ADX: Measure Trend Strength Before You Trade It

ADX (Average Directional Index) tells you how strong the current trend is — not which direction it is going. It solves the most common problem in technical analysis: entering a trend-following indicator signal in a market that is actually ranging.

Section 1: Core Mechanics

ADX is part of the Directional Movement System developed by J. Welles Wilder in 1978. It always moves between 0 and 100. Rising ADX = trend getting stronger. Falling ADX = trend weakening. ADX does not tell you if the trend is up or down — only how strong it is.

The system comes with two companion lines: +DI (Positive Directional Indicator) and -DI (Negative Directional Indicator) that show direction. ADX shows strength.

Formula

Directional Movement:

True Range and Directional Indicators:

Directional Index and ADX:

Where Wilder smoothing uses (heavier than standard EMA) and requires a seed of N bars before producing valid values.

Inputs

• High: Bar high price
• Low: Bar low price
• Close: Previous close for True Range calculation
• Period (N): Smoothing period for all Wilder calculations. Default: 14.

Parameters

Output

Three series: ADX (0-100), +DI (0-100), -DI (0-100). All plotted in a separate sub-panel below price. ADX is typically the thicker line; +DI and -DI are thinner.

Visual Behavior

• ADX rising from below 20 toward 25 = trend building
• ADX above 25 and rising = strong trend — trend-following strategies are valid
• ADX above 40 = very strong trend (less common, often near exhaustion)
• ADX falling from above 25 = trend weakening, watch for reversal or consolidation
• +DI above -DI = upward directional movement dominant
• -DI above +DI = downward directional movement dominant

Section 2: Interpretation & Signals

ADX Strength Levels

DI Crossover Signal

The +DI/-DI crossover gives direction alongside ADX's strength reading:

• +DI crosses above -DI AND ADX > 25 → bullish entry signal
• -DI crosses above +DI AND ADX > 25 → bearish entry signal
• Crossover when ADX < 20 → ignore — no trend behind it

ADX as a Filter for Other Indicators

The primary use of ADX is not to generate signals itself — it is to validate signals from other indicators:

• Before acting on an EMA crossover → check ADX > 25
• Before acting on a MACD bullish cross → check ADX > 20
• Before entering any Supertrend or Ichimoku signal → confirm ADX is rising

Extreme ADX Readings

When ADX peaks above 50 and begins to fall, it signals that the trend is exhausting itself. Price may continue moving in the same direction (momentum carries it), but the rate of change is declining. This is a warning to tighten stops, not necessarily to reverse.

Section 3: Pass vs. Live — Real-Time Reliability

ADX uses Wilder smoothing, which converges more slowly than standard EMA. A 14-period ADX typically needs 28+ bars before the reading is stable. Always use at least 2× the period in warmup bars. On bar close, ADX locks — no repaint.

Section 4: Practical Use Cases

Real example: Any sustained market trend — when ADX(14) daily rises through 25, it signals the start of a tradeable trend. During SPY's 2023 bull run, ADX cleared 25 in January 2023 and remained elevated throughout the year. Markets where ADX stayed below 20 for weeks (like late 2022 consolidation) crushed trend-following strategies with repeated whipsaws.

Section 5: Pseudo Code

INPUT: high[], low[], close[], period=14

PROCESS:
  Step 1: Compute True Range for each bar i:
            tr[i] = max(high[i]-low[i], abs(high[i]-close[i-1]), abs(low[i]-close[i-1]))

  Step 2: Compute raw DM:
            plus_dm[i]  = high[i]-high[i-1] if high[i]-high[i-1] > low[i-1]-low[i] and > 0 else 0
            minus_dm[i] = low[i-1]-low[i]   if low[i-1]-low[i] > high[i]-high[i-1] and > 0 else 0

  Step 3: Wilder smooth TR, +DM, -DM using period N:
            Seed: wilder_tr[N]   = sum(tr[1:N+1])
            For i > N: wilder_tr[i] = wilder_tr[i-1] - (wilder_tr[i-1]/N) + tr[i]
            (same formula for wilder_plus_dm and wilder_minus_dm)

  Step 4: Compute DI lines:
            plus_di[i]  = 100 * wilder_plus_dm[i]  / wilder_tr[i]
            minus_di[i] = 100 * wilder_minus_dm[i] / wilder_tr[i]

  Step 5: Compute DX:
            dx[i] = 100 * abs(plus_di[i] - minus_di[i]) / (plus_di[i] + minus_di[i])

  Step 6: Wilder smooth DX to get ADX:
            Seed: adx[2*N] = average of dx[N : 2*N]
            For i > 2*N: adx[i] = (adx[i-1] * (N-1) + dx[i]) / N

OUTPUT: adx[], plus_di[], minus_di[]
EDGE CASES:
  - Need at least 2*N bars for valid ADX: 28 bars minimum for period=14
  - If plus_di[i] + minus_di[i] = 0: set dx[i] = 0 to avoid division by zero
  - NaN in high/low/close: propagate NaN through entire chain

Section 6: Parameters & Optimization

Standard ADX Conventions

Parameter Impact

Market-Specific Adjustments

• Crypto: ADX(14) on daily works well. On 4H, use ADX(10) for faster regime detection
• Equities earnings: ADX spikes hard around earnings — ignore ADX for 2 days before and after reports
• Forex: Standard ADX(14) daily — pairs like USDJPY tend to trend strongly; ADX stays elevated longer

Section 7: Synergies & Conflicts

Section 8: Common Mistakes

Section 9: Cheat Sheet