Slippage, Latency Arbitrage, and Adverse Selection
Three related costs, each arising from the gap between the price you saw and the price you got.
Slippage
The difference between expected and actual execution price. It has three sources:
Walking the book. Your order is larger than the size at the touch, so it consumes worse levels. Predictable from the book itself.
Price movement. The market moves between your decision and your fill.
Queue failure. Your passive order does not fill and you end up crossing later at a worse price.
Slippage grows with order size, volatility and urgency, and shrinks with liquidity. It is why execution algorithms exist.
Latency arbitrage
When the same instrument trades on multiple venues, quotes update at slightly different times. A participant who sees venue A move can trade against a stale quote on venue B before it updates.
The profit is small per trade, certain, and taken directly from whoever was slow. For a market maker this is a pure cost, and it is the main reason firms invest in speed: not to run this strategy, but to avoid being its victim.
Adverse selection
The most important of the three, because it defines the economics of market making.
Your quotes are available to everyone. Whoever chooses to trade with you does so because your price is attractive to them, and their reasons may be better than yours. So the fills you get are not a random sample of the market: they are biased toward the trades that are good for the other side.
This is the winner's curse. Winning a trade is evidence that everyone else declined it at that price. In market making, getting filled is mildly bad news by construction.
The mechanism, concretely. You quote 99.98 at 100.02. News arrives that the instrument is worth 100.50. Everyone who saw it first lifts your offer at 100.02, and you are short at a price that is now 48 cents wrong. You did nothing incorrect except being slower than the information.
Measuring it
Track price movement after your fills. If prices systematically move against you in the seconds afterwards, you are being adversely selected:
signed by whether you bought or sold. A market maker with consistently negative markouts is being picked off, however healthy the spread capture looks in isolation. Markout analysis is one of the standard diagnostics on a market making desk.
The defences
Speed. Cancel stale quotes before someone reaches them.
Wider spreads. Charge enough that the informed flow is covered by the uninformed. This is why spreads are wide when uncertainty is high.
Smaller size. Limit the damage from any single bad fill.
Flow selection. Quote where flow is less informed. Retail-heavy venues are attractive for exactly this reason, and it is why wholesalers pay for retail order flow: it is flow that is unlikely to know something.
Fast fair value updates. Much adverse selection is really a stale price. Updating from the leading instrument rather than your own book removes a large share of it.
"Why do market makers make money if anyone can trade against them?" Because most flow is uninformed. The business is charging the uninformed enough to cover the losses to the informed, and the whole craft is telling them apart.
That balance can be made exact. Adverse selection and the spread derives the width a competitive market maker must quote from the fraction of flow that is informed and how much it knows.