An ATR trailing stop sits a fixed number of Average True Ranges below a price you choose: the close (close minus 2 x ATR) or the recent high (the 22-session highest high minus 3 x ATR, the "chandelier" version), and for a long position it only ever moves up. The point is to let each stock's own noise set the distance instead of a round number. At the October 16, 2026 close, that distance ran from 1.84% of price for $SPY to 9.02% for $SHOP.TO on the same 2 x ATR rule, and two of ten names had already closed below their chandelier level.
Below: the formula, the October 16 levels for ten US and Canadian tickers, one stop computed by hand from raw prices, and the mistakes that quietly break the method. More explainers live in the Analysis section.
The formula: true range first, then Wilder's average
Average True Range was developed by J. Welles Wilder and published in his 1978 book New Concepts in Technical Trading Systems, according to StockCharts ChartSchool. It is built in two steps.
- True range (TR) for each session is the largest of three numbers: high minus low, the absolute value of high minus the previous close, and the absolute value of low minus the previous close. The last two exist to capture gaps that a plain high-low range misses.
- ATR(14) starts as the simple average of the first 14 true ranges. After that, Wilder smooths it: today's ATR = (yesterday's ATR x 13 + today's TR) / 14.
Two stops are then built on top of ATR:
- Fixed-multiple stop: close minus 2 x ATR (or 3 x ATR).
- Chandelier-style stop: highest high of the last 22 sessions minus 3 x ATR. ChartSchool credits the chandelier exit to Charles Le Beau and notes it was featured in Alexander Elder's books. Its textbook setting uses ATR(22); the figures below use ATR(14), so treat them as chandelier-style rather than a carbon copy.
ATR stops for ten tickers at the October 16, 2026 close
Computed from Financial Modeling Prep end-of-day data. Canadian tickers are in Canadian dollars. A negative value in the last column means the close was already below the chandelier level.
| Ticker | Close | ATR(14) | ATR % | 2 x ATR stop | 22-day high | Chandelier | Close vs chandelier |
|---|---|---|---|---|---|---|---|
| $SPY | 763.25 | 7.01 | 0.92% | 749.24 | 781.62 | 760.60 | +0.3% |
| $QQQ | 709.80 | 10.55 | 1.49% | 688.71 | 762.86 | 731.22 | -3.0% |
| $NVDA | 214.13 | 5.97 | 2.79% | 202.19 | 243.37 | 225.46 | -5.3% |
| $AAPL | 343.46 | 6.70 | 1.95% | 330.06 | 348.15 | 328.05 | +4.5% |
| $TSLA | 354.72 | 11.85 | 3.34% | 331.02 | 388.56 | 353.02 | +0.5% |
| $JPM | 339.72 | 6.53 | 1.92% | 326.67 | 354.50 | 334.92 | +1.4% |
| $CNQ.TO | 76.10 | 1.79 | 2.36% | 72.51 | 76.25 | 70.87 | +6.9% |
| $SHOP.TO | 239.71 | 10.81 | 4.51% | 218.09 | 255.95 | 223.52 | +6.8% |
| $RY.TO | 280.01 | 4.06 | 1.45% | 271.90 | 290.56 | 278.39 | +0.6% |
| $XIU.TO | 54.17 | 0.62 | 1.14% | 52.94 | 54.57 | 52.72 | +2.7% |
What the table says, without spin:
- $QQQ and $NVDA had already tripped. $QQQ closed at 709.80, 3.0% below its chandelier level of 731.22. $NVDA closed at 214.13, 5.3% below 225.46. A long position run on this rule would already have been stopped out.
- $SPY was hanging by a thread. It closed 0.3% above its 760.60 chandelier. $TSLA (0.5%) and $RY.TO (0.6%) were barely clear as well.
- $CNQ.TO and $SHOP.TO had the most room, both near 7% above their chandelier levels. $CNQ.TO closed within 0.15 of its 22-session high.
Worked example: the $NVDA stop, step by step
Here is the October 16, 2026 $NVDA stop rebuilt from raw daily bars.
Step 1, true range. On October 14, $NVDA traded between 219.64 and 226.00 after closing at 228.52 the session before. High minus low was 6.36. High minus the prior close was 2.52. Low minus the prior close was 8.88. True range = 8.88, because the stock gapped lower: it opened at 225.53 and never got back to 228.52. A plain high-low range would have understated that day's move by 2.52.
On October 16, the bar ran 213.32 to 218.68 against a prior close of 217.23. The three candidates were 5.36, 1.45 and 3.91, so TR = 5.36.
Step 2, ATR. ATR(14) at the October 15 close was 6.0151. Applying Wilder's update: (6.0151 x 13 + 5.36) / 14 = 83.556 / 14 = 5.968, which rounds to 5.97.
Step 3, ATR %. 5.97 / 214.13 = 2.79% of price.
Step 4, the stops.
- 2 x ATR from the close: 214.13 - 2 x 5.97 = 202.19, about 5.6% below the close.
- Chandelier: the highest high of the 22 sessions from September 17 to October 16 was 243.37, set on October 6. 243.37 - 3 x 5.97 = 225.46.
The close of 214.13 sat below 225.46, so the chandelier exit had fired. On a closing basis it first did so on October 12, when $NVDA closed at 226.32 against a level of 226.59, then recovered for a day before closing at 221.30 on October 14.
2 x or 3 x ATR: choosing the multiplier
The multiplier is a straight trade-off between getting shaken out and giving back gains. At the October 16 close, a 2 x ATR stop sat 5.57% under $NVDA; 3 x ATR sat 8.36% under. On $XIU.TO the same choices meant 2.28% versus 3.41%.
- 2 x ATR exits faster, which suits shorter swing trades. It also gets triggered by ordinary noise more often.
- 3 x ATR survives more routine pullbacks but surrenders more open profit before it fires.
Even 3 x is no shield against whipsaw. $NVDA closed below its chandelier level from September 14 to 16, 2026 (210.96 against 213.63 on the 14th), then rallied to a 239.24 close on October 6. Anyone stopped out in mid-September missed that leg. No multiplier fixes this; it only moves where the trade-off lands.
Because the stop distance sets the risk per share, it also sets how many shares fit a fixed dollar risk. That arithmetic is covered in our position sizing formula guide.
Why ATR % beats ATR in dollars
ATR is measured in dollars, so on its own it says little about how jumpy a stock is. ChartSchool makes the same point: raw ATR values are not comparable across securities at very different prices. Divide by the close and they are.
On October 16, $XIU.TO, a roughly $54 ETF, had an ATR of 0.62, or 1.14% of price. $AAPL, at $343.46, had an ATR of 6.70, or 1.95%. Measured in percent, $AAPL was about 1.7 times as volatile, not the 11 times the dollar figures suggest. A flat $2 stop works out to about 3.2 ATRs on $XIU.TO but only 0.3 ATR on $AAPL. The first is a loose stop. The second is tight enough to trip on an ordinary afternoon.
ATR % also flags when volatility shifts. $QQQ's ATR climbed from 8.84 on October 13 to 10.55 on October 16, helped by a 20.77-point true range in the final session. $SHOP.TO's rose from 9.23 to 10.81 over the same days.
The ratchet rule, and the mistakes that break it
For a long position, the stop moves up or stays put. It never moves down. That matters more than it sounds, because the raw chandelier formula itself can fall: when ATR expands, the subtracted 3 x ATR gets bigger. $QQQ's raw chandelier level slid from 736.35 on October 13 to 731.22 on October 16 with the 22-day high unchanged at 762.86. That drop came entirely from the higher ATR. A ratcheted stop would have stayed at 736.35. $NVDA's raw level slipped from 226.59 to 225.46 over the same stretch.
The common mistakes:
- Comparing ATR in dollars across different prices. A 6.70 ATR on $AAPL and a 0.62 ATR on $XIU.TO tell you nothing until both are turned into percentages.
- Moving the stop down. Recomputing the formula every day and following it lower when volatility rises turns a trailing stop into a retreating one. Keep the highest level reached.
- Ignoring gaps. Building "ATR" from high-low ranges leaves out the gap component, as $NVDA's October 14 bar showed. Gaps also mean a stop order can fill well below the stop price. On September 14, $NVDA opened 3.2% below the prior close.
- Mixing up close and intraday triggers. A rule based on the close and a resting stop order based on intraday price will fire on different days. Choose one and backtest it.
- Treating a backtest as costless. Any test of these rules that leaves out commissions, slippage and taxes overstates the result.
For daily coverage of the tickers in the table, see the Traders On Trend homepage.
Bull/Bear Verdict
Bull Case: Where it helps: ATR stops scale to each stock's real volatility, so $SHOP.TO got a 9.02% cushion and $SPY 1.84% from the same 2 x rule on October 16, 2026. The ratchet locks in a mechanical exit before emotion gets a vote.
Bear Case: Where it misleads: ATR describes past noise, not future risk. Gaps can blow straight through any stop, and $NVDA's mid-September 2026 whipsaw shows even a 3 x chandelier can sell right before a rally. The multiplier is a judgment call, not science.
FAQ
What is the best ATR multiplier for a trailing stop?
There isn't one. The standard chandelier setting is 3 x ATR off the 22-day high, and 2 x from the close is a common tighter version. On October 16, 2026, those two choices put $NVDA's stop 5.57% or 8.36% below the close. Tighter means more whipsaw; wider means bigger give-back.
Should ATR be 14 or 22 periods?
Wilder's ATR is commonly run at 14 periods. The chandelier exit as published by StockCharts uses ATR(22). Longer settings react more slowly. Pick one and use it consistently.
Does an ATR trailing stop work for short positions?
Yes, mirrored: lowest low of the lookback plus a multiple of ATR, and the stop only moves down. Gaps against a short can be just as violent as gaps against a long.
Why did my ATR stop move lower?
If you recompute the formula each day, rising ATR pushes the level down. That is what happened to $QQQ between October 13 and 16, 2026. A proper trailing stop keeps the highest value it has reached.