What Is the ADX Indicator, and What Does It Actually Measure?

The ADX, or Average Directional Movement Index, is a single line that reports how one-sided a market's recent price movement has been. It takes the gap between two other lines, a positive and a negative directional indicator, as a share of their sum, strips the sign off it, and averages the result, so a high reading says movement has been lopsided and says nothing at all about which side it was lopsided toward.

It belongs to a system Welles Wilder published in 1978, in a book the reference list of Park and Irwin's AgMAS report on futures trading rules gives as New Concepts in Technical Trading Systems. The arithmetic is documented by the platforms that draw it, the reading conventions come from vendor documentation rather than from any test, and the futures evidence read for this post covers the +DI/−DI crossover rule from the same Wilder system, which is a different thing from the index itself.

What does the ADX indicator measure?

The ADX measures the strength of directional movement in a market over a chosen number of bars, without reporting direction. Quantower's documentation puts it forward as a gauge of how strong a trend is overall, calls it "an average of expanding price range values", and files it inside Wilder's directional movement system, beside the DMI+ and DMI− lines.

Park and Irwin, writing up the system as one of twelve trading rules they tested on US futures markets, give its objective as telling a market that is probably trending apart from one in a trading range. In their description the system produces two directional indicators running from 0 to 100%, and generates its signals by comparing those two against each other. They attribute the system to Wilder, and file it with the momentum oscillators, the family in that report whose members read a trend off how large recent price changes have been rather than off the price level itself.

The line is a practitioner's convention documented by software vendors. It is not a defined term in the CFTC's futures glossary, which carries no entry for the ADX and none for directional movement. The nearest thing it carries is Directional Trading, defined there as speculating on where the underlying market goes, and set against volatility trading.

How is the ADX calculated?

Four quantities are built in sequence, and the ADX is the last of them. Sierra Chart's documentation sets out the chain for its own studies, and Wilder's original specification as written up by Park and Irwin follows the same shape with plain sums in place of the platform's smoothing.

It starts with two differences per bar. The up move is today's high minus yesterday's high; the down move is yesterday's low minus today's low. Only the larger of the two survives, and only if it is positive: the positive directional movement takes the up move when the up move exceeds the down move and is above zero, otherwise zero, and the negative directional movement mirrors that. Park and Irwin note the consequence for quiet bars — on an inside day, where yesterday's range covers today's, both come out as zero.

Each of those two series, plus the true range of each bar, is then accumulated over the length you set, using what Sierra Chart calls a Welles sum. Dividing the accumulated positive movement by the accumulated true range and multiplying by 100 gives the positive directional indicator; the negative one is the same division on the other series.

The last two steps are where the ADX appears. The directional index, DX, is 100 times the absolute difference between the two indicators divided by their sum — a measure of how far apart they sit, with the sign discarded. The ADX is then DX put through a Wilder moving average over a second length. Sierra Chart's ADX study therefore takes two separate inputs, a DX length and a DX moving-average length, and states that its first value appears only once enough bars exist to fill both: at the index equal to the two lengths added together minus one. With both set to 14 that is index 27, the 28th bar of the series.

What is the difference between ADX, +DI and −DI?

The two directional indicators carry direction and the ADX carries strength. They are introduced below in the order the arithmetic builds them, since the ADX cannot exist until the indicators do.

Sierra Chart also publishes a combined study, DMI & ADX & ADXR, which draws all three in one chart region and maps its own length and moving-average inputs onto the equivalents in the separate studies. So the same four numbers can arrive on a chart as one study or as three, and the setting called Length in the combined version is the DX length elsewhere.

What ADX reading counts as a strong trend?

The familiar answer is that above 25 is a trend and below 20 is not, and it comes from vendor documentation attributing that to Wilder: Quantower's ADX page states it in those terms. The same page adds several readings of the line's shape: a rising ADX as a strengthening trend, a declining one as a market becoming less directional, a turn down from high values as a trend that may be ending, and a rise of four or five units off a long low stretch as a possible signal for trading the trend already in place.

Two qualifications belong next to those numbers. The first is that they are conventions published by a platform, not results from a test: the page names Wilder for the 25 and the 20 and cites no test behind any of them. Treat the thresholds as somebody's reading of the line.

The second is that the other vendor documentation read for this post declines to interpret the line at all. Across the four Sierra Chart pages — ADX, ADXR, Directional Movement Index, and the combined study — the words trend, signal, strong and default do not occur once, and no threshold or default length is given. Those pages carry the formulas, the inputs and the spreadsheet files, and stop there. The CFTC glossary offers nothing to check a threshold against either, having no ADX entry at all.

Why do two platforms show a different ADX?

Two platforms show different ADX values because each one documents a different averaging step. Sierra Chart's ADX study specifies a Wilder moving average of DX, and its page for that average gives the recursion: each new value is the previous average plus one-nth of the gap between the new DX and that average, so old readings fade without ever dropping out. Quantower's ADX page describes the calculation as a plain mean of the DX values over the specified period. Run the same DX series through both and they agree exactly on the first bar, because Sierra Chart seeds its recursion with precisely that simple mean, then separate from the next bar onward on any DX series that moves. Both lines are labelled ADX.

The inputs differ too. Sierra Chart's ADX takes two lengths, its Directional Movement Index study takes one, and the ADXR study takes three, while Quantower's pages speak of a single period and name 14 periods for the DMI lines. There is also a documented difference inside one vendor's own set: when accumulated true range comes out at zero, the Directional Movement Index study carries the previous bar's indicator forward, and the ADX study sets the indicator to 100 instead, which the page flags by saying the indicators are calculated slightly differently there.

None of that makes one chart wrong. It means the number is a property of the software and the settings as much as of the market, which is the same lesson as the averaging choice behind the RSI — and the reason a threshold copied from one chart does not transfer cleanly to another.

What has been tested on futures data?

What has been tested on futures data is the crossover rule from the same directional movement system, rather than any threshold on the averaged line itself. Park and Irwin specify Wilder's directional movement system as a crossover: when +DI crosses above −DI, a buy stop goes in the next day at the crossing day's high and stays until filled while +DI remains above; the mirror applies on the other side, with an extreme point rule that fixes the reverse point at the crossing day's low for a long and at its high for a short. Across the whole of that report the letters ADX do not appear.

Their test is a strict one. It covers twelve US futures markets: corn, soybeans, world sugar and cocoa; silver and copper; the British pound, the Deutsche mark and T-bills; live cattle, lumber and pork bellies. The prices were daily, running from 1975 to 2003 and a little shorter for the three financial contracts, with 1978 onward kept out of sample. Transaction costs were charged at $100 per contract per round turn, commission and bid-ask spread together. Parameters were re-chosen every year from the previous three years' data and then traded forward, so each year's results came from settings fixed before that year began. Significance was assessed by a stationary bootstrap with 1,000 resamples.

Did Wilder's directional movement rule make money?

Wilder's directional movement rule earned 8.09% a year in the first of the two windows Park and Irwin report, and lost money in the second. That return came on an equally weighted portfolio of the twelve markets over 1978 to 1984, or 1980 to 1984 for the three financial contracts, and was significant at the 5% level. Its Sharpe ratio of 0.76 was the highest of the twelve systems tested, and three of the twelve markets carried individually significant returns: world sugar, the Deutsche mark and T-bills.

Over 1985 to 2003 the same rule returned −7.42% a year on the same portfolio, with a Sharpe ratio of −0.92 and no statistical significance anywhere. Park and Irwin report that in that later window no trading system of the twelve earned a positive return on the twelve-market portfolio, and the aggregate across all markets and systems was −5.82%, improving only to −3.80% when they re-ran it with lower costs. Over the full 1978 to 2003 period the directional movement rule came out at −3.25%, keeping significant returns in two markets only. The moving average crossover was one of the other eleven systems in the same report, and its portfolio return moved the same way between the two windows. The annual re-optimisation behind these numbers is there so that no result comes from settings chosen with the answer already visible.

What does an ADX reading mean on a funded account?

An ADX reading means nothing in dollars, and that is the practical limit of it on a funded account. The ADX is an index built from ratios, so a reading of 32 is not a quantity of money, a stop distance or a position size, and none of the platform documentation behind it converts the number into any of those. What converts a trade into dollars is stop distance in ticks multiplied by tick value, worked against the risk you have available — the arithmetic in our post on futures position sizing.

That matters more on a simulated funded account than on a chart, because the constraint there is a published loss line. Accounts here are simulated accounts that pay out real money on the profits, and the distance to the drawdown line is a dollar figure you can read off; what that line is and how it moves is set out in how the drawdown works. A trend-strength reading cannot widen that distance.

Which markets the futures evidence above covers matters here too, because several of its twelve are contract families on the instrument table: corn and soybeans at CBOT, silver and copper at COMEX. The results belong to those markets, over those years, on daily prices. An equity index contract on a five-minute chart is outside all three of those bounds, and the study says nothing about how the same crossover behaves inside a session.

FAQ

Is 14 the standard ADX setting?

Not as a published default. Quantower's documentation names 14 periods for the DMI lines. None of Sierra Chart's four pages for the ADX, the ADXR, the Directional Movement Index and the combined study states a default length at all: neither the word default nor the number 14 appears on any of them. In the futures test, Wilder's rule was run across thirteen candidate lengths from 3 to 39 in steps of three, with the length re-chosen annually.

Does the ADX tell you whether to buy or sell?

No. The DX step takes the absolute difference between the two directional indicators, so the sign is discarded before the averaging begins, and a strong uptrend and a strong downtrend can produce the same ADX value. Quantower's documentation puts the buy and sell side on the directional indicators crossing each other, and uses the ADX and ADXR to gauge the strength of that crossover rather than its direction.

Why is my ADX line blank at the start of the chart?

Because both lengths have to be filled before a value exists. Sierra Chart's ADX study states that its first value appears at the index given by the DX length plus the moving-average length minus one, so with both set to 14 the line begins on the 28th bar. The Directional Movement Index study needs fewer: its indicators display from the DX length onward.

Do those futures results tell you anything about the ADX line itself?

No. What Park and Irwin specify and trade is the +DI/−DI crossover with stop entries and the extreme point rule; the averaged index is never computed in that report, and the letters ADX appear nowhere in it. So the 8.09% over 1978 to 1984 and the −7.42% over 1985 to 2003 are evidence about a crossover rule. No source read for this post tests a 25 or 20 threshold on the ADX.