What it actually computes
A 20-period simple moving average is the arithmetic mean of the last twenty closes, recalculated each period. That is the entire definition. Add twenty numbers, divide by twenty, plot the result, repeat.
It follows that the line cannot contain information that was not already in those twenty closes. It is a summary of recent price, drawn as a line so the eye can see the middle of the data without being distracted by every wick.
The lag is the feature
A moving average is always late. People say this as a criticism; it is actually the point.
Because the line includes twenty closes, it takes several new closes to shift it meaningfully. That is what filters out noise. If the line reacted instantly to price it would be price, and it would tell you nothing you could not already see.
The trade-off is fixed and unavoidable: shorter period, less lag, more noise. Longer period, more lag, fewer false signals. There is no setting that gives you both, and searching for one is how people spend years optimising parameters that will not survive the next market.
Slope over crossover
The most commonly taught use — buy when price crosses above, sell when it crosses below — is also the least reliable, for a reason that is obvious once you see it.
The more durable reading is the slope. A rising average means the recent mean of closes is climbing — which is a plain, checkable statement about what has happened. A flat average means it is not, and that is the market telling you it has no trend to summarise.
Ask the average what the middle of recent price is doing. Do not ask it what price will do next. It does not have that number.
Why price "respects" it
Price does sometimes bounce off a moving average, and it is worth being honest about why.
Partly it is circular: enough participants watch the same common periods — 20, 50, 200 — that orders cluster near those lines, which makes the lines behave like levels. That is real, and it is a statement about crowd behaviour, not about mathematics.
Partly it is an illusion. In any steady trend, price oscillates around its own mean by construction. An average drawn through it will therefore appear to "support" pullbacks — but so would a hand-drawn line through the middle of the same candles. Attributing the bounce to the indicator mistakes a description for a cause.
Adding a second and third average to get more signals. Two averages crossing is just a slower version of price crossing one average, and it inherits every weakness of the original in a sideways market — while looking more sophisticated. If one average is not telling you anything, the problem is the market condition, not the number of lines.
Read it yourself
One average over a full cycle from decline to advance.
- Where does price first cross above the average?
- Where does the slope of the average turn from falling to rising?
- Which of those two events would have been the more reliable read, and why?
Show answers
1. Around the seventh candle, while the line itself is still pointing downward.
2. Around the twelfth, several candles later. The average needs new closes to work through the window before its direction changes.
3. The slope change, on this chart and in general. The crossing happened while the average was still falling — which is exactly the situation that produces the eight whipsaws in the sideways example above. The slope turning is slower and gives up some of the move, and in exchange it does not fire in conditions where there is no trend to catch. That trade — later entry, far fewer false ones — is the whole argument for reading slope instead of crossings.
Terms used
- Simple moving average
- The mean of the last N closes, recalculated each period.
- Exponential moving average
- The same idea with recent closes weighted more heavily. Less lag, more noise.
- Period
- How many closes go into the average. The only real setting.
- Lag
- The delay between price changing and the average reflecting it. Unavoidable, and the source of its usefulness.
- Whipsaw
- Repeated false signals in a sideways market.