Slippage, and why it is not a fee

Between the moment an order leaves a platform and the moment it is matched, a small amount of time passes. Prices continue to move during it, resting orders are consumed by other participants, and the book that the order arrives at is not exactly the book that was on screen when it was sent. Slippage is the arithmetic difference between the price that was expected and the price that was achieved, and it is a measurement rather than a charge.

The distinction matters because slippage is routinely described as a hidden cost, and that description gets the mechanism backwards. A fee is a known amount computed from a schedule and always runs in one direction. Slippage has no schedule, no fixed sign and no recipient. When it runs against an order the difference is not paid to anybody, and when it runs in favour of an order the difference is not paid by anybody. It is the market having moved.

What produces a different price 

Three separate causes produce fills away from the expected price, and they behave differently enough that treating them as one obscures what happened.

1. Latency. The interval between an order being sent and being matched is short but not zero, and a fast moving market can print several updates inside it. The order is matched against the book as it stands on arrival, which may be a level or two away from the one that was displayed.
2. Depth. An order larger than the quantity resting at the best price fills across several levels, and the resulting average price is worse than the top of the book by construction. This is not a market movement at all: the book was always that shape.
3. Discontinuity. Where the price moves from one level to another without printing anything in between, an order can only be filled on the far side of the gap. This is what happens across a session break and around a scheduled release, and it is the cause capable of producing the largest differences.

The first two are symmetrical in principle. A market moving during the latency interval is as likely to move in an order's favour as against it, and a book whose depth sits above the current price improves an order as readily as one whose depth sits below it worsens the same order. The third is not symmetrical in the same way, because a discontinuity has a direction and every order arriving during it is filled on the same side of it.

The favourable half has its own name 

When a fill lands on the better side of the expected price, the industry calls it price improvement rather than slippage, and the naming asymmetry is the source of a good deal of confusion. Both are the same measurement with opposite signs. A venue that reports slippage statistics without also reporting price improvement is describing one tail of a distribution and calling it the distribution.

This is why execution reporting is done as a distribution rather than as an average. An average of a symmetrical variable tends towards zero and says almost nothing, while the shape of the distribution says a great deal: how often fills land exactly at the expected price, how far the tails reach on each side, and whether the two tails are the same length. A distribution with a long tail on one side only is describing something other than latency.

The net line is the one worth holding on to. A trader who notices the nineteen unfavourable fills and not the nineteen favourable ones will conclude that slippage is a cost, and the arithmetic in front of them will not support it. Attention is asymmetrical even when the measurement is not.

When it stops being symmetrical 

Two situations break the symmetry, and both are structural rather than accidental. The first is a stop order, which by definition is triggered by the price moving in one particular direction. The order enters the market during a move it is on the wrong side of, so its fills are distributed asymmetrically for a reason that has nothing to do with how it is handled once submitted.

The second is a discontinuous market. Around a scheduled release, at the reopening after a session break and during a disorderly move, prices travel a distance without trading at the levels in between. Every order arriving during that travel is filled on the far side of it, so all of them slip in the same direction at once. The distribution in those windows is not the distribution of an ordinary session, and averaging the two together describes neither.

Separating it from spread and commission 

A statement records the price an order filled at, not the price that was on screen when it was sent, so slippage does not appear as its own line. Measuring it requires the intended price, which is why platforms record a requested price on the deal ticket alongside the executed one. The difference between those two fields, across a representative sample of orders, is the only honest measurement of it available to an account holder.

Confusing it with the spread is the more common error. The spread is present in every fill by construction, because a position opens on one side of a two way price and closes on the other. Slippage is the additional difference between where a fill was expected on that side and where it landed. Subtracting the spread first is the step that makes the remainder interpretable, and a comparison that skips it will attribute the spread to execution quality.