# ADX (Average Directional Index) > ADX is Wilder's average directional index, a 0–100 gauge of how strongly a stock is trending in either direction — it measures the trend's strength while saying nothing about which way it points. Canonical: https://patternsradar.com/learn/adx ADX starts from directional movement: how much of each session's range pushed above the previous high versus below the previous low. Those become the +DI and −DI lines, and ADX is the smoothed, absolute difference between them — high when one side consistently dominates, low when they trade the lead. The double smoothing, all Wilder's own, is why the 14-period version needs roughly twice that in history to settle, and why it turns deliberately. The reading is a regime dial, not a signal. Below 20 the stock is drifting and trend-following tools — moving average crosses, Supertrend flips, breakouts — mostly produce noise; above 25 there is a trend worth joining, and above 40 a strong one. Its commonest use on daily NSE bars is therefore as the filter on other scans: a DI cross or a Supertrend flip with ADX above 20 is a change of control inside a market actually going somewhere. Two honest caveats. ADX is direction-blind — a reading of 35 describes a powerful downtrend as readily as a powerful uptrend, and needs the DI lines or a price filter beside it to have an opinion. And it lags badly at turns: the double smoothing means ADX is still falling when a new trend starts and still elevated after one has died, so a rising ADX confirms a trend in progress rather than announcing a fresh one. Against Aroon, the other trend-strength gauge here, the difference is the raw material: Aroon only asks how recently the highs and lows were made, while ADX measures how much directional ground is being covered — magnitude versus recency. ADX is the slower and sterner of the two. ## In Sift Written as `adx(14)`. A working scan — stocks in a measurable trend with the moving averages stacked beneath price: ```sift where adx(14) > 25 and close > sma(50) > sma(200) ``` 1 of the 500 most-traded NSE stocks match today, as of 2026-08-20. ## Scans that use it - [Supertrend flips bullish](https://patternsradar.com/screener/supertrend-flip.md): The ATR trailing stop turning up, with the trend strong enough to be worth trusting. - [Parabolic SAR flips bullish](https://patternsradar.com/screener/psar-flip.md): Price crossing above the parabolic stop, in a market with a trend to catch. - [DI bullish cross](https://patternsradar.com/screener/di-bull-cross.md): +DI overtaking −DI while ADX confirms there is a trend worth joining. - [DI bearish cross](https://patternsradar.com/screener/di-bear-cross.md): −DI taking control with directional movement building — the exit-or-avoid list. - [Ten sessions above the 20-day](https://patternsradar.com/screener/persistent-trend.md): Price that has not closed below its 20-day average in two weeks. Persistence, not a single good day. - [Strong directional trend](https://patternsradar.com/screener/adx-strong-trend.md): ADX above 30 with buyers in control and the averages stacked — a trend worth trading. ## Common questions ### What is a good ADX value? The working conventions: below 20 means no meaningful trend, 25 and above means one worth trading, above 40 a strong one. But ADX is direction-agnostic — 40 can be a powerful decline. It answers whether the market is going somewhere, and needs the DI lines or price itself to answer where. ### Does a falling ADX mean the trend is reversing? No — it means directional movement is weakening, which is just as often a trend pausing to consolidate as one ending. A stock can digest a big advance sideways, let ADX bleed down, then resume. Reversal is a claim about direction, and ADX by construction holds no opinion on direction. ### Why do so many scans pair signals with ADX above 20? Because most trend-following signals — DI crosses, Supertrend flips, average crossovers — fire constantly in sideways markets, where nearly all of them fail. Requiring ADX above 20 confines the signal to stocks where a trend exists to change. It is the cheapest known filter against chop, and its effect on any scan's hit rate is directly testable here.