BUILD: COURSE 1 | LESSON 4
Volatility: ATR and Bollinger Bands
Learning objectives
Calculate and interpret ATR, and use it to set stops and position sizes that match current market conditions
Explain what Bollinger Bands measure, why a band "touch" is not a signal, and what the squeeze does and doesn't tell you
Convert volatility awareness into concrete risk decisions rather than directional predictions
Volatility tools predict nothing — and that's their strength
Every tool so far tried to describe direction. Volatility tools give up on direction entirely and measure something more honest: how much the market moves, regardless of which way. That sounds less exciting. It's arguably more useful, because volatility is one of the few market properties with genuine persistence — volatile periods tend to be followed by volatile periods, quiet by quiet (statisticians call it volatility clustering). Direction is hard to forecast; conditions are somewhat forecastable.
That's why this lesson's tools feed straight into risk management. If Lesson 1's theme was "structure describes context", this one's is "volatility sets the size of everything": your stop distance, your position size, your profit targets, even whether your strategy should be running at all.
ATR: the market's average step size
The Average True Range (Wilder again, 1978) starts from the true range of each period — the largest of:
- High − Low (today's range)
- |High − Previous Close| (gap up captured)
- |Low − Previous Close| (gap down captured)
The gap terms matter for CFDs on indices and stocks that jump over weekends and news; for major FX pairs term 1 usually dominates. ATR is then a smoothed average (default 14 periods) of true range — Wilder used his own smoothing, ATR(today) = (ATR(yesterday) × 13 + TR(today)) / 14, which behaves like a slow EMA.
Intuition: ATR is the market's average step size. If EURUSD's daily ATR(14) is 0.0065, the pair has been covering about 65 pips per day lately. If gold's daily ATR is 28 dollars, a 10-dollar stop on a swing trade is not a stop — it's a donation, because normal daily noise will hit it with no directional information whatsoever.
Trade-offs of the settings: ATR(14) daily is the standard; shorter (5–7) adapts faster after a volatility shift but jumps around; longer (20–50) is stabler but slow to notice that conditions changed. ATR is also an absolute number, so you can't compare 65 pips on EURUSD with 28 dollars on gold directly — divide by price (ATR%) if you need cross-market comparison.
The three uses that pay rent:
- Stops. Place stops a multiple of ATR beyond your invalidation point — commonly 1.5–2× ATR for swing trades. Long EURUSD at 1.0750 with daily ATR 0.0065: a 2×ATR stop sits ~130 pips away at 1.0620, deliberately outside one normal day's noise.
- Position sizing. This is the crown jewel, and Course P2 builds a whole lesson on it. Risking 1% of a $10,000 account ($100) with a 130-pip stop means $100 / 130 pips ≈ $0.77 per pip ≈ 0.077 lots. When volatility doubles, the ATR stop doubles, the position halves — and your dollar risk stays constant across calm and storm. That single habit removes more blow-ups than any entry signal ever will.
- Regime awareness. EURUSD ATR drifting from 65 to 95 pips means yesterday's stop distances, targets and expectations are stale. Around major news, ATR-based expectations are the difference between a plan and a surprise.
What ATR does not do: it says nothing about direction, and a rising ATR is not "bullish" or "bearish" — it's just louder.
Bollinger Bands: a moving average wearing standard deviations
John Bollinger's bands wrap a volatility envelope around price:
Intuition: the standard deviation measures how spread out the last 20 closes have been. Multiply by two and, if returns were well-behaved statistically, roughly 95% of price action would sit inside the bands. Real markets have fat tails, so treat "about 90%" as the honest expectation, and treat excursions outside the bands as "unusual", never "impossible".
The bands therefore do two jobs at once: the middle band tracks the short trend (it's your 20-SMA from Lesson 2), while the band width breathes with volatility — expanding in storms, contracting in calm.
The touch trap. The most common misuse is treating an upper-band touch as "sell" and a lower-band touch as "buy". Bollinger himself is blunt about this: a band tag is not a signal. In a trend, price can "walk the band" — close after close riding the upper band for weeks, each one a fresh fake sell signal to the touch-fader. On a US500 CFD grinding from 5,600 to 5,900, the upper band gets tagged a dozen times; fading each tag with no filter is a strategy for donating margin.
Legitimate readings, with their caveats:
- In a confirmed range, band tags at range edges make coherent mean-reversion setups — same regime logic as stochastic in Lesson 3, and it fails the same way the moment a range becomes a breakout.
- The squeeze. When band width contracts to multi-month lows, volatility is compressed — and since volatility clusters, expansion often follows. Worked example: GBPUSD spends three weeks in a 90-pip coil around 1.2650; band width hits its lowest since spring. The squeeze says a bigger move is likely brewing. It does not say which way — the first thrust out of a squeeze is also a classic head-fake ("false break, then the real move"), so squeeze traders typically wait for a close outside the bands plus structural confirmation, with a stop on the far side of the coil.
- %B and band-relative closes. Closing outside a band on expanding width is a statement of strength (continuation evidence), not exhaustion — the opposite of the touch-fader's read. Context decides which interpretation applies, and "context" here means the trend structure you built in Lesson 1.
Settings trade-offs: (20, 2) is the default and the one everyone watches. Tightening to (10, 1.5) generates more signals and more noise; widening to (50, 2.5) turns the bands into a slow regime tool. As always, parameters tuned until history looks perfect are a promise history won't keep.
Putting volatility to work: a checklist
Before any Practitioner-level trade, volatility should answer four questions:
- How far is "noise"? → stop = invalidation point + 1.5–2× ATR buffer.
- How big can I be? → size = account risk ÷ stop distance (ATR-scaled).
- Is my target realistic? → a 200-pip target on a 65-pip-ATR pair is a multi-day thesis, not a day trade.
- Does the regime fit my strategy? → squeezes favour patience and breakout plans; expanding bands mid-trend favour trend-following; band-walks forbid fading.
Notice that none of the four is a prediction. Volatility tools have the best claim to statistical validity of anything in this course precisely because they aim lower: they don't call direction, they calibrate risk. The hit rate of your entries will stay stubbornly modest whatever you do — ATR is how you make sure the losers are survivable and the winners are given room to pay for them.
Key takeaways
Volatility clusters, so ATR and band width are among the few genuinely persistent, forecastable market properties — unlike direction.
ATR(14) = smoothed average true range (including gaps); read it as the market's average step size, and never set a stop inside one normal step.
ATR-based sizing (risk ÷ ATR-scaled stop) keeps dollar risk constant as volatility changes — the single highest-value habit in this course.
Bollinger Bands = 20-SMA ± 2 standard deviations; band tags alone are not signals, and trends "walk the band" against faders.
The squeeze flags compression that often precedes expansion — but gives no direction, and the first break is frequently the false one. Confirmation and stops still do the real work.