What is the ATR indicator?

Almost every indicator a retail trader meets is trying to tell them something about direction: whether to be long, short or out. ATR does not participate in that conversation at all. Every one of its inputs is an absolute distance, a size with the sign thrown away, which means the number is structurally incapable of pointing anywhere. A market that rises 3 percent and a market that falls 3 percent produce the same contribution to ATR.

That sounds like a limitation. It is the opposite. Because ATR refuses to answer the direction question, it answers a different one cleanly: how far does this instrument ordinarily move? And that turns out to be the question sitting underneath most of the decisions that go wrong.

A trader who says "I use a 50 point stop" has not made one decision. They have made a different decision on every instrument they apply it to, and they do not know which ones.

Consider what a distance decision normally looks like. Someone picks a stop of 50 points, or a target of 100, or a position of 200 units, and the number comes from habit, from a round figure, or from what the account can absorb. None of those sources knows anything about the instrument. The market, meanwhile, has a very definite opinion about how far it moves in a session, and that opinion is not constant across instruments and not constant across time within one instrument.

ATR is the translation layer between the two. Once you know that one bar of ordinary travel is worth 18 rupees on this instrument this month, then 50 points stops being a number and becomes a statement: it is 2.8 bars of ordinary travel. That is a claim you can evaluate. And critically, it is a claim you can carry to a different instrument without re-deriving it, which a rupee figure can never be.

This is why ATR belongs to the risk layer of a method rather than the signal layer. It does not decide whether there is a trade. It decides how much room the trade needs and, through that, how large it can be. Everything in this guide follows from holding that distinction firmly, because almost every misuse of ATR comes from forgetting it and trying to read a volatility measure as a directional one.

ATR is built on true range, and true range is deliberately more careful than the obvious answer to "how far did it move today". For each bar it takes the greatest of three distances: (a) the high minus the low, (b) the absolute value of the high minus the previous close, and (c) the absolute value of the low minus the previous close. The bar's own high-to-low range is only one candidate. The other two reach back to where the last bar finished.

The reason is the gap. When a market opens sharply away from its previous close, on results, an overnight move in global markets, a policy surprise, the new bar's own high minus low can be perfectly modest while the real distance travelled from the prior close was large. Terms (b) and (c) are what catch that jump. Because Indian equities and indices routinely gap between sessions, this is not a technicality that can be waved away. It is the entire reason true range exists rather than plain range.

Notice what that failure mode implies. A naive range measure does not degrade gracefully. It is accurate on ordinary days and badly wrong on exactly the days that matter, which is the worst possible error profile for a risk instrument. It would systematically tell you the market is calmer than it is, precisely when it is not.

The three-way maximum also handles cases that are less obvious than a simple gap. If a bar gaps up and then fills the gap during the session, the plain range can legitimately be the largest of the three again. If a market gaps and then keeps running, the gap term can dwarf the bar's own range by a wide margin. The definition does not need to know which case it is in; it just takes the largest, and the largest is always the honest distance from where the last bar left off.

Two rows in that table are worth sitting with. The gap that fills back shows that the previous close terms do not always win, so the definition is not a gap-detector bolted on; it is a genuine maximum. And the gap that keeps running shows the error at its most extreme: an 8 point candle that actually represents 30 points of travel from the prior close. A measure that reported 8 there would be describing a different market from the one that traded.

One bar of true range is noisy. ATR is what you get when you smooth it, and Wilder specified the smoothing precisely: seed the first value as the simple average of the first fourteen true ranges, then update recursively. Today's ATR equals yesterday's ATR multiplied by thirteen, plus today's true range, all divided by fourteen.

That last point is the one worth paying for in attention. The smoothing is not free, and the price is denominated in bars of lateness. On this recursion, ATR closes half of any change in the underlying volatility level in about 9.4 bars and ninety percent of it in about 31. Those are not estimates of market behaviour; they follow directly from the arithmetic of decaying 13/14 of the gap each bar.

This has a direct practical consequence that is easy to state and easy to forget. Immediately after volatility expands, ATR is too low. A stop sized off it will be tighter than the new regime warrants, and a position sized off it will be larger than the new regime warrants. Both errors point the same way, which is the dangerous direction. The mirror case is milder: after volatility contracts, ATR is too high, so stops are wider and sizes smaller than necessary. That costs opportunity rather than capital.

Shortening the period reduces the lag and increases the noise, and there is no setting that escapes the trade off, because it is the trade off. Wilder's 14 is a convention that has survived because it is a reasonable compromise, not because it is optimal. What matters far more than the number is knowing which way the error points when a regime turns.

Here is the part that changes how the number is used. A stop distance has to do one job: sit far enough away that the ordinary noise of the instrument does not reach it, and close enough that being wrong is affordable. Ordinary noise is a property of the instrument. A fixed rupee figure knows nothing about it.

The clearest way to see this is to hold the noise constant and change only the scale. The figure below does exactly that: both panels are driven by one normalised price path, drawn at two different volatilities. The shape of the two charts is identical by construction. The only thing that differs between them is how much a unit of that shape is worth in rupees.

That bottom strip is the whole argument in one line. The pullback marker lands on the same mark on both rows, because in ruler units the two instruments did the same thing. The ₹40 marker lands in two completely different places. A single rule stated in ATR, the 3 ATR mark in the figure, lands on the same place on both rows and survives on both, without anyone having to know in advance what either instrument was worth.

The four instruments in that table are close to each other in proportional terms: their ATR sits in a narrow band between 1.68 and 2.31 percent of price. By any sensible reading they are comparably volatile. Yet the same ₹50 stop is worth 7.81 ATR on one and 0.35 ATR on another, a spread of twenty two times in what the identical number actually means. On W it is so far away that the position size it permits collapses to almost nothing. On Z it will be removed by an unremarkable session.

One clarification that prevents a common misreading. None of this says an ATR based stop is safer. It says it is consistent. A 2 ATR stop on a violent instrument is a large rupee number and will produce a large rupee loss when it is hit. The consistency is in the decision, not in the outcome, and the rupee consequence is handled downstream by size.

Once the stop distance is expressed in ATR, position size stops being a matter of preference and becomes arithmetic. Risk per unit is the stop distance. So if you have decided in advance how much money a single trade may lose, the number of units is that budget divided by the per-unit stop distance, and the stop distance is a multiple of ATR.

Written out: units = risk budget ÷ (ATR × multiple). ATR is in the denominator, and that single structural fact produces the behaviour that most retail traders get backwards.

Read the lower panel carefully, because it contains the real finding. The trader holding 185 units did not decide to take four times more risk in the stressed regime. They decided nothing at all. They held their size constant, which feels like discipline, and the market silently rewrote their exposure for them. Holding units fixed does not hold risk fixed. It holds risk hostage to volatility.

The ATR-sized version does the opposite, and it does it without requiring any forecast. As volatility rises the denominator rises, so the position falls; as volatility falls the position grows back. Nobody has to predict the regime. The size reacts to the regime that has already arrived, which is precisely what a backward-looking measure is entitled to do.

This is also the correct reflex, and it is the opposite of the common one. The instinct in a wild market is to trade bigger, because the moves look larger and the opportunity looks better. The arithmetic says the wild market is exactly where the position should be smallest, because the distance to being proven wrong has grown. The chain that runs from an invalidation level, through a volatility-scaled stop, to a size derived from a fixed risk budget, is the spine of the method we teach, and it sits inside the wider discipline covered in the guide to risk management in trading.

The third thing a ruler fixes is expectation, and it is the one that gets the least attention because it produces no order and no chart marking. A target is also a distance, and it is subject to exactly the same problem as a stop: chosen in rupees, it carries an implicit and usually unexamined claim about how far the instrument can travel in the time available.

Put the claim in ruler units and it becomes checkable. If one bar of ordinary travel is ₹18.90 on a ₹1,050 instrument, then a target five ATR away is asking for about ₹94, or 9 percent of the instrument's price. Stated that way, the question answers itself: that is not a target, it is a forecast of five sessions of pure one-way movement with no pullback, on an instrument that ordinarily gives back part of every leg.

An instrument that moves one ATR a day will not deliver a five ATR move because your position size requires it to.

This matters most where it is least visible: in the reasoning that happens before a trade. A trader who needs a certain rupee outcome will, without noticing, choose a target that delivers it, and then construct a story about why that level is reachable. The ruler interrupts that sequence by making the demand explicit in the instrument's own units, before the story gets written.

There is a second-order effect worth naming. Because the same ATR sets both the stop and the sensible target, the ratio between them is a ratio of ATR multiples, which is finally comparable across instruments and across time. A 1.5 ATR stop against a 3 ATR objective is the same shape of trade whatever you are trading, whereas a 20 point stop against a 40 point objective is a different shape of trade on every instrument you apply it to. That is the difference between a rule and a habit that resembles one.

None of this makes a target reachable. ATR cannot tell you whether a move will happen, and nothing in this section should be read as suggesting otherwise. It rules out the demands that were never plausible in the first place, which is a smaller claim and a more useful one.

There is one confusion around ATR that costs real money, and it is worth isolating. ATR is realised volatility: it is computed entirely from bars that have already printed. Implied volatility is implied: it is solved backwards out of live option prices and describes movement that has not happened yet, which the market is currently willing to pay for.

They are not two measurements of the same thing, and they are not interchangeable. One is a record and the other is a price. Reading a low ATR as "the market expects calm" is a straightforward category error, because ATR has no mechanism for expecting anything.

The sequencing in that figure is the mechanism, not a coincidence of the example. Implied volatility can rise before an event because it is a price and prices move on expectations. ATR cannot, because it is an average of distances that have already been travelled. The two peak one session apart, in a fixed order, for structural reasons.

The gap between them is itself information, and it is the reason the distinction is worth holding rather than merely knowing. When implied sits far above realised, the market is paying a premium for movement that has not yet appeared in the bars: something is expected. When implied sits below realised, the market is pricing a calming that the bars have not yet delivered. Neither reading is a signal, and neither tells you direction. Both tell you something about the disagreement between what has happened and what is being paid for, which is a genuinely different piece of information from either number alone.

For the mechanics of how implied volatility is extracted from option prices, what moves it, and how it is quoted, the guide to implied volatility covers the ground properly. The one thing to carry away here is the boundary: use ATR to size a distance, and implied volatility to judge whether movement is expensive. Swapping them produces confident, well-reasoned, wrong conclusions.

Because ATR is a smooth line, its slope carries information that its level does not. A rising ATR marks an expanding regime: bars are covering more ground, stops need more room and positions should be smaller. A falling ATR marks contraction: the market is quietening, and the same rule set produces tighter stops and larger positions. This is context, not instruction. It changes the size of what you do without saying anything about what to do.

The level, by contrast, is nearly meaningless in isolation, because it is denominated in the instrument's own units. "ATR is 42" is not a fact anyone can act on without knowing the instrument and its history. ATR is a comparison instrument, and the only two comparisons it supports honestly are against its own recent past, and against another instrument once both have been converted to a percentage of price.

That distinction is worth one more sentence because it is where most of the confusion between the two indicators lives. Size and persistence are independent properties of a move. Knowing that bars are large tells you nothing about whether they are going anywhere, and knowing that a trend is orderly tells you nothing about how much room it needs. Using either one to answer the other's question produces exactly the sort of error that looks well-reasoned in a journal entry.

A last practical note on the slope reading: it inherits every bit of the lag established earlier. A rising ATR tells you volatility has expanded, not that it is about to. By the time the slope is unambiguous, the expansion is already several bars old and has already been paid for. That is not an argument against reading the slope. It is an argument against treating the reading as early.

The honest limits of ATR all descend from a single property: it is a backward-looking average of absolute distances. Everything it does well and everything it cannot do follows from that one sentence.

The last item on the right deserves more than a line, because it is where a well-built ATR framework can still fail completely. An ATR based stop is a distance, and a distance turns into an order, and an order is only an instruction to transact at a price if a price is available. A gap opens past the level and fills wherever the market chooses to open. A circuit halt removes the ability to transact at all until it lifts, and when it lifts the price is where it is. Neither case is a malfunction of ATR. Both are situations in which the ordinary, which is all ATR measures, has stopped applying.

ATR belongs in the risk layer of a method and nowhere else. It does not generate trades, confirm trades or time them. It calibrates how much room a trade needs and, through the size formula, how large that trade is allowed to be, given how far this instrument is currently travelling. Read plainly, it turns "the market feels wild" into a number that can be put into a rule, and it does so with a definition careful enough to count the gaps a naive measure would silently drop.

The narrowness is the feature. ATR will not tell you direction, will not forecast the next shock, and cannot be compared across instruments in its raw form. What it does instead is make three of the most commonly botched decisions in retail trading, how far away the stop goes, how many units to hold, and what it is reasonable to expect, into arithmetic that is consistent from one instrument to the next and from one month to the next.

The judgements that matter most still sit upstream of it and always will. Where the idea is genuinely wrong, and how much of the account may be put behind it, are decisions ATR has no opinion on. It only makes them measurable, and measurable is the precondition for everything else.

• J. Welles Wilder, New Concepts in Technical Trading Systems (1978). The original source, which defines true range as the greatest of the three distances and introduces the average true range smoothed over a default of 14 periods. Establishes the definition, the author and the date, and is also the origin of ADX.
• StockCharts ChartSchool, Average True Range (ATR) and Average True Range Percent (ATRP). Sets out the three-way true range maximum, the Wilder smoothing recursion with its 14 period seed, and ATRP as ATR divided by price times 100 for cross-instrument comparison. chartschool.stockcharts.com
• Macroption, ATR calculation. Details the Wilder smoothing formula, the first-value seeding convention, and the two conventions for expressing ATR as a percentage of price. The 1 ÷ 14 weighting used throughout this guide follows directly from that recursion. macroption.com
• StockCharts ChartSchool, Chandelier Exit. Documents a published volatility-based exit that trails a stop a multiple of ATR below the highest high since entry, which is a worked example of ATR feeding an adaptive stop rather than a signal. chartschool.stockcharts.com