Supertrend Settings Tested: All 16 Lost on EURUSD M5, and the Default 10/3 Ranked 7th

Supertrend is one of the few indicators readers have found this site through from Google, via my Supertrend guide for MT4. That guide gives the standard advice: 10/3 is the real default, and tighten toward 14/2 for scalping. It did not have a backtest behind it.

This is the backtest. Sixteen settings, the same 16,926 bars, one rule, every flip counted. If one setting is meaningfully better than the default, it should show up here.

How was Supertrend tested?

The data is 16,926 EURUSD 5-minute bars from June 3 to August 26, 2026 — the same bars as my Parabolic SAR vs Supertrend comparison and every other EURUSD test here.

Supertrend is calculated in the standard way: a band of Average True Range — the same volatility measure behind my ATR trailing stop guide — around the bar’s midpoint, with the line only allowed to tighten until price closes through it. The rule is the simplest way to trade it — always in the market. When the line flips below price, close any short and buy. When it flips above, close any long and sell. Every flip is a trade, every trade pays 1 pip, the cost measured in my spread and swap test, and the direction is taken on the close of the bar, never on a forming candle.

The grid is ATR period 7, 10, 14 and 20, crossed with multiplier 1.5, 2, 3 and 4.

Which Supertrend setting did best on EURUSD M5?

Sorted by total pips. Not one row is positive, and — more important — not one row is positive in the gross column either. Before any spread, every Supertrend setting on EURUSD M5 lost money. A cheaper broker would make the losses smaller. It would not make any setting profitable.

The four multiplier-4 settings take the top four places. The default 10/3 sits seventh, in the middle of the multiplier-3 block. The four multiplier-1.5 settings take the bottom four.

Does the ATR period or the multiplier matter more?

Moving the multiplier from 1.5 to 4 cut the number of trades by about three quarters. Moving the period from 7 to 20 changed it by 20 trades. If you only change one Supertrend input, the multiplier is the one that does anything. The period is close to decoration.

That also explains the ranking. Because every setting loses per trade, the total is mostly a count of how often the setting trades. The multiplier-4 rows are at the top not because their trades are good, but because there are fewer of them.

Is 14/2 better than 10/3 for scalping?

It depends on what you count. Per trade, 14/2 lost 0.28 pips less. In total, it lost 827.8 pips more, because it flipped 540 more times and paid the spread on every one. For a scalper paying per trade, that is the number that reaches the account.

So the advice in my own MT4 guide does not survive the test. Tightening toward 14/2 made Supertrend busier, not better.

Would a different exit rescue the signal?

Always-in trading forces an exit on the next flip, which can be minutes later. So here is the default 10/3 flip used only as an entry, with a fixed hold instead:

Only one version was positive before costs: holding four hours, at +0.31 pips per trade, which the spread turned into -0.69. The flip on its own carries almost no information about the next few hours of EURUSD on a 5-minute chart.

The median Supertrend trend on these bars lasted 17 bars — 85 minutes. 37.1% of trends were over within an hour and only 10.1% lasted beyond four. That is the whole problem in two numbers: the line flips faster than the moves it is supposed to catch.

Does any setting hold up in both halves of the data?

The bars were split on July 15, 2026. EURUSD fell from 1.1609 to 1.14548 in the first half and rose to 1.16645 in the second, so each half has a direction of its own.

One positive half out of 32. The best setting overall, 14/4, owes its first place to a first half at +0.06 pips and then lost 2.84 pips per trade. That is what a best-in-grid result looks like when there is no edge underneath it: the winner changes when the data does. My in-sample vs out-of-sample test and optimised vs default settings found the same pattern in other indicators.

So is Supertrend useless?

As a stand-alone always-in system on EURUSD M5, in this data, yes. As something else, not necessarily:

• As a filter it helped. In my ADX vs Supertrend test, Supertrend 10/3 made a losing EMA system less bad, and in the Parabolic SAR comparison it improved an EMA crossover from -0.72 to -0.35 pips.
• It does not repaint on closed bars. I recalculated 450 points using only the bars available at the time; the direction changed later on 0 of them.
• On gold it did better and still failed. In my gold settings test all 16 settings lost in the month gold fell, and 10/3 lost $0.54 per trade.
• The timeframe moves the result more than the settings do. The same 10/3 behaves very differently once the bars get longer — that is the next test in this series, and the same question my MT4 vs MT5 comparison asked of another rule.

Frequently asked questions

What is the best Supertrend setting?

On 16,926 EURUSD 5-minute bars, 14/4 lost the least: -1.41 pips per trade and -622.5 pips over 440 trades after a 1-pip cost. No setting of the 16 tested was profitable, before or after costs, and 14/4 was only positive in one half of the data.

Is 10/3 the best Supertrend setting for forex?

It ranked 7th of 16 on EURUSD M5, losing 2.21 pips per trade over 738 trades. Every multiplier-4 setting finished above it. It is the most common default, not the best-tested one.

What Supertrend settings are best for scalping?

None in this test. The common scalping choice, 14/2, lost 1.93 pips per trade over 1,278 trades — less per trade than 10/3 but 827.8 pips more in total. Lower multipliers make Supertrend flip more, and every flip pays the spread.

What does the Supertrend multiplier do?

It sets how far the line sits from price in ATR units, and therefore how often it flips. At period 10, raising the multiplier from 1.5 to 4 cut trades from 1,778 to 458. Changing the ATR period from 7 to 20 at multiplier 3 changed trades by only 20.

Does Supertrend repaint?

Not on closed bars. Recalculating 450 points using only the data available at the time gave the same direction as the full history every time. On the bar that is still forming, the line can move until the bar closes, which is why every trade in this test used closed-bar signals.