Hidden orders and displayed depth

A level showed 1.2 units, you sent an order for 5.0 expecting to walk through it, and the whole thing filled at that one price. Or the reverse: a level looked thick and the fill was worse than the depth implied.

Displayed depth is not a measurement of resting quantity. It is a report of the resting quantity the venue has been asked to display, which is a subset — sometimes a small one. The difference has a name and several implementations, and knowing the shape of them is the only defence against reading the book as though it were complete.

Why a venue supports concealment at all

Posting a large order is informative. The quantity, the price and the side are all published the moment it rests, and other participants can act on that publication before the order has filled.

The consequence is mechanical: someone who needs to trade a size well above what sits at the touch faces a choice between revealing the size and paying the walk up the book. If neither option is acceptable, they may not post at all — and quantity that is never posted is worse for the venue than quantity posted invisibly.

So venues offer order types that rest and match normally while displaying less than their full size. The quantity is real; only the display is reduced.

Iceberg orders

An iceberg order has two quantities: a total and a displayed portion, sometimes called the peak or the tip. The book shows the displayed portion. The remainder rests, matchable, unshown.

The refill behaviour is the part that generates the surprise. When the displayed portion is consumed, another slice of the hidden remainder is disclosed and the level appears to be replenished. Repeated, this looks exactly like new participants arriving to defend a price, and it is one order being eaten.

Two implementation details vary and both matter. Whether the refill takes a new timestamp — on many venues each disclosed slice joins the back of the queue, so an iceberg does not hold priority for its whole size. And whether the hidden portion matches before displayed orders at the same price, which is usually no: displayed quantity generally has priority over hidden quantity at the same price, because display is the thing the venue is rewarding.

Fully hidden orders

Some venues support orders that display nothing at all, and some support reserve variants where the displayed portion can be zero.

Mechanically these behave as resting orders that the market data feed does not mention. They match when crossed, they consume incoming aggressive quantity, and they print to the tape when they trade — which is the only moment they become visible, and by then the trade has happened.

The priority treatment is the same principle taken further: hidden quantity generally yields to displayed quantity at the same price, so a fully hidden order sits behind every visible order at its level regardless of when it arrived. It has bought invisibility with priority.

What this does to the book you are reading

Three separate distortions, all in the same direction.

Displayed depth understates resting depth. By an unknown amount, at unknown levels, on the side that happens to be concealing. There is no calculation that recovers the true figure from a snapshot.

Apparent resilience can be one order. A level that refills three times may have been replenished by three participants or by one iceberg disclosing three slices. The feed does not distinguish them, and the distinction is exactly what resiliency is supposed to measure.

Fills can be better than the visible book allows. An aggressive order can complete at a price whose displayed quantity was smaller than the fill, which reads as an error and is hidden quantity matching normally.

The mechanism

THE MECHANISM — display versus reality

  · Iceberg order rests
                    → displayed portion shown, total
                      matchable. The book understates
                      the level.

  · Displayed portion consumed
                    → next slice disclosed. Looks like
                      replenishment by new arrivals.

  · Each disclosed slice
                    → often takes a NEW timestamp and
                      rejoins the back of the queue.

  · Fully hidden order
                    → matches when crossed, appears in
                      the tape only after trading.

  · Hidden quantity at a displayed price
                    → generally YIELDS PRIORITY to
                      displayed orders at that price.
                      Invisibility costs queue position.

  · Displayed depth as a capacity estimate
                    → a LOWER BOUND ONLY. NO GUARANTEE
                      the shortfall is small, and no way
                      to size it.

  · Which types exist, minimum peak sizes,
    refill timing and priority treatment
                    → VENUE-SPECIFIC. Some venues have
                      none of this; some have several
                      variants.

Worked example

Illustrative figures, synthetic throughout. Suppose the ask side displays 40,004 for 1.2 units and 40,010 for 2.5, and behind the 1.2 there is an iceberg with a total of 8.0 and a displayed peak of 1.2.

You send a market buy for 5.0 units. Reading the display, you expect 1.2 at 40,004 and 3.8 at 40,010, an average of about 40,008.6.

What happens instead: the 1.2 fills, the iceberg discloses another 1.2, that fills, and so on until your 5.0 is complete — entirely at 40,004. Your average is 40,004, better than you calculated, and the book afterwards still shows 40,004 with 1.2 displayed as though nothing happened. The level has 3.0 of hidden remainder left.

Now the reverse case. Suppose instead you had rested a limit buy at 40,004 to join that price. You are behind the displayed 1.2 and, on a venue where display takes priority, ahead of the hidden 6.8 — so your position is better than the total quantity at the level suggests. A seller crossing 2.0 units fills the displayed 1.2 and then reaches you, not the iceberg.

Flip the priority convention and the same order at the same price fills at a completely different point. The book display is identical in both cases.

Finally: consider what the tape shows for the first scenario. A sequence of prints at 40,004 totalling 5.0 units, against a level that never displayed more than 1.2. Anyone reconstructing depth from the tape sees quantity that was never visible in the book.

Interest that was never an order at all

Hidden order types are the formal version of a larger and unfixable problem: the book contains only orders that have been placed.

A participant willing to sell at a price but not willing to publish that willingness is indistinguishable from a participant who does not exist — until the price arrives and the order appears. This is not a venue feature and no venue can remove it. It means the book is a snapshot of disclosed current intentions, and disclosure is voluntary.

The practical form: depth can materialise at a price the instant the price is reached, and depth can evaporate as the price approaches, and neither event requires anything unusual.

The failure mode

The trap is symmetric and the two halves fail differently.

Treating displayed depth as complete makes you overestimate cost when hidden quantity is present, which is the harmless direction — the fill is better than expected. Treating displayed depth as reliable is the harmful direction, because the same reasoning that lets you predict a good fill lets you predict a bad one, and the shortfall in the other direction is not bounded. A level can be displayed and cancelled before you arrive; a level can be thick and yield you nothing because you were at the wrong end of the queue.

The only statement the display supports is a lower bound on quantity that existed at the moment the snapshot was published. It is not a capacity, not a commitment, and not a count of participants — and any estimate of execution cost built on it inherits every one of those gaps.