What it measures

An exponential moving average is a running estimate of where price has been, weighted so that the latest bars count most and old bars fade out rather than drop off a cliff. It answers one question: is price above or below its recent centre of gravity, and which way is that centre moving. It is the most used calculation in the strategies the desk screened, by a wide margin.

The calculation

Each new value blends the latest price with the previous average:

EMA today = α × price today + (1 − α) × EMA yesterday, with α = 2 / (length + 1)

A 20-period EMA uses α ≈ 0.095, so today's price carries about a tenth of the weight and the rest is yesterday's average. The series is seeded with a simple average of the first length bars, which is why two charting packages can disagree slightly in the first few hundred bars of history.

Inputs and defaults

  • Length: the only real parameter. 9 and 21 are the day trader's pair, 50 and 200 the position trader's, because they were the defaults in the first charting packages and everyone has watched them since.
  • Source: close by default; some strategies feed it a typical price (high + low + close) / 3 or another indicator.

How it is read

Price above a rising EMA is read as an uptrend; a shorter EMA crossing above a longer one is read as a trend starting. Two EMAs of different length are the simplest possible trend-following rule, and a large share of the screened universe is some elaboration of that rule with filters added.

Where it fails

An EMA lags price by roughly (length − 1) / 2 bars, so a 200 EMA tells you about a trend that is already a hundred bars old. In a range, two EMAs cross back and forth and every cross is a losing trade; that is the whipsaw cost every trend rule pays, and the EMA does nothing to reduce it. The shorter the length, the more crosses, the more fees.

How the desk reads it

The desk treats any moving-average rule as a hypothesis about trend persistence and reads it across coins and timeframes before trusting it on one. The lengths are parameters, whatever their tradition, and a rule that only works at 9 and 21 is a rule that works on the data it was fitted to.

In the desk's screen

Of the 614 screened strategies in the library, 325 (52.9%) use this indicator somewhere in their logic.

130 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.