MEASURED · 21 SEPTEMBER 2026
Should a gold stop be fixed, or scaled to volatility?
Scale it to volatility, and expect a small gain. Everyone is told to size stops with ATR, and almost nobody is shown what it buys. We ran 55,678 trades with both rules risking the same average distance. The adaptive stop won on every measure except consistency at a 2:1 target, by less than you would expect, and the difference showed up in the fastest markets.
How far gold’s ATR swings
On 15-minute bars, ATR(14) had a median of $2.08 per ounce across twelve years. In the calmest tenth of the time it sat near $0.90, and in the fastest tenth it was above $7.04.
Calmest 10%
$0.90
ATR(14), per ounce
Median
$2.08
ATR(14), per ounce
Fastest 10%
$7.04
ATR(14), per ounce
That is roughly an eight-to-one range. A stop that gives a trade sensible room in a quiet market sits inside the noise in a fast one. We measured whether that costs you anything.
Most comparisons only measure stop width
A wider stop wins more often because it has more room before it is hit. Comparing “1.5 × ATR” against “$3” only tells you which one is wider. Almost every comparison you will read has this problem.
Here the fixed stop is calibrated. It is set to the mean ATR-scaled distance across the same entries, so both rules risk the same average amount: $4.78 in the headline run. Any difference left over comes from adapting to volatility.
The entries are deliberately worthless: one every two hours, long and short at the same time so any trend cancels. The stop rule is the only thing that varies, so the result can be put down to it. We checked the method first. The simulation returns exactly 50.0% on a random walk with a 1:1 target, and hand-built cases confirmed that the long, short and losing paths resolve correctly.
The result at 1.5 × ATR
Win rate by volatility decile. Decile 1 is the calmest market the sample contains, decile 10 the fastest.
| STOP RULE | D1 | D2 | D3 | D4 | D5 | D6 | D7 | D8 | D9 | D10 |
|---|---|---|---|---|---|---|---|---|---|---|
| ATR | 49.8% | 49.7% | 49.8% | 49.8% | 49.8% | 49.7% | 49.8% | 49.9% | 49.9% | 50.0% |
| Fixed | 49.9% | 49.8% | 49.7% | 49.7% | 49.8% | 49.8% | 49.8% | 49.9% | 49.4% | 44.5% |
For eight of the ten deciles the two rules are indistinguishable, both at roughly 49.8%. In the fastest tenth the fixed stop drops to 44.5% while the ATR stop holds at 50.0%.
So a fixed stop is fine almost all of the time and fails when the market is moving fastest, which is when the stop matters most and when you are least able to react.
Single candles swallow fixed stops in fast markets
An ambiguous trade is one where a single 15-minute bar covers both the stop and the target, so there is no way to know which was hit first. We count those as losses under both rules. How often they happen measures how small the stop is next to how far price travels in one bar.
| CONFIGURATION | ATR AMBIGUOUS | FIXED AMBIGUOUS | RATIO |
|---|---|---|---|
| 1.5 x ATR, RR 1.0 | 0.41% | 1.54% | 3.7× |
| 1.5 x ATR, RR 2.0 | 0.30% | 0.72% | 2.4× |
| 1.0 x ATR, RR 1.0 | 0.91% | 3.27% | 3.6× |
A fixed stop produces two to four times as many of these in every configuration. Those are fixed stops swallowed by single candles in fast markets, the same effect the decile table shows from another angle.
What held in all three configurations
We tested three configurations. On the headline measures, the adaptive stop won all three times:
| CONFIGURATION | WIN RATE (ATR / FIXED) | EXPECTANCY (ATR / FIXED) |
|---|---|---|
| 1.5 x ATR, RR 1.0 | 49.79% / 49.23% | -0.2350 / -0.2946 |
| 1.5 x ATR, RR 2.0 | 33.39% / 33.00% | -0.1933 / -0.2687 |
| 1.0 x ATR, RR 1.0 | 49.55% / 48.37% | -0.2468 / -0.3254 |
The win-rate gain is between 0.4 and 1.2 percentage points. It is consistent, and far smaller than the change ATR is usually sold as.
Every configuration lost money. The entries are random and every trade pays the spread. ATR sizing lost less and still lost. A stop rule changes the shape of your results and leaves their sign alone, which is what our risk-reward study found about target placement too.
At a 2:1 target the consistency result reverses
At a 2:1 target, the consistency advantage reverses. The fixed stop varied 1.5 points across the deciles against 2.6 for ATR, the opposite of the headline result on the same data.
The obvious suspect is trades running out of time before they resolve, which is what distorted the high-ratio rows in our risk-reward study. Essentially every trade resolved in all three runs, so truncation is ruled out.
We have no confirmed mechanism. Leaving this run out and reporting “ATR is more consistent” would have made a cleaner article and a false one. The claim that holds is narrower: ATR sizing gives a better win rate, better expectancy and far fewer single-bar wipeouts in every configuration tested. The large consistency gain is real at a 1:1 target and does not generalise.
The stop rule in our own system
In our own bot, the order-block strategy sets its stop at 1.5 × ATR with a 2:1 target, and any strategy signal whose stop would exceed 3.5 × ATR is rejected. That is the configuration in the second row of the tables above, chosen before this study existed and left unchanged by it. The other three strategies place the stop just beyond the signal candle or at the far side of the opening range.
In dollars: at gold’s median 15-minute ATR of $2.08, 1.5 × ATR is about $3.11 per ounce. In the fastest tenth of conditions the same rule gives roughly $10.56. A fixed stop cannot be both, and the gap between them is where the 5.4-point drop in the table comes from. You can work out the lot size for either with the lot size calculator. If you keep a fixed stop anyway, check how often a $3, $5 or $10 stop is hit within an hour in each session before you pick the number.
Limits
The data is one instrument at one broker, 2014-01-14 to 2026-09-21, on 15-minute bars. Ambiguous bars count as losses. That is pessimistic, and it applies to both rules the same way, so it cannot favour either. The entries are random by design. That isolates the stop rule cleanly and says nothing about how either rule works with an entry that has a real edge, which is probably the more useful question and a much harder one.
Only two multipliers and two target ratios were tested, and one of the three combinations run contradicted the consistency result. Treat the win-rate and expectancy findings as the durable ones. The full decile tables for all three runs are published with our other studies.
Common questions
What is the ATR of gold?
Measured on 15-minute XAUUSD bars from 2014-01-14 to 2026-09-21, ATR(14) had a median of $2.08 per ounce. It swings widely: near $0.90 in the calmest tenth of the time and above $7.04 in the fastest tenth. That range of roughly eight to one is the whole reason the fixed-versus-adaptive question exists.
Should I use ATR for my stop loss on gold?
On this evidence, yes, though the gain is smaller than usually implied. In all three configurations tested, an ATR-scaled stop gave a higher win rate, a better expectancy per trade and far fewer trades where one bar covered both the stop and the target. At 1.5 × ATR with a 1:1 target the win rate was 49.79% against 49.23% for a fixed stop risking the same average distance.
What ATR multiplier should I use for gold?
This study tested 1.0 and 1.5. The tighter 1.0 multiplier made the gap between the two approaches larger: a tight fixed stop is punished hardest when volatility rises, because it is the one most likely to sit inside the noise. In our own bot, the order-block strategy sets its stop at 1.5 × ATR, and the other three of the four strategies place it just beyond the signal candle or at the far side of the opening range. Any strategy signal whose stop would exceed 3.5 × ATR is rejected. That is what we settled on for our bot, and we do not offer it as a recommendation for your trading.
Does an ATR stop make a losing strategy profitable?
No. Every configuration tested had negative expectancy, because the entries were deliberately random and every trade still pays the spread. ATR sizing lost less, 0.2350 per trade against 0.2946, and it still lost. Position sizing changes the shape of your results. It does not change their sign.
When does a fixed stop actually fail?
In the fastest markets. With a 1:1 target, the fixed stop matched the ATR stop almost exactly through the calmest eight tenths of market conditions, then fell to 44.5% in the fastest tenth while the ATR stop held at 50.0%. Most of the time a fixed stop does as well. It falls behind when the market is moving fastest, which is when position sizing matters most.
Is the result the same at every risk-reward ratio?
No. At a 2:1 target the consistency advantage reversed: the fixed stop varied 1.5 points across volatility deciles against 2.6 for ATR. We checked whether trades running out of time explained it, as they did a similar artefact in our risk-reward study. They did not, because essentially every trade resolved. We have no confirmed explanation, so we report the result as it came out.