ATR Indicator Guide: Turning Volatility Into Stops and Size

ATR is a ruler, not a compass. It plays no part in deciding whether to go long or short. It answers one question only: over the recent stretch of trading, how far does a single candle typically travel? J. Welles Wilder introduced it in his 1978 book New Concepts in Technical Trading Systems (Open Library record: Trend Research, 1978). The motivation was plain: commodity futures contracts differ wildly in how restless they are, so before you talk about entries and exits you need one consistent way to measure how much a market is thrashing around right now.

CoinView has already covered RSI, MACD, Bollinger Bands and OBV, plus liquidation price math and the Kelly criterion. The first group answers "which way," the second answers "how much can I lose." ATR is the missing piece in between. It translates "how choppy is this market" into a distance measured in price, and that distance is exactly the interface that plugs a directional call into a position-sizing formula. So this guide isn't about copying out the formula one more time. It's about two things: how that distance becomes a stop level, and how it becomes the size you should actually trade.

One thing up front: ATR won't stop you from getting stopped out. It lags, it has no immunity to wicks, and even a well-reasoned stop can be taken out by a single liquidity gap. What it can do is give the "how far away is my stop" decision something to stand on, instead of a round number picked on a hunch.

1. What ATR measures: a ruler, not a compass.

ATR's output is different from most indicators you already know: it carries price units. RSI gives you a position between 0 and 100, MACD gives you a positive or negative value, but ATR gives you an amount of money. Say BTC/USDT's daily ATR reads 1,200 USDT on a given day. That means over roughly the past couple of weeks, BTC has been swinging about 1,200 USDT a day on average.

That sounds like a minor distinction. In practice it isn't. A number in price units can be subtracted straight from your entry price. Whether RSI reads 68 or 32, you can't calculate a stop level from it. If ATR is 1,200, you can immediately work out that a 2× ATR stop sits 2,400 below your entry. Not many indicators can do that.

The second property to remember: ATR is always greater than or equal to zero. It's an average of ranges and has no sign. A day that rallies 5% and a day that drops 5% feed ATR the same number. So it is completely silent on direction, and that's not a flaw, it's a division of labor (Investopedia: Average True Range).

So when do you actually need to look at it? Realistically, at three moments. Before entry, to set the stop distance and position size. While you're in the trade, if you use a trailing stop, because ATR decides how closely it follows. When the market changes character: ATR lifting off a long-term low means the market has switched from quiet to restless, and the stop scale you've been using may no longer fit. Beyond that, staring at ATR doesn't achieve much. It doesn't generate signals in the first place.

2. True range takes the largest of three: why high minus low isn't enough.

ATR's raw material is the true range (TR). It isn't simply the high minus the low. It's the largest of three candidate values (Wikipedia: Average true range):

• Current high − current low
• | Current high − previous close | (absolute value)
• | Current low − previous close | (absolute value)

Why the detour? Take a hypothetical. An asset closed at 100 yesterday, opened today straight at 108, and spent the whole day between 108 and 110. Today's high minus today's low is only 2, so the day looks calm. But measured from yesterday's close, price has actually moved 10. The three candidates are 2, |110−100| = 10 and |108−100| = 8. The largest wins, so TR = 10.

That example shows what TR is really measuring: not how long today's candle is drawn, but how far price actually traveled from the last known fair price. Nothing traded inside the gap, but the risk was real. A stop order sitting in that zone still gets filled after the gap, at a worse price. Counting the gap is what makes ATR measure the distance you actually have to absorb.

Taking the largest of three has a side benefit. TR is always at least as large as today's high minus today's low, and it's never negative. So ATR can't go negative, and it won't understate a day's real displacement just because the candle happened to be short.

Does crypto even need this design? It trades 24/7, so overnight gaps would seem not to exist. It still matters; the dominant term just changes:

• Around contract maintenance and the resumption of trading, and in the instant an order book gets blown through during an extreme move, price does jump, and the last two candidates take over.
• CME Bitcoin futures used to have genuine weekend gaps. Those are gaps in the traditional sense, covered separately in our CME gap article. Since May 29, 2026, CME crypto futures have traded around the clock (with a weekend maintenance window), so those weekend gaps no longer form the way they used to.
• On intraday candles, a single deep wick makes the first candidate (high minus low) the largest on its own. In wick-heavy crypto markets, watch this one closely.

Put differently: in traditional markets that close overnight, large TR values mostly come from the last two terms (gaps). In crypto they mostly come from the first term (wicks). That difference carries all the way through to the issues in section 6.

3. Wilder's smoothing: it says 14, but it's slower than you think.

Once you have a TR for every candle, the next step is to average them. Here Wilder used not an ordinary moving average but his own recursive smoothing, which many platforms label RMA or SMMA. The process: take the simple average of the first 14 TR values as the seed, then roll forward with each new candle:

ATR(today) = [ ATR(yesterday) × 13 + TR(today) ] ÷ 14

This recursion is laid out clearly in Wikipedia's ATR entry. The general form is ATR_t = [ATR_t−1 × (n−1) + TR_t] ÷ n. Expand it and it's equivalent to an exponential smoothing line with a smoothing factor of α = 1/14.

The EMA you normally use has a smoothing factor of α = 2/(n+1). The two are easy to lump together, but put them side by side and solve, and you see how far apart they are: set 2/(n+1) = 1/14 and you get n = 27. In other words, Wilder's 14-period ATR responds roughly like a 27-period EMA. Anyone can derive this from the two formulas, so there's no need to take our word for it.

Knowing that it's slow has two very practical consequences:

• ATR can't keep up with sudden moves. When a huge red candle slams down, the new TR enters ATR at only 1/14 weight, and it takes several candles in a row to bring the level up to where it should be. At the most dangerous moment, ATR is still handing you the calm-period reading, which means a stop calculated from it is too tight right when it matters.
• It's just as slow on the way down. After a wick pushes ATR up, even if the market calms down straight away, ATR takes several more candles to come back down. During that hangover, stops set from ATR are too wide and the position size it produces is too small. You keep paying for a burst of volatility that's already over.

As for why the default period is 14: in Wilder's 1978 book, RSI, DMI and ATR mostly all use 14, which works out to just under three weeks of trading days. It's the author's rule of thumb; the book doesn't present it as a proven optimum. So don't treat 14 as gospel, but don't fiddle with it casually either. Section 7 explains why.

One more trap worth checking before you read the chart: many platforms let you pick the smoothing method in the ATR settings panel.

On the same candle, the three methods give different ATR values. Switching the smoothing method means switching rulers, and the multiple you set earlier no longer corresponds to the same distance. If you calibrated your settings on one platform, check this setting before you move to another.

4. Turning ATR into a stop distance: what k should be, and what decides it.

This is ATR's main job. The formula itself needs little explanation:

Long stop = entry − k × ATR
Short stop = entry + k × ATR

The hard part is choosing k. Let's be blunt about it: there is no universal optimal multiple, and any claim that hands you one specific number and calls it optimal deserves suspicion. Public textbooks and common templates mostly land between 1.5 and 3. That's a conventional range, not a conclusion. What actually decides k is the following.

The table points you in a direction; it isn't a recipe. So how do you settle on your own number? Not by guessing, but by going back and counting your own trades. Pull up your past fills and measure two ratios:

1. The share of trades that exited at the stop;
2. The share of stopped-out trades where, within the next several candles, price went back in your original direction and reached your original target.

If the second number is high, your k is too small and you keep paying tuition to everyday noise. If it's the other way around, and you almost never get stopped but every loss is heavy, k is too large and the stop is barely doing its job. Anyone can calculate both ratios from their own trade history; you don't need to borrow someone else's settings. Your k should be set by your own trade history, not by an article.

One common variant is the trailing stop. You place the stop at the highest price since entry minus k × ATR, and you only ever move it up, never down. The upside is that the stop distance flexes with volatility automatically. When the market gets choppier, ATR rises and the stop trails further behind the high; when things calm down, it trails closer and locks in more of the profit. Because the stop only ever moves up, a sudden jump in ATR won't pull an existing stop lower. It just stops climbing for a while, so a burst of volatility can still reach it. The downside is that once the trend is done, it hands back part of your open profit. That's the built-in cost of this kind of stop, not a sign that the parameters are badly tuned.

Finally, three hard reminders, written down here so none of this gets treated as a cure-all:

• ATR lags. On the candle where the market suddenly breaks, it still holds the old value, so a stop set from it at that moment is inherently too tight.
• Don't put your stop exactly where everyone can calculate it. Just below a round number, just below an obvious prior low: those are precisely the places liquidity gets actively hunted, which is the mechanism covered in our liquidation map guide. ATR gives you the distance; you still have to step away from the crowded levels.
• However sensible your stop, it can still get hit. ATR lowers the odds of being shaken out by everyday noise. It can do nothing about the instant liquidity dries up.

5. Turning ATR into position size: fix the loss first, then work back to size.

This section flips the usual order. You don't pick a leverage multiple first and then see how much you could lose. You decide the most this trade is allowed to lose, then work back from the stop distance to the size you should open. Leverage is the result of the calculation, not the starting choice.

Three steps, one line of arithmetic each:

① Risk amount = account equity × risk per trade
② Stop distance = k × ATR
③ Position size = risk amount ÷ stop distance

The "risk per trade" in step one is a rule you set for yourself. A common choice is around 1%; conservative traders use 0.5%, aggressive ones go up to 2%. ATR doesn't decide this number. What decides it is how deep a run of consecutive drawdowns your equity curve can take, which is exactly the question our Kelly criterion article deals with. ATR only handles steps two and three.

There's a detail here that's easy to miss: the 1,250 USDT notional value above and how much you stand to lose are two different things. How much you stand to lose is already capped at 50 USDT by the stop, provided the stop actually triggers. That proviso leads straight to the failure mode this whole method has to guard against.

While we're at it, let's separate the jobs of ATR and the Kelly criterion, because the two articles often get read together and confused:

• Kelly answers how big the risk budget should be, meaning the percentage in step one above. Its inputs are win rate and payoff ratio, and its output is a fraction. Keep in mind the conclusion of that article: in live trading, hardly anyone uses full Kelly.
• ATR answers how many units that budget buys at current volatility. It's a unit converter that turns a percentage into a concrete quantity.

The two don't conflict, and the order can't be reversed: the budget comes first, the conversion second. The site's position size calculator and Kelly calculator map onto these two steps. Enter your ATR reading and you get the quantity directly, no hand math needed.

6. Four ways crypto is different.

ATR was designed in 1978 for commodity futures. Carry it over to a crypto market that trades 24/7 and it bites in four places.

① There is no close, so "yesterday's close" is an artificial cut. Traditional markets have a clear closing price every day. Perpetual futures never stop trading, so the so-called daily close is just a slice taken out of a continuous price stream at some chosen moment. Cut it at 00:00 UTC or at midnight in your own time zone and you get a different previous close, and that price feeds straight into the second and third TR candidates. If two apps show different ATR(14) values for the same coin, the first thing to check is the time zone the daily candles are based on.

② Fewer gaps, more wicks. The largest-of-three design was originally aimed mainly at overnight gaps. Crypto has almost no overnight session, but it gets something else in exchange: deep wicks caused by liquidity vanishing in an instant. A wick enters TR through the bluntest term, the first one (high minus low). One wick can push TR to several times its normal size, and then the slow smoothing from section 3 drags that out over several candles. The result: for a while after a wick, ATR is inflated, and the size you calculate from it will come out noticeably too small. That isn't wrong, but you should know what you're paying for.

③ Same coin, different ATR on each exchange. ATR's inputs are each candle's high, low and close, and every exchange produces those three numbers from its own order matching. Different depth means wicks reach different depths in extreme moves, so the highs and lows naturally differ. It gets messier with perpetual futures, which run on two prices: the chart usually plots the last traded price, while liquidation is decided on the mark price. The gap between the two is widest exactly during a wick, which is exactly when you need them to agree. There's only one rule you can act on: calculate ATR from the candles of the exchange you actually trade on, and don't take exchange A's number to place orders on exchange B. It also means that before you start setting stops and size by ATR, you need to settle on which exchange you use. Switch venues and the same multiples and sizing formula need recalibrating. If you haven't decided yet, the data on this site mostly uses Binance candles as its reference, but you can pick any exchange. The point is to fix one and stop switching back and forth.

④ The quote currency decides whether ATR can be compared across assets. On USDT-margined contracts, ATR is in USDT and can be used directly as a dollar distance. On coin-margined contracts, profit and loss are counted in the coin, so the dollar risk behind the same ATR floats with the coin's price. And when you compare across assets, absolute values are never comparable: BTC's ATR runs to four digits, while a coin priced at 0.01 may have an ATR in the fourth decimal place. To compare them, normalize first:

ATR% = ATR ÷ current price × 100%

This percentage form, often written ATRP, is the right basis for comparing volatility across assets and across time. If you want to know how many times choppier an altcoin is than BTC, this ratio is the only thing that can tell you.

One last structural blind spot: ATR only sees price. It can't see leverage. When open interest is piled high and funding rates are at extremes, tail risk in the market has already risen, but as long as price hasn't started moving, ATR stays low. Quiet plus crowded is the combination most likely to end badly, and ATR is sensitive to the first half while being completely blind to the second. That's why ATR needs to be paired with other indicators, which is what section 8 lays out.

7. Five ways to misuse ATR, and how to avoid them.

① Using it as a directional signal. Rules like "buy when ATR crosses above its moving average" fall apart at the root. TR takes absolute values and ATR is always positive, so a violent rally and a violent sell-off look identical to it. A rising ATR only says the environment has gotten noisier. It says nothing at all about which way the noise is heading.

② Comparing absolute values across assets. "This coin's ATR is only 0.0004, so it's far safer than BTC at 1,200." That sentence means nothing, because the two numbers aren't even in the same units. To compare, convert to ATR% from the previous section first. A low-priced small coin, once converted to a percentage, is often a lot choppier than BTC.

③ Reading a shrinking ATR as "safe." This is the one that hurts most. Low volatility usually doesn't mean risk has gone away; it means energy is building up. The squeeze described in our Bollinger Bands guide is the same phenomenon seen through a different measure. A tight stop set from the reading during an ATR contraction is exactly the kind that gets swept out in one go on the candle where the breakout starts. The right way to read low ATR is "stops are cheap right now," not "it's safe right now." Those two statements are very far apart.

④ Using the same k for take-profit and stop-loss. Plenty of templates say "stop at 2× ATR, target at 2× ATR," which pins your reward-to-risk ratio at 1:1. That 1:1 didn't come from any analysis; it's a by-product of lazy parameters. Take-profit levels should come from price structure: prior highs, liquidity clusters, key round numbers. ATR's only job there is to set a floor for the target. If the target is less than 1× ATR away, the trade may not even have enough room to cover fees and slippage.

⑤ Changing the period whenever trades go badly. Changing 14 to 20, then to 10, is the easiest way to fool yourself. Changing the period means changing the ruler. Every stop distance and sizing formula that relied on ATR is now on a different scale, your past records are no longer comparable, and you'll never be able to tell whether the problem was the parameter or the strategy. If you're going to change it, lock it in before you start trading, and rerun your own trade history once you've changed it.

8. Where ATR sits in your decision process.

ATR isn't a signal source. It's a parameter at the execution layer. That's also why it's hard to backtest on its own: you can't backtest a ruler, only whether the stops and position sizes measured with that ruler made the whole strategy better or worse. Line up the indicators CoinView has covered by where they sit in the process, and the division of labor becomes clear:

ATR and Bollinger bandwidth deserve a closer look, because they often move together and get mistaken for the same thing. The difference in how they're calculated is covered in the FAQ below. Here's the part that's more useful in practice: when they split. The first case is a day with a wick but a close that barely moved. ATR ticks up while bandwidth hardly budges; that's when your stop should widen, and bandwidth alone won't show it. The second case is a market where price steadily shifts to a new level while daily ranges stay small. Bandwidth expands, and ATR won't necessarily follow. When the two split, use ATR to set your stop distance, because your stop has to survive the actual range, not the standard deviation of closing prices.

One principle about order is worth remembering: ATR never takes part in the decision to trade or not. It only takes part in how big and where to get out. Direction comes from the first two steps, ATR takes over at step three, and you always have to come back to step four and check the liquidation price. Break any link in that chain and all the work before it is wasted. For a fuller look at combining indicators, read our five indicator stacks article.

9. FAQ.

What to read next.

• BTC leverage liquidation case: why margin disappears before the chart looks fatal — once you've sized a position with ATR, come back here and check whether the liquidation price is closer than your stop.
• Kelly Criterion for crypto position sizing — Kelly sets the risk-budget percentage; ATR turns that percentage into a quantity.
• Bollinger Bands for crypto: the squeeze, the walk and why bands alone fail — bandwidth and ATR both measure volatility, one relative and one absolute, and the places where they diverge carry the most information.
• Open Interest Guide: The Leverage Still on the Table — ATR can't see leverage, and quiet plus crowded is the most dangerous combination.
• Position Size Calculator — enter your ATR reading and risk percentage and get the position size directly.
• Liquidation Calculator — for the cross-check in section 5: put the two prices side by side.
• Full disclosure and disclaimer

Crypto assets are volatile and not suitable for every investor. This page is editorial analysis, not financial advice.