Trend indicators: what a moving average smooths away
A moving average answers one question — what has the average price been? Everything else people read into it is inference.
A moving average is the average of the last n closes, recalculated each bar. That
is the entire idea. The variants differ only in how they weight the bars inside the window.
The simple average treats a bar from thirty sessions ago exactly like this morning's. The exponential average never fully discards old data but weights it down geometrically, so it reacts faster to recent bars. The weighted average sits between them.
What gets removed
Smoothing removes variation. That is the point — but the variation removed includes the extremes, and extremes are frequently the part of the chart that matters.
A day with a violent reversal and a day that drifted quietly can produce identical closes and therefore identical contributions to the average. The average cannot distinguish them. Anything you wanted to know about how the day traded is gone by the time it reaches the line.
This is why averages of the close and measures of range answer different questions. An average tells you where price has been centred. A range measure tells you how far it travelled to get there. Neither substitutes for the other.
The period is a memory setting
Choosing n is choosing how much history you want the line to remember. There is
no correct value, and the conventional ones — 20, 50, 200 — are conventions rather
than findings. They are widely watched, which is a genuine reason to know where they sit, but
it is a different reason from the one usually given.
Crossovers, and where they fail
A faster average crossing a slower one is the oldest trend rule there is. It works in the sense that it will always get you pointed the right way during a sustained move, and it fails in the specific and common case of a sideways market, where the two lines cross repeatedly and each crossing is immediately reversed.
The failure is structural. A rule built to detect persistence will misfire when there is no persistence to detect. No parameter fixes this, because the problem is the market condition, not the setting.
What a moving average is genuinely good for
- Stating the direction of the last
nbars without argument - Giving a consistent reference that means the same thing on every chart
- Making the difference between a drifting market and a trending one visible at a glance
Those are modest and real. Treating the line as support, as a forecast, or as a reason on its own is where the trouble starts.