Indicator library · Volume
SBV Histogram, Volume by Trade Size
Splitting a session’s volume by the size of the tickets rather than by the direction of the close. It was once a staple measure, and the inference it rested on has been substantially undermined by how orders are executed today.
The idea
Every other volume measure in this library splits a session’s volume by direction: up or down, by the close, by the typical price, by where the close sat in the range. This one splits it by size: how much traded in small lots, and how much in blocks.
The premise was that ticket size proxies for the participant. Small tickets were individuals; ten-thousand-share blocks were institutions. If that holds, the split answers a question no price-derived measure can reach: not how much was traded or which way it went, but who was doing it, and a change in the mix during an advance is the distribution argument in its strongest possible form.
What changed underneath it
The measure’s inference depended on a market structure that no longer exists in the same form, and any honest page about it has to lead with that rather than bury it.
| Change | Effect on the inference |
|---|---|
| Order slicing | Execution algorithms split a large order into hundreds of small child orders deliberately, so institutional flow now prints as small tickets. |
| Decimalisation and smaller lots | The typical trade size fell across the whole market, so a fixed "small block" threshold means something different in each era. |
| Off-exchange execution | A substantial share of volume is not printed the way it once was, so the visible mix is a sample rather than the market. |
| Market-making flow | Roughly half of consolidated volume is mechanical two-sided liquidity provision, carrying no directional view at any size. |
The last row is measured rather than asserted: the short-volume research on this site finds short-marked executions running at roughly half of all consolidated volume across every session examined, with a tight distribution, the signature of plumbing rather than opinion. Any measure that infers participants from prints is reading that plumbing along with everything else.
What survives
A weaker claim, and one worth keeping. The trade-size distribution still varies, and a sustained shift in it is still a change in how an instrument is being traded, a different mix of execution strategies, a different set of participants, or both. What no longer follows is the translation of that shift into "retail" and "institutional".
Read that way the histogram is a description of execution rather than of intent, which is less exciting and considerably more defensible. It also keeps the measure’s one genuine advantage: the input is entirely independent of price, so a shift in the mix is not another view of the same closing prices every other indicator in this library is built from.
Reading the histogram without the old inference
Stripped of the retail-versus-institutional translation, the chart still supports two readings, and both are about change rather than level.
A drift in the small-block share over weeks says the execution mix is changing, more slicing, a different set of venues, or a genuine change in who is trading. It does not say which, and it is still a fact about the instrument that no price-derived measure would show. A sharp break in the series is more often a data artefact than a market event: check the vendor's block threshold and the venues included before interpreting it, because a change in either produces exactly that shape.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Size read as participant | Algorithms slice large orders into small prints on purpose. A small ticket is weak evidence of a small participant. |
| Thresholds quoted absolutely | The block definition, the instrument, the venue and the era all change the mix. Only comparison with its own history means anything. |
| Single sessions read | A holiday week or an expiry produces an unusual mix for reasons unrelated to anyone's intent. |
| Historical charts trusted | Published examples predate order slicing, so their inference held better then than the same chart would today. |
| Data source changed mid-series | Vendors differ on the block threshold and on which venues they include; a splice produces a step that looks like a market event. |
| Mechanical volume included | Market-making flow is roughly half the tape and carries no view at any ticket size. |
Why the page is written this way
A great deal of published commentary rests on this measure, much of it written when the inference still held. The temptation with a long-established measure is to present it at its strongest and leave the erosion out.
The reason not to is the same one that governs every "where it misleads" table here. An indicator whose limits are stated can be used carefully; one presented without them cannot be used at all, because there is no way to know when it has stopped working. In this case the answer is unusually specific, the inference weakened as execution changed, and that is more useful to a reader than another confident chart would be.
Frequently asked questions
What is small block volume?
The share of a session’s volume that traded in small lots — historically under a few hundred shares — as distinct from block trades of ten thousand shares and up. The premise is that trade size proxies for who is trading: small tickets were retail, large blocks were institutions. A histogram of the small-block share over time was for years a staple of trade-size analysis, and a good deal of the older commentary rests on it.
What was the reading?
A rising small-block share during an advance was read as the later buying coming increasingly from small participants, the institutions who started the move having stopped adding to it. That is the distribution argument this whole site is organised around, stated in terms of who rather than how much, and it is the strongest form the argument can take when it holds.
Does the inference still work?
Much less well than it did, and the page has to say so. Order handling changed: large orders are routinely split into hundreds of small child orders by execution algorithms precisely so they do not announce themselves, and a great deal of volume never appears on the public tape as a block at all. A small print today is far weaker evidence of a small participant than it was in the 1990s.
So is the measure useless now?
Not useless, and much weaker than its historical presentations imply. What survives is that the trade-size distribution still changes, and a sustained change in it is still a change in how the instrument is being traded, but the interpretation "small trades mean retail" no longer follows. Reading the shift as a fact about execution rather than about participants is the honest version.
Where does the data come from?
Trade-level data with sizes, which is a considerably heavier requirement than the daily bars most measures here need. Consolidated tape data provides it; free end-of-day feeds do not. That cost is part of why the measure never became widespread, and why the version most people encounter is a vendor’s chart rather than something they computed.
How is it different from money flow measures?
They ask different questions. Money-flow measures weight volume by where the close sat or by the direction of the typical price, inferring intent from price behaviour. This ignores price entirely and looks at the size of the tickets. In principle that is a genuinely independent input, which is what made it attractive; in practice its inference has been eroded by execution technology while the money-flow inference has not.
What is the histogram form for?
Plotting the small-block share as bars against the price above makes a slow drift visible that a table of daily figures hides. The reading is never a single session (a quiet holiday week will show an unusual mix for no reason), but a share that has moved by ten or fifteen points over a month is a change in how the instrument is being traded, whatever the cause.
Should thresholds be quoted for it?
No, and this is stronger advice than for most measures here. The mix depends on the instrument, the exchange, the era, and the vendor’s block definition, so a figure like "60 per cent small block" carries no meaning outside its own series. Everything about this measure is comparison against its own recent history, which is also why a long, consistently-sourced series matters more here than the exact threshold ever could.