Exponential moving average (EMA)

A moving average in which each observation carries a weight that decays geometrically with age. Where a simple average gives every price in its window the same weight and drops the oldest one entirely each time it advances, an exponential average keeps a diminishing share of everything that came before, so nothing ever leaves the calculation completely.

The calculation is recursive and needs only two inputs: the latest price and the previous value of the average. A smoothing factor is derived from the chosen number of periods, the latest price is multiplied by that factor, the previous average is multiplied by one minus it, and the two are added. A shorter period means a larger factor, which means the average tracks price more closely and lags it less.

Because it responds faster it also reverses more often, and that is the trade rather than an improvement: an exponential average of a given length crosses a price series more frequently than a simple one of the same length, which produces earlier signals and more of the signals that immediately reverse. A second point catches people comparing charts: the recursion needs a starting value, and platforms seed it differently, some with the first close and some with a simple average of the opening window, so two charts can print marginally different lines for the same setting. Like every moving average it is computed from prices that have already happened, so it describes what a series has been doing, not what it will do.