Traders who optimise their fee tier to the last decimal often ignore a cost several times larger, because nobody sends them an invoice for it. Slippage is the gap between the price you decided on and the price you got. It has causes you can name, effects you can measure, and a component you can genuinely reduce.
Slippage is the difference between a reference price and the price you actually paid. Everything interesting follows from which reference you choose. Against the price at the moment you clicked, it measures how much the market moved while your order travelled. Against the mid price when you decided, it measures the whole cost of turning an intention into a position, which is the number that matters to a strategy.
The word gets used as though it described an event, something that happens to an order like a delay happens to a train. It does not. It is arithmetic performed after the fact, and the same fill can show large slippage or none at all depending on the reference you subtract it from. Traders who complain about slippage without saying against what are describing a feeling rather than a cost.
Slippage is a subtraction. Name the reference price before you name the number, or the number means nothing.
Lumping them together is what makes slippage feel arbitrary. Separated, each one has a different cause, a different remedy, and a different relationship to your size.
Any order that crosses pays half the spread relative to the mid price, by construction. This is not a surprise and not a market moving against you, it is the standing price of immediacy, published continuously and visible before you send anything. On a liquid pair in a normal hour it is the smallest of the three. On anything thinner it stops being small very quickly.
If your size is larger than what rests at the best price, the remainder fills at the next level, and the one after that. You are not a victim of this, you are the cause of it: the book returns a worse price because you consumed the good one. Impact grows faster than your size, which is why doubling an order more than doubles this component.
Between the moment you decide and the moment your order reaches the matching engine, the price can move for reasons that have nothing to do with you. This is the component people mean when they call slippage bad luck, and it is the only one where the description is accurate. It is also, for most retail sizes on liquid instruments, the smallest of the three.
If movement were random, slippage would help you as often as it hurts you and would average to zero over enough trades. It does not, and the reason is worth understanding because it is structural rather than a matter of luck.
Your limit order gets filled when someone is willing to trade against it. The moment they are most willing is the moment the price is about to move through your level. So the fills you get are disproportionately the ones you would have preferred not to get, and the ones you miss are disproportionately the ones you wanted. This is adverse selection, it applies to every resting order in every market, and it is the reason passive execution has a hidden cost that no fee schedule shows.
The same asymmetry applies to stops, more brutally. A stop is triggered by the market reaching a price, which means it fires exactly when momentum is against you and the book is thinnest. The distribution of stop fills is not centred on the trigger price, it leans to the wrong side of it, permanently.
Passive orders fill more often when you would rather they had not. That is not a defect of the venue, it is what being on the other side of an informed trade looks like.
For a trader sending small orders into a deep book, the fee is usually the dominant execution cost and impact is negligible. There is a size at which those two swap places, and the whole exercise of managing execution cost is about knowing where that crossing point sits for the instruments you trade.
It moves constantly. Depth at three in the afternoon on a major pair is a different market from depth at four in the morning on the same pair, and the size that would have executed invisibly in the first case walks several levels in the second. Traders who size positions in units of currency and never in units of visible depth are exposed to this without measuring it.
The practical version is simple and takes ten seconds: before sending anything above your usual size, look at how much is resting within a tolerable distance of the touch. If your order is a meaningful fraction of that, it will move the price, and the only question is whether you split it or accept the cost.
Slippage is not distributed evenly through the day. It concentrates in identifiable windows, and knowing them is worth more than any clever order type.
Market makers widen or withdraw before events whose outcome they cannot price. Depth thins in the minutes before, not after, so an order sent just ahead of a known release meets a book that has already emptied. The move afterwards is what everyone watches; the disappearance beforehand is what actually costs them.
Overnight for the dominant region of an instrument, the same nominal size represents a much larger share of the available depth. Nothing is wrong with the market, there is simply less of it, and orders sized for the busy session behave very differently.
On leveraged instruments, forced closes arrive as market orders that must execute regardless of price. They consume depth in one direction while discouraging anyone from replacing it, which is the specific condition under which slippage stops resembling its usual distribution entirely.
Most advice on this subject is about order types, which addresses only one of the three sources. The measures below are ordered by how much they typically save.
This is unglamorous and it dominates everything else. The same order at the same size costs a fraction as much during active hours as during quiet ones, and no order type recovers that difference. If a strategy can choose its hour, choosing it well is the single largest saving available.
Slicing an order into equal parts at fixed intervals is better than sending it whole, and worse than sizing each slice to what is actually resting. The book tells you how much it can absorb; using that number instead of a schedule is what separates execution from ritual.
Resting rather than crossing removes the spread component and earns the maker side of the fee schedule. It also introduces adverse selection and missed trades. The net is positive for entries you can abandon and negative for anything that has to happen, which is a rule you can apply per order rather than per strategy.
The honest measurement is the one institutions call implementation shortfall, and it needs only two numbers per trade: the mid price at the instant you decided, and your average fill. The difference, plus the fee, is your total cost. Not the price when the order arrived, which conveniently hides everything that happened while you hesitated.
Recording it for a month produces two findings almost everybody is surprised by. The first is that the total is larger than the fees, often by a multiple. The second is that it varies far more by hour and by instrument than by order type, which reorders the entire list of things worth optimising.
It also settles arguments that are otherwise unsettleable. Whether post-only entries are worth their missed fills, whether splitting orders pays for the extra fees, whether a strategy that looks profitable on closing prices survives contact with a real book: all three are empirical questions, and none can be answered without this measurement.
The fee is published. The spread is visible. Impact and movement are yours to measure, and together they are usually the larger half of the bill.
A backtest fills at a price that existed. A live order fills at a price it helped create. The gap between those two statements is where a large share of strategies that looked viable stop being viable, and it widens with frequency: a system trading once a week can absorb a cost that destroys the same system trading forty times a day.
This gives a test worth applying before committing capital. Take the strategy's measured edge per trade, subtract the fee, then subtract a realistic slippage figure drawn from your own records rather than from an assumption. If what remains is thin, the strategy is not a strategy, it is a way of transferring money to whoever is on the other side. That conclusion is unpleasant and it is much cheaper to reach on paper.
No, the spread is one of its three components. Slippage against the mid price includes half the spread by construction, plus the impact your own size has on the book, plus whatever the price did between your decision and your fill. Traders who equate the two are measuring the smallest and most visible part and calling it the total.
Not on price, by definition, since it will not execute worse than its limit. It slips in the other currency: it fills late, partially, or not at all. Measured against the price you decided at, an unfilled limit order can be the most expensive outcome of the day, and it will never appear as slippage in any report.
Because the mechanism is asymmetric. Passive orders fill preferentially when the price is about to move through them, and stops trigger precisely when the book is thin and momentum is against you. A distribution that leans one way is not evidence of a rigged venue, it is what adverse selection does to anybody trading against better-informed flow.
It reduces one of the three components and only the smallest one for most traders. Latency matters enormously to strategies competing for queue position in microseconds; it matters very little to somebody sending a mid-size order into a deep book. Spending money on speed before measuring which component dominates your own fills is a common and expensive misdiagnosis.
Split them against the depth that is actually there rather than as a habit. Splitting reduces impact and increases the number of fees you pay, and it lengthens your exposure to price movement while the remaining slices wait. It pays clearly when your size is large relative to the top of the book and pays nothing when it is not.
There is no general answer, and any number offered without an instrument, a size and an hour attached is meaningless. The useful benchmark is your own history: measure your fills for a month, and you have a baseline against which a bad day is identifiable and a change of venue or of hour is testable.
Far less, and this is one of the clearest arguments for lower frequency. A cost paid once on entry and once on exit is a rounding error on a position held for months, and the same cost paid forty times a day is the difference between a viable system and a losing one. Frequency multiplies execution cost while it does nothing to the size of the underlying move.
No, and any product that claims to remove it has moved it somewhere else, usually into a wider quoted price. Somebody has to bear the cost of immediacy and of moving a book. What you can do is measure which of the three components you are actually paying, and stop paying for the ones that are avoidable at your size.