Reference · Volume
Average Volume as a Baseline
Nearly every claim that a session was heavy is a comparison against an average, usually an unstated one. This is the page about the denominator: which window, whether today belongs in it, and why the mean sits above the typical day.
The figure only matters as a denominator
An average volume of 1.2 million shares tells you nothing by itself. It becomes useful the moment it is divided into a session: today was 1.7 times the recent norm, or 0.4 times it. That ratio, relative volume, is the form every volume claim on this site reduces to, and it is worth stating explicitly because so much commentary quotes the raw count instead.
The reason is comparability. A share count can only be interpreted by someone who already knows the instrument, and it cannot be compared with the same instrument five years ago, when its float and its shareholder base were different. A ratio around 1 means the same thing in every context, which is what makes it possible to say anything general about volume at all.
Four decisions hidden inside one number
Two platforms reporting different average volume for the same instrument are usually making different choices from this list, and none of the four is stated on a chart.
1. Whether today is in the window
It should not be. A session compared against a baseline that includes it is being compared partly against itself, which flattens exactly the extreme readings the comparison exists to find. On a twenty-session window an included current bar pulls the ratio in by roughly five per cent, small in ordinary conditions, and largest precisely on the days that matter.
2. The window length
Twenty sessions is the standard and a sound default, because it contains each weekday four times and so averages out the weekly pattern. Five sessions does not, and a five-day comparison partly measures which days of the week it happens to span. Fifty or more is steadier and slower to notice that an instrument’s liquidity has genuinely changed.
3. Mean or median
Convention says mean, and platforms report the mean, so use it for the ratio, but know that on a skewed series the mean sits above the typical session. When the question is "what does a normal day look like here", the median answers it and the mean does not.
4. What the window contains
One rebalance or expiry session inside the baseline raises the denominator for the whole window, making the following weeks read as quieter than they were. The dates are published in advance; the fix is to know them rather than to quietly remove the bars.
Reading a relative volume figure
| Reading | What it says | Check first |
|---|---|---|
| Below about 0.6 | A quiet session. Price moves made on it required little stock to change hands, and gaps in a thin book move price further per share. | A holiday, a half-session, or a market-wide quiet spell rather than anything specific to the instrument. |
| Around 1 | Ordinary participation. This is where most sessions sit, and volume adds nothing to the reading of the price. | Nothing. The useful conclusion is that there is no conclusion. |
| 1.5 to 3 | A genuinely busy session: the move, in whichever direction, involved a lot of stock. This is the range where volume earns its place in the reading. | Whether the whole market was busy. A sector or index-wide expansion is context, not a finding about this instrument. |
| Above 5 | Something specific happened, and it is usually identifiable: results, an index change, an expiry, a corporate event, or forced selling. | The news and the calendar, before any indicator. A reading like this that stays in the baseline distorts the next twenty sessions. |
The row worth arguing with is the third. A ratio of two is the range in which most of the volume literature makes its claims, and it is also the range where the market-wide check matters most, a session that is busy everywhere says nothing about the instrument in front of you, and a great deal of published volume analysis omits that comparison entirely.
Intraday, where the pattern is stronger than the signal
Relative volume computed intraday needs a different baseline. Trading is concentrated at the open and the close, so an average across the whole day is a poor comparison for any particular half-hour: measured against it, every opening looks extraordinary and every lunchtime looks dead, on every instrument, every day.
The correct comparison is the same time of day across previous sessions. This half-hour against the same half-hour over the last twenty days. That removes the shape of the trading day and leaves what is actually unusual about now. It is more work, and it is the difference between a figure that describes the instrument and one that describes the clock.
Where this baseline is used on this site
Almost everywhere, and mostly by implication, which is the reason for stating it once here. The volume moving average is this number as an indicator. Volume spread analysis asks whether a bar’s range was justified by its volume, which requires a normal volume to compare against. The money flow index and every other volume-weighted oscillator inherit whatever baseline is embedded in their period.
Which means a criticism of this figure is a criticism of all of them. If the window is too short, every measure built on it partly measures the calendar; if today sits inside the baseline, every one of them understates the extremes. Getting the denominator right is unglamorous and it propagates further than any single indicator choice on the page.
Frequently asked questions
What is average volume?
The mean number of shares traded per session over some recent window, most often twenty or fifty sessions, sometimes three months. It exists to answer one question: was today heavy or light. On its own the figure says very little, because a share count means nothing without knowing the instrument; as a denominator it is the most useful number on this site, because the ratio it produces is comparable between instruments and across time.
Why is relative volume the better form?
Because a ratio is scale-free. Two million shares is enormous for one company and a quiet morning for another, so a raw count can only be read by someone who already knows the instrument. Dividing today by the recent average gives a number around 1 that means the same thing everywhere: 2.4 is two and a half times the recent norm whether the instrument trades thousands of shares or tens of millions. Every "heavy volume" claim in this reference is a ratio underneath.
Should the average include today?
No, and this is a real defect in several published implementations. If today is inside the window it is being compared against itself, which drags the baseline towards the current session and understates how unusual it is, most visibly on exactly the extreme days you care about. Use the previous n sessions and exclude the current one. On a twenty-session window the difference is around five per cent of the reading, which is enough to move a session from notable to ordinary.
Why is the mean the wrong summary for volume?
Because the distribution is skewed. Volume has a floor at zero and no ceiling, so the occasional enormous session pulls the mean above the typical one, on the series in the figure below, well over half of all sessions fall below the mean. If you want to know what a normal day looks like, the median is the better answer. The mean remains the right denominator for a ratio, because it is what convention uses and what your platform reports; just do not read it as typical.
What window length should I use?
Twenty sessions is the standard for daily data and a good default: it spans four weeks, so it contains each weekday four times and averages the weekly pattern out. Shorter windows measure the calendar as much as the market. A five-session average is partly a statement about which weekdays it happens to contain. Longer windows, fifty or more, are steadier and slower to acknowledge a genuine change in an instrument’s liquidity after an index inclusion or a corporate event.
Does volume really have a weekday pattern?
Yes, consistently enough to matter for short windows. Midweek sessions tend to be heavier than Mondays and Fridays, and the effect is large enough that a five-day comparison can report a change that is entirely the calendar. There are equally strong intraday patterns — the first and last half-hours carry a disproportionate share of the day’s trading — which is why intraday relative volume has to be compared against the same time of day rather than against a flat daily average.
How do index rebalances and expiries distort the average?
Severely, and predictably. A quarterly index rebalance or an options expiry concentrates enormous volume into one session, often several times the normal level, and that session then sits inside the baseline for the whole window afterwards, raising the denominator and making the following weeks look quieter than they are. The dates are published in advance, so the honest fix is to know when they were and to say so rather than to quietly drop the bars.
Is a heavy session bullish or bearish?
Neither, and the question is the most common mistake made with this number. Volume is a measure of participation, not of direction: every share bought was sold. A heavy session says the move mattered enough for a lot of stock to change hands, which makes the price move more meaningful in either direction, and the interpretation always requires the price alongside it. The measures in the library that appear to give volume a sign, on-balance volume and its relatives, do so by assigning the whole session’s volume to the direction of the close, which is a convention rather than a measurement.