A stock spends three weeks bumping into $48.20 and finally closes above it at $48.55. You buy the breakout. Twenty minutes later it is back at $47.90 and you are out for a loss. The next week the same stock clears the same level again, and this time it runs to $53 without looking back.
The price action was nearly identical. What differed was how many shares changed hands to produce it: the failed attempt traded 1.9 million shares, the one that ran traded 7.9 million. Volume is what separates those two charts, and it is the only major input on a chart that is not just another arithmetic transform of price.
What a volume bar actually counts
Volume is the number of shares — or contracts, or coins — that changed hands during a bar’s time window. A daily volume of 2.4 million means 2.4 million shares were bought and sold that day; each transaction is counted once, not twice.
That last point kills the most common beginner misreading. You will hear that a green candle on high volume means “more buyers than sellers.” That is impossible. Every share sold was bought by someone. Buyers and sellers are exactly equal at all times, by definition.
What changes is urgency, and volume measures the size of the crowd expressing it. A 3% advance on 500,000 shares means few participants agreed on a higher price. A 3% advance on 8 million shares means many did, including institutions that cannot move that size without leaving a footprint. Same price outcome, very different breadth of agreement.
So the honest definition is: volume measures participation, which serves as a rough proxy for conviction. It says how many people showed up. It never says which way they were leaning — the point where most volume mistakes originate.
Why a raw volume number means nothing without a baseline
“The stock traded 7.9 million shares” tells you nothing. That is enormous for a small-cap which normally trades 300,000, and a dead session for a mega-cap which normally trades 60 million.
The fix is relative volume: today’s volume divided by an average of recent volume for the same instrument, commonly a 20-day simple average. Suppose those 20 sessions sum to 48.0 million shares. The baseline is 48.0M ÷ 20 = 2.4 million shares per day. Now the two breakout attempts are comparable.
| Session | Volume | Relative volume | Read |
|---|---|---|---|
| Failed breakout | 1,900,000 | 1.9 ÷ 2.4 = 0.79x | Below normal participation |
| Successful breakout | 7,900,000 | 7.9 ÷ 2.4 = 3.29x | Roughly triple normal |
A breakout on 79% of normal participation was produced by fewer people than a boring Tuesday. A breakout on 3.3x was produced by a genuine influx. Neither guarantees anything, but they are not the same evidence.
Intraday, one further adjustment: volume is not spread evenly across the day. The first and last 30 minutes routinely carry several times the volume of midday, so comparing 10:30 a.m. cumulative volume against a full-day average is meaningless. Compare it against typical cumulative volume by 10:30 a.m. If a stock has traded 1.8 million shares by then and normally trades 600,000, that is 1.8 ÷ 0.6 = 3.0x — an unusually busy morning.
Volume confirming a trend versus diverging from it
The classical framework is simple. In a healthy uptrend, volume expands on the advances and contracts on the pullbacks. Buyers are the motivated side; sellers are just taking breaks.
Concretely: across a six-week advance, the up days average 4.1 million shares and the pullback days average 1.6 million. The advances draw 4.1 ÷ 1.6 = 2.56 times the participation of the retracements. Money arrives on strength and merely drifts on weakness.
Divergence is the opposite. Price grinds to higher highs while each new high draws less volume than the last — 5.2M, then 3.8M, then 2.3M. The trend is intact on price, but fewer participants will buy at each successive level. That is a warning, not a signal: divergences persist for months, and plenty resolve by the trend simply continuing. Treat it as a reason to tighten risk, never as a reason to short.
Breakouts, reversals, and levels: where volume changes the read
A quiet breakout is a suspect breakout
To clear a level that three weeks of sellers defended, someone has to absorb all of that supply, and absorbing real supply generates real volume. When price pokes above a well-tested level on 0.79x volume, the arithmetic says the supply was not absorbed — the sellers simply stepped away briefly. That is the failed breakout in the opening example.
A common working rule is to require 1.5x to 2x average volume on a breakout bar before treating the break as meaningful. The exact threshold is not sacred; the discipline of having one is the useful part. What those levels are and why they break is covered in support and resistance levels.
Capitulation spikes at the end of a decline
Sometimes an enormous volume bar means the opposite of continuation. A stock slides from $62 to $41 over six weeks, and the final session drops 9% on 22.6 million shares — 22.6 ÷ 2.4 = 9.4x the baseline. That is not a normal down day. It is a large fraction of everyone who wanted out getting out at once.
The logic is exhaustion: once the forced sellers have sold, the supply driving price down is gone, and it takes little buying to lift the price. That is capitulation. The catch is that it is far easier to identify weeks later than on the day — a 9.4x spike can equally be the start of an institutional exit. Nobody rings a bell.
Volume at support and resistance
A level tested on heavy volume that holds is more informative than one tested on light volume. Heavy volume at support means buyers defended it in size; light volume means it was never seriously challenged.
The corollary helps with level-drawing: prices where a lot of volume changed hands historically tend to matter later, because that is where the most participants have a cost basis to defend or an underwater position to escape. It is closely related to why VWAP works as a reference level — it is literally the price the average share transacted at.
How to read the volume histogram beneath a price chart
The standard presentation is a histogram in a thin panel below the price chart, one vertical bar per candle, usually colored to match whether the candle closed up or down. Bar height is scaled relative to the largest bar on screen, not to any absolute number.
That has two consequences. You are always reading relative heights, which is what you want — one bar towering over its neighbors is the signal, not any pixel height. But the scale shifts as you scroll or change timeframe, so a bar that looked huge disappears into the noise once a genuine 9x spike enters the window.
Read it in two passes. Scan for outliers that visibly dwarf the surrounding twenty bars, then check what price did on exactly those bars: a wide-range candle, a reversal, a thrust into a level. Volume without the corresponding candlestick structure is only half the information.
Where volume data misleads you
It is fragmented across venues. A US equity trades on more than a dozen exchanges plus off-exchange venues, and a meaningful share never prints on a lit exchange at all. Your platform’s number depends on which feeds it consolidates, so two charting tools can show different volume for the same stock on the same day.
Crypto volume is frequently fictitious. Wash trading — the same entity buying and selling to itself to manufacture apparent activity — has been documented at large scale on multiple exchanges, and no consolidated tape or regulator enforces a single standard. Decentralized venues make it cheaper still. Treating a crypto volume figure with the trust you would give an NYSE print is a mistake.
Futures and forex count differently. Futures volume is contracts, and rolls between expiries distort it. Spot forex has no central tape, so “volume” there is usually tick count from one broker’s feed — a proxy for activity, not a transacted quantity.
And the fundamental limit: volume is directionless. A 9.4x bar tells you a crowd showed up, not whether they were accumulating or dumping. Every interpretation above is an inference layered on top of price, and only ever as good as the price read underneath it.
Why Stockade’s volume histogram cannot teach volume analysis
This needs saying plainly, because volume is the one topic in this series that Stockade genuinely cannot help you practice.
Stockade draws each candle’s volume from a uniform random number that knows nothing about the candle. Worse, it uses two different draws. The seeded history loaded when you pick a symbol uses Math.floor(Math.random() * 600_000 + 80_000) — roughly 80,000 to 680,000, expected value about 380,000. Candles that print live while you watch use Math.floor(Math.random() * 500_000) instead — 0 to 500,000, expected value 250,000. So the histogram silently changes distribution at the seam between the history you loaded and the bars generated in front of you, and neither figure derives from the candle’s range, its direction, or its position in a trend. Under both, the expected volume on the strongest breakout candle equals the expected volume on a flat doji. Two arbitrary distributions, neither of them about price.
Every relationship described above is therefore absent. A breakout on a tall volume bar and a breakout on a short one carry exactly the same information here, which is none. No confirmation, no divergence, no capitulation spike, no accumulation footprint — those patterns cannot exist in random numbers. It also means the VWAP overlay is weighted by random numbers, so it behaves close to an unweighted average of typical price rather than a true volume-weighted one.
What you can do here is learn the plumbing: where the histogram sits, how it rescales as you scroll, how to eyeball relative bar heights against their neighbors quickly. That muscle memory transfers. The interpretation does not.
Practice the mechanics here, then take volume to real charts
Stockade is useful for the parts of this that are execution rather than signal: reading candle structure, marking levels, placing stops, reviewing the trade journal. Build the habit of glancing at the volume panel before every entry — the routine is worth having even where the data behind it is meaningless.
For the actual skill, pull up real charts on any free charting site, compute the 20-day average yourself, and go back through breakouts that worked and breakouts that failed to compare relative volume on each. Then bring the trade management back to Stockade’s stock market simulator, where $100,000 in virtual trading capital lets you rehearse everything except the one thing random numbers cannot show you.