What is the Supertrend indicator?

Strip the colour away and Supertrend has exactly three components, and once you can name them the indicator is transparent. The first is a band: a level placed at a volatility-scaled distance from the middle of the bar, using the Average True Range to decide how far away that should be. The second is a ratchet: a rule that permits the band to move toward price and forbids it from moving away. The third is a flip: one condition, a close on the far side of the band, that reverses the state and throws the line to the opposite side of price. Everything else, the green and red, the arrows some platforms paint, the alerts, is presentation layered on top of those three parts.

That anatomy already tells you what the tool can and cannot answer. Because the band is built from ATR, it knows roughly how much this instrument moves and nothing whatsoever about why. Because the ratchet only tightens, the line is a stop rather than a forecast: it marks the level at which the trend it is currently tracking would be declared over. And because the flip requires a close, the state is always a statement about a bar that has already finished. Supertrend is a summary, delivered late, of a question that has been restricted to two permitted answers. It is not a prediction and was never designed as one. Its originator built it as an "always-in" stop-and-reverse, a system that is never flat and always carries a live exit for whichever side it happens to be on.

The figure below is that machine, computed. The Supertrend line drawn on it was not placed by hand to make a point: the ATR, the basic bands, the final bands and the flips were all calculated from the price series shown, at the standard settings of a 10-bar ATR and a multiplier of 3. Both regimes share a scale for the simplest possible reason, which is that they are one series. The same instrument, the same indicator, the same settings run straight through the divider. Only the market changes.

Read the two halves against each other. Across the 46 trending bars the line produced zero flips. It ratcheted upward in steps, held dead flat through four separate pullbacks, and never came close to being broken: the tightest of those four pullbacks still closed almost 25 points clear of the line. This is the tool doing precisely what it was built for, and doing it well. Then the trend ends, and across the following 75 directionless bars the same line at the same settings flips four times. The first flip was honest, because the trend really had finished, though it arrived 52 points below the high. The other three were noise. Every one of the four legs in that stretch lost money, and between them they gave back nearly twice what the trend leg had captured.

Supertrend takes two inputs and runs them through four steps. The inputs are an ATR period, conventionally 10, and a multiplier, conventionally 3. ATR is a volatility measure, the average size of a bar's true range over that period, so it is large when the instrument is swinging and small when it is calm. Feeding ATR into the band is what makes Supertrend volatility-adaptive: the line automatically sits wider from price in turbulent conditions and tighter when things settle, without anyone touching a setting. That property is genuinely useful and is a large part of why the indicator has survived.

The steps run as follows. First the bar's midpoint, HL2, the average of the high and the low. Then two basic bands, one above and one below, at the multiplier times ATR from that midpoint. Then the final bands, which is where the ratchet lives and which most descriptions skip or garble. Then the line itself, which is the final lower band while the state is up and the final upper band while the state is down, and the flip that switches between them. The table below gives each step alongside what it produced on the series in the figure above, so the formula and its output can be read together.

Two details in that table are worth pausing on, because they are where the indicator's character actually comes from. The first is that the flip requires a close, not a touch. An intrabar spike straight through the line that does not hold into the close leaves the state completely unchanged, which is a deliberate and sensible noise filter. The second is the pair of final-band rules in step 4. Left as raw basic bands, the line would wobble up and down with every bar's midpoint and every twitch in ATR, which is useless as a stop and is not what any charting platform draws. The max and its mirror are what convert the envelope into something a trader can actually lean on, and they deserve their own figure.

The ratchet is one function call and it is the whole idea. While the state is up, the final lower band is set to the higher of the new basic lower band and its own previous value. It can therefore rise or hold flat, but it can never fall. The upper band is the exact mirror: while the state is down it takes the lower of the new basic upper band and its previous value, so it can fall or hold but never rise. The band tightens toward price as the trend runs and then locks. This is the behaviour of a trailing stop, and Supertrend is best understood as one that has been automated and given a reverse gear.

Described in prose that sounds like a technicality. Watched on real bars it is the entire difference between a volatility band and a usable exit, so the figure below zooms in on twelve bars of the same series and shows both lines at once: the basic band that ATR alone would give you, and the final band the rule actually allows.

Follow the two lines through the pullback. At bar 28 the basic band and the final band agree at 1036.21, because price has been rising and the ratchet has been stepping up with it. Then price pulls back. ATR ticks up slightly, the midpoint falls, and the basic band does what an unconstrained volatility band does: it drops, all the way to 1023.48 by bar 32, a full 12.7 points below where the indicator's line actually sits. The final band ignores every one of those bars. It holds at exactly 1036.21 for seven consecutive bars, and only at bar 36 does a higher basic band arrive and lift it again.

All of that downward movement was thrown away on purpose, and the reason is worth stating plainly: a stop that can move down is not a stop. If the line were allowed to follow ATR lower, then every pullback would widen your risk at the exact moment the trade was going against you, which is the single most common way a discretionary trader turns a small loss into a large one. The ratchet makes that structurally impossible. Whatever else you conclude about Supertrend, this part of it is well built. It also explains the flat stretches that beginners often read as the indicator being broken or lagging. Those flat stretches are the tool working: it is not that the line has failed to update, it is that the rule has refused an update that would have loosened your risk.

Note what the ratchet does not do. It does not decide whether the trend is real, it does not measure the trend's strength, and it does not know that a pullback is a pullback rather than the beginning of a reversal. It has exactly one opinion, held with total consistency: the exit may improve and may never worsen. Only a close through the line releases it, and when that happens the state flips and a fresh band starts ratcheting the other way.

The flip is where the bill arrives. Because the band sits roughly three ATRs from the midpoint, and because the state only changes on a close beyond it, the line cannot possibly confirm a turn until price has already travelled about that distance. That lateness is not a defect to be tuned out. It is the price of the noise filter, and it is the same distance that let the line hold through four pullbacks in the trend. You are buying the right to ignore small adverse moves, and the currency you pay in is a portion of every large one.

On the series above, that arithmetic is exact rather than rhetorical. The move topped at 1128. The line did not flip until a close at 1076, seven bars later. The 52 points in between, about 4.6% of price, is what the indicator conceded to find out that the trend was over. Divide that giveback by the ATR at the flip bar and you get 3.63, against a multiplier of 3: the giveback is the band, plus the overshoot of whatever bar finally broke it. The relationship is not a coincidence, it is the definition. Whatever multiplier you choose, you are choosing how many ATRs you will hand back at every turn.

The table below is the whole always-in system on that series, leg by leg, as the recursion produced it. It is worth reading as an argument rather than as data, because the shape of it is the shape of every mechanical trend follower ever built.

One leg made 67 points. Four consecutive legs then lost 34.5, 27.0, 34.6 and 30.0. This is not an unlucky sample or a badly chosen setting; it is the mechanism. A trend follower is a machine that pays a small, certain toll over and over in order to be present for the rare large move, and it is solvent only when the large moves are large enough and frequent enough to cover the tolls. Supertrend inherits that arithmetic wholesale, with one aggravating feature of its own: because it is always in, it cannot decline to trade. A discretionary trader looking at those 75 directionless bars would very likely have stood aside. Supertrend structurally cannot. It has no flat.

Supertrend exposes two settings, and they are usually presented as a matched pair to be tuned together. Measuring them says otherwise. Sweeping each dial across the same series, holding the other fixed, shows that one of them does nearly all of the work and the other barely registers.

The left panel is the multiplier, and the finding is as clean as this kind of thing ever gets. As the band widens from 1.0 to 5.0, the flip count falls monotonically from 13 to zero, and the average giveback climbs the other way from 14.9 points to over 40. The two curves move in opposite directions across the entire sweep, without a single exception, and that is not a quirk of this series: it is forced by the geometry. A tighter band is closer to price, so it is crossed sooner, which means more flips and less distance conceded at each one. A wider band is further away, so it is crossed later, which means fewer flips and more distance conceded. Whipsaw and giveback are the same quantity viewed from opposite ends. You cannot reduce one without increasing the other, because they are both just the band.

The right panel is the ATR period, and it is the more surprising result. Sweeping it from 5 to 30, a six-fold change, moves the flip count only from 4 to 6, and the mean ATR from 11.49 to 11.58, which is less than a tenth of a point. The reason is straightforward once seen: the period governs how quickly ATR responds to a change in volatility, not what its average level is. Since the band's distance is a multiple of that level, changing the period reshuffles the timing of small adjustments while leaving the band essentially where it was. The practical consequence is worth stating bluntly, because it is the opposite of what a lot of tinkering assumes: if you are trying to fix a whipsaw problem by adjusting the ATR period, you are turning the dial that is not connected to it.

There is also a dead zone at the right of the multiplier sweep that deserves more attention than it usually gets. Past about 4.25 on this series the line stops flipping altogether. That is not the whipsaw problem solved; it is the indicator switched off. A band wide enough never to be hit is a band that will hold a stale state for months and tell you the trend is up while the instrument grinds down through it. This is the failure mode that gets discussed far less than whipsaw and is frequently more expensive, because whipsaw announces itself with a run of small losses whereas a stale state is silent. It also destroys the naive intuition that quiet markets are where Supertrend suffers. A genuinely quiet range does not whipsaw the default settings at all: it leaves the band outside price entirely and the line simply goes to sleep. What actually shreds a 3-ATR band is a market that is volatile and directionless at the same time, making swings long enough to reach the band and reversing before they go anywhere.

Every flip is a stop-and-reverse, which means a completed round trip: one entry, one exit, both sides of the levy stack. It is common to hear from this that a choppy market is a fee-generating machine. The claim is worth checking against real rates rather than repeating, and when you do check it, it turns out to be true only in a specific place and false in the place it is most often asserted.

An always-in system has to be able to go short, which in practice means it trades futures rather than cash. Running an illustrative futures round trip of ₹5,00,000 a leg through the same rate model as our cost estimator, the full statutory stack comes to about ₹330: STT of ₹250 on the sell side, brokerage of ₹40, exchange transaction charges of ₹18.30, GST of ₹10.67 on the correct base, stamp duty of ₹10 on the buy, and the SEBI turnover fee of ₹1. That is 0.066% of the position, or about 0.73 points on an instrument trading near ₹1,100. All figures illustrative; statutory rates are as of 17 July 2026 and can change, so verify current rates with your broker before relying on them.

Now set that against what the flip actually cost. The giveback averaged 40.2 points per leg. The toll is 0.73. The statutory bill is under 2% of what a flip took out of you, and the other 98% is the band, the three ATRs of travel the indicator demanded before it would concede the turn. On a daily chart, the fee is simply not the problem. Anyone who tells you Supertrend fails in chop because of brokerage and taxes has not done the arithmetic; it fails in chop because the band is wide and the market keeps crossing it without going anywhere.

But there is a real version of the cost argument, and it is a ratio rather than a slogan. The toll is a fixed fraction of turnover. The band is not. The toll is 0.066% of the position whether you trade once a quarter or thirty times a day, because STT and exchange charges do not care about your holding period. The band, meanwhile, is three ATRs, and ATR shrinks as the bar shrinks. So the identical flip that surrenders about 2% of its band to the exchequer on a daily chart surrenders about 22% of it on a 5-minute chart. The lower panel of the figure is that ratio, and it is the whole reason low-timeframe mechanical trend following is so much harder than it looks: the edge per flip contracts with the clock while the toll sits exactly where it is.

The instrument choice moves the same lever. Because the toll is proportional to turnover but the rates differ sharply by segment, the identical strategy pays very different tolls depending on where it is expressed.

An always-in system on options pays roughly 3.7 times the futures toll on the same turnover, which is the arithmetic behind a regulatory point worth keeping in view: about 93% of individual traders in equity derivatives made net losses over FY22 to FY24, aggregate net losses exceeding ₹1.8 lakh crore (SEBI, September 2024). That figure is not an indictment of any indicator. It is a reminder that the segment where mechanical flip-trading is most popular is also the segment where the cost wall is highest, and that a tool which cannot decline to trade is a poor fit for a market that charges you for every opinion.

Everything above is mechanics, and on the mechanics Supertrend is a decent piece of engineering. The ratchet is sound. The volatility scaling is sensible. The close-based flip is a real noise filter. If the tool has a deep problem it is not in the arithmetic, it is in what the arithmetic is asked to represent, and that is a problem the indicator cannot fix because it is the reason people like it.

Supertrend has two states and no third. On every bar it must be either up or down, and it must commit on the close. Consider what that means when the market is genuinely undecided. There is no direction to report, so the indicator reports one anyway, at full confidence, in the same green or red it uses when a trend is running cleanly. The line does not get paler in a range. It does not widen, hedge, or flag low conviction. The 75 directionless bars in the first figure look, at any single moment, exactly as authoritative as the 46 trending ones. That is not a rendering choice that could be improved with a better colour scheme. It is structural: a binary variable has no vocabulary for uncertainty.

An indicator that always has an opinion is not more informed. It is less honest.

This is why the appeal and the danger are the same feature. Traders reach for Supertrend because it removes the agony of deciding, and it genuinely does remove it, but it removes the agony without removing the uncertainty that caused it. The doubt you felt looking at a choppy chart was not a personal failing that a better indicator could cure. It was accurate. It was the correct perception of a market that was not offering a trade. What the line does is overwrite that accurate signal with a false one, and it feels like progress because certainty is pleasant and ambivalence is not.

Compare it to a moving average for a moment, not to score a point but because the contrast is instructive. A moving average will go flat. It has a gradient, and the gradient can be approximately zero, and a flat one is a legible statement that there is no trend to speak of. That is a worse-looking output and a more honest one. Supertrend deleted the flat state, and the deletion is exactly what makes it feel decisive. The information content did not go up when the ambiguity went away. It went down. Something that was in the chart, the fact that direction was absent, is now not representable at all.

The practical translation is a rule about what you are allowed to conclude from the line. "The line is green" means one thing only: the last close was above a band computed from the last ten bars of volatility, and no close has broken it since. It does not mean the trend is strong, that it is likely to persist, that the market is trending at all, or that a long has an edge here. Every one of those is a further claim requiring further evidence, and the flat, confident colour of the line supplies none of it. Read that narrowly and Supertrend is useful. Read the colour as a verdict and you have outsourced a judgement to something that is incapable of making it.

Faced with the trade-off in the dial figure, the natural next move is to go looking for the setting that would have worked. Run the multiplier across a grid, check the result on your history, keep the winner. This is such an obvious thing to do that most Supertrend discussion online consists of little else: which multiplier for Bank Nifty, which for a 15-minute chart, which for a stock. It is also, precisely and unavoidably, the curve fitting that an honest back-test exists to catch, and running it on our own series shows exactly how it fails.

This is the same sweep as the previous figure. Same series, same ATR period, only the multiplier moving, but plotted as the net result rather than the mechanism. It swings from +131 points at a multiplier of 1.0 to −128 at 2.5 and back to +79 at 4.25. That is a spread of 259 points generated purely by a parameter, across a range every trader would call a matter of taste. Neighbouring settings do not agree with each other. There is no region of the curve you could stand on and say that anything nearby would also have been fine.

Now put the two figures side by side, because the pairing is the actual argument. In the dial figure, every quantity moved smoothly and monotonically: a quarter point of multiplier always bought fewer flips and more giveback, in the same direction, every single time. The mechanism is not noisy at all. It is close to a straight line. Yet the outcome of that same smooth mechanism is the jagged, sign-flipping mess above. When a smooth cause produces a jagged effect, the jaggedness is not coming from the cause. It is coming from the specific accidents of these 121 bars: which particular wiggle happened to clear the band by half a point and which fell short. Pick the tallest column and you have not found a setting, you have found the place where this sample's noise happened to pile up highest.

Two tells in that chart are worth learning to recognise, because they generalise well beyond this indicator. The first is that the winner sits at the very edge of the tested grid. When your optimum is at the boundary, the honest reading is not that you found the peak but that the peak is somewhere you did not look, which usually means the search was not the search you thought it was. The second is that the flat tail is a mirage. The +79 points at 4.25 and above look like a broad, stable, attractive plateau, which is exactly what people are told to look for instead of a needle. But those settings produce zero flips: the system entered once and never traded again, so its result is just the opening long held to the end. It is not a robust parameter region. It is the absence of a strategy, wearing a plateau's clothing.

Pull the threads together and the shape of the thing is clear. Supertrend is superb at one job and structurally incapable of another. Given a trend, it will track it patiently, hold through pullbacks that would shake out a tighter exit, and hand you a mechanical level that only ever improves. Given a directionless market, it will flip repeatedly, lose on each turn, and report every one of those losses in the same confident colour it used when it was right. Its output is therefore only ever as good as the answer to a question it does not itself ask: is this a trending regime?

That is a specific gap, and it has a specific remedy. The reason Supertrend is so often paired with a separate trend-strength filter is that the filter answers the question the indicator cannot. An ADX reading, for example, is explicitly a measure of whether a directional move exists, and it is deliberately silent about which way. Gate the flips behind a regime check and you restore the third state that Supertrend deleted: not "up" or "down" but "not now". The pairing works because the two tools have opposite blind spots. One knows direction and nothing about conviction; the other knows conviction and nothing about direction. Neither is complete, and the incompleteness is why they are usually seen together.

None of which makes the combination a system. A filter narrows when you are allowed to act on a flip; it says nothing about the questions that actually decide outcomes. Where is the level that would genuinely invalidate this idea, as opposed to the level ATR happened to put a band at? How large should the position be given the distance to that level, and given that the giveback at the exit is going to be roughly three ATRs whatever you hope for? What happens to the account across a run of the four consecutive losing legs we measured above, which is a completely ordinary thing for a trend follower to produce? Those are sizing and risk questions, they sit upstream of every indicator, and they are what the method we teach is built around. The line is the easy half. Knowing what it is worth in the regime you are actually standing in, and how much to risk on it, is the rest.

• Origin and design intent. Supertrend was devised by the French trader and author Olivier Seban around the year 2000 as an "always-in" stop-and-reverse system built on the Average True Range. The design intent matters for reading it correctly: it was built as a trailing stop that reverses, never as a predictive tool, and the absence of a neutral state is a deliberate property of a stop-and-reverse rather than an oversight.
• Construction and defaults. The formula HL2 plus or minus the multiplier times ATR, the final-band rules that only tighten in-trend, the close-based flip, and the standard settings of a 10-period ATR with a multiplier of 3 are the specification documented by charting platforms. The recursion used to compute every figure on this page follows that specification exactly. tradingview.com
• Average True Range. ATR, the volatility input that scales the band, comes from J. Welles Wilder's New Concepts in Technical Trading Systems (1978), which also introduced the smoothing that the ATR period relies on. The figures here use Wilder's smoothing, seeded with a simple mean of the first ten true ranges, which is why the first nine bars of the series carry no state at all.
• The cost model. Every toll on this page is computed with the same rate model as our cost estimator: STT at 0.05% of the sell side for futures, 0.15% of the sell premium for options, 0.1% both sides for delivery and 0.025% on the sell for intraday; NSE exchange transaction charges; the SEBI turnover fee of ₹10 per crore; stamp duty on the buy side; and GST at 18% on brokerage plus exchange charges plus the SEBI fee only. Rates are as of 17 July 2026 and can change. Verify current rates with your broker before relying on them.
• The derivatives loss data. About 93% of individual traders in equity derivatives made net losses over FY22 to FY24, aggregate net losses exceeding ₹1.8 lakh crore (SEBI, September 2024). sebi.gov.in
• On the figures. The price series used throughout is authored and illustrative, not market data. Its swing structure was written deliberately and its per-bar noise comes from a fixed seed chosen once and not searched over. Everything drawn on top of it, the bands, the flips, the flip counts, the givebacks, the leg results and both parameter sweeps, is output computed from that series by the recursion described above. The purpose is to show the shape of the tool's behaviour and the direction of its trade-offs, which are properties of the mechanism. No figure here is evidence about any instrument, and none of the point counts should be read as a result anyone obtained or could obtain.