The Candle and Its Noise: Range and ATR

Range is movement, ATR is the habit

One session’s range is a single observation, and like any single observation it is noisy: a stock can range 1% on a quiet Tuesday and 8% on an earnings day without contradicting anything about itself. ATR fixes that by averaging the true range over a window — conventionally fourteen periods, which on a daily chart is three trading weeks — so the number describes a habit rather than an incident. The construction detail matters more than it looks: the true range of a period is the high-to-low distance, or the distance from the prior close to the current high or low if that is larger, which is how gaps get counted as movement. A stock that gaps 5% overnight and then trades in a 1% range all day has a true range near 5%, not 1%, and a stop placed without that knowledge is placed against a fiction. Two uses follow and they are the only two this lesson cares about. The first is the noise floor: if the average day moves $1.60, a stop $1.00 away is inside the range of a normal session, so it will be triggered by ordinary movement, and the frequency of the trigger will be a fact about the instrument’s volatility rather than about the thesis. The common rule of thumb — at least two ATR, and often more in fast names — comes from that observation, and it has an uncomfortable consequence: in a volatile instrument a structural stop may be genuinely wide, which means the position has to be small. There is no way to hold a name that moves 8% a day, risk 1% and own a meaningful number of shares; the three cannot all be true, and the one that gives is the size. The second use is sizing, and this is where chart reading produces a number. The stop is placed at the price that invalidates the idea — a level, a structure, a point where the thesis is simply wrong — and only then is the distance measured. Risk budget, usually 0.5% to 2% of capital, divided by that distance in dollars gives the share count. Notice the order: the chart chooses the stop, the stop chooses the distance, the distance chooses the size. Inverting it is the most common beginner error in the whole subject, because choosing a size first and then placing a stop where the loss is tolerable is choosing where you are wrong from your comfort rather than from the chart. And a stop placed by comfort is one that will be moved. The same idea at two volatilities — Name A: entry $62.40, level $57.80, ATR $1.60: 4.60 risk = 2.88 ATR — outside the noise ← · $300 budget ÷ $4.60: 65 shares, about $4,070 of stock · Name B: same level, ATR $3.90: 4.60 risk = 1.18 ATR — inside the noise · Name B sized safely (2.5 ATR = $9.75): $300 ÷ $9.75 = 31 shares, a third of the notional ← ATR is usually quoted in dollars on the instrument’s own chart, so the conversion to percent is ATR ÷ price. Compare ATR percentages, not dollars, when ranking two names for risk: a $2 ATR is small on a $400 stock and large on a $25 one.

What ATR cannot do

ATR describes ordinary movement and says nothing about the shape of the tail, which is the risk that actually ends accounts. A name that moves 1.5% most days and 20% on earnings has an ATR that reports the calm days, and a position sized from it will be five times too large twice a year. That is why the trading calendar belongs beside the ATR: a scheduled event is a known change in the distribution, and the honest responses are to reduce size into it, to widen the stop and accept less stock, or to avoid holding through it — not to pretend the average describes the day of the report. Nor does ATR know where liquidity is. In a thin instrument, the ordinary range can be small while the gap between the bid and the offer is wide, so the same 4% stop costs much more to realise: you exit into a spread and the recorded loss is the stop distance plus the spread plus any slippage. That combination — a quiet average and a wide spread — is exactly the profile of the small-cap names retail traders find attractive, and it is why the effective risk of a position is the stop distance plus the cost of getting out. The practical version: check the spread as a share of price, and if it is more than a small fraction of the ATR in percentage terms, the position is more expensive to own than its chart suggests. Finally, ATR is backward-looking by construction. It is an average of what happened, so it adapts to a volatility regime change only after several sessions, and in a fast market it will understate what is coming. That is an argument for using it as a floor rather than a forecast: the ATR tells you the minimum distance a stop needs to be honest about ordinary movement, and judgement — plus the calendar — tells you whether today is ordinary. A learner who uses it as a ceiling, sizing to exactly two ATR and calling the risk fixed, has replaced one fiction with a more arithmetic-looking one. • ATR averages true range, so gaps count as movement — which is what you want from a risk measure. • Use it as a noise floor for stops and a divider for size: at least two ATR, then budget ÷ distance. • It is silent about event days, so size for the calendar as well as the average. • It is silent about spreads, so add the cost of getting out to the recorded risk. • It adapts slowly, so treat it as a minimum rather than a forecast. One trap worth naming: making the stop wider to keep the position size up. If the structural stop is far away and the position it implies is unwelcome, the answer is a smaller position or a different instrument, not a stop that sits inside the noise. The size is the variable that moves; the stop is the fixed point.

One unit of risk, across names that move differently

ATR in dollars is what sizes a stop on one chart. The moment there is more than one position, the units have to become comparable, because a dollar of ATR means something different on a twenty-dollar stock than on a two-thousand-dollar one — and a portfolio where every position carries the same dollar risk is not a portfolio of equal risk at all. The normalisation is simple and worth doing once: divide the ATR by the price, and you have the daily movement as a percentage. A name with a two-dollar ATR at forty dollars moves five percent a day; a name with a twenty-dollar ATR at two thousand moves one percent. The first is five times the risk per dollar invested, and if both positions are the same dollar size, the portfolio’s returns will be dominated by whichever of them is moving — which is usually the one the trader was least thinking about. Expressing it that way also makes ATR comparable with a number the learner meets elsewhere in the curriculum. A daily true range as a percentage annualises approximately with the square root of the number of trading days in a year, so a one-and-a-half percent average daily range corresponds to something in the low twenties as a percentage a year, in the same units as the volatility quoted in an option chain. The conversion is rough — the smoothing in Wilder’s average and the shape of the distribution both bend it — but it is close enough to answer a question that matters: is this stock’s movement cheap or expensive against what the option market is charging for it? That comparison belongs in the options subject, and it starts here, because the two disciplines are measuring the same thing in different clothes. The practical consequence for a portfolio is a single sizing rule rather than a per-name feeling. Pick the risk each position should carry in account terms — half a percent, say — and let the stop distance in ATR decide the size, so that a quiet name gets a larger position and a jumpy one gets a smaller. This is the reason a professional can hold a utility and a biotech at the same dollar risk without the biotech dominating the P&L: the sizes are not equal, deliberately, and the volatility is what equalises them. Run the same book with equal dollar sizes and the portfolio’s behaviour is set by its most volatile holding, which is a decision made by accident. There is one caveat the normalised number cannot fix, and it is worth stating plainly. ATR describes recent movement, so a name that has been calm can re-rate violently — a biotech before a readout, an acquisition target, a stock with a scheduled event — and the position sized off the quiet ATR will be far too large for the volatility that arrives. Normalising across names makes the comparison honest today; the scheduled-event adjustment is what makes it honest tomorrow, which is why the trading calendar sits beside the ATR and not separate from it. • ATR ÷ price puts every name’s movement in the same units. • Equal dollar risk is not equal risk: volatility decides which position dominates. • A daily ATR percentage approximates an annualised volatility, so it can be compared with what options imply. • Size from the stop distance, not from a dollar amount you like — quieter names earn larger positions. Once positions are sized this way, the portfolio has an ATR of its own: the weighted sum of the positions’ movements, adjusted for how often they move together. That is the number the account’s daily swing is made of, and it is the reason two calm positions and one wild one is a different portfolio from three calm ones.

Sources

• Average true range, construction and useWilder, “New Concepts in Technical Trading Systems” (1978)
• Volatility-based position sizingStandard risk-management practice; Van Tharp, “Trade Your Way to Financial Freedom”
• Stops inside versus outside the noise bandMarket microstructure literature on stop clustering