Home / Strategies / Supertrend EA

Does the Supertrend Expert Advisor (EA) work? 10/3 backtested on 8 markets

By Jesse Lau · 2026-08-10 · Backtest period Aug 2024 – Aug 2026 (24 months)
Short answer: on H4 in a trending market, sometimes. Everywhere else, no. Supertrend(10, 3) lost on six of eight markets on H1 — US500 −23.79%, GBPUSD −21.67%. On H4 the same rules produced four grade-C results, led by USDJPY +27.19% (PF 1.61) and BTCUSD +21.27% (PF 1.37). But the out-of-sample profit factor came in below 1.0 on most of the winners, and every "improvement" we tried — the ADX filter everyone recommends, a blunter setting, a faster setting — made the results worse.

Supertrend is probably the most-shared indicator in retail trading groups: one line, green or red, no interpretation required. That clarity is exactly why it ends up inside so many Expert Advisors. Here is what it does when you run it unchanged across a basket of markets.

The rules we tested

Same rules on every market, no per-symbol tuning.

Results: 8 markets × 2 timeframes

MarketTF · configTradesReturnPFMax DDWinOOS trades / PFGrade
EURUSDH1 · Supertrend 10/3304-2.05%0.9919.32%33.9%69 / 0.93D 40.0
EURUSDH4 · Supertrend 10/382+1.71%1.0313.43%34.1%19 / 0.53D 38.2
GBPUSDH1 · Supertrend 10/3320-21.67%0.8727.48%31.6%65 / 0.84D 36.1
GBPUSDH4 · Supertrend 10/393+1.68%1.0313.53%34.4%23 / 0.55D 38.4
USDJPYH1 · Supertrend 10/3296+29.59%1.1619.58%37.2%55 / 1.07D 48.9
USDJPYH4 · Supertrend 10/378+27.19%1.619.08%43.6%17 / 0.6C 53.7
GBPJPYH1 · Supertrend 10/3240-16.04%0.8825.08%30.0%48 / 0.83D 37.0
GBPJPYH4 · Supertrend 10/366-13.48%0.6621.11%27.3%17 / 0.48F 33.2
XAUUSDH1 · Supertrend 10/3270-20.47%0.8530.41%30.0%55 / 0.92D 36.3
XAUUSDH4 · Supertrend 10/376+18.36%1.574.21%40.8%13 / 0.69C 55.1
BTCUSDH1 · Supertrend 10/3393+0.94%1.026.01%33.6%74 / 1.2D 42.8
BTCUSDH4 · Supertrend 10/3102+21.27%1.376.03%41.2%22 / 0.81C 53.4
NAS100H1 · Supertrend 10/3337-3.87%0.9823.9%33.2%73 / 1.2D 40.0
NAS100H4 · Supertrend 10/391+1.44%1.0315.99%33.0%16 / 0.84D 41.8
US500H1 · Supertrend 10/3329-23.79%0.8727.33%31.3%64 / 0.78D 35.0
US500H4 · Supertrend 10/389+13.35%1.257.37%39.3%14 / 1.44C 58.6

Backtest period Aug 2024 – Aug 2026 (24 months) · Dukascopy M1 data, UTC · $10,000 starting balance · final 20% of each run held out of sample · spread charged; commission and swap not modelled.

The same H1 / H4 split as every other strategy we've tested

On H1 the strategy takes 240–390 trades and loses on six of eight markets. On H4 it takes 66–102 trades and turns positive on six of eight. We saw the identical pattern in our RSI study — and that is a mean-reversion strategy, the structural opposite of this one.

Two opposite strategies failing on the same timeframe is worth pausing on. The common factor is not the logic; it is how often the logic is asked to be right. Supertrend's win rate sits at 30–34% on H1 and 33–44% on H4. At a 2:1 reward-to-risk target, break-even is 33.3%. H1 sits right on or under that line, and pays spread 300+ times to stay there.

The one big winner is a currency story, not a strategy story

USDJPY is the standout: +29.59% on H1 and +27.19% on H4 — the only market that worked on both timeframes. That period contained one of the strongest sustained yen trends in years. A trend-following system printing money during a historic trend is not evidence that the system is good; it is evidence that the trend was strong. Notice that GBPJPY, also a yen pair, was the worst H4 result in the table at −13.48%.

Read the out-of-sample column before you get excited

Every run holds back the final 20% of the period and scores it separately. Among the four grade-C H4 winners, the out-of-sample profit factor came in at 0.60 (USDJPY), 0.69 (XAUUSD), 0.81 (BTCUSD) and 1.44 (US500). Three of the four best results were losing in the segment the strategy had not been shaped around.

That is the single most important number on the page, and it is the one most strategy articles never show.

Do the popular "improvements" help?

MarketTF · configTradesReturnPFMax DDWinOOS trades / PFGrade
USDJPYH4 · + ADX > 2523+13.89%2.353.88%56.5%3 / n/aC 50.1
USDJPYH4 · 10 / 4.064+5.94%1.158.28%37.5%13 / 0.87D 48.2
USDJPYH4 · 7 / 2.0138+3.89%1.0515.68%37.0%27 / 0.3F 34.3
XAUUSDH4 · + ADX > 2536+2.57%1.136.17%38.9%5 / 0.06F 32.3
XAUUSDH4 · 10 / 4.051+7.25%1.296.12%39.2%12 / 0.82D 48.5
XAUUSDH4 · 7 / 2.0149+17.52%1.2310.37%41.6%34 / 0.71D 48.7
BTCUSDH4 · + ADX > 2539+5.92%1.287.26%43.6%9 / 0.45D 40.4
BTCUSDH4 · 10 / 4.068+3.85%1.111.1%36.8%12 / 0.18D 35.3
BTCUSDH4 · 7 / 2.0192-0.06%1.013.16%34.9%38 / 0.86D 40.0
US500H4 · + ADX > 2526+0.75%1.044.9%34.6%2 / n/aF 28.0
US500H4 · 10 / 4.066-6.94%0.8417.09%33.3%8 / 0.9D 40.0
US500H4 · 7 / 2.0167-11.24%0.8720.91%32.3%34 / 0.67D 36.1

The ADX filter hurt in all four cases. "Only take Supertrend flips when ADX is above 25" is the most commonly recommended add-on. It cut USDJPY from +27.19% to +13.89%, XAUUSD from +18.36% to +2.57%, BTCUSD from +21.27% to +5.92% and US500 from +13.35% to +0.75%. The profit factor often improved while total return collapsed — the filter removed trades, including winning ones.

A blunter setting (10 / 4.0) was worse in three of four. Fewer, larger swings, and the 2:1 target got hit less often.

A faster setting (7 / 2.0) was worse in three of four too, and it doubled or tripled the trade count while doing it — the classic way to convert a thin edge into a fee bill.

In other words: the stock 10/3 settings beat every variation we tried. That is not because 10/3 is magic. It is because tuning on the same data you evaluate on rarely produces improvement you can keep.

What we would actually take away

FAQ

What are the best Supertrend settings for an Expert Advisor (EA)?

On this data the default 10 / 3.0 beat both a blunter (10 / 4.0) and a faster (7 / 2.0) setting on three of four markets. That is a useful reminder that the widely used default is widely used for a reason — and that settings you discover by trying variants on the same history usually do not survive contact with new data.

Is Supertrend better than moving average crossover?

On our numbers they are cousins with the same weakness. The MA crossover study also topped out at grade C and also failed on the lower timeframe. Both are trend-following systems that need a sustained trend; neither has an edge in chop.

Why does Supertrend lose with a 2:1 reward-to-risk ratio?

Because 2:1 needs a 33.3% win rate just to break even before costs, and Supertrend's H1 win rates were 30–34%. It sits exactly on the line, then pays the spread several hundred times.

Can I build a Supertrend EA without coding?

Yes — describe the rules in a sentence, edit the numbers on a flowchart, run the backtest, and download the MetaTrader 5 source if the report holds up. The report is never hidden behind payment.

Try your own Supertrend variant.
Sweep the period and multiplier across a whole range in one run, and check the out-of-sample column before you trust any of it.