Analysis · Volume
Volume at Market Turns
The most durable observation in this field is also the least useful as a rule: extreme volume clusters at lows, not at highs. The reason is mechanical; selling can be forced and buying almost never is.
One asymmetry, and everything that follows from it
Almost everything this reference says about volume at turning points comes from a single structural fact: selling can be involuntary and buying almost never is.
A margin call, a fund redemption, a risk limit breached. Each forces a sale at whatever price exists, on a schedule set by someone else. Nothing symmetrical exists on the buying side. Nobody is compelled to buy an instrument by close of business, which means that when a market falls hard enough for forced selling to begin, an enormous quantity of stock changes hands in a very short period, and no comparable mechanism operates when a market rises.
Every observation below is a consequence of that, and each is weaker than its popular version.
One detail in the same computation is worth noting, because it is the part most often turned into a rule that does not work: the heaviest single session of that synthetic market falls 8 sessions after the low rather than on it. The model has nothing to say about which side of a turn the peak falls on (its volume responds to the size of a move and not to the direction of time), so read that only as the general point: extreme volume lands in the neighbourhood of a turn, not at its precise point. In real declines the documented tendency is for the heaviest sessions to come in the final phase, with the closing low itself often arriving later on much lighter trade, and either ordering defeats a rule that needs the exact session.
The three claims, ranked by how well they hold
Forced selling is visible and indiscriminate
The strongest of the three, and the one with a mechanism behind it. When holders sell because they must, they sell without regard to merit, and every measure records it at once: advance/decline ratios collapse, new lows expand across the list, TRIN spikes, and volume goes to several times its baseline. This is a description of a state, it is checkable on any long series, and it says nothing about when the state ends.
The heaviest volume of a decline arrives near its end
A consequence of the first claim rather than an independent finding: forced selling happens after a decline is well advanced, because that is what triggers the margin calls. It holds well enough to be worth knowing and it is not a timing tool, because "near the end" is only identifiable once the end has happened.
Tops are quiet
The weakest, and still worth stating. A large seller distributing stock has every reason not to be noticed, so the volume signature of a top is an absence, and an absence is compatible with almost anything. What can be measured at tops is narrowing participation rather than volume, which is why the breadth measures carry more weight there than any volume reading does.
How to run this study on real data
| Step | What to do | The trap it avoids |
|---|---|---|
| Define the turns first | A stated rule, a decline of at least some size reversed by a rise of at least some size, applied mechanically. | Choosing turns by eye, which selects the episodes that look like the conclusion. |
| Use a volume ratio | Each session against a trailing average that excludes it, not raw share counts. | Listing and float growth, which makes raw volume rise across decades for no market reason. |
| Fix the window before looking | A symmetric window of a stated length around each turn, the same for lows and highs. | Widening the window until the asymmetry appears, which it eventually will. |
| Report both sides | The highs as well as the lows, and the turns where the pattern failed. | A gallery of confirming episodes, which is what most published volume studies are. |
| Exclude the known distortions | Note index rebalances and expiry sessions inside the window rather than silently dropping them. | Attributing a scheduled volume event to capitulation. |
What the site publishes instead of a dated study
This reference carries one volume study computed from real published data, the short-volume measurement, built from FINRA’s daily consolidated files, which are free and republishable. It exists partly to demonstrate the standard this page describes: it states its window, names its public source, sets out the aggregation step by step, and reports a result that contradicts the popular reading of the number.
Everything else here is either arithmetic or a clearly labelled synthetic illustration, because the alternative (a dated study of licensed index history, printed on a static page) decays into something misleading the moment the market moves. A study that cannot be rerun is an anecdote with a chart attached, and the published literature on volume already has more of those than it needs.
Frequently asked questions
Is volume higher at lows or at highs?
At lows, consistently, and by a wide margin. It is the most durable observation in this whole field and it has a mundane explanation: selling under pressure is synchronised. Margin calls, redemptions and risk limits force it regardless of price, so it concentrates enormous volume into a few sessions. Buying is almost never forced in the same way, so accumulation spreads out and never produces the same spike.
Why does that make tops hard to identify?
Because a top is not a volume event. A large holder reducing a position wants to do it without being noticed, which means spreading the selling over weeks and into whatever demand appears, the opposite of a spike. So the volume signature of distribution is an absence rather than a presence, and what marks a top, when anything does, is participation narrowing: fewer issues advancing, fewer new highs, while the index still rises.
Does the heaviest volume mark the exact low?
Near it, not at it, and the ordering matters. Capitulation volume tends to arrive during the final phase of a decline rather than on the precise closing low, which frequently comes days or weeks later on much lighter trade. That is why "heaviest volume equals the bottom" fails as a rule while "the heaviest volume of a decline usually arrives near its end" survives. The second is a description, the first is a timing claim.
How would I check this on my own data?
Take a long daily series, mark the major lows and highs by whatever definition you prefer — stating it first — and compute average volume in a window around each, as a ratio to the volume average of the surrounding period. Use a ratio rather than raw share counts, because listing and float changes make raw volume incomparable across decades. The asymmetry shows up on almost any long series and does not need statistical machinery to see.
Does the pattern hold for individual stocks?
Broadly, with more noise and one additional cause. A single company’s volume is dominated by its own events (results, guidance, index inclusion), which can produce spikes unrelated to any market turn. The forced-selling mechanism still operates, and a market-wide decline produces heavy volume in almost everything, but on a single name a volume spike is more likely to be company news than capitulation.
What about volume during the advance?
It usually declines as an advance matures, and this is where the honest version of the claim is much weaker than the popular one. Volume contracting during a long rise is often described as distribution or as a warning, and it has also accompanied advances that continued for years. Rising markets simply need less trade than falling ones: the buying is voluntary and patient. Reading a quiet advance as fragile is an interpretation, not a measurement.
Is any of this a trading rule?
No, and the distinction is the point of the page. "The heaviest volume of a decline usually arrives near its end" describes the record; "buy the heaviest-volume session" is a rule that requires knowing the decline has ended, which is the information you do not have at the time. What the observation is good for is context, recognising the character of a session as it happens, and knowing that the same reading appears in declines that continued.
What would change your mind about it?
Evidence that the asymmetry disappears once forced selling is accounted for, that is, that the extra volume at lows is not attributable to margin calls, redemptions and risk-limit selling, but is instead an artefact of how lows are selected after the fact. Since lows are identified retrospectively, selection is a genuine concern, and a properly specified test would define the turns in advance from a rule rather than by inspection.