If you've spent any time on TradingView, you've seen it: a chart covered in green and red Supertrend labels, each Buy printed near a bottom, each Sell near a top. It looks like a cheat code.
We wanted the real number, not the impression. So we ported the classic Supertrend script (ATR period 10, multiplier 3 — the default settings used on millions of charts) and ran it over six months of minute-level broker data on three markets: Bitcoin, Gold (XAU/USD), and GBP/JPY. Every flip taken, no cherry-picking, spread costs charged on every trade.
This article shows everything — including the parts that make the indicator look bad. That's the point of how we test.
How we tested (so you can judge the method)
- Data: real broker minute candles, December 2025 → June 2026, aggregated to 1-hour, 4-hour, and daily bars.
- Rules: the exact Pine v4 Supertrend logic — hl2 source, Wilder ATR(10), multiplier 3. A Buy flip opens a long at that bar's close; the next Sell flip closes it and opens a short. No discretion, no filters.
- Risk accounting: every trade is measured in R — profit or loss relative to the indicator's own initial stop distance (the Supertrend line at entry). +1R means you made one unit of risk; −0.5R means you lost half.
- Costs: a round-trip spread was charged on every trade.
- Replication check: we split the six months into two halves. A result that only shows up in one half is treated as noise, not signal. This discipline regularly kills "great" results — ours included.
The results
| Market | Timeframe | Trades | Win rate | Avg R / trade | Total R | Survives split-half test? |
|---|---|---|---|---|---|---|
| Bitcoin | 1H | 95 | 41.1% | +0.10R | +9.7R | Yes (weakly) |
| Bitcoin | 4H | 30 | 33.3% | −0.28R | −8.4R | Consistently negative |
| Gold | 1H | 68 | 35.3% | +0.01R | +0.9R | No — flipped sign between halves |
| Gold | 4H | 10 | 70.0% | +0.84R | +8.4R | Yes — but only 10 trades |
| GBP/JPY | 1H | 73 | 38.4% | +0.03R | +2.2R | No — flipped sign between halves |
| GBP/JPY | 4H | 22 | 18.2% | −0.48R | −10.6R | Consistently negative |
Daily timeframe — the one where Supertrend charts look most impressive — produced only 2 to 4 completed trades per market in six months. That's not enough to conclude anything, which tells you something about those beautiful daily charts: you're looking at a handful of labels, placed by hindsight, on one trending stretch.
What the numbers actually say
1. Supertrend is not "accurate" — and it isn't trying to be. Win rates ran 18–41% on the timeframes with enough trades to judge. That's normal for a trend-following tool: it gets stopped into many small losses and aims to catch the occasional large move. Anyone selling Supertrend as a high-accuracy signal is describing a chart, not a track record.
2. The pain is real even when the total is positive. Bitcoin 1H finished at +9.7R — and along the way endured 16 consecutive losing trades. Most people abandon a system long before loss number sixteen. A backtest total you couldn't have emotionally survived is a fiction.
3. The best-looking result is the least trustworthy. Gold 4H: 70% win rate, +0.84R per trade. Exciting — and it's 10 trades. We've watched too many 10-trade hot streaks evaporate on the next sample to present this as anything but "interesting, unproven."
4. Several results did not replicate. Gold 1H and GBP/JPY 1H both flipped from profitable in one half of the window to unprofitable in the other — same settings, same market, adjacent months. This is the single most underrated fact in retail backtesting: a result that doesn't replicate isn't a result. It's noise wearing a costume.
5. Where it worked, it worked for a mundane reason. Bitcoin's positive 1H total came entirely from the short side: shorts earned +14.4R while longs lost 4.7R, during a window where Bitcoin fell ~31%. A trend follower riding a falling market is doing its one job — just don't confuse that with the indicator "knowing" anything.
So should you use Supertrend?
As an automatic money machine — no, the data above is clear enough. As a trend-state descriptor — it's a reasonable, transparent tool: it tells you which side of recent volatility price is on, with a built-in trailing stop. Some disciplined traders use it purely to define which direction they're allowed to trade, not as an entry signal.
But here's the more useful takeaway: the test mattered more than the indicator. Method beats opinion. Every "is X accurate?" debate on trading forums could be settled in seconds with the same procedure we used here — all flips, real costs, R-multiples, split-half replication.
Test your own strategy the same way
This study came out of the same research engine that powers our platform. The same honesty rules apply to everything on it: our own AI agents make their calls on a public, hash-stamped record — wins and losses both shown, nothing deleted — and our Research Lab applies the same baselines and sample-size warnings to your ideas that we applied to Supertrend above.
If you think you've got a strategy, put it on the record: try it free, no sign-up — or see the live track record our agents are building in public, right now.
Nothing here is financial advice. Past performance — the indicator's or anyone's — does not guarantee future results. That's rather the point of this article.
