How the ADX Indicator Works: Trend Strength and the +DI and -DI Lines

What the ADX indicator shows—and what it does not show

Fidelity describes ADX as a bounded oscillator whose values fluctuate between 0 and 100. TradingView's documentation and IG's educational materials confirm the same range. The scale has a fixed upper bound, so a value of 40 is always stronger than a value of 20, regardless of the instrument on which it appears.

Reference sources describe the indicator's role in equally unequivocal terms. StockCharts writes that ADX measures “the strength or weakness of a trend, not the actual direction.” TradingView puts it even more succinctly: “It does not take direction into account at all.”

Here is a short list of what ADX does and does not provide:

1. It provides a single-number estimate of the strength of the current move.
2. It shows how that strength is changing: a rising line means that the move is gaining momentum, while a falling line means that it is losing steam.
3. It provides a common language for different instruments because the scale is the same everywhere.
4. It does not indicate direction: ADX alone cannot tell you whether the price is rising or falling.
5. It does not provide an entry point or set a stop-loss order.
6. It does not provide a price target or tell you how long a move will last.
7. It does not warn of a reversal: strength can also increase during a decline.

Because ADX is non-directional, it is rarely used on a chart by itself. If you are just putting together your toolkit, start by learning which indicators make sense for beginners, then return here when you need to assess the strength of a move.

How ADX, +DI, and -DI are related

Wilder designed Directional Movement as a system of three values, each of which is incomplete on its own. The two DI lines answer the question “which direction?” while the third line answers “how strongly?” Fidelity describes the combination as follows: the ADX line determines trend strength, while the two DMI lines complement it by helping determine direction.

Fidelity summarizes the classic interpretation of crossovers in one sentence: a buy signal occurs when DMI+ crosses above DMI-, while a sell signal occurs when the opposite happens. This is where sources begin to disagree, and the disagreement is genuine. IG writes that most traders ignore the DI lines because they “tend to cloud the issue.” StockCharts, by contrast, bases its approach specifically on DI crossovers while adding external confirmation.

My conclusion from this debate is simple: the DI lines are useful for describing the balance of strength, not as an entry button. A single crossover proves nothing when the strength of the move is near zero.

How True Range, +DM, and -DM are calculated

The calculation begins with the true range. True Range is the largest of three values: the candle's high-to-low range, the distance from the high to the previous close, and the distance from the previous close to the low. The same value underlies a separate indicator that measures volatility. If this is new to you, see the guide explaining how true range helps set a stop-loss order.

Directional movement is calculated next. Wilder's rule is presented in public sources as follows:

1. UpMove is today's high minus yesterday's high.
2. DownMove is yesterday's low minus today's low.
3. If UpMove is greater than DownMove and greater than zero, +DM equals UpMove; otherwise, +DM equals zero. The rule for -DM is the mirror image.

The crucial detail is hidden in the third point. Only one side is counted for each candle: either upward movement or downward movement, but never both at once. StockCharts describes the edge case directly: when both calculations are negative or equal, both values are set to zero. The reference TA-Lib implementation examines seven possible relative positions of two candles and arrives at the same rule.

An inside candle that lies entirely within the range of the previous candle produces no directional movement in either direction. This is precisely why ADX drifts lower in a narrow sideways market: the numerator is set to zero more often than the denominator.

How DX and ADX are derived from directional movement

Wilder uses his own smoothing method, and this is the first reason why values differ across descriptions. The first value equals the sum of the first 14 periods. After that, a recursive formula is used: subtract one fourteenth of the previous value from it, then add the current value. Tulip Indicators explains that this is an exponential average with a coefficient of 1/n instead of the conventional 2/(n+1), and that Wilder's smoothing over a period of 14 is equivalent to a standard exponential moving average with a period of 27.

The DX formula explains the indicator's defining feature. DX equals one hundred multiplied by the absolute difference between +DI and -DI, divided by their sum. StockCharts, the TA-Lib reference, and the English-language Wikipedia all provide the same formula. The absolute value removes the sign, and direction disappears along with it: a 30-point difference between the lines produces the same DX regardless of which line is higher.

The final step is to average DX using Wilder's method. The first ADX value equals the average of the first fourteen DX values. Each subsequent value is calculated by multiplying the previous ADX by 13, adding the current DX, and dividing by 14. This second smoothing makes the line smooth while also adding lag.

How ADX differs from DMI in TradingView and MetaTrader

TradingView's documentation answers the question directly: the built-in ADX indicator does not plot the +DI and -DI lines by itself. To see all three lines, you need the separate Directional Movement indicator. DMI has two parameters in its Inputs section: ADX, the smoothing period for the final line, and DI Length, the period used to calculate the DI lines themselves. Both default to 14. The ADX documentation page has no Inputs section at all, so the official pair of 14 and 14 is specifically confirmed for DMI.

The situation in MetaTrader 5 is more nuanced. The terminal includes two separate built-in indicators: Average Directional Movement Index and Average Directional Movement Index Wilder. According to MetaQuotes' description, the second strictly follows the algorithm in Wilder's book and uses a smoothed moving average, while the first uses a different smoothing method. The values of these two lines on the same chart do not match, even though their names are almost identical.

MetaTrader 4 warrants a separate row in this table: its iADX function has an applied price parameter—the price used in the calculation—while the MT5 version has no such parameter. Thinkorswim goes even further, allowing the averaging type to be selected from five options, including simple, exponential, and Wilder's smoothing. The same ADX on the same instrument produces different values there depending on the option selected.

Pine Script has no separate function for ADX. Everything is returned by ta.dmi, with ADX provided as the third element:

[diPlus, diMinus, adx] = ta.dmi(14, 14)
plot(adx, "ADX")

This call pattern can be seen in published TradingView scripts. You cannot assign the result to a single variable because the compiler expects it to be unpacked into three variables.

What the DI period and ADX smoothing change

Many people overlook the separation between these parameters, but it has practical implications. Reduce DI Length, and the directional lines will cross more often because they respond to shorter bursts. A shorter ADX smoothing period makes the final line more erratic, causing it to jump across thresholds more frequently. Wilder used 14, and this remains the default in TradingView and MetaTrader.

A much more important question is how much history is needed before a value becomes meaningful at all. The first ADX(14) value does not appear until the 28th bar, which follows from the structure of the calculation. Fourteen bars are used for the initial smoothing of the true range and directional movement, so the first DI values appear on the fifteenth bar. Another fourteen DX values are then needed to calculate the first ADX average.

Another mechanism then comes into play, although reference materials rarely mention it. Wilder's recursive formula never forgets old values completely: their weight decreases but never reaches zero.

Consequently, two terminals with different depths of loaded history will show slightly different ADX values on the same bar. StockCharts estimates that around 150 periods of data are required for stable figures. TA-Lib documentation calls this the initial unstable period and gives the same explanation: early values depend on how much history precedes them.

Important. A discrepancy between values for the same instrument on two charts is more often explained by different history depths or smoothing methods than by a platform error. Check the settings and the length of the loaded data segment first.

Why the 20 and 25 thresholds do not provide a ready-made signal

The most common pair of boundaries is as follows: above 25 indicates a strong trend, below 20 indicates no trend, and the area between them is uncertain. StockCharts, Fidelity, and TradingView all cite these boundaries, while IG confirms only the upper one: a trend is considered strong above 25. Fidelity adds a useful detail: values above 60 are relatively uncommon.

The discrepancies begin immediately beyond this pair. StockCharts attributes the threshold of 25 to Wilder himself, then notes that many analysts use 20 instead. The English-language Wikipedia provides another scale: a weak trend below 20, a strong trend above 40, and an extremely strong trend above 50. That is a twofold difference, and the chosen scale changes every other decision.

TradingView offers the most candid explanation for this variation: acceptable values differ from one instrument to another, so it is sensible to test an instrument against its own history. It also notes that Wilder developed the system for currencies and commodities, which are more volatile than stocks and exhibit stronger trends.

IG provides the best counterexample to a rigid scale: during a Nasdaq rally of more than 1400 points, ADX remained below 25 for most of the move. Formally, there was no trend; in reality, there was. IG's conclusion is worth memorizing verbatim: “No indicator is correct at all times.”

How to check ADX manually on closed candles

This check takes an evening and eliminates half of your future arguments with yourself. Follow these steps:

1. Choose one instrument and one timeframe, then fix a segment containing at least 30 closed candles. Do not include the open candle in the calculation because its high and low can still change.
2. Record the high, low, and close of each candle in a table. Adjacent candles must be consecutive, with no gaps.
3. Calculate the true range for each candle as the largest of the three values.
4. Calculate UpMove and DownMove, apply the selection rule, and set the weaker side to zero.
5. Add the first 14 values for true range, +DM, and -DM, then continue using Wilder's recursive formula.
6. Divide the smoothed movement by the smoothed range and multiply by 100 to obtain +DI and -DI.
7. Calculate DX and average its first 14 values. This is your first ADX value.

Next, add the DMI indicator with parameters of 14 and 14 to the same chart and compare the three numbers on the most recently closed candle. A discrepancy of a few hundredths is normal: libraries do not use the integer rounding found in the 1978 book, and differences in history depth also shift the result.

This educational diagram illustrates the very reason for performing a manual check. The price moved upward on the seventh candle, but the line did not cross above 25 until the eleventh. The indicator remained silent for four candles of the move. This is how double smoothing works; nothing is broken.

How to test an ADX filter without overfitting to historical data

A typical task is framed like this: I have a setup and want to skip trades when the market is going nowhere. A threshold of 25 is used as the starting point because it is the most common boundary in platform documentation. From there, the integrity of the test depends entirely on the order of the steps and your discipline.

The fifth step is the most important. A decline in the overall result proves nothing by itself if the number of trades fell threefold at the same time.

StockCharts explicitly warns about the main pitfall in this exercise: an ADX requirement does reduce the number of signals, but this heavily smoothed indicator filters out just as many good signals as bad ones. That is exactly what IG's Nasdaq example demonstrates. The decision must therefore be based on both figures together, not on whichever one looks more appealing.

If the filter does not help after testing, that is a valid result. It tells you that the strength of the move is not the main limiting factor for your rule and saves you months of work. Another way to filter out sideways markets is to look at the slope of a moving average and the price's position relative to it; how to interpret a moving average is covered separately.

Which mistakes most often distort the interpretation of ADX

Six situations occur more often than the rest:

1. Reading direction from ADX. The line rises, and this is mistaken for a rising price. It rises in exactly the same way during a strong decline.
2. Treating the threshold of 25 as universal. TradingView explicitly warns that acceptable values depend on the instrument and require historical testing.
3. Trading on DI crossovers alone. Wilder himself supplemented them with the extreme point rule: the crossover bar establishes the point, and an entry occurs only when its level is broken. The rule was introduced to eliminate false signals.
4. Comparing figures from different platforms. ADX and ADX Wilder in MetaTrader 5 are calculated differently, while thinkorswim even lets users select the averaging type manually.
5. Reading the value on an open candle. The high and low can still change, altering the entire calculation chain along with them.
6. Looking at a short segment of historical data. Values calculated from thirty bars differ from those that appear when one hundred and fifty bars are available.

A seventh mistake is more methodological: using the ADX indicator as the only tool on a chart. In its limitations section, TradingView recommends combining it with indicators that examine the price itself and help filter signals. If you are assembling such a combination, see the guide on how to build a set of indicators for a specific task.

Risks and limitations

• The indicator does not guarantee a profit. A value above the threshold is a reason to analyze the chart, not an instruction to enter a trade.
• Markets are volatile, and trading involves the risk of losing capital.
• Past results do not guarantee future results.
• ADX lags by design: it uses two consecutive smoothing operations, so confirmation arrives several candles after a move begins.
• The indicator misses slow trends. IG's example of a Nasdaq rally of more than 1400 points shows a move during which the value remained below 25 most of the time.
• There is no universal threshold standard: different authoritative sources provide scales that vary twofold.
• Values depend on the platform, smoothing method, and depth of the loaded history, so it is incorrect to compare figures from two different charts.

Disclaimer. This is not personalized investment advice. Markets are volatile, and trading involves the risk of losing capital. Past results do not guarantee future results.