Indicators, and what the readings mean

Moving averages: SMA, EMA, and what a cross really tells you

A moving average is the simplest indicator there is, which is why it is the most misread. It does exactly one thing: it tells you what the average price used to be.

About 10 minutes to read. Free, no account.

What a moving average actually is

A simple moving average, or SMA, is the mean of the last N closing prices. Nothing is weighted, nothing is predicted. Each new bar closes, the newest price joins the window, and the oldest price drops out of it.

A 5-period SMA, by hand

Closes:   100, 102, 101, 105, 107
Sum:      515
SMA(5):   515 / 5  =  103.0

Now the next bar closes at 109. The 100 leaves the window and the 109 joins it:

One bar later

Closes:   102, 101, 105, 107, 109
Sum:      524
SMA(5):   524 / 5  =  104.8    (+1.8)

Look at where that +1.8 came from. The new bar added 109/5 = 21.8 and the departing bar removed 100/5 = 20.0. Half of the move in the average was caused by a price from five bars ago leaving the window - something that happened in the past, not today.

SMA versus EMA: where the difference comes from

An exponential moving average, or EMA, weights recent prices more heavily. Instead of a window it uses one multiplier, usually written k, applied to the newest close:

The EMA recurrence

k            =  2 / (N + 1)
EMA(today)   =  close x k  +  EMA(yesterday) x (1 - k)

For N = 5, k = 2/6 = 0.3333. Take the same numbers: yesterday’s average was 103.0 and today closes at 109.

A 5-period EMA, by hand

109 x 0.3333      =  36.33
103 x 0.6667      =  68.67
EMA(5)            =  105.00

the SMA(5) above  =  104.80

The EMA is higher, because it gave today’s close a third of the weight instead of a fifth. That is the whole difference between the two, and it has two consequences worth knowing.

An EMA reacts sooner, and therefore turns on noise sooner. Faster is not the same as earlier-and-right.
An EMA has no window, so old prices never fully leave - they fade. A 20-period EMA gives the newest bar k = 2/21 = 9.5% of its value and spreads the rest backwards forever. It cannot produce the drop-out jump an SMA can, and in exchange it never completely forgets a spike.

How each of the last ten bars is weighted. The SMA gives every bar in its window the same weight and everything older a weight of zero; the EMA fades, and never quite reaches zero.

Neither is more correct. They answer slightly different questions, and on a chart of any length they mostly agree. If a decision of yours flips depending on which one you drew, the decision was never resting on the average.

What the slope and the position tell you

There are only two readings on a single moving average, and both are statements about the past.

Position - price above the average means the current price is higher than the recent mean. That is a definition, not a forecast. In a sideways market price crosses its own average constantly and means nothing by it.
Slope - a rising average means the recent mean is higher than it was. On a long average the slope is the closest thing an indicator gives you to the word "trend", and it is a description of bars that have already printed.

One practical point that catches people out: a moving average is only final when the bar closes. The value you see halfway through a 4-hour candle is computed from a close that has not happened. It can move, and so can a cross you think you just saw.

The cross, and the arithmetic of its lag

A fast average crossing above a slow one - the "golden cross" when it is 50 over 200 - is the most quoted signal in trading. Here is what it literally means: the average of the last 50 closes has just become larger than the average of the last 200. That is all it says. It is a fact about history, and it is late by construction.

You can put a number on the lateness. Suppose a coin trades flat at 100 for a long time and then steps up to 120 and stays there. Each new bar at 120 replaces a bar at 100, so:

How fast each average climbs after the step

SMA(50)  gains  20 / 50   =  0.40 per bar
SMA(200) gains  20 / 200  =  0.10 per bar
the gap opens at          =  0.30 per bar

If the 50 was sitting 6 points under the 200 when the move began, the cross happens 6 / 0.30 = 20 bars later. On a daily chart that is twenty days after the entire 20% move already happened. Nothing is wrong with the indicator; that is what averaging 200 closes does.

So the cross is useful as a slow description of what regime you are in, and close to useless as an entry trigger. Those are different jobs, and most of the disappointment with moving averages comes from hiring one for the second.

What a moving average cannot do

It cannot give you a level. "Price bounced off the 50" is a sentence about a line whose position is recomputed every bar. Round numbers and prior highs sit still; an average moves to wherever the last N closes put it.
It cannot tell you about size. An average says nothing about whether the move is large relative to how much this coin normally moves. That is what a volatility measure is for.
It cannot be optimised honestly on a chart. If you try 10, 20, 50, 100 and 200 and keep whichever would have worked best on the chart in front of you, you have chosen a number using the answer. That is curve fitting, and it is the single most common way a plan looks excellent in hindsight and fails in front of you.
It does not know what happened. An average has no idea whether the move came from a listing, a liquidation cascade or a quiet grind. It is one number summarising a window of closes.

Using an average alongside a signal

The defensible use is as context: before you act on anything, you want to know whether price is above or below its recent mean and whether that mean is rising, falling or flat. That is one glance, and it frames everything else on the chart.

What it should not be is the reason. A signal - from us or anyone else - already contains an entry, a stop and a time limit. A moving average cannot improve those numbers, and overriding a stop because an average looks supportive is how a small planned loss becomes an unplanned large one. See how a signal is built.

The terms, defined

Simple moving average (SMA): The arithmetic mean of the last N closes. Every bar in the window counts equally; everything older counts zero.
Exponential moving average (EMA): A weighted average in which the newest close gets k = 2/(N+1) of the weight and all earlier values keep the rest, fading but never disappearing.
Period (N): How many bars the average is computed over. It is a choice, not a property of the market, and the chart timeframe changes what it means.
Golden cross / death cross: A fast average crossing above / below a slow one, conventionally 50 and 200. A statement about two averages, not about the next bar.
Whipsaw: Repeated crosses back and forth while the two averages sit close together, which is what happens in a range.
Curve fitting: Choosing a setting because it would have worked on the data you are looking at. It makes the past fit perfectly and tells you nothing about the future.

What this means in practice

Read the slope first and the cross second. The slope of a long average is the cheapest summary of the regime you are trading in.
When an average turns, ask which end of its window caused it - a new price arriving or an old one leaving.
Treat a cross as a slow description, never as an entry. By the time 50 crosses 200 the move it describes is already dozens of bars old.
Ignore crosses that happen while the two averages are tangled together.
Pick a period before you look at the chart, and keep it. Shopping for the period that would have worked is the mistake, not the period.
Remember the value is provisional until the candle closes.

Nothing in this lesson is financial advice or an instruction to trade. See the risk disclaimer. We never recommend a leverage setting - that decision is yours.