Adverse selection
Adverse selection is the risk, for whoever provides liquidity, of trading with a counterparty who knows more about the next price move. A resting limit order is filled most readily when the market is about to move against it, so passive fills carry a systematic cost that the bid-ask spread partly compensates.
Senzoukria · Glossary · Updated September 2026
The idea
A market maker quoting both sides does not know which incoming orders are informed. Uninformed orders arrive on both sides and let the maker earn the spread; informed orders arrive on the side the price is about to move toward, and the maker loses on them. In Glosten and Milgrom's 1985 model, the spread is set wide enough that the gains on uninformed flow pay for the losses on informed flow. Adverse selection is that loss, and it is the reason spreads widen when informed trading is more likely, for example just before scheduled releases.
What it means for a trader using limit orders
The same logic applies to anyone resting orders. A bid at the best price is filled when sellers are active enough to reach it. It is filled completely when they are active enough to trade through the level, which is also when price is most likely to keep falling. Orders that are not filled are often the ones that would have worked: price moved away before the queue was reached. The result is a skewed sample of fills, better on paper than in practice.
How to measure it
Record the mid-price at each fill and again a fixed time later, for example 30 seconds or 1 minute. For a buy, the markout is the later mid minus the fill price; for a sell, the fill price minus the later mid. An average markout below zero measures adverse selection on those fills. With 100 passive buy fills and an average markout of minus half a tick, the fills cost on average half a tick relative to where the market went next, which has to be set against the spread the orders captured by not crossing it.
- Measure over many fills; single outcomes say nothing.
- Compare passive and aggressive fills: aggressive fills pay the spread but are not selected against in the same way.
- Separate quiet periods from active ones; adverse selection is concentrated when information arrives.
Consequences for backtests
A backtest that fills a limit order whenever price touches its level ignores both the queue and the selection: it records the fills that would have come too late or not at all, and it drops the fills that were hit on the way through. Tick replay that assumes the order was first in the queue is optimistic for the same reason.
In Senzoukria
The backtest panel's transaction costs apply slippage against the entry of every trade; a strategy built on limit orders and tested on bars should still be read with its fill assumptions in mind, since bars do not record where an order would have stood in the queue. The journal's 'Review in Replay' button reopens a trade in the Replay workspace, where the price path after each fill can be examined, which is the manual version of a markout study.
Related
In the same section
- Implementation shortfall
- Penny jumping
- Aggregated vs order-by-order
- Activation fee
- Aggregated trades
- Acceptance and rejection
- Aggressor
- Absorption
This page in other languages
Frequently asked questions
- Is adverse selection the same as slippage?
- No. Slippage is paid when an order fills at a worse price than expected. Adverse selection is paid after a fill that happened at the expected price, when the market then moves against the position because the counterparty knew something. Limit orders avoid slippage but are exposed to adverse selection.
- Why do spreads widen before news?
- Because the chance that an incoming order is informed rises around information events. Liquidity providers protect themselves against adverse selection by quoting wider or smaller, or by stepping back, which is visible on the DOM as a thinner and wider book before scheduled releases.