Volume Is the Corroboration

Relative volume is the only usable form

An absolute number is meaningless without its baseline: 4.3 million shares is enormous for a small-cap and a rounding error for a mega-cap. Relative volume fixes that by dividing the period’s volume by a recent average — conventionally twenty sessions, which is about a month of trading — and the ratio is what gets read. A day at 2.4 times average says that participation was unusual for that instrument; a day at 0.6 says the session was thin. Two properties make the measure honest: it is self-referential, so it needs no cross-sectional comparison, and it is backward-looking by design, which is exactly right for a qualifier — you are asking whether the move was taken seriously relative to this market’s own recent habits, not forecasting anything. Applied to a breakout, relative volume is the corroboration that separates a decision from a drift. A close above a level on 2.4 times average volume means many participants transacted at the moment the level gave way, which is what a repricing looks like; the same close on 0.6 times means a handful of orders pushed the price, and prices moved by few participants are reliably moved back. The practical consequence is a rule that is checkable in advance: state the volume bar before the event — “confirming means at least 1.5 times the 20-day average” — so that the read of participation is not negotiated after the outcome is known. Applied to a pullback, relative volume reads the character of the trend rather than the direction of a single day. Healthy advances are usually built with heavier volume on the up legs and lighter volume on the declines, because the participants who were present are staying while the ones who wanted out are finishing. When that inverts — heavy volume on a decline, light volume on the recovery — the market is saying the pace of supply changed even if price has not yet broken the structure, which is the earliest of the pre-break warnings. And the extremes at the end of a move have their own signature: the largest volume of a trend often prints near the end of it, because a blow-off is the moment when the maximum number of participants are in the same trade at the same time. That is not a sell signal by itself. It is the moment when the ordinary pullback gets larger, which is a risk statement. Reading the tape against its own average — Breakout close, 4.32M against a 1.80M average: 2.4× average — confirmed, above a 1.5× bar ← · Pullback leg at 0.7× average: Participation leaving as price dips — healthy · Recovery leg at 1.8× average: Participation returning as price turns — healthy ← · Advance at 0.6× average: Thin: the same price achieved by fewer participants, least reliable The average you compare against should exclude the event you are measuring, or the ratio compresses itself: a 20-day average that already contains an earnings day will make the next ordinary day look quiet. Using a median rather than a mean is a common fix, because one 40-million-share day can double a mean for a month.

What volume cannot tell you

Volume counts shares traded, not who traded them and not on which side. A large print could be one institution buying from another, a fund meeting a redemption, a market maker hedging an option book, or an index rebalancing on an effective date — four completely different facts about the future, all showing up as the same number on the same chart. That is why “volume shows what the institutions are doing” is not something the data supports: it shows that somebody transacted at size, and the identity and motive are inference. The second limit is structural. A large and growing share of US equity volume prints away from the lit exchanges, in off-exchange venues and internalisers, and not all of it is reported in the same place with the same timing. The tape a learner sees is an incomplete census, and on quiet days the off-exchange share can be the majority of the total. None of this makes relative volume useless — it is consistent, it is comparable to itself, and the corner of the market it does show behaves like the whole in normal conditions — but it does mean the measure is a proxy rather than a ledger. The third limit is that the same expansion means different things at different places on the chart. A volume spike at a level, after a base, in the direction of the higher frame is participation arriving to reprice; a spike at the end of an extended trend is participation arriving to capitulate or to complete. The number is identical; the meaning comes from location, which is the lesson of TR2 applied to the tape. So the honest hierarchy for volume is: measure it relative to its own average, use it to qualify a move that location has already made interesting, and decline to draw conclusions about intent, because the chart cannot support them. • Volume counts shares, not buyers: the same print can be accumulation, distribution, hedging or rebalancing. • Off-exchange prints mean the tape is a sample rather than a census of the day’s trading. • The same spike means repricing at a level and completion at an extreme — location decides which. • A stated bar in advance (“1.5× the average”) stops the read from being negotiated after the outcome. • Blow-off volume is a risk statement about the ordinary pullback, not a reversal signal. Beware of a volume indicator that hides the raw number behind a smoothed line. OBV and its relatives accumulate volume by the sign of the close, which means a week of small down closes can outweigh one enormous up day, and the resulting line looks authoritative while being an arbitrary weighting. If you use one, know that it is a rule of thumb rather than a measurement, and check the underlying daily bars before acting.

Half the volume is off the tape

The consolidated tape shows every trade with a print, but it does not show where the trade happened or who was on the other side, and a large and growing share of volume never touches an exchange at all. Retail orders are internalised by wholesalers, institutional blocks cross in dark pools and via alternative trading systems, and only a fraction of the total reaches a lit book. On many large-cap names the off-exchange share runs near half of all volume, which means “the tape” is a partial record of a market that is mostly elsewhere. This changes how a volume reading should be used. A relative-volume spike on the lit tape still tells you that activity increased — that part survives, because the same venues are being compared to themselves. But it does not tell you *who* is trading, and a spike driven by routing changes or a new venue’s share can look like conviction when it is plumbing. The specific signal a stock chart most understates is the block trade: a two-million-share cross in a dark pool may never print in a way that shows up as a daily-bar anomaly. So treat volume as a measure of *participation intensity on the visible venues*, not of total demand. For a liquid mega-cap it is a reliable activity gauge. For a thin small-cap it can be a large fraction of the real market, which is precisely why volume signals work better there — and why a print you can see matters more than flow you cannot. A volume spike with no direction is just participation. “More shares traded” tells you a disagreement intensified, not who won it.

The volume indicators, and what they restate

Volume matters, and the indicators built on it mostly repackage it. Knowing what each one computes is the difference between using three of them as three confirmations and seeing them as the same observation drawn three ways — which is the same redundancy problem the oscillators in this subject have, applied to a different input. **On-balance volume** is the simplest: add the day’s volume if the close rose, subtract it if the close fell, and accumulate. It is a running total of direction-signed volume, and its notable limitation is that it discards magnitude — a session that gained a hundredth of a percent contributes exactly as much as one that gained ten percent, so a large part of the information in a volume series is thrown away at the first step. **Accumulation/distribution** keeps more by weighting each bar’s volume by where the close finished inside the bar’s range, then accumulating. It adds the intraday location that on-balance volume ignores, and it still reduces each session to a single signed number. The **money-flow** family — the index, the Chaikin measures — applies a similar location weight and then bounds the result into an oscillator, so the money-flow index is essentially an oscillator computed on a volume-weighted price position rather than on price alone. **Volume-price trend** takes the opposite trade: it multiplies volume by the percentage price change, so it preserves magnitude and is dominated by a single extreme session. What they share is the thing to hold onto. Each is a function of price and volume together, so when one of them diverges from price, the divergence is a statement about how the price path relates to the volume path — which can be read directly by plotting relative volume against the price. Three cumulative indicators agreeing is not three pieces of evidence; it is one relationship with three names, and treating agreement as confirmation is how a redundant toolkit produces false confidence. The genuinely non-redundant uses of volume are elsewhere and simpler. **Relative** volume — the session against its own average — says whether a move attracted participation. The **location** of the volume inside the bar says whether the business of the day was done at the highs or the lows. **Volume at price** shows where the market reached agreement, which is the level-grading tool rather than a price indicator. And the change in participation around an event says whether a move was an institutional repricing or a thin drift. None of those is price times volume restated. The practical allocation follows: if a cumulative volume line helps, keep one, and spend the effort that a second and third would have taken on relative volume and the volume profile, which answer questions the price series cannot. • On-balance volume and friends are all functions of price and volume together. • That makes their agreement a restatement rather than three confirmations. • Relative volume, volume location and volume at price are not redundant with price. • Keep at most one cumulative line, and check it against the raw relationship. A quick test of any volume indicator: plot it against price and ask what it shows that the two series do not. If the answer is a smoothed version of the same relationship, the indicator is a preference rather than a measurement.

Sources

• Relative volume and breakout studiesBulkowski, “Encyclopedia of Chart Patterns” (volume-confirmation sections)
• Off-exchange and dark-pool share of volumeMarket structure literature on US equity trading venue share
• VWAP, volume profile and level formationStandard execution and market-profile practice