Moving Averages: Simple and Exponential
The most used indicator there is. An average of recent closes, redrawn each bar, which converts a jagged series into a slope.
MadStockAlerts Research · Updated August 28, 2026
What to take away
- A simple moving average weights every period equally; an exponential moving average weights recent periods more.
- The period chosen determines the horizon being described, and nothing else.
- Averages describe trend well and are worse than useless in a sideways range.
- Widely watched periods matter partly because they are widely watched.
- Price distance from an average measures stretch against that security's own history.
MAD Academy Training Video · 0:43
The Average Is Always Late — On Purpose
Lag is not a defect in a moving average. It is the feature you are paying for, and knowing the price of it is the whole skill.
This lesson is part of a Stock Alerts + Tools plan.
The calculation
SMA(n) = (C1 + C2 + ... + Cn) / n
- C is the closing price of each of the last n periods
Each bar the oldest close drops out and the newest enters, which is what makes it move. A consequence worth knowing is that a simple moving average can turn because of what left the window rather than what entered it: a very large bar rolling off the back end moves the line with no new information at all.
An exponential moving average applies a decay so that recent closes count for more and older ones fade rather than drop off. It turns sooner than a simple average of the same length and, for the same reason, changes direction more often.
EMA today = (close today x k) + (EMA yesterday x (1 - k)), where k = 2 / (n + 1)
Choosing a period
| Period | Conventional use |
|---|---|
| 9 / 10 | Very short-term momentum; used heavily on intraday charts |
| 20 / 21 | About a month of trading; the common short-term trend reference |
| 50 | The medium-term trend on a daily chart |
| 100 | An intermediate reference between the 50 and the 200 |
| 200 | The long-term trend, and the most widely watched line in equities |
There is nothing mathematically special about 50 or 200. They matter because enough participants watch them that behaviour clusters around them, which makes them partly self-reinforcing rather than predictive.
The self-reinforcement is real but limited. A level that many people watch attracts orders, and orders are supply and demand. It does not follow that price must respect the level, only that more decisions are being made near it than elsewhere.
What they describe well
- Direction: a rising average means the recent window is higher than the one before it.
- Structure: price consistently above a rising average is what an uptrend looks like numerically.
- Reference levels: pullbacks in strong trends frequently stall near widely watched averages.
- Separation: price far above its average is stretched relative to its own recent history.
- Agreement: several averages of different lengths sloping the same way describes a trend at several horizons at once.
The separation reading is the one most often underused. Expressing the gap between price and an average as a percentage turns a visual impression into a comparable number, and shows when a security is further from its own trend than it usually gets.
Where they fail
In a sideways range a moving average sits in the middle of the noise and crosses price constantly. Every crossing looks like a change of trend and none of them is. This is the single largest source of losses from mechanical moving-average systems, and it is inherent rather than fixable by parameter choice.
Scroll the chart sideways to see all of it.
- SMA 20
- SMA 10
The usual mitigation is not a better average but a filter that decides whether a trend exists at all before the average is consulted, which is the job ADX is normally given.
The variants, and what each trades away
Every moving average is a compromise between responsiveness and stability, and the variants differ only in how they weight the past. None of them removes the compromise; they move along it.
| Type | Weighting | Character |
|---|---|---|
| Simple | Equal across the window | Stable. A single large bar leaving the window moves it |
| Exponential | Decaying, most weight on the newest | Faster to respond, and noisier |
| Weighted | Linearly decaying | Between the two, and rarely used |
| Hull and similar | Composite, tuned for lag reduction | Faster still, at the cost of overshoot |
| Volume-weighted | By volume rather than by time | Reflects where trading happened, not just when |
The first row contains an effect that is easy to miss. A simple moving average changes when a bar drops out of the back of the window, which means the line can move sharply on a day when nothing much happened, because of what occurred exactly one window ago.
No smoothing method can be both fast and stable, because the two properties are the same trade-off viewed from either end. A variant advertised as removing lag has removed stability instead.