What is the ADX indicator?

Start with the claim that trips up newcomers. ADX is a single line that answers one question, how strong is the current move, and flatly refuses a second, which way is it going. A sharp rally and an equally sharp collapse produce the identical ADX value. The move having conviction and the move being bullish are two entirely different facts, and ADX carries only the first. This is why the indicator never travels alone: it is the top layer of a three-line system Wilder called the directional movement system, in which the two directional indicators, +DI and minus-DI, supply the direction, and ADX distils their disagreement into a single strength reading.

Most explanations stop at asserting this, which invites the reasonable suspicion that it is roughly true rather than exactly true. It is exactly true, and there is a clean way to show it. Take any price history and reflect it in a horizontal mirror, so that every high becomes a low and every up-move becomes a down-move of precisely the same size. Under that reflection Wilder's arithmetic does something very specific: each +DM becomes a minus-DM and each minus-DM becomes a +DM, while the true range, which measures the span of a session without caring which end is which, does not change at all. The DX is built from the absolute difference of the two directional lines over their sum, so swapping them leaves it untouched, and the ADX built from that DX is unchanged to the last decimal. The reading is not approximately the same under a bull move and a bear move. It is the same number.

That is the whole thesis in one picture, and it is worth being precise about what it does and does not say. It does not say that a bull market and a bear market are the same thing, which would be absurd. It says that this particular instrument cannot tell them apart, because the quantity it reports has no sign. Read the two panels above and every difference that matters is carried by the two thin directional lines at the bottom: on the left the green +DI leads, on the right the coral minus-DI leads. The thick gold ADX line, the one most people actually look at, is identical in both. If you take one idea from this page, take that: the ADX line is a magnitude, and asking a magnitude which way the market is going is a category error, not a subtle misreading.

You do not need to compute ADX by hand to use it, but you do need the shape of the calculation, because the shape is what causes both of its defining properties. Everything ADX gets wrong, it gets wrong for a reason visible in the four steps below, and the two most important reasons live at different points in the chain. The blindness to direction enters at step three. The lag accumulates at every step and is worst at step four.

Wilder begins with directional movement, which is pure measurement. Each session he compares today's high and low to yesterday's: if today reached higher, the excess above yesterday's high is the up-move; if today fell lower, the shortfall below yesterday's low is the down-move. Then comes the rule that surprises people. The larger of the two wins, and the loser is set to zero outright, not averaged, not kept as a smaller contribution. A session is counted as an up-session or a down-session, never as both. The winning move is then divided by the true range, the session's full span including any overnight gap, which is the same quantity the ATR indicator averages. Dividing by range is what makes the output a percentage rather than a rupee amount, and therefore what lets an index and a modestly priced stock be read on one 0-to-100 scale. Both series are then put through Wilder's smoothing over a look-back, conventionally 14 periods, producing +DI and minus-DI.

Up to this point the indicator still knows the difference between a rally and a crash: two separate numbers, one for each side, and you can read the market's direction straight off which is larger. The third step ends that. The DX is the absolute difference between +DI and minus-DI, divided by their sum, times 100. Those absolute-value bars are the exact place where the sign dies. A +DI that leads by sixteen points and a minus-DI that leads by sixteen points produce the identical DX, because the operation deliberately discards which of the two was larger and keeps only by how much. The fourth step, ADX, then puts that DX through Wilder's smoothing a second time. So the line on your chart is a smoothing of a quantity that was itself computed from two already-smoothed series, and it is built downstream of the operation that threw the direction away. No amount of study of the ADX line can recover what step three deleted.

The worked session above is a useful one because it is an outside bar: the market made both a higher high and a lower low, so both an up-move and a down-move genuinely existed. Wilder's rule kept the larger, 7.78, and discarded the smaller, 0.65, entirely. That is worth pausing on, because it shows the indicator is not a gentle average of pressure from both sides. It is a contest with a single winner per session, and the losing side contributes nothing at all. Set the three lines side by side and the division of labour is unambiguous.

Read the third column downwards and the design of the indicator becomes obvious. Direction is fully decided by the first two rows, which is to say by the DI pair, and ADX is a deliberate summary that strips the sign out on purpose in order to answer a different question cleanly. That is not a defect. A tool that measures one thing well and declines to guess at a second is more honest than one that blurs both. The mistake is not in the design, it is in reading the third row as though it were the first two.

Every explanation of ADX mentions that it lags. Very few say by how much, and the number is what turns a mild caveat into a rule about how the indicator may be used at all. All that smoothing buys something real: the ADX line is stable and legible, and it does not flap about on a single noisy session the way a raw reading does. But stability and speed are the same dial turned in opposite directions, and Wilder turned it hard toward stability. The price is paid in sessions.

The chart below measures it on a deliberately simple case: a flat range that goes nowhere, then one clean break into a sustained trend. Both lines in the lower panel are computed from exactly the same bars, so the comparison is fair by construction. The raw DX, the quantity that exists one step before ADX, cleared 25 on the very session the trend began. The ADX, which is nothing but that same DX smoothed a second time, read 15.8 on that same bar, still deep in the zone the convention calls no trend, and did not reach 25 for another 16 sessions. The information was there on day one. The indicator most people watch simply had not finished averaging it yet.

Sixteen sessions is roughly three trading weeks, and by then a substantial part of the move has already happened. Wilder's smoothing has a second, quieter consequence in the same direction: because each value carries a decaying memory of everything before it, the series needs a long run-up, on the order of 150 bars, before it fully settles. Feed a chart only 30 bars and the ADX you are shown is not really the ADX, it is a number still converging toward it. Both effects push the same way, and together they settle the question of how the indicator may honestly be used.

A number that reports a trend sixteen sessions after it began is not a trigger. It is a confirmation, and confirmations are for deciding whether to keep believing something, not for deciding when to act.

This is why the common instruction to wait for ADX to cross 25 before taking a trend trade is self-defeating when read as an entry rule. It is not that the advice is wrong about the market. It is that the rule is late by construction, and no choice of look-back repairs it, because the lag is the smoothing and the smoothing is the indicator. Shorten the period and you get a faster line that crosses your threshold more often in both directions, which is a different problem rather than a solution. The productive response is not to tune the delay away. It is to stop asking the line to do a job whose whole value is timing, and to give it one where being three weeks slow is tolerable. That job exists, and the last section of this guide is about it.

ADX is usually read against a small set of thresholds: under about 20 is a weak or absent trend, over about 25 is a trend worth respecting, over 40 is a strong and comparatively rare move, and the gap between 20 and 25 is a grey zone Wilder himself acknowledged rather than resolved. Those numbers are genuinely useful as a shared vocabulary, and they are repeated in this guide because they are the language every chart platform and every other explanation uses. But it is worth being exact about their status, because almost nothing else you read will be.

Here is what those numbers are not. They are not derived from any property of markets, they were not fitted to data, and nothing about the arithmetic makes 25 a boundary between two different states of the world. They are conventions, and they became conventions the ordinary way: Wilder used them, the books repeated him, the platforms drew the line at 25 by default, and forty years of repetition turned a sensible starting point into something people quote as though it were measured. Twenty-five is a reasonable place to begin. It is not a fact about the market.

The practical consequence is that the same reading does not mean the same thing everywhere. ADX is a normalised quantity, which makes it comparable across instruments in the narrow sense that it always lands between 0 and 100, but that is a statement about the scale and not about the distribution. A broad index and a volatile single stock do not produce the same spread of ADX readings, and neither does the same instrument on a daily chart and a five-minute chart. A market that spends most of its life between 15 and 30 makes 25 a demanding threshold that fires rarely; a market that routinely runs into the 40s makes the same 25 a low bar that is cleared during ordinary noise. Carrying a threshold from one to the other is borrowing a number, not measuring one.

Given that the levels are conventions and the conventions drift, a fair question is what is left to read. The answer is the slope, and for most purposes it is the better half of the indicator. The level tells you where conviction has been over the look-back. The direction of travel tells you what is happening to it now. A rising ADX says the prevailing move, whichever way it points, is gaining strength; a falling ADX says it is losing steam, and it can say that while price is still drifting the same way it has been for weeks.

The slope has a quiet advantage over the level, and it is the same advantage that survives the previous section. A threshold is a comparison against a number someone else chose, and that number does not transfer between instruments or timeframes. A slope is a comparison of the line against itself, a few sessions ago, on the same instrument, on the same timeframe. Nothing is borrowed. That makes it the more portable reading of the two, and it is why an ADX rising from 18 to 24 is often telling you more about a market waking up than a flat ADX sitting at 26 is telling you about a trend, even though only the second one has cleared the famous threshold.

None of which makes the slope a forecast. Both readings are computed from the same doubly smoothed series and inherit every session of its lag, so a rising ADX reports strength that has already arrived, and a falling one reports conviction that has already drained. The specific error worth naming is treating a falling ADX as a reversal signal. It is not one. It says the move is losing intensity, and a move can lose intensity by turning around, by pausing, by grinding sideways for a month, or by resuming at a gentler pace. Those are four different futures and the indicator does not distinguish between them, because distinguishing between them would require exactly the directional information the DX threw away.

The section a careful reader should read twice. Everything so far has been mechanical: the indicator is blind to sign, and it is slow. Put those two facts together at a market top and they produce a specific, expensive failure, and it is the one most explanations skip entirely. The trap is not that ADX is inaccurate. It is that ADX is accurate about a quantity that behaves identically on the way up and on the way down, and a trader watching that quantity through a reversal receives no signal at all, while feeling thoroughly informed.

Work through what the arithmetic must do at a top. A strong rally produces a wide gap between +DI and minus-DI, so the DX is high and the smoothed ADX climbs. Now the market rolls over and collapses. The gap between the directional lines re-opens just as wide, only with the other line on top, so the DX climbs again and so does the ADX. The line cannot fall merely because the market inverted, because the inversion is invisible to it. The only thing that could make ADX drop is the two lines converging, which happens when neither side is winning, which is precisely what does not happen in a violent sell-off. A crash is not the absence of a trend. It is a trend, and ADX scores it accordingly.

The computed result is worse than the argument. Across the whole episode above, the rally's best reading was 54.6 and the sell-off that erased it reached 50.6: to a strength gauge those are the same answer. At the top itself, with the advance finished and the collapse about to begin, the ADX was still reading in the forties. And across all 62 sessions the line never once printed below 39.5, so it never came within fifteen points of the 25 threshold and never spent a single bar suggesting that anyone should stand aside. A trader reading that gold line alone, and reading it correctly, would have concluded throughout that a strong trend was in force. That conclusion was true. It was also useless, because the trend it was describing had changed sides.

Collect the limitations and the correct use of ADX falls out of them rather than having to be argued for. It cannot say which way to trade, so it cannot be a directional signal. It arrives weeks late, so it cannot be a trigger. What remains is the question that comes before any entry, and it happens to be a question ADX is well suited to answer: is there a trend here at all? That is a regime question, it changes slowly, and a reading that lags by a few weeks is a far smaller handicap when the thing being measured persists for months.

The distinction that makes this work is architectural. A regime filter sits in front of a strategy and decides whether the strategy is allowed to speak; it does not sit inside the strategy vetoing individual entries. Get that backwards and you inherit all of the lag with none of the benefit, because you are asking a slow line to rule on a fast decision. Used properly, the logic is plain: when ADX is low and flat the market is ranging, and trend tools, moving-average crossovers, breakout entries, a trailing stop that follows the move like the Supertrend indicator, are all in the exact environment that shreds them. When ADX is higher and rising, a trend has room to run and those same tools are back on home ground. You are not asking ADX where to enter. You are asking it which toolbox to open.

The chart above runs a plain 5-over-20 moving-average rule across a settled range and then a trend, with ADX as the switch in front of it, and it is honest about both halves of the bargain. Inside the range the rule crossed five times in about thirty sessions while price finished four points from where it started: five invitations to trade a trend in a market that had none, and every one of them was refused, because ADX sat between 9.8 and 14.0 throughout and the gate never opened. That is the entire case for the filter, and it is a strong one. Then the bill arrives. Price broke out, and ADX did not clear 25 for another 23 sessions, so the first three weeks of the move happened with the rule still switched off. The gate is late at both ends, and there is a wrinkle worth noticing: the deader the range, the lower ADX sits, and the further it has to climb before the gate opens. The filter is slowest exactly where it has been working hardest.

Notice that not one row of that table is a trade instruction, and that the second and fourth rows are costs rather than benefits. That is the correct register for this indicator: it classifies the weather, it is slow to notice the weather changing, and it cannot tell you whether the wind is at your back or in your face. Deciding the regime before choosing the tactic, and being clear-eyed about how much the tool doing the deciding actually knows, is the upstream judgement that the method we teach is built around. The instrument you reach for matters far less than whether it fits the market in front of you, and far less again than knowing what your instrument is blind to.

Run the whole account together and a narrow but genuine picture emerges. ADX belongs to the context layer of chart reading, the step that sizes up the market before any tactic is chosen, and it is good at exactly one thing: giving a single stable number for how trending or how rangebound conditions have recently been, so that trend tools are only deployed where they have a chance. That is a real service, and it is not a small one, because a great deal of the damage that trend-following does to newcomers happens in the flat markets a low reading would have flagged. Learning to ask what kind of market is this before asking where do I enter is one of the more durable habits in technical analysis for beginners, and ADX is a reasonable instrument for the first question and a poor one for the second.

What it cannot do is longer, and every item on the list is structural rather than fixable. It cannot tell you direction, because the DX deletes the sign. It cannot give you a level, a target or a stop, because it measures nothing about price. It cannot warn you at a top, because a reversal is a trend and it scores it as one. It cannot be sped up, because the delay is the smoothing and the smoothing is the whole indicator. And its famous thresholds cannot be trusted as transferred, because they are conventions whose meaning shifts with the instrument and the timeframe. An indicator summarises past price. It does not foresee the next bar, and ADX, doubly smoothed, summarises the past more slowly than most.

ADX answers one honest question: is there a trend here at all. Ask it that and it earns its place. Ask it anything else, and you are asking a magnitude to have an opinion about direction.

Which leaves one sentence worth carrying away, and it is the same shape as the sentence that governs every other indicator. A tool is only dangerous when you believe it is answering a question it was never built to answer, and ADX is unusually easy to misread that way, because a single confident number invites you to read more into it than it contains. The line is honest. It reports a magnitude, it reports it late, and it reports it on a scale whose landmarks are borrowed. Keep it to the one question and it is a useful part of a process; stretch it past that, and the failure will not announce itself, because the number will look exactly as convincing while it is misleading you as it does while it is helping.

• Origin and purpose. J. Welles Wilder introduced the directional movement system, including +DI, minus-DI, the DX and the ADX, in his 1978 book New Concepts in Technical Trading Systems, defining ADX as a non-directional measure of trend strength on a 0 to 100 scale. The 14-period look-back used throughout this guide is Wilder's own convention. en.wikipedia.org
• Construction and smoothing. The step-by-step build set out in the second section, directional movement normalised by true range into +DI and minus-DI, the DX as the absolute difference over the sum times 100, and ADX as Wilder's smoothing of the DX, follows the StockCharts ChartSchool reference, which also documents the long data run-up Wilder's smoothing requires before values settle. chartschool.stockcharts.com
• The reading bands, and their status. The conventions that ADX under about 20 indicates a weak or absent trend and over about 25 a trend worth respecting, with a grey zone between them and comparatively rare readings above 40, follow Wilder's guidance as documented in the ChartSchool reference above. They are presented here as conventions rather than as measured thresholds because that is what the sources support: they record what Wilder used and what practice repeats, not a derivation.
• How the charts on this page were made. Every figure is computed rather than illustrated. The price paths are deliberately authored bar by bar, with irregular swing sizes and real pullbacks inside each trend, and Wilder's 14-period calculation is then run over those bars to produce the +DI, minus-DI, DX and ADX series that are drawn and quoted. Each series is computed over a long warm-up and only the settled portion is shown, so the readings are not artefacts of a short run-up. The figures are illustrative of the mechanics and are not records of any real instrument.
• Indian retail derivatives context. The Securities and Exchange Board of India study of profit and loss of individual traders in the equity derivatives segment, published September 2024, reports the loss figures quoted in the closing section. Regulatory findings and any figures drawn from them are stated as of 17 July 2026; verify current figures at the source before relying on them. sebi.gov.in