Skip to main content

ADX (Trend Strength)

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,117 words

The Average Directional Index (ADX) is J. Welles Wilder's gauge of how strongly a market is trending — in either direction — distilled into a single line that oscillates between 0 and 100. Its defining tension is that it is direction-blind: a high ADX tells you a trend is powerful but says nothing about whether it is up or down (that job belongs to the two companion lines, +DI and −DI). For a swing trader the value proposition is narrow but real: ADX is a context filter that separates "trade the breakout / ride the move" markets from "fade the chop / stand aside" markets. It is not, by itself, an entry signal, and treating it as one is the most common way traders lose money with it.

How it's calculated / formed

ADX is the last of three layers Wilder published in New Concepts in Technical Trading Systems (1978), all using a default 14-period smoothing:

1. Directional Movement. +DM is today's high minus yesterday's high; −DM is yesterday's low minus today's low. Only the larger of the two counts (the other is set to zero), and neither counts if price simply stayed inside the prior range. 2. Directional Indicators. Smooth +DM and −DM over 14 periods (Wilder's running smoothing, not a simple MA), divide each by the smoothed True Range, and multiply by 100 to get +DI and −DI. These are the directional lines. 3. DX, then ADX. DX = 100 × |(+DI) − (−DI)| / ((+DI) + (−DI)). ADX is the 14-period Wilder-smoothed average of DX.

Because DX is smoothed twice (the DIs are already smoothed, then DX is averaged again), ADX is a heavily lagged line. StockCharts notes roughly 150 periods of data are needed before ADX values fully stabilize, and that the double-smoothing gives it "a fair amount of lag."

How it's used in practice

For swing and short-term trading, three uses dominate:

  • Regime filter (the primary use). Per Wilder/StockCharts convention, ADX > 25 = trending, ADX < 20 = no trend / range, with 20–25 a gray zone. A trader takes trend-following setups (pullback buys, breakout continuation) only when ADX is above ~20–25 and rising; in a sub-20 reading they switch to mean-reversion tactics (fade band extremes, sell resistance / buy support) or simply pass.
  • Rising-ADX as the real signal. The slope matters more than the level. A trend gaining conviction shows ADX turning up from the teens through 20–25; that is the swing trader's green light to hold winners and add. A high but falling ADX (e.g. rolling over from 40+) warns that an established trend is losing steam — time to tighten stops, not initiate.
  • DI crossovers for direction/entry. +DI crossing above −DI is a bullish trigger; −DI above +DI bearish. Wilder validated these only when ADX confirms strength, and used the "extreme point rule" — place the initial stop at the signal bar's low (long) or high (short) and hold until penetrated even if the lines cross back, which suppresses whipsaw exits.

A practical swing playbook: filter → direction → trigger → confirm. Require ADX rising above ~20–25 (filter), +DI > −DI for longs (direction), a pullback/breakout entry on the price chart (trigger), and an independent confirmation such as volume expansion or a chart pattern (confirm — StockCharts explicitly says ADX/DI signals need outside confirmation). Caution on the top end: readings above ~40–50 mark very strong trends but often arrive late, near exhaustion, so they argue for trailing rather than fresh entries.

Adoption, debate & evidence

ADX is one of the most widely shipped indicators — standard in virtually every charting platform — and well respected as a regime classifier. But its standalone predictive edge is weak, and the honest evidence is mixed.

  • The headline weakness is lag: by construction it confirms trends that already exist, so the cleanest part of a move is often gone before ADX rises.
  • An independent multi-asset backtest (quant-signals.com) of 24 variations across ~4,236 trades (forex, crypto, gold, indices; daily and hourly) found 9 of 24 produced negative expectancy. Critically, ADX used as a filter on other strategies reduced trade count 60–80% without reliably improving profitability — e.g. a BTCUSD daily system's profit factor fell from 1.56 to 1.16 when ADX filtering was added despite cutting trades 81%. The same study's better outcomes came from DI crossovers as direct signals on daily trending assets, not from filtering. Treat these as a single non-peer-reviewed source, not a verdict — but the direction of the finding matches the well-documented lag critique.

The fair synthesis: ADX is a credible descriptive tool (it does measure trend strength) but a poor predictive one. It earns its keep as a regime context layer feeding other decisions, not as a trade generator.

Strengths & limitations

Works when: a market is in a sustained, persistent trend and the trader's goal is to avoid range-bound chop or to decide whether to apply trend-following vs mean-reversion logic. It is genuinely useful for classifying the environment.

Fails when: markets chop or whipsaw with no net direction (when +DI and −DI stay close, DX/ADX stays low and gives no edge), at trend turns (lag), and at extreme highs (late, near exhaustion). A sharp directional spike followed by reversal can also briefly lift ADX into "trending" territory just as a move ends. On low-volatility individual stocks, Wilder's default parameters may never generate reliable signals.

The #1 misuse: treating a high ADX as a buy signal. ADX is direction-blind — a high reading during a downtrend is strength to the downside. Always read +DI/−DI (or the price chart) for direction before acting.

Sources