Supertrend indicator: ATR formula, flips, accuracy and how to test it
Supertrend is a trailing line set a multiple of ATR away from the bar midpoint. It can only move toward price, so it holds through long trends and flips repeatedly in ranges. Judge it by payoff ratio after costs and by a grid of settings, not by one run's win rate.
By Soren Huang · Published · updated · Editorial policy
Contents
How does the Supertrend indicator work?
The Supertrend indicator is a trailing line built from Average True Range (ATR). It sits below price in an uptrend and above price in a downtrend, and it flips when the close crosses it. Because the line only moves toward price, it holds through long trends and flips back and forth in sideways markets.
Think of a dog on a lead that can be shortened but never lengthened. While the dog walks forward, you keep taking up slack. The moment the dog turns around and walks past you, you switch sides and start again. The Supertrend indicator works the same way: the line is the lead, price is the dog, and a flip happens only when the close walks through the line. That is why a Supertrend strategy belongs to the trend-following family, together with moving-average crosses and channel breakouts, and not to the oscillator family of the KD stochastic indicator or the RSI, which are built to fade moves inside a range.
The rest of this article makes one claim and defends it: the result of a Supertrend rule comes from how much of the tested history was trending, not from the indicator itself. The formula, the flip rule, the win rate and the settings grid all follow from that claim. For the wider picture of which tools belong in which family, the guide to technical analysis tools and indicators you can test sorts them by what they measure.
How is Supertrend calculated from ATR?
The Supertrend line is the bar midpoint plus or minus a multiplier times ATR, so the ATR period and the multiplier are the only two settings, and both control the distance between price and the line. ATR is the average, over a chosen number of bars, of each bar's true range, where the true range is the largest of three distances: high to low, high to the previous close, and low to the previous close. Wilder (1978) introduced it so that gaps between bars count as part of the day's movement. The ATR page covers the calculation in full; here it matters only as a measure of how far price typically travels in one bar.
The one formula you need:
upper band = midpoint + multiplier × ATR; lower band = midpoint − multiplier × ATR
Here the midpoint is half of the bar's high plus low, ATR is the Average True Range over the chosen period, and the multiplier is a plain number, often 3. In an uptrend the lower band is the line you see; in a downtrend the upper band is. Common default settings are an ATR period of 10 and a multiplier of 3, but the next sections explain why the defaults are a starting point for a test and not an answer.
Hypothetical example
A bar has a high of 120 and a low of 100, so the midpoint is 110. ATR over the last 10 bars is 8 points and the multiplier is 3, so the band distance is 3 × 8 = 24 points. The upper band is 110 + 24 = 134 and the lower band is 110 − 24 = 86. If the market is in an uptrend, the Supertrend line for this bar is the lower band at 86, as long as the ratchet rule in the next section does not hold it at an earlier, higher level.
Both settings do the same job from two directions. A longer ATR period smooths the band distance, so one wild bar changes it less. A larger multiplier pushes the line further from price. Either change produces fewer flips and later flips; either change in the other direction produces more flips and earlier flips. There is no third knob, which keeps the settings grid small.
How does Supertrend flip and trail?
On every new bar the active band may tighten or hold but never loosen, and the direction changes only when the close crosses that band, which is why flips arrive late and whipsaws come in clusters. The rule that stops the band from loosening is the ratchet rule. A ratchet is a tool that turns one way and locks against the other, and the band behaves the same: in an uptrend the lower band can rise toward price or stay where it is, but it cannot drop just because ATR grew or the bar midpoint fell.
- The bar closes. Nothing is decided on an open bar; the flip test uses the close.
- Compute ATR and both raw bands from the formula above.
- Apply the ratchet. In an uptrend, keep the higher of the new lower band and the previous lower band. In a downtrend, keep the lower of the new upper band and the previous upper band.
- Compare the close with the active band. In an uptrend the active band is the lower one; in a downtrend, the upper one.
- Hold or flip. If the close stayed on the trend side, the direction holds and the line trails. If the close crossed the band, the direction flips and the other band becomes the line, starting fresh from its raw value.
Two consequences follow directly from this order. First, flips are late by design: price has to travel the full band distance and then close beyond it, so the first part of every new trend is spent inside the old position. Second, whipsaws arrive in groups. A whipsaw is a flip that is reversed by the next flip before price has gone anywhere. After a flip the new band starts from its raw value, close to price, so a small move the other way crosses it again. In a range, price keeps crossing a line that keeps resetting near it, and one flip breeds the next.
Does Supertrend work: trend or chop?
A Supertrend rule takes its result from a few long trends and gives part of it back through many small whipsaw losses in ranges, so the net outcome depends on how much of the tested history was trending, not on the indicator. This is the claim the article rests on, and it is not special to Supertrend. Moskowitz, Ooi and Pedersen (2012) document that rules which stay with the recent direction of futures prices earn their result in the stretches where a move persists; any flip rule that holds a position until price reverses by a set distance shares that profile.
| What you look at | Trending stretch | Sideways stretch |
|---|---|---|
| Number of flips | Few; one position held for weeks or months | Many; a flip every few bars |
| Size of a single trade | Large, because the line trails far behind price | Small, roughly the band distance plus costs |
| Where the result comes from | One or two trades carry the stretch | Losses add up in small, regular steps |
| Effect of a wider multiplier | Gives back more at the end of the trend | Fewer whipsaws, each one larger |
| Effect of costs | Small share of a large trade | Large share of a small trade |
This is why the question "does Supertrend work" has no fixed answer and why a single backtest on a single period can mislead in both directions. A test that happens to cover a long directional stretch flatters the rule; a test that covers mostly ranges condemns it. The practical response is to test across more than one kind of market and to read the equity curve by stretch, not just the final number. It also explains why traders who work on short bars often rank oscillators above trend tools: the article on the best indicator for intraday trading shows that intraday bars spend more of their time in ranges, where a flip rule pays most of its whipsaw tax.
How accurate is Supertrend: win rate or payoff?
A trend-following flip rule produces many small losses and a few large wins by construction, so the number to check is the payoff ratio against the break-even line, not the win rate, and any single run's win rate is only that run's result. The payoff ratio is the average winning trade divided by the average losing trade. Expectancy is the average result per trade after wins and losses are both counted. The free win rate and expectancy calculator turns a win rate, an average win and an average loss into that per-trade figure, which is the number to compare with your round-trip cost.
The reason win rate alone misleads here is the shape of the trades. Every whipsaw is a loss of about the band distance plus costs, and there are many of them. Every trend trade is a win of many band distances, and there are few of them. A rule like this can lose more often than it wins and still have a positive expectancy before costs, as long as the wins are large enough. The break-even condition is simple to state in words: with a payoff ratio of 2, you break even when one trade in three wins; with a payoff ratio of 1, you need one in two; with a payoff ratio of 3, one in four is enough. Above that line the rule has a positive expectancy before costs; below it, a negative one.
Hypothetical example
A Supertrend rule produces 100 trades: 40 winners averaging $600 and 60 losers averaging $300. The payoff ratio is 600 ÷ 300 = 2.0 and the win rate is 40 percent. Winners total $24,000, losers total $18,000, so the expectancy before costs is $60 per trade. With a round-trip cost of $50 per trade (commission plus slippage on both sides), only $10 per trade remains. The rule is above the break-even line before costs and barely above it after them, which tells you the costs setting decides this test, not the indicator.
The second half of the section's claim matters just as much: a win rate is a property of one run. Change the market, the bar size, the period or the multiplier and the win rate moves with them. Quoting "Supertrend accuracy" as a single percentage treats a run's output as a property of the rule. The next section shows a real run with its own win rate, and it is quoted as exactly that, one run's result.
How sensitive are Supertrend settings?
Test a grid of ATR periods and multipliers and look for a broad region of similar results rather than one sharp peak, then judge the surviving setting after commission and slippage on enough trades. A parameter plateau is a block of neighbouring settings that all behave alike; a parameter peak is one setting that does well while its neighbours do not. Pardo (2008) treats the plateau as the sign that a setting describes a repeatable pattern and the peak as the sign that it describes the noise of one history. Bailey, Borwein, López de Prado and Zhu (2014) show the same thing from the other side: the more settings you try, the more the best one is inflated by chance alone.
- Write the grid before you run anything. For Supertrend a 3 × 3 grid of ATR period (7, 10, 14) by multiplier (2, 3, 4) is a reasonable start; a 5 × 5 grid is still small.
- Run every cell with the same costs, sessions and next-bar execution. A cell that only works with free fills is not a result.
- Read the shape, not the maximum. Shade the grid by net result or by profit factor and look for a block of similar cells.
- Pick from the middle of the plateau, not from the best cell, so that a small shift in market behaviour moves you to a neighbour that behaves the same.
- Count the trades in that cell. A wide multiplier on daily bars can leave you with a few dozen trades, which is too few to separate a pattern from luck.
A grid like this is also a search constraint in the plain sense: it fixes which entry and exit conditions are allowed and which parameter ranges they may take before any candidate is evaluated, which is the job of strategy generation in AlphaPrime (Valynox). The run below shows what a grid search produces when it is read honestly. It did not test Supertrend; it tested a stand-in Linear Regression rule, built from standard indicators, as the only entry and exit on ES daily bars from 2019-01-02 to 2025-06-30, long and short, one contract, with $20 commission and $25 slippage per trade. It is the one candidate the search kept out of the 1,024 it tried (AlphaPrime run of 2026-10-31), chosen for its return against its drawdown, and picking the kept candidate after the fact flatters it. The purpose here is only to show what a trend-style equity curve looks like over several years, including the stretches a reader has to sit through.
Source: AlphaPrime product engine (CPU), run 2026-10-31. ES daily bars, 2019-01-02 to 2025-06-30, long and short, 1 contract, $20 commission and $25 slippage per trade.
| Number | Result |
|---|---|
| Net profit (closed trades) | $88,920 |
| Max drawdown | $64,710 |
| Trades | 129 |
| Win rate | 55.04% |
| Profit factor | 1.331 |
| Average trade | $689.3 |
| Time in market | 86.29% |
Three things in that box are worth reading the way this article has argued. The time in market is 86.29 percent, which is what a rule that is always long or short looks like: it has a position almost all the time, and the flat stretches of the curve are positions that went nowhere. The win rate is 55.04 percent, and that figure belongs to this run of this stand-in rule on these bars; a Supertrend grid cell on the same data would show its own number. And 129 trades in six and a half years is enough to read the shape of the curve but not enough to trust a per-trade figure tightly, which is the trade count problem the daily-bar cells of any Supertrend grid will share. The same reading applies to the RSI indicator and how to use it or any other rule: one run's win rate is that run's property, the grid tells you whether the setting is stable, and the costs line tells you whether the average trade survives.
What to do next: load one market you trade, write down a 3 × 3 grid of ATR period and multiplier, run all nine cells with your real commission and slippage and next-bar execution, and shade the result. If the nine cells look alike, pick the middle one and count its trades. If one cell stands alone, treat the whole grid as a result about that history rather than about Supertrend.
Frequently asked questions
How is the Supertrend indicator calculated?
Supertrend takes the midpoint of each bar (half of high plus low) and moves it up and down by a multiplier times the Average True Range (ATR). The upper value is the band used in a downtrend and the lower value is the band used in an uptrend. A ratchet rule keeps the active band from loosening, and the direction flips when the close crosses it.
Does the Supertrend indicator work?
It follows long trends well and flips repeatedly in sideways markets, so its result depends on how much of the tested history was trending. A fair test runs the rule across several kinds of market with realistic commission and slippage and reads the equity curve by stretch. One run on one period can flatter or condemn the rule for reasons that have nothing to do with the indicator.
How accurate is the Supertrend indicator?
Accuracy, meaning win rate, is the wrong number for a trend-following flip rule, because the rule takes many small whipsaw losses and a few large trend wins by construction. Check the payoff ratio (average win divided by average loss) against the break-even line and then the average trade after costs. Any win rate you see is the result of one run on one market and period.
What are the best Supertrend indicator settings?
There is no single best pair. Common defaults are an ATR period of 10 and a multiplier of 3, but the right question is whether a block of neighbouring settings behaves alike on your market and bar size. Run a small grid such as periods 7, 10 and 14 against multipliers 2, 3 and 4, pick from the middle of the plateau, and confirm the cell has enough trades to judge.
References
- Wilder, J. W. (1978). New Concepts in Technical Trading Systems. Trend Research.
- Moskowitz, T. J., Ooi, Y. H., & Pedersen, L. H. (2012). Time Series Momentum. Journal of Financial Economics, 104(2), 228–250.
- Pardo, R. (2008). The Evaluation and Optimization of Trading Strategies (2nd ed.). John Wiley & Sons.
- Bailey, D. H., Borwein, J., López de Prado, M., & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the American Mathematical Society, 61(5), 458–471.