ADX and DMI — measuring trend strength
The answer
ADX, the Average Directional Index, is a number from 0 to 100 that measures how strongly price is trending. It says nothing at all about which direction.
That sounds like a weakness and it is the whole value. Almost every other tool in this cluster is either a trend tool or a range tool, and each one fails badly in the conditions the other was built for. ADX is the reading that tells you which kind of market you are in today, and therefore which of your other tools to trust.
Why this costs you money
Most traders run one toolkit in every market.
They have an oscillator that tells them when something is oversold. They have a moving average crossover that tells them when a trend has started. They use both, all the time, on every chart.
Here is what that produces.
In a strong trend, the oscillator destroys them. RSI goes below 30 and they buy. Price falls further. RSI stays below 30 for 3 weeks. They buy again. This is covered in the oscillators article, and the summary is that an oscillator measures speed, not value, and a falling market has genuine speed.
In a range, the trend tool destroys them. The moving averages cross, they buy, price returns to the middle of the range, the averages cross back, they sell at a loss. Then it happens again. The moving averages article shows that a ranging chart produces several times more crossovers than a trending one, which means the tool fires most often exactly where it works least.
Both failures have the same cause. The trader had no way to tell which market they were in before they acted.
That is the specific gap ADX fills. It does not tell you what to buy. It tells you whether the question "which direction is the trend" is even a sensible question today, or whether there is no trend to have a direction.
The cost of not knowing is not a single dramatic loss. It is a year of trades in which roughly half were taken with the wrong instrument, and the winners paid for the losers with nothing left over. That pattern is extremely common and it looks like bad luck from inside.
How it works
ADX is built in 3 stages. Wilder introduced all of it in New Concepts in Technical Trading Systems in 1978, alongside RSI, ATR and Parabolic SAR.
Stage 1. Directional Movement
For each bar, compare it to the previous bar.
- Up move = today's high − yesterday's high
- Down move = yesterday's low − today's low
Then apply Wilder's rule. Only the larger of the 2 counts, and only if it is positive.
- If the up move is larger and positive,
+DM = up moveand−DM = 0. - If the down move is larger and positive,
−DM = down moveand+DM = 0. - If neither is positive, both are 0.
Read that again, because it is the part everyone skips. On any given bar, at most 1 of the 2 directional movement values is non-zero. The system assumes a bar is either extending upward or extending downward, never both.
Stage 2. The DI lines
Smooth +DM and −DM over 14 bars, and divide each by the smoothed True Range over the same 14 bars. Multiply by 100.
- +DI = the share of recent range that came from upward extension
- −DI = the share of recent range that came from downward extension
Dividing by True Range is the important step. It converts an absolute price distance into a proportion, so a ₹2,000 stock and a ₹40 stock give comparable readings.
Stage 3. ADX itself
First compute DX, which measures how unequal the 2 DI lines are:
DX = |+DI − −DI| ÷ (+DI + −DI) × 100
If +DI is 40 and −DI is 10, DX is 60. If +DI is 25 and −DI is 24, DX is about 2. DX is high when one side clearly dominates and low when the 2 sides are balanced.
ADX is then a 14-bar smoothed average of DX.
That last step is where the lag comes from and it is worth stating plainly. The DI lines already contain 14 bars of smoothing. ADX smooths those again over 14 bars. A 14-period ADX reflects roughly 27 bars of history. It is one of the slowest indicators in common use, and any signal derived from it arrives well after the event it describes.
What it tells you, and what it does not
ADX tells you how one-sided the recent directional movement has been. High ADX means one side has been extending the range consistently. Low ADX means the 2 sides have been taking turns.
Seven limits, and several of them are routinely ignored.
It has no direction. ADX rises in a strong advance and rises just as high in a collapse. A reading of 45 says a powerful trend exists. It does not say whether you should be long or short. The DI lines carry the direction; ADX carries only the strength.
It lags heavily. Roughly 27 bars of information, as shown above. By the time ADX confirms a trend, a meaningful part of the trend has happened.
The thresholds are conventions, not findings. "Below 20 is a range, above 25 is a trend" comes from Wilder's book and from long repetition. It is a reasonable starting point. It is not a result from a study, it is not the same on every instrument, and it is certainly not the same on every timeframe.
A high ADX often marks the end, not the beginning. ADX peaks when directional movement has been most one-sided. That frequently happens near the point of maximum enthusiasm. Buying because ADX just reached its highest reading of the year is a real way to buy the last day of a move.
Falling ADX does not mean the trend has ended. It means directional movement has become less one-sided. A trend that pauses and consolidates for 3 weeks will show a falling ADX and then resume. Reading every fall as a reversal produces constant, expensive exits.
DI crossovers are noisy. Wilder described entries using +DI crossing −DI, and those crossings happen often, especially in the ranges where the whole system is least useful. Treating each one as a trade is the same error as treating each moving average crossover as a trade.
It is not a signal. This is the single most important sentence in the article. ADX belongs in the filter, not in the trigger.
The decision rule
Use ADX to decide which of your other tools is allowed to speak today.
If ADX is low and flat — under about 20 on a daily chart, using the common convention — treat the instrument as a range. In a range, support and resistance levels and oscillators carry more information, and breakout signals and moving average crossovers carry less. Reduce your size or stand aside.
If ADX is above about 25 and rising — treat the instrument as trending. In a trend, stop fading overbought and oversold readings. Use the DI lines or a moving average for direction, use ATR for your stop, and let the position run.
If ADX is high but falling — the trend is losing its one-sidedness. This is not a reversal signal and it is not a range. Tighten management, stop adding.
Unless the instrument is illiquid, or the recent bars include days limited by a circuit band, or the period contains a large overnight gap. In each case the directional movement input is distorted, and the ADX reading is describing the distortion.
The word that matters in that rule is filter. ADX never tells you to buy anything. It tells you which other instrument on your chart is currently qualified to have an opinion.
Try this now
Five minutes, and it uses your own trades rather than a chart you were shown.
- Open your trade history. Pick your last 20 closed trades, or a year of trades if you trade less often. Write down the instrument, the entry date and the result of each one in rupees or dollars.
- For each trade, open the daily chart of that instrument and add ADX with the default period of 14.
- Find the entry date on the chart and read the ADX value on that day. Write it next to the trade.
- Sort your list by ADX. Split it into 2 halves: the 10 trades with the highest ADX at entry, and the 10 with the lowest.
- Add up the result of each half separately. Write down 2 totals.
What you should see. This depends on what kind of trader you are, and both outcomes teach something.
If you mostly trade breakouts, trend continuations and moving average signals, you will usually find that the high-ADX half holds most of your profit, and that the low-ADX half is close to flat or negative. Your method works, and it works in a specific condition you were not checking for.
If you mostly trade reversals, oversold bounces and range edges, you will usually find the reverse. Your winners cluster in the low-ADX half, and the high-ADX trades are where you were repeatedly early against a strong move.
Either result is the same lesson in 2 forms. Your method has a market it works in, and you have been running it in both markets. The split is the measurement that shows it.
Now do 1 more step. Look at the low-ADX half and count how many of those trades you would simply not have taken if you had checked ADX first. Multiply that count by your average loss. That number is what the filter would have been worth last year.
Three real cases
1. J. Welles Wilder Jr., 1978 and 1987 — the author's own later position Wilder introduced the Directional Movement system and ADX in New Concepts in Technical Trading Systems in 1978. The book also introduced RSI, ATR and Parabolic SAR, which is why a single 1978 publication sits behind so much of a modern chart. Nine years later, Wilder published The Adam Theory of Markets, a short book arguing for a much simpler approach to reading price direction . The point is not that his indicators were abandoned. It is that the person who built the most widely used trend-strength measure in the world went on to argue publicly for something simpler. Anybody presenting ADX as a complete method is claiming more for it than its author did.
2. The S&P 500, September to November 2008 (United States) — high ADX, falling price Through the autumn of 2008 the S&P 500 fell heavily and persistently, with the sharpest part of the decline in late September, October and November. Directional movement in that period was overwhelmingly one-sided to the downside, so ADX rose to high readings while price collapsed. Anybody who had learned "ADX above 25 means a strong trend, and strong trends are good" had a strong trend and a catastrophe at the same time. This case is the cleanest available proof that ADX has no direction inside it. The strength reading was correct. It was strength in the direction that removed capital.
3. The NIFTY 50, roughly October 2021 to June 2022 (India) — low ADX, and what it protects you from After a long advance through 2021, the NIFTY 50 spent the following months moving broadly sideways within a wide band, with several sharp advances and declines that each reversed. Directional movement in such a period is not one-sided, so ADX falls and stays low even though individual weeks feel dramatic. That is exactly the market in which moving average crossovers, breakout entries and trend-following stops produce a long sequence of small losses. A trader watching ADX had a warning available for months. A trader watching only price saw what looked like a series of new trends starting.
The question that resolves it
A novice looks at ADX and asks: is it above 25?
An expert looks at ADX and asks: which of my indicators should I switch off today?
The first question treats ADX as a signal with a threshold. The second treats it as what it is, which is a statement about the type of market, and therefore about the validity of everything else on the chart.
There is a second question underneath, and it separates 2 readings of the same number. Is ADX rising or falling? A reading of 30 on the way up from 15 and a reading of 30 on the way down from 55 are the same number describing 2 completely different situations. The level alone is not the reading. The level and the slope together are.
What would make this wrong
The central claim is that ADX identifies trending and ranging conditions well enough to be useful as a filter, and that the reader's own results are already split along that line without their knowledge.
You would falsify it directly with the exercise above. Sort your trades by entry ADX and add up each half. If the 2 halves are roughly equal, the filter adds nothing to your method, and this article is wrong about your trading. That is a real possibility and it is worth knowing.
You would falsify the threshold claim by testing several thresholds on your own instruments. If 25 turns out to be no better than 18 or 33 on your charts, then 25 was a convention that got repeated, which is exactly what this article says it is.
Four honest limits.
First, ADX is slow. By the time it confirms a trend, a large part of the move may be finished. Using ADX as an entry trigger is using a 27-bar measurement to time a decision.
Second, the split test measures correlation, not cause. Your high-ADX trades may have worked for reasons that happen to coincide with trending markets. Twenty trades is a small sample, and a year of trades is not a study.
Third, ADX behaves differently on different timeframes and instruments. An ADX of 20 on a 5-minute index chart and an ADX of 20 on a weekly single-stock chart are not describing the same thing.
Fourth, nothing here shows that trading only in high-ADX conditions is profitable. It shows that the tools you use should match the condition, and that ADX is the cheapest available way to identify the condition.
In India
ADX is available on every Indian broker platform and it is most often used as a companion to Supertrend, which is the country's most popular trend tool. That pairing is sound in principle, because Supertrend is a trailing stop that whipsaws in ranges, and ADX identifies the ranges.
Three Indian features matter, and 2 of them are distortions.
Price bands truncate directional movement. Most Indian cash-market stocks have daily bands of 5%, 10% or 20%. On a day when a stock is locked at its band, the recorded high and low are capped, so the computed directional movement is capped as well. Feed several such days into the calculation and ADX rises less than the actual move deserves. On Indian stocks, ADX understates trend strength during the most violent episodes, which is exactly when a trader most wants to know. The correction is simple: look at the last 20 candles, and if any closed at or near a band, do not trust the ADX reading for that period.
The index has no band, so index ADX is cleaner. NIFTY 50 and BANK NIFTY have no equivalent daily cap. Index-level circuit breakers exist at 10%, 15% and 20% and halt the whole market, but they are rare. So ADX on an Indian index is a more honest number than ADX on an Indian single stock.
Weekly expiries produce intraday chop that suppresses ADX. Indian index derivatives have a heavy short-dated expiry structure. On intraday index charts, the settlement-driven movement around expiry is frequently two-sided, which pushes ADX down even on days that felt violent. That is the indicator working correctly. Two-sided movement is genuinely not a trend, however large each swing was.
In the United States
ADX is standard on every United States platform and it has a longer history in professional use, because it came out of commodity futures trading and that is where systematic trend following developed.
Three United States features matter.
No daily bands on cash equities. United States equities have Limit Up-Limit Down bands that pause trading briefly around rapid moves rather than capping the day. So the daily high and low are a fairer record, and the directional movement computed from them is not truncated the way it is on a band-limited Indian stock.
Overnight gaps enter directional movement directly. Most United States companies report results outside market hours. A stock that gaps up 9% at the open produces a very large up move under Wilder's definition, because today's high is far above yesterday's high. ADX will rise sharply. A single scheduled event can therefore produce a high ADX reading with no trend behind it. Check the chart for a gap before treating a jump in ADX as evidence of a developing trend.
Systematic trend following uses this family professionally. Managed futures funds have used directional movement and related trend measures across currencies, bonds and commodities for decades. They apply them across dozens of markets at once and accept a low win rate in exchange for a small number of very large winners. That context matters. The indicator was designed for a portfolio approach, not for picking 1 trade on 1 stock, and most retail use ignores the difference.
Where they differ, and what that tells you
Two real differences, each with a direct consequence.
1. India censors the input; the United States inflates it. An Indian stock at its circuit band prints a capped high and low, so its directional movement is smaller than the day deserved and ADX under-reads. A United States stock that gaps on results prints an enormous directional movement in a single bar, so ADX over-reads. The same indicator, on the same settings, has opposite biases in the 2 markets.
What that tells you is what to check before you trust the number. In India, check for circuit days. In the United States, check for gaps. Both checks take 10 seconds on the chart and both change what the reading means.
2. Index ADX means different things because the indices are different portfolios. The NIFTY 50 holds 50 companies and a large share of its weight sits in financial services. The S&P 500 holds 500 companies across the economy, though it too has become concentrated in a small number of very large names. A high ADX on a concentrated index can be produced by a strong move in 1 sector. A high ADX on a broad index is more likely to reflect a broad move.
That is a warning against a specific and common mistake: reading index ADX as a statement about "the market". On a concentrated index, it may be a statement about 1 industry. The breadth article in this cluster is the tool that answers that question, and ADX is not.
Carry this
- ADX measures how one-sided recent movement has been. It carries no direction at all.
- The level and the slope together are the reading. ADX at 30 rising and ADX at 30 falling describe different markets.
- Use ADX to choose which of your other tools is valid today. Oscillators and levels in low ADX, trend tools in high ADX.
- Check for circuit days in India and for gaps in the United States before you believe a reading.