Skip to content
Menu
Finabase Learn

Strategies · Trend and momentum

ATR trailing stop: a volatility-based exit rule, tested

An ATR trailing stop sells when the price falls a set multiple of its average true range below its best close. Exact rules, a QQQ backtest with costs, and where it fails.

On this page
  1. The idea
  2. The rules
  3. What the backtest shows
  4. Where it fails: the losing periods
  5. Compared with its neighbours
  6. How Finabase does it
  7. Frequently asked questions
  8. Summary
  9. Check your understanding

Most strategy rules say when to get in. A trailing stop is a rule for getting out: it follows the price up, never moves down, and sells when the price falls back to it. The open question is how far below the price to put it. A fixed distance such as 8% is too tight for a volatile stock and too loose for a quiet one. An ATR trailing stop sets the distance from the average true range, the typical size of one day's price movement, so the stop widens when the market is rough and tightens when it is calm.

This lesson writes the rule out, pairs it with a simple entry so it can be tested, and runs it on two years of daily prices. The test is short, and one of its results looks better than it deserves to. The page shows why.

The idea

A stop has two ways to be wrong. Too close, and ordinary daily movement sets it off before the trend has a chance. Too far, and it gives back most of the gain before it acts. What counts as "ordinary movement" differs between symbols and changes over time, so a stop measured in percent is right only by accident.

The average true range measures that ordinary movement directly. Each day's true range is the largest of three distances: the day's high minus its low, the high minus the previous close, and the previous close minus the low. The ATR is a smoothed average of the last 14 of them. J. Welles Wilder introduced it in 1978 and used it in a stop-and-reverse system. The version most traders use today hangs the stop from the highest price since entry, like a chandelier from a ceiling, which is why it is often called the chandelier exit.

The claim behind the rule is narrow. It does not say prices trend. It says that if you are going to exit on a pullback, the size of pullback that matters should be judged against how much the price normally moves.

The rules

On daily bars, checked after each close:

  1. Compute the 14-day ATR, with Wilder's smoothing.
  2. Buy at today's close if you hold cash and today's close is the highest of the last 20 closes.
  3. On the day of the buy, set the stop at the close minus 3 × ATR.
  4. Each day after, compute today's close minus 3 × today's ATR. If that is higher than the current stop, raise the stop to it. Never lower the stop.
  5. Sell at today's close if it is below the stop. Then wait for the next 20-day high.

The entry is the plainest breakout available, the same idea as the Donchian channel breakout on a shorter clock. It is there because an exit rule cannot be tested alone. The stop is the subject.

Two details differ from a broker's trailing stop order. The rule checks closing prices only, so an intraday dip below the stop that recovers by the close does nothing. And it trails the highest close, not the highest intraday price.

QQQ daily, 2024-12 to 2026-11: the trailing stop (3 × ATR below the highest close); shaded days are in cash $400 $500 $600 $700 QQQ close Trailing stop 5 10 15 20 ATR (14), dollars Jan 2025 Mar May Jul Sep Nov Jan 2026 Mar May Jul Sep Nov
QQQ daily, December 2024 to November 2026. The stop sits three ATRs below the highest close since the buy and only moves up. Shaded days are in cash. Bottom: the 14-day ATR in dollars.

A worked example

On May 1, 2025 QQQ closed at $481.68, its highest close in 20 days. The ATR was $14.81, which was large because April had been violent. Three ATRs is $44.43, so the first stop was $437.26, 9.2% below the price.

As QQQ rose, the stop rose with it, and as the market calmed, the ATR shrank and the stop drew closer. By early October the highest close was $611.44, the ATR was near $7, and the stop had ratcheted up to $593.35, about 3% below the peak. On October 10 QQQ closed at $589.50. That was below the stop, and the rule sold for a gain of 22.1% after costs.

The same trade shows the range of the stop's width. Over the test, the ATR ran from 0.9% to 4.4% of the price, so a three-ATR stop sat anywhere from under 3% to 13% below the best close.

What the backtest shows

The test, stated in full:

  • Data. QQQ daily bars from November 18, 2024 to November 17, 2026. The first 19 days build the 20-day high, so the test runs from December 16, 2024: 482 trading days, about 1.9 years.
  • Trading. The signal and the trade both use the day's close.
  • Costs. 0.10% of the amount on every buy and every sell.
  • Left out. Dividends, interest on cash and taxes.
  • What the period is. It contains one sharp fall, of 22.9% from February to April 2025, and a strong recovery. Two years is too short to show whether a rule has an edge. It shows how the rule behaves.
QQQ daily, 2024-12 to 2026-11: 20-day high with a 3-ATR trailing stop against buy and hold, 0.10% per trade $0.8 $1 $1.2 $1.4 Growth of $1 Buy and hold 3-ATR trailing stop 0% −10% −20% −30% Fall from the peak Jan 2025 Mar May Jul Sep Nov Jan 2026 Mar May Jul Sep Nov
QQQ daily, December 2024 to November 2026: a 20-day high entry with a 3-ATR trailing stop against buy and hold, 0.10% per trade, dividends not included. Top: growth of $1. Bottom: the fall from the highest value so far.
Measure 3-ATR trailing stop Buy and hold
Total return 24.8% 22.1%
Return per year (CAGR) 12.3% 11.0%
Deepest fall from a peak −9.2% −22.9%
Longest time below a peak 120 days 113 days
Time invested 57% 100%
Round-trip trades 7 1
Trades with a gain 29% n/a

All seven trades:

Bought Price Sold Price Result after costs
Dec 16, 2024 $538.17 Dec 18, 2024 $516.47 −4.2%
Jan 22, 2025 $531.51 Feb 25, 2025 $513.32 −3.6%
May 1, 2025 $481.68 Oct 10, 2025 $589.50 +22.1%
Oct 20, 2025 $611.54 Nov 7, 2025 $609.74 −0.5%
Dec 3, 2025 $623.52 Feb 4, 2026 $605.75 −3.0%
Apr 9, 2026 $610.19 Jun 5, 2026 $705.06 +15.3%
Aug 13, 2026 $732.07 Oct 14, 2026 $728.95 −0.6%

Two trades carry the whole result. Five of seven lost. This is the usual shape of trend following: many small losses paid for by a few large gains. A trader who skipped the May 2025 buy because the last two had lost would have ended well behind.

The stop was out before the fall. The rule sold on February 25, 2025 at $513.32 and bought back on May 1 at $481.68. In between, QQQ fell to its April low. That one stretch in cash is the reason the rule's deepest fall is 9.2% and not 22.9%, and the reason it finished ahead.

The first trade shows what a stop cannot do. On December 18, 2024 the stop was $519.76. QQQ fell 3.6% that day and closed at $516.47. The rule sold at the close, below its stop. A stop limits a loss only to the next price at which a trade can happen.

Costs were small. Fourteen trades at 0.10% took about 1.5 percentage points. With no costs the rule gains 26.4%.

The two halves of the period

Period 3-ATR trailing stop Buy and hold
December 2024 – December 2025 12.2% 15.9%
December 2025 – November 2026 11.2% 5.4%

These are total returns for each half. In the first half the rule avoided the spring fall but re-entered 6% below where it had sold and then stood aside for parts of the recovery, and it trailed. In the second half it sat out much of February to April 2026 and sold on October 14, 2026, before most of a 14.8% fall in October and November, and came out ahead.

Where it fails: the losing periods

The result is a peak among its neighbours. The same entry with other stop widths, on QQQ over the same days. Buy and hold gained 22.1%.

Stop distance Total return Deepest fall Trades Time invested
1.5 × ATR 8.7% −9.9% 14 40%
2 × ATR 4.6% −12.9% 13 46%
3 × ATR 24.8% −9.2% 7 57%
4 × ATR 17.4% −11.0% 6 63%
5 × ATR 12.0% −12.6% 6 70%

Three is the best row by a wide margin, and it is the only row that beat holding. When a result drops sharply on either side of the chosen setting, the setting has probably fitted the particular days in the sample. With seven trades, moving one exit by a day changes the outcome. The honest reading of the table is "somewhere between 5% and 25%, with a smaller deepest fall than holding in every row", not "24.8%". The overfitting lesson explains why the best row is the one to trust least.

Other funds. With the three-ATR stop on the same days, SPY gained 11.0% against 20.7% for holding, and IWM gained 15.4% against 14.7%. On five years of weekly bars, with a 20-week high and a three-ATR stop, the rule trailed on all three funds: 36.2% against 66.6% on SPY, 45.5% against 87.9% on QQQ, and 14.9% against 33.7% on IWM. In each case its deepest fall was smaller.

Individual stocks. The same daily rule on each of the 30 stocks in the Dow Jones Industrial Average as of late 2024:

Stop Average return, rule Average return, hold Stocks where the rule did better Trades with a gain
2 × ATR 5.6% 20.7% 8 of 30 41%
3 × ATR 7.6% 20.7% 8 of 30 39%
4 × ATR 11.3% 20.7% 11 of 30 42%
Fixed 8% 10.9% 20.7% 11 of 30 40%

With three ATRs, results ran from −40.8% on Salesforce to +112.5% on Caterpillar, and the rule's deepest fall was smaller than holding's on 19 of the 30. These are today's index members, so both columns benefit from hindsight.

The ATR did not beat a plain percentage. On the stocks, a fixed 8% stop did about as well as four ATRs and better than three. On QQQ, fixed stops of 3%, 5%, 8% and 10% gained 9.1%, 18.9%, 16.9% and 2.8%. Nothing in this data shows that scaling the stop by volatility adds return. What it does is make one setting usable across symbols with very different volatility, which matters for position sizing: if the stop is three ATRs away, the amount at risk per share is known when the trade is opened.

Sideways markets. Each new 20-day high that fails costs the distance to the stop plus costs. From October 2025 to February 2026 the rule bought twice near the top of a range and was stopped out twice.

What research says about stops. Kaminski and Lo (2014) show that a stop-loss rule can add return only when returns are predictable, for example when they have momentum (falls tend to be followed by more falls) or switch between calm and stressed regimes. If price changes are independent, a stop lowers the expected return by the time spent out of the market, and it adds trading costs.

Other limits:

  • Taxes. Most of these trades lasted under a year, so gains are short-term in a US taxable account (IRS Topic 409).
  • Orders. A trailing stop order resting with a broker triggers on intraday prices and becomes a market order, which can fill well below the stop on a gap (see the SEC bulletin in the sources). The tested rule acts on closes only and behaves differently.

Compared with its neighbours

Exit rule Sells when Adjusts to volatility In this test, QQQ
ATR trailing stop (3 ×) Close falls three ATRs below the highest close Yes 24.8%
Fixed-percent trailing stop (8%) Close falls 8% below the highest close No 16.9%
Channel exit Close sets a new low of the last so many days Partly Not run here
Moving average exit Close falls below an average No Not run here

The moving average rule and the channel rule use the same signal family for entry and exit. A trailing stop separates the two decisions, so it can be attached to any entry.

How Finabase does it

No Finabase tool places or manages a stop for you. These show the ingredients.

  • Chart. The link above opens two years of QQQ. Add the ATR study from the indicator list to see the value the stop is built from; the ATR lesson explains the reading.
  • Backtest. The Backtest tool trades the signals of the indicators it offers on one symbol. It does not run this trailing-stop rule. Its numbers also differ in kind from this page's: each signal buys a fixed $10,000 and no trading cost is charged.
  • Strategy. The Strategy tool replays if/then allocation rules over two years of daily prices, the same span as the main test here.

Frequently asked questions

Why three times the ATR?

Convention. Multiples between two and four are the common choices, and three is widely quoted for daily bars. In this test three happened to do best on QQQ and was not best on the 30 stocks, which is a reason to treat the multiple as a rough choice and not a tuned one.

Does a trailing stop limit my loss to the stop distance?

No. It limits the loss to the next price at which the position can be sold. In the first trade of this test the close was 0.6% below the stop. On a single stock after an earnings report the gap can be far larger.

Is the ATR stop better than a percentage stop?

Not on return in this data. Its advantage is consistency: three ATRs means "three typical days of movement" on any symbol, while 8% is a wide stop for a utility and a narrow one for a small technology stock.

Should the stop use the highest close or the highest high?

Both versions are in use. Trailing from the highest high puts the stop slightly higher and sells slightly sooner. This page uses closes throughout so that every number can be checked against a closing-price chart.

Summary

An ATR trailing stop sells when the price closes more than a set multiple of its average true range below the best close since entry, and it never moves down. Paired with a 20-day-high entry on QQQ from December 2024 to November 2026, a three-ATR stop gained 24.8% after costs against 22.1% for holding, with a deepest fall of 9.2% against 22.9%, on two winning trades out of seven. Neighbouring multiples gained between 4.6% and 17.4%, the rule trailed on SPY, on five years of weekly bars and on the average Dow stock, and a fixed percentage did as well. What held up across every version was a smaller deepest fall on the funds, paid for with time out of the market.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. A stock is bought at $100 when its 14-day ATR is $2. Where does a three-ATR trailing stop start?

    1. $98
    2. $94
    3. $97
    Show the answer

    B. $94 Three times $2 is $6, and $100 minus $6 is $94. The stop then follows the highest close upward and never moves down.

  2. The stock rises to a highest close of $120 and its ATR shrinks to $1.50. Then it closes at $117. What happens?

    1. The rule sells, because $117 is below the first stop plus the gain
    2. Nothing: the stop is near $115.50 and the close is above it
    3. The stop is lowered to give the trade room
    Show the answer

    B. Nothing: the stop is near $115.50 and the close is above it Three ATRs is now $4.50, so the stop has ratcheted up to about $115.50. A close of $117 is above it. The stop is never lowered.

  3. On QQQ the three-ATR stop gained 24.8%, while two ATRs gained 4.6% and four ATRs 17.4%. What does that pattern suggest?

    1. Three is the correct multiple
    2. The best result probably fits this sample; with seven trades, small changes move the outcome a lot
    3. Larger multiples are always worse
    Show the answer

    B. The best result probably fits this sample; with seven trades, small changes move the outcome a lot A result that falls away sharply on both sides of the chosen setting is a warning sign of fitting to the sample. On the 30 Dow stocks, three was not the best multiple.

  4. On December 18, 2024 the stop was $519.76 and the rule sold at $516.47. Why below the stop?

    1. The price fell through the stop during the day, and the rule sells at the next price it can, the close
    2. The rule made an error
    3. Costs moved the stop
    Show the answer

    A. The price fell through the stop during the day, and the rule sells at the next price it can, the close A stop is a trigger, not a guaranteed price. The sale happens at the next available price, which after a fast fall or a gap can be well below the stop.

Sources and further reading

  1. J. Welles Wilder Jr., New Concepts in Technical Trading Systems (Trend Research, 1978) — average true range and the volatility system
  2. Alexander Elder, Come Into My Trading Room (Wiley, 2002) — describes the chandelier exit and credits it to Chuck LeBeau
  3. Kathryn M. Kaminski and Andrew W. Lo, “When Do Stop-Loss Rules Stop Losses?”, Journal of Financial Markets 18, 2014, pp. 234–254
  4. SEC Investor Bulletin — Stop, stop-limit and trailing stop orders
  5. Finabase chart data — daily bars for QQQ, SPY, IWM and 30 Dow stocks, November 2024 to November 2026, and weekly bars for five years; rules and results computed by Finabase