Moving Averages and What They Actually Say

A moving average is an average of recent closes. That is all it is. It cannot see the future, it contains no information the price does not already contain, and everything it tells you is about the past — which turns out to be useful, as long as you are not expecting anything else.

Simple and exponential

A simple moving average adds the last N closes and divides by N. Every close counts the same, and the one that drops off the back counts for nothing the moment it does.

An exponential moving average weights recent closes more heavily and never fully forgets the old ones. It turns sooner after a sharp move, and it is noisier for the same reason.

The difference between them is smaller than the debate about them. On most charts a 20 SMA and a 20 EMA disagree by a fraction of the spread most of the time, and any method whose results depend on which one you picked is fitted to noise.

Choosing a period on purpose

The period sets the trade-off, and it is the only parameter that really matters. Short means responsive and noisy; long means stable and late.

Choose it from a purpose, not from a list. "I want a line that summarises roughly the last month of daily closes" gives you 20 on a daily chart. "I want the level everyone else is watching" gives you the 200. "I optimised it and 37 backtested best" gives you a number fitted to history that will not repeat.

Slope, not crossover

The crossover is the famous signal and the weakest thing an average offers. Two averages crossing is an arithmetic consequence of a move that has already happened — by construction, it arrives after the move.

In a trend, crossovers look superb in review. In a sideways market they produce a continuous stream of signals in alternating directions, each one late, and that is where a crossover system gives back what the trend gave it.

What the average is actually good for

• Slope. Rising, falling or flat is a mechanical trend classification, in the sense lesson 4 demanded, and it needs no crossover.
• Position. Price persistently above a rising average is a different market from price oscillating across a flat one. This is a classification, not a signal.
• As a dynamic level. In a strong trend, pullbacks often stall near a well-watched average — partly because many participants are watching it. Treat it as a zone, like any other level.
• As a filter. "Only take long setups while the 50 is rising" removes a category of trade. That is a legitimate and testable use.

Every one of those uses reads the average as a summary of conditions. None of them treats it as an instruction to enter, which is the distinction that separates an average that helps from one that generates losses at a steady rate.

Where averages fail

The failures are predictable, which makes them manageable.

Notice that the first two are the same property seen twice. The lag that makes an average stable in a trend is what makes it wrong at a turn — you cannot have one without the other, and no variant of the calculation escapes it. Everything marketed as a "zero-lag" average has simply traded lag for noise somewhere else.

Using it honestly

• One or two averages, not five. Five lines produce a signal at all times and information at none.
• Do not use it to set the stop. The stop goes where the idea is wrong; an average is not that place.
• Do not use it as the reason for a trade. It is context. The reason comes from structure, a level, or a tested setup.

Read that way, a moving average earns its place on a chart: one line, chosen for a purpose, answering one question about conditions — and silent on everything else.