Why price moves against you right after a fill
Posted: Fri Oct 30, 2026 1:27 pm
Why price moves against you right after a fill
A common complaint is that price turns just after an entry. The feeling is that the market knows. The cause is more mundane and has a name: adverse selection.
When a liquidity provider fills an order, the provider's quote reflects its view of fair value. A buy order at the ask gets filled only if the provider is willing to sell at that price. If the provider has information, or a model suggesting the price is about to fall, it's happy to sell. The trade that gets filled is, on average, the one that works out slightly worse for the buyer.
Last look amplifies this. Orders that would have been profitable get rejected more often, and the unprofitable ones get filled. The fills that remain skew toward losers.
The effect is small, a fraction of a pip on average, but it's systematic. It explains why live results often trail backtests, which assume every signal is filled.
Measuring it requires data: compare the mid-price at the moment of fill with the mid-price one second and ten seconds later. If the price tends to move against the position after the fill, adverse selection is at work.
One second after a fill, which way does the average price lean in the logs?
A common complaint is that price turns just after an entry. The feeling is that the market knows. The cause is more mundane and has a name: adverse selection.
When a liquidity provider fills an order, the provider's quote reflects its view of fair value. A buy order at the ask gets filled only if the provider is willing to sell at that price. If the provider has information, or a model suggesting the price is about to fall, it's happy to sell. The trade that gets filled is, on average, the one that works out slightly worse for the buyer.
Last look amplifies this. Orders that would have been profitable get rejected more often, and the unprofitable ones get filled. The fills that remain skew toward losers.
The effect is small, a fraction of a pip on average, but it's systematic. It explains why live results often trail backtests, which assume every signal is filled.
Measuring it requires data: compare the mid-price at the moment of fill with the mid-price one second and ten seconds later. If the price tends to move against the position after the fill, adverse selection is at work.
One second after a fill, which way does the average price lean in the logs?