Indicator library · Volume
Volume Moving Average
Every claim about heavy or light volume rests on a yardstick, and this is the yardstick. It is also the number most often quoted without saying which window, or which kind of average, produced it.
Which VMA this is
The abbreviation carries three meanings and they are unrelated. This page is about a moving average of the volume series: the sense this reference uses throughout. Elsewhere, VMA means a variable moving average of price, whose period adapts to volatility, and occasionally a volume-weighted average of price, which is closer to VWAP. Before quoting anyone else’s VMA, establish which of the three they plotted.
The calculation is unremarkable: VMA = mean of the last n volume figures. What makes it worth a page is that it is the denominator underneath every statement about participation anywhere on this site. "Heavy volume", "thinning volume" and "three times average" are all claims about a comparison, and this is the thing being compared against.
Why the mean overstates an ordinary session
Volume is strongly right-skewed. Most sessions cluster in a band; a few (expiries, index rebalances, earnings, a news shock) are several times larger. On a symmetric series the mean is the typical value; on this one it is pulled upward by the tail, and the pull lasts exactly as long as the outlier stays inside the window.
The consequence is circular and easy to miss. If a heavy session raises the average, the next heavy session is measured against a yardstick the previous one inflated, so a genuine cluster of activity reads as progressively less unusual. Sessions immediately after a spike look quiet when they are ordinary, and the effect ends abruptly when the spike leaves the window, exactly like the window-exit artefact on any simple moving average.
| Yardstick | Behaviour |
|---|---|
| Simple mean | The convention. One outlier lifts it for the whole window, then drops out and steps it back down. |
| Rolling median | Barely moved by one or two extreme sessions. Answers what people usually mean by "typical". |
| Mean with mechanical dates excluded | The most work and the most honest: drop expiries and rebalances, whose volume expresses no view. |
Choosing the window
Window length matters more here than for a price average, because volume has a rhythm price does not: a weekly pattern, a monthly one around expiries, and a quarterly one around index reconstitution. A window shorter than those cycles reports the cycle rather than the market.
Twenty sessions covers a month and averages the weekly pattern out. Fifty covers a quarter and absorbs a monthly expiry. Five does neither. Whichever is used, the window belongs beside any multiple that gets quoted, and if a figure is being compared with someone else’s, the window and the kind of average have to match before the comparison means anything.
Relative volume, and why the denominator travels with it
The form most screens actually use is relative volume, today's figure divided by this average, and it inherits every choice made here. That is worth spelling out because relative volume is quoted constantly and almost never with its provenance.
Three numbers are needed before a reading like "2.4× average" can be checked by anyone else: the window, whether the average is a mean or a median, and whether mechanical sessions were excluded. Change any one and the same market produces a different multiple. It is the same standard this site applies to a smoothing constant or a pivot formula. The arithmetic is trivial and the convention is the part that has to be stated.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Mean after a spike | The yardstick is inflated by the event being measured, so the sessions that follow read as quiet when they are ordinary. |
| Window exit | When a heavy session leaves the window the average steps down on a day when nothing happened. |
| Short window | Anything under about ten sessions sits inside volume's own weekly rhythm and reports the calendar. |
| Unadjusted volume | A split doubles the share count permanently; every average spanning the date is meaningless. |
| Mechanical sessions included | Expiries and rebalances contribute volume that carries no opinion, and they raise the baseline for weeks. |
| Flat average on intraday bars | Volume clusters at the open and the close, so every opening bar looks extraordinary against a flat average. |
Why this page exists at all
Most of this library describes measures that produce a reading. This one describes a measure that produces a baseline, and it is the least examined number in volume analysis: every statement about a surge, a dry-up or a divergence in participation is a comparison against it, and the comparison inherits whatever was wrong with the average.
The volume spread analysis page is the clearest case. Its definitions turn on phrases like "volume lower than the two previous bars", a deliberately local yardstick, chosen precisely because a longer average would have been distorted by whatever happened three weeks ago. Knowing what a volume average does under skew is what makes that choice look like judgement rather than arbitrariness.
Frequently asked questions
What is a volume moving average?
An ordinary moving average computed on the volume series instead of on price, usually simple, usually 20 or 50 sessions. Its job is to turn a vague claim into a measurable one: "heavy volume" means nothing until there is a number to compare against, and this is the number almost every volume statement on this site is implicitly using.
Is VMA the same as a volume-weighted moving average?
No, and the abbreviation is genuinely overloaded. This page describes an average OF volume. Elsewhere "VMA" means a variable moving average of price, whose period adapts to volatility, and sometimes a volume-weighted average of price, which is closer to VWAP. This reference uses VMA to mean the volume moving average throughout; when reading anyone else’s chart, establish which of the three they plotted.
What window should I use?
Twenty sessions for a month of context, fifty for a quarter, and the choice matters more than it does for a price average, because volume has a strong weekly and monthly rhythm. A five-day window sits inside that rhythm and will call the same weekday quiet or busy for no reason. State the window whenever you quote a multiple: "three times average" over 10 sessions and over 50 are different claims about different things.
Why is the median often better than the mean?
Because volume is strongly right-skewed. Most sessions cluster in a band and a few are several times larger, so one expiry or index-rebalance session lifts a 10-day mean substantially and keeps it there for the whole window. The yardstick you are measuring "above average" against has then been moved by the very event you were trying to measure. A rolling median is barely affected by one outlier and answers what people actually mean by typical.
How does this relate to relative volume?
It is the denominator. Relative volume — today’s volume divided by the average of the last n sessions — is the form most screens use, and it inherits every property of the average underneath it: the window length, mean against median, and whether mechanical sessions were excluded. A relative-volume figure quoted without those three choices cannot be reproduced by anyone else.
Does volume need adjusting for splits?
Yes, and it is the mirror image of the price adjustment: a two-for-one split doubles the share count for the same money, so an unadjusted series steps up permanently on the split date and every average spanning it is meaningless. Most providers adjust volume alongside price. When they do not, a long history contains steps that look like changes in interest and are arithmetic.
Should intraday volume use the same approach?
Not against a flat average, no. Volume within a session is strongly shaped, a disproportionate share arrives in the first and last few minutes, so comparing a mid-morning bar with a flat daily average makes every opening bar look extraordinary and every midday bar quiet. The minimum adjustment is to compare each bar against the average for that same part of the session. This distortion is the problem this site is built around.
What multiple of average volume is meaningful?
It depends on the instrument, and it can be computed rather than guessed. Take a few years of that instrument’s own relative-volume readings and find the percentiles: its 95th and 99th are meaningful thresholds, whereas "twice average" is a rule of thumb that is routine on one stock and a once-a-year event on another. Every threshold on this site gets the same treatment.