Supertrend Strategies

The Supertrend is a volatility-anchored trend overlay built entirely from Average True Range: a single line that sits below price while the indicator reads the market as rising, and above it while it reads the market as falling. On each bar the calculation takes a midpoint — normally (high + low) / 2 — and offsets it by a multiple of ATR to form a provisional upper and lower band. Those bands are then locked: the active band may only move toward price, never away from it, until price closes through it. When that close occurs the state flips and the line jumps to the other side.

That locking rule is what the indicator actually is. Supertrend is not an oscillator and has no notion of overbought or oversold; it is a ratcheting trailing stop with a direction label attached. Two consequences shape most catalogued versions. The indicator is always in one of two states — there is no neutral reading — so the literal implementation is permanently in the market, long or short. And the ATR multiple decides how much adverse movement is tolerated before a flip, making it a trade-frequency control as much as a sensitivity setting.

## Parameters and what they control

Two numbers define the classic version: the ATR lookback and the multiplier, most commonly 10 and 3. The lookback governs how quickly band width responds to volatility — a short one reacts to individual spikes, a long one adapts slowly after a regime change. The multiplier sets the distance: a small multiple hugs price, flips often and exits early; a large multiple absorbs deep pullbacks, flips rarely and gives back more at turning points. The two interact, so adjusting one without re-checking the other is rarely meaningful.

## Main variants

- **Signal role** — the flip as a stop-and-reverse entry; the line as a trailing exit for entries generated elsewhere; or the state as a directional filter only. - **Multi-timeframe** — a higher-timeframe Supertrend supplies bias, a lower timeframe supplies the trigger. - **Stacked instances** — two to four Supertrends with different parameters, requiring agreement or scored as a consensus. - **Alternative input series** — Heikin Ashi candles, Renko bricks, or a smoothed series such as a DEMA replacing the raw midpoint. - **Adaptive and clustering versions** — the multiplier chosen dynamically from a volatility measure, or many multipliers evaluated in parallel and grouped by recent behaviour, which is the basis of the "AI" or k-means family. - **Oscillator conversions** — the distance between price and the line normalised into a bounded series read like a momentum gauge. - **Filter combinations** — trend-strength measures such as ADX/DMI, other trailing systems such as Parabolic SAR or SSL, moving averages, volume and range filters, volatility-compression measures. - **Non-directional applications** — the trend state used to select an options structure or scale exposure rather than open a spot position.

## What typically differentiates implementations

Far more than the two visible parameters. The **ATR smoothing method** — Wilder's recursive average, a simple average of true range, or an exponential one — produces different lines from identical nominal settings, and platforms disagree on the default; the **source series** matters for the same reason. **Confirmation** is decisive: an intrabar band cross, a confirmed close and a next-bar open are three different systems. So is whether the version stays always in the market or flat between signals. The **exit model** — opposite flip, fixed risk multiple, ATR target, partial scaling at the line — often dominates results more than the entry does. And the filter stack decides where the version is allowed to trade at all.

## Common mistakes

Backtesting on Heikin Ashi or Renko output is the most consequential: those series are derived, so orders fill at prices the market never traded. Acting on an unconfirmed intrabar state, or on a higher-timeframe series without correct offsetting, produces results that cannot be reproduced live. Treating the line as support to buy against inverts its meaning — it is an invalidation level, not a target. Stacking trend filters that measure the same thing adds complexity without information, since they fail together. Tuning the multiplier per instrument until an isolated peak appears fits the sample, not the market. And ignoring costs is especially damaging here: the multiplier controls trade count, and low multiples push a version into a frequency where spread and slippage dominate.

## How to evaluate and backtest a version

Pin the definition down first — ATR method, lookback, multiplier, source, confirmation rule, execution timing — then reproduce the line independently before trusting any result. Benchmark against the plain single Supertrend flip on the same data: every added filter, timeframe or instance should visibly earn its complexity. Examine the parameter surface across lookback and multiplier rather than one setting, and prefer a broad plateau to a spike. Segment by year and by volatility regime, and check how much of the equity curve comes from a handful of trades — trend-following rules concentrate their results in a minority of periods. Apply realistic commissions and slippage at the flip, and re-test any Heikin Ashi or Renko version on standard candles. Then validate out of sample or with walk-forward, on more than one instrument, examining long and short results separately.

The 23 decoded versions linked from this page differ along exactly these axes: role of the flip, timeframe structure, input series, adaptivity, filter stack and exit logic. Read side by side, they make it easier to separate the choices that change behaviour from those that only change the chart.

Strategies in this concept (42)

Frequently asked questions

What are the standard Supertrend settings?

The conventional configuration is an ATR lookback of 10 with a multiplier of 3, and it is a starting point rather than a rule. The two are coupled: shortening the lookback makes the ATR estimate noisier, so it usually calls for a different multiple to produce a comparable flip frequency. Because the multiplier also controls how many trades a version takes, changing it changes the cost profile as well as the sensitivity. Prefer a setting that sits inside a broad region of similar behaviour over an isolated optimum.

Does the Supertrend indicator repaint?

The closed-bar value does not change, but the value on the bar currently forming can flip and flip back before that bar closes. A version that acts on the live state is therefore not the same system as one that waits for a confirmed close, and only the second is straightforward to reproduce in a backtest. Multi-timeframe versions add a second source of the same problem: if the higher-timeframe series is read without correct offsetting, the strategy sees a value that was not yet available at that moment. Check which convention a version uses before comparing its results with anything else.

Why do two Supertrend implementations draw different lines with the same settings?

Usually because of the ATR calculation underneath. Wilder's recursive average, a simple average of true range and an exponential average all produce different band widths from the same lookback, and platforms and scripts do not agree on a default. The source used for the midpoint is a second cause, and running the indicator on Heikin Ashi or Renko series rather than standard candles is a third. When comparing versions, reconcile these before concluding that a parameter difference explains the gap.

Can Supertrend be used on Heikin Ashi or Renko charts?

It can be plotted on them, and several catalogued versions do, because the smoothed series produces fewer flips and a visually cleaner trend. The problem is testing and execution, not display: Heikin Ashi opens and closes are averages, and Renko bricks are constructed from a price threshold, so neither represents a price at which an order could actually have been filled. Signals derived from them should be executed and evaluated against real candle prices, and any version built this way is worth re-testing on standard bars to see how much of its behaviour was a charting artefact.

What is an adaptive or 'AI' Supertrend?

It is a family of versions in which the multiplier is not fixed. The simpler form ties it to a volatility or trend-strength measure, widening the band in fast conditions and tightening it in quiet ones. The form usually labelled 'AI' computes several Supertrends with different multipliers in parallel, scores each one on its recent behaviour, and groups those scores — commonly with k-means clustering — to select which setting is currently active. The label refers to that clustering step, not to a learned model of the market, and the approach introduces its own questions: the scoring window is itself a parameter, and the selection is made on recent data, so out-of-sample validation matters more here, not less.

Why does Supertrend struggle in sideways markets?

Because it has no flat state. The indicator is always labelled up or down, so a version that trades every flip is always positioned, and in a range the price repeatedly crosses the band without following through, reversing the position each time. Raising the multiplier reduces how often that happens but also delays exits when a real trend ends. This is why so many catalogued versions add either a trend-strength or volatility condition to stand aside, or use the line only as an exit while the entry comes from elsewhere. When evaluating a version, segment the results by regime — if the gains come entirely from a few trending stretches, the range behaviour is the part that needs the work.

All strategy concepts · Explore all strategies