I went back through a stretch of my own breakout trades once where nearly every entry looked textbook and they all died the same way. Price cleared the level, I got filled, and within a couple of sessions it sagged back into the range and stopped me out. My first instinct was to blame the entries, but the entries were fine. The market had quietly stopped trending a few weeks earlier and nothing on my chart was measuring that. ADX is the tool I lean on for exactly this problem, and most of the frustration with it comes from expecting a directional signal from an indicator built to ignore direction.
What ADX actually computes
Welles Wilder built ADX in the late 1970s for commodity markets, and the mechanics are simpler than the acronym soup suggests. For each bar, did the high push further above yesterday's high than the low pushed below yesterday's low, or the other way around? The bigger push gets recorded as directional movement, up or down, and the smaller one is zeroed out. Those raw values are divided by true range, so quiet and violent markets become comparable, then smoothed, typically over 14 bars. That produces the two directional lines, DI+ for upward movement and DI- for downward movement.
ADX itself is one step further removed. Take the gap between DI+ and DI-, divide it by their sum, and smooth that ratio again. You get a single line between 0 and 100 that only cares how lopsided the fight between buyers and sellers has been. A market grinding higher every day prints a high ADX. A market collapsing every day also prints a high ADX. A range prints a low one, because up moves and down moves keep roughly cancelling out. Strength is the whole product, and if you want direction you read the DI lines.
The double smoothing matters more than people expect. On a daily chart with default settings, ADX is digesting roughly a month of price action before it fully reflects a change in character. That lag drives both its main use and its main trap.
Why 20 and 25 keep coming up
The classic reading goes like this. Below 20 there is no trend worth the name, above 25 there is a trend with enough persistence to trade, and the zone in between is ambiguous. These numbers are conventions that have held up reasonably well rather than laws of nature, but the logic is sound. Below roughly 20, upward and downward movement are close to balanced, which is another way of saying the market is behaving like a range. Moving average crossovers whipsaw there, breakouts have historically failed back into the range, and trend entries mostly end up buying the top of a swing. Above roughly 25 the imbalance is real, and pullbacks in the direction of the move have historically been rewarded with follow-through.
The gap between the two thresholds is useful on its own. Treat it as hysteresis, the same way a thermostat avoids switching on and off twenty times an hour. I flip into trend mode when ADX rises above 25 and only flip back into range mode when it drops below 20. A single threshold has you flip-flopping constantly right at the boundary, and the boundary is exactly where the indicator is least informative.
The DI lines add the direction that ADX leaves out. DI+ above DI- means upward movement is dominating, and the reverse means down. On their own, DI crossovers are among the noisiest signals in the standard toolbox, hence their bad reputation. Gated by ADX they change character: a DI+ cross above DI- while ADX sits at 15 is chop and should be ignored, while the same cross with ADX pushing through 25 has an established, strengthening move behind it.
Two playbooks, one switch
Here is where the indicator earns a permanent spot for me. The most practical way to run ADX is as a switch that decides which of two playbooks is allowed to trade, with the actual entries coming from somewhere else. Mean reversion and trend following are both legitimate, and each becomes a slow leak in the other regime. Fading extremes during a strong trend means shorting into strength over and over, and chasing breakouts inside a range means buying the high of the range over and over. A lot of strategies that look broken are simply running in the wrong regime.
- ADX below 20: range playbook only. Fade pushes toward the range extremes, target the middle, and treat breakouts as suspect until proven otherwise.
- ADX between 20 and 25: gray zone. Stand aside or trade at reduced size, and demand extra confirmation from price structure.
- ADX above 25 and rising with DI+ on top: trend playbook, long side. Buy pullbacks rather than fading strength, trail stops rather than taking quick profits.
- ADX above 25 and rising with DI- on top: the same playbook mirrored. Sell rallies, respect the downtrend, and do not average into longs because something looks cheap.
- ADX above 40 and rolling over: mature trend. Tighten stops, stop adding, and expect the next regime to be a range, since strong trends typically decay into consolidation before they reverse.
The value of the switch is subtractive. It will not find you a single new trade, but it benches the breakout playbook during the weeks it is more or less guaranteed to bleed, which in my experience removes more losing trades than any entry refinement I have tested.
The ways it will burn you
The lag cuts both ways. Because ADX is smoothed twice, a fresh breakout from a long quiet range often launches while ADX still reads 15 or 18. If your rule says never take a trend trade below 25, you will sit out the first and often best leg of every new trend. The practical fix is to weight the slope more than the level early on. ADX turning up from below 20 while price clears a well-defined level and one DI line pulls away from the other is an earlier read than waiting for the cross of 25. The level judges whether a trend is established, and the slope catches one being born.
The mirror-image trap is that ADX stays elevated after a trend has effectively ended, again because of the smoothing. A reading of 35 tells you the last few weeks were strongly directional and says nothing about tomorrow. High and rising is a very different environment from high and falling, and a naive threshold check reads them as identical, hence the rolling-over condition in the list above.
One more thing worth checking rather than assuming. The 20 and 25 conventions were shaped on daily charts of trending commodity and equity markets. They carry over roughly to crypto majors and intraday timeframes, but roughly is doing real work there. A 24/7 market with thin weekend liquidity spends more time in low-ADX chop than an index future does, and some altcoins basically live below 20 until they violently do not. Pull a year of data and see where your instrument spends its time before trusting anyone's thresholds, including these.
For a low-effort starting point, leave your live rules alone. Go back through your last few dozen losing trades and note what ADX read at each entry. If the breakout losses cluster below 20 and the fading losses cluster above 25, you have found the leak, and it was never in your entries.