How many imbalances make a stack?

Three consecutive is the usual convention, and it is a convention rather than a finding. Four or five give fewer zones and wider bands. What matters is consistency: the count, the ratio and the volume floor must be fixed before you look, or you are choosing the pattern after seeing the outcome.

Does a gap in the run break the stack?

Under a strict definition, yes — adjacency is the whole claim. Some platforms allow one unflagged price inside the run. Both are defensible and they produce different zones on the same data, which is why the setting belongs in writing next to the ratio and the floor.

Is a sell stack the mirror image of a buy stack?

Mechanically yes: consecutive bid-side imbalances instead of ask-side. Interpretively, the asymmetry of the instrument outweighs the symmetry of the definition. Index futures rise and fall with different volume signatures, so a sell stack and a buy stack of equal size are not interchangeable readings.

Glossary

Stacked imbalances

Also called: imbalance stack · consecutive imbalances · imbalance cluster

Stacked imbalances are three or more consecutive footprint imbalances on the same side, printed at adjacent prices. They mark a band of the range where one side kept paying to get filled without pausing. Traders use the band as a reference level on a later retest, not as an entry by itself.

What stacking adds

A single footprint imbalance is one price, one bar, one moment of uneven aggression — it happens dozens of times a session. A run of three or more at consecutive prices is a different object: aggression stayed one-sided while price travelled through a band, rather than at a single tick.

That persistence across prices is the whole informational content. It is harder to produce by accident than a lone flag, and it localises where the initiative came from. It is not evidence of size, intent, or institutional participation, and nothing in footprint data can supply those.

The stack is a band, not a line

Record the lowest and highest price of the run. That band is the level. Reducing it to a single line throws away the only measurement you made, and on a four-tick stack the distance between the edges is a real fraction of a stop.

The count threshold is arbitrary in exactly the way the ratio is. Three is the convention; four or five give fewer and wider stacks. Decide, write it down, keep it stable — a stack found at three and a stack found at five are not the same object and cannot be compared.

Where the claim stops being testable

Hit rates for stacked-imbalance zones circulate in order flow teaching, and none of them has been replicated publicly. None specifies the dataset, what counts as "hold", which ratio and volume floor defined the imbalances, how many stacked levels were required, or the trend the stack printed in. Six unspecified variables produce whatever number you want.

That is not a reason to discard the concept. It is a reason to measure it yourself, on your settings and your instrument, with a written definition of hold and invalidation — the only version of the number that can apply to your trading.

A worked example

In a synthetic ES bar, four consecutive buy imbalances print: 388 against 71 at 5 312.25, 341 against 84 at 5 312.50, 502 against 118 at 5 312.75, and 274 against 88 at 5 313.00 — ratios of 5.5, 4.1, 4.3 and 3.1 to one, each above a 150-contract floor.

The band is 5 312.25 to 5 313.00: three ticks, 37.50 $ per contract from edge to edge. That range is the level. Price returns to 5 312.50 two bars later and trades straight through. The stack recorded where buyers paid up; it said nothing about whether they would defend it.

The trap

A buy stack marks the prices at which aggressive buyers paid the offer. If the move fails, that band is precisely where the late longs sit — the same coordinates read as support by one trader and as a pocket of trapped traders by another. The data does not choose between those readings. Only the subsequent reaction does.

The structural version: in a sustained trend, stacks print continuously. Mark them all and your chart carries a band every few points, so "price held a stack" becomes unfalsifiable — in a rising market it will hold something on most pullbacks. A level that is everywhere is not a level.

Frequently asked

How many imbalances make a stack?: Three consecutive is the usual convention, and it is a convention rather than a finding. Four or five give fewer zones and wider bands. What matters is consistency: the count, the ratio and the volume floor must be fixed before you look, or you are choosing the pattern after seeing the outcome.
Does a gap in the run break the stack?: Under a strict definition, yes — adjacency is the whole claim. Some platforms allow one unflagged price inside the run. Both are defensible and they produce different zones on the same data, which is why the setting belongs in writing next to the ratio and the floor.
Is a sell stack the mirror image of a buy stack?: Mechanically yes: consecutive bid-side imbalances instead of ask-side. Interpretively, the asymmetry of the instrument outweighs the symmetry of the definition. Index futures rise and fall with different volume signatures, so a sell stack and a buy stack of equal size are not interchangeable readings.