give their investors completely different experiences — one a gentle climb, the other a ride that wipes out a third of the money before coming back. If you only look at the average, you cannot tell them apart; and if you cannot tell them apart, you cannot decide how much of either you can live with.

This chapter is about describing that randomness with numbers. Statistics gives the tools: the standard deviation of returns becomes volatility and is scaled from days to years with a square-root rule; skewness and kurtosis tell us whether the shape of the return histogram is the familiar bell curve or something with "fat tails"; the running maximum of a NAV series gives the drawdown, the investor's real experience of risk. Then two series at once: covariance and correlation, the regression line of a stock on the index (whose slope is beta), and ratios — Sharpe, Sortino, Treynor, information ratio — that put return and risk on one scale so that funds can be compared fairly. Finally probability turns these descriptions into a model: a random walk for prices, simulated thousands of times by Monte Carlo, to answer "what is the chance my SIP reaches the goal?". Every data set on these pages is simulated in your browser with a seeded random-number generator, so you can regenerate it and see which conclusions survive.

Daily returns→Volatility & $\sqrt{\text{time}}$→Skewness & fat tails→Drawdowns→Volatility drag→Correlation & beta→Sharpe, Sortino, IR→Random walks & Monte Carlo
🎯 By the end of this chapter you can…
  • compute daily and monthly returns, their mean and volatility, and annualise them correctly with the 252-day convention and the $\sqrt{\text{time}}$ rule;
  • read a return histogram, compute skewness and kurtosis, and explain why the normal curve underestimates how often large moves happen ("fat tails");
  • find the maximum drawdown and recovery time of a NAV series, and explain why a 50% fall needs a 100% gain to recover;
  • explain volatility drag — why the compound growth rate is roughly the average return minus $\sigma^2/2$;
  • compute covariance, correlation, beta, alpha and $R^2$ of a stock or fund against an index, and split its risk into market and specific parts;
  • compare funds with Sharpe, Sortino, Treynor and information ratios, and judge whether a difference is larger than the noise;
  • simulate random-walk price paths and use Monte Carlo to estimate the probability of reaching a target or suffering a loss.
Teaser — Same Average Return, Three Different Journeys

Three simulated funds receive the same random monthly "news" and have the same average monthly return; they differ only in volatility: low (8% a year), medium (18%) and high (30%). The dashed line is what the average return would give with no ups and downs at all. Each path starts with ₹1,00,000 and runs for 15 years. Press Play to draw a new simulated market every few seconds and watch which fund ends highest — and how deep each one falls on the way. (Simulated data, not real funds.)

no-volatility path ends at—
low-vol fund: end · worst fall—
medium-vol fund: end · worst fall—
high-vol fund: end · worst fall—
simulated market no.—

Where this shows up in real life

Equity mutual-fund factsheets in India commonly report a fund's standard deviation, beta and Sharpe ratio, and many also report its Sortino ratio, $R^2$ against the benchmark, and maximum drawdown; the riskometer on the same page is a regulator-mandated attempt to summarise risk in one label. Fund rating services rank funds by risk-adjusted return. Portfolio managers and research analysts estimate betas by regression to hedge with index futures (§6.1) and to apply CAPM (§4.2); risk teams at banks and brokers use volatilities and correlations to set margins and compute value at risk (§7.4); India VIX, published by NSE, is a market-implied estimate of the Nifty's annualised volatility (§6.4); and a trader's backtest is judged by its Sharpe ratio and maximum drawdown (§7.3). The certification examinations for mutual-fund distributors and investment advisers test exactly these measures, and "why is the average return not the compound return?" and "what does a beta of 1.3 mean?" are standard interview questions. Most of all, an investor who knows in advance how deep a drawdown a fund can have is far less likely to sell at the bottom.

Sections in this chapter

✅ Before you start

From Chapter 1 you need compound growth, CAGR and the difference between the arithmetic and geometric mean, and log returns (all in §1.3 Measuring Returns); Chapter 2 supplies the instruments — index funds, NAVs and the Nifty 50 (§2.1, §2.2). From school statistics you need the mean, the standard deviation and the idea of a histogram; we recall the normal distribution and its "68–95–99.7" rule when we use it, and introduce everything else (skewness, kurtosis, covariance, the regression line) from scratch. A spreadsheet (Excel, LibreOffice or Google Sheets) with its AVERAGE, STDEV.S, SKEW, KURT, CORREL and SLOPE functions, or Python with numpy, is enough for every computation. No market data are needed: each widget generates its own simulated prices, and the "🛠️ Do it yourself" projects show how to repeat the analysis on a fund's published NAV history.