What it measures

Wilder's directional movement system has three lines. +DI measures upward pressure, −DI measures downward pressure, and the ADX measures how far apart they are, smoothed, which is a reading of trend strength with no sign: a strong downtrend and a strong uptrend both give a high ADX. It is the most common trend filter in the screened universe, usually written as "only trade when ADX is above 20 or 25".

The calculation

+DM = today's high − yesterday's high, when that is larger than yesterday's low − today's low and positive; otherwise 0 −DM = yesterday's low − today's low, under the mirror condition +DI = 100 × RMA(+DM, n) / ATR(n); −DI = 100 × RMA(−DM, n) / ATR(n) DX = 100 × |+DI − −DI| / (+DI + −DI) ADX = RMA(DX, n)

Two rounds of Wilder's smoothing make the ADX slow: with n = 14 it reacts like a 27-period EMA applied twice.

Inputs and defaults

  • DI length: 14.
  • ADX smoothing: 14.
  • Thresholds: 20 or 25 for "there is a trend"; 40 and up for a strong one.

How it is read

Rising ADX means the trend is strengthening, whichever way it points; falling ADX means it is fading or the market is ranging. +DI crossing above −DI is a bullish signal in Wilder's system, and the reverse bearish. Most strategies use only the ADX, as a gate in front of some other entry rule.

Where it fails

It is late by construction: the ADX confirms a trend after the move that started it, and it keeps reading "strong trend" for a while after the trend has ended. A rule gated on ADX above 25 skips the first leg of every move. The threshold is a tuned parameter like any other, and the same strategy with ADX above 20 and ADX above 30 can have very different trade counts.

How the desk reads it

An ADX gate is read by what it removes: the desk counts the trades a strategy skipped because of it, and in which months. A gate that only trims the sideways stretches is doing its job; one that also skips the first leg of every trend is a tuned number, and the threshold deserves the same suspicion as any other fitted parameter.

In the desk's screen

Of the 614 screened strategies in the library, 80 (13.0%) use this indicator somewhere in their logic.

23 of those were gross-profitable on enough cells to reach the after-fee replay.

The count is read from the compiled code of each strategy in the library; a strategy that computes the indicator by hand under another name is not counted. The desk's own rules are not part of the count.