Technical Analysis · Chapter 7 of 12
Volatility: ATR, Bollinger, and the family that actually earns its keep
The least glamorous indicators and the most useful, because position sizing is where returns are actually determined.
If you take one practical thing from this book, take this chapter. Volatility measures do not predict direction and do not claim to. What they do is tell you how far price normally moves, and that single number is what makes stops, targets and position sizes non-arbitrary. It is also the family with the strongest empirical foundation, because volatility clustering — volatile periods following volatile periods — is one of the most robust regularities in all of finance.
ATR, and why it is not just the range
True Range for a bar is the greatest of: high − low; |high − previous close|; |low − previous close|. ATR is a Wilder-smoothed average of that over n periods, usually 14.
The reason for the two extra terms is gaps. If a stock closes at 100 and opens at 110, the bar's own high–low range might be small, but the move was large. Plain range would understate it badly, and in a market with overnight gaps — which is every equity market — that understatement is systematic rather than occasional. This is a small definitional detail with large practical consequences, and it is a good example of why reading an indicator's construction is worth the twenty minutes.
Volatility-based position sizing, which is the highest-value idea in this book
The standard approach: risk a fixed fraction of capital per trade, and let volatility determine the position size.
position size = (capital × risk fraction) ÷ (ATR multiple × ATR)
Worked: ₹10,00,000 capital, risking 1% (₹10,000) per trade, stop at 2 × ATR, and the stock's ATR is ₹25. Stop distance is ₹50, so the position is 10,000 ÷ 50 = 200 shares. If a different stock has an ATR of ₹5, the same ₹10,000 of risk buys 1,000 shares.
Why this matters more than the entry. Both positions now risk the same amount in rupees. Without volatility scaling, a fixed rupee position size means your volatile stocks dominate your P&L entirely and your quiet ones contribute nothing — so your returns are determined by which stocks happened to be volatile rather than by your analysis.
Most people spend their effort on entries and use a fixed quantity. The reverse allocation of effort would improve most trading records more.
Bollinger Bands, and the misuse to avoid
A 20-period moving average with bands at ±2 standard deviations of price over the same window. Roughly 90% of bars fall inside the bands — not 95%, because financial returns are fat-tailed and the normal-distribution intuition does not hold.
- The misuse: selling at the upper band and buying at the lower. In a trend, price walks the band for extended periods. This is the same error as selling RSI 70 and it has the same cause.
- The genuinely useful reading is the width. Bandwidth is a direct measure of current volatility relative to recent volatility. Narrow bands — a ‘squeeze’ — indicate compression, and compression resolves into expansion. That is volatility clustering doing the work, and it is one of the few technical observations with a solid statistical basis.
- But note what it does not tell you: a squeeze predicts that a large move is coming, not which way. Anyone selling you a squeeze indicator as a directional signal is adding a claim the statistics do not support. The correct structures for this are ones that profit from movement in either direction, or a squeeze used as a filter on a separate directional decision.
Volatility regimes, and adapting rather than optimising
The most sophisticated use of this family is to switch behaviour by regime rather than to search for a parameter that works everywhere:
| Regime | Signature | What tends to work | What tends to fail |
|---|---|---|---|
| Low and falling volatility | Narrow bands, low ATR relative to its own history | Mean reversion, range trading, selling premium | Breakout systems — most breaks fail |
| Expanding volatility | Bandwidth rising sharply from a low base | Breakout and momentum entries | Mean reversion, which now catches falling knives |
| High and sustained volatility | ATR well above its own average, persistent | Trend following, with sizes reduced | Anything with a fixed rupee stop — you will be stopped constantly |
| High and falling | ATR declining from a spike | Re-entry into the prior trend | Continuing to trade reduced size after the risk has passed |
The practical implementation is simpler than it sounds: normalise ATR by its own long-run average to get a regime score, and use that score to scale position size and to gate which rules are allowed to fire. That is adaptive without being over-fitted, because it has one parameter rather than twenty.
Implied versus realised, for anyone touching options
Everything above is realised volatility — computed from what price actually did. Implied volatility is derived from option prices and represents what the market is charging for future movement.
The durable observation is that implied is systematically higher than subsequently realised, across markets and across decades. That gap is the volatility risk premium and it is compensation for bearing the risk of a large move, not a free lunch. The trap — and it is a well-trodden one — is concluding that selling options therefore has an edge: capturing the premium requires hedging the directional exposure, and hedging costs consume a large share of it. An unhedged short volatility position is a directional bet that wins most of the time and occasionally ends the account.
Check yourself