Resource · Public
ADX: Separate Trend Strength from Trend Direction
Compare opposite DI relationships with identical DX, work one ADX smoothing update and distinguish strength, direction and selected threshold conventions.
Public summary
ADX can be high in either an upward or downward trend. A paired numerical example gives the same directional-strength calculation with opposite +DI and −DI readings, then works one smoothing update and shows why thresholds are conventions rather than trading instructions.
Image credit
Hypothetical already-smoothed inputs; one DX reading is not a complete ADX history.
FavAcademy · Original FavAcademy educational illustration; public use authorized
Separate strength from direction first
The average directional index, or ADX, belongs to a family of calculations based on directional movement and true range. Its central interpretive job is to summarise trend strength under that method. It does not encode upward direction merely by rising. A strong decline can accompany a rising ADX, and a rising price can accompany a falling ADX as the measured directional imbalance becomes less pronounced.
Fidelity’s ADX education makes the strength-versus-direction distinction, while TradingView documents the related directional calculations. The positive and negative directional indicators, written +DI and −DI, provide separate directional information. Read their labels and settings rather than treating every line in an indicator panel as interchangeable. A platform may display ADX alone or combine it with directional lines in a DMI panel.
Understand the input chain
The calculation begins with successive highs and lows to identify positive or negative directional movement under the chosen convention. True range incorporates the current high–low span and distances from the previous close, using absolute distances for the latter comparisons. These quantities are smoothed, then directional movement is scaled by smoothed true range. Different implementations and initialisation choices can produce small differences, especially near the start of a series.
This guide does not manufacture a complete fourteen-bar history. Instead, it supplies hypothetical already-smoothed inputs to isolate the next arithmetic steps. That boundary matters: a worked ratio is not a fully reconstructed production indicator. If you need to reconcile an actual platform value, obtain the full price history, smoothing length, seed convention and treatment of missing observations before assuming the displayed number is incorrect.
Calculate direction-scaled values
Suppose smoothed true range is 20 units, smoothed positive directional movement is six and smoothed negative directional movement is two. The corresponding +DI is 100 times six divided by twenty, or 30. The −DI is 100 times two divided by twenty, or 10. These are scaled indicator values rather than currency profits, probabilities or percentages of time that price must rise.
The directional index DX is 100 times the absolute difference between the two DI values divided by their sum. Here that is 100 times 20 divided by 40, giving 50. The absolute value removes the sign of the imbalance. That is why the strength calculation cannot by itself tell you which direction dominates: the sign information has been deliberately discarded at this stage.
Reverse the direction without changing strength
Now swap the two directional-movement inputs while leaving smoothed true range at 20. +DI becomes 10 and −DI becomes 30. The DX calculation is still 100 times the absolute difference of 20 divided by the sum of 40, again 50. The illustration presents the two cases side by side, highlighting opposite directional order with identical DX values.
Do not call those two single DX values ADX. ADX is the smoothed series built from DX observations. If the two hypothetical histories had the same prior ADX and mirrored directional inputs, the next ADX update would also match. The exercise demonstrates a structural property of the formula, not evidence that two real markets have identical risk, volatility or future opportunity.
Work one smoothing update
For an already-initialised fourteen-period Wilder-style series, suppose the prior ADX is 25 and the new DX is 50. The next update is thirteen times 25 plus 50, divided by fourteen. That equals 375 divided by fourteen, approximately 26.7857. The value rises because the new DX exceeds the previous smoothed level, regardless of which DI line is on top in the paired example.
The initial ADX seed requires its own set of DX observations; the update above does not replace that initialisation. If the DI sum is zero, the displayed DX treatment also needs an explicit implementation rule because ordinary division is undefined. Avoid inserting an arbitrary value without checking the platform’s documented convention. These edge cases are important when hand calculations and chart software appear to disagree.
Treat thresholds as selected conventions
Levels around 20 or 25 are commonly discussed in ADX education as descriptive filters. They are not universal boundaries between safe and unsafe trades, and crossing them does not establish a profitable entry. The lookback period, instrument and timeframe affect how the series behaves. A historical threshold selected after viewing outcomes should be treated as an exploratory choice and evaluated on separate observations.
A falling ADX does not necessarily predict a price reversal; it can indicate that measured strength is easing while price continues in the same direction. A low ADX does not mean price is motionless or risk-free. Keep the indicator’s descriptive role separate from execution, position size and costs. Adding several indicators derived from the same prices may also add less independent information than a busy panel suggests.
Use a three-part reading routine
First identify the settings and verify that the series is sufficiently initialised. Second read the DI relationship for its defined directional information. Third read the ADX level and change as a smoothed strength description. Write those observations as separate sentences. This prevents “ADX rose” from quietly becoming “price will rise,” which is precisely the inference the paired example disproves.
For the fictional inputs, the valid conclusions are +DI 30 versus −DI 10 in one case, the reverse in the other, DX 50 in both, and a next ADX near 26.7857 given the same stated prior value. No return, order or forecast follows automatically. Understanding the separation of direction, strength and smoothing makes the indicator easier to read—and makes its limitations harder to overlook.
Continue learning with Moving Averages: Smooth the Chart Without Losing Context and Leading and Lagging Indicators: Understand the Trade-Off.
Risk disclosure
Inputs are hypothetical and already smoothed. ADX thresholds are conventions, not forecasts, probabilities or guarantees of a suitable trade.