Adverse Selection in Trading: Why Your Fills Go Bad

Sold. A little too quickly.

You list a used camera for $300. Seconds later, someone offers to collect it immediately. No negotiation. No questions. Then three more messages arrive.

At first, it feels like a successful sale. Then you wonder: did I price it too cheaply?

An immediate buyer does not prove you made a mistake. But their eagerness is information. They chose your offer from the alternatives—and they may know something about its value that you do not.

The market maker’s winner’s curse

A market maker offers to buy and sell, hoping to earn the difference between those prices. The catch is that everyone else gets to decide when to accept the offer.

Imagine you quote $99.98 to buy and $100.02 to sell. New information moves the market’s value toward $100.10. Someone buys at your old $100.02 offer before you update it. You got the trade, but sold eight cents below the new value. Being filled was the bad news.

This is the market-making version of the winner’s curse: winning the trade can mean you offered terms that better-informed or faster traders were happy to accept. Adverse selection describes that tendency for your fills to arrive when trading is less favorable for you.

The reverse can happen to a buy quote before a fall. When your quote would have been profitable, nobody may trade with it. That is why a high fill rate, or apparently earning the spread, says little on its own. What happens after the fill matters.

Research: information and adverse selection — Federal Reserve

Taking liquidity has a selection problem, too

You might think the answer is to stop waiting and take an existing offer. But the trades you manage to take are selected, too.

Your signal spots an attractive ask. You send a buy order. Other traders see the same opportunity and consume the available quantity first—or the seller withdraws it. Your order arrives to find little or nothing left at that price.

The strongest opportunities can attract the most competition. Meanwhile, an offer that remains easy to hit may still be there because its advantage has disappeared. The fills you miss can therefore be better than the ones you get.

That is a related execution-selection problem, rather than the identical mechanism faced by a resting market maker. It does not make every missed trade a winner. It means missing trades are not necessarily random.

Research: races to trade and cancel — FCA

The price tag on an empty shelf

Think back to the toilet-paper shortages during COVID. Now picture a supermarket that still has its usual price on the shelf, but only a few packs left. More people want them than the shelf can supply.

Shoppers who anticipate the rush and come early, or happen to arrive just after restocking, get a pack. Others see the same price tag and an empty shelf. In this example, the posted price has stayed the same; the chance of buying at it has changed.

The analogy is about access to limited stock, not a claim that every retailer held prices unchanged. Knowing a good offer exists is different from being able to take it. An order book can lose its available quantity much faster than a supermarket shelf.

Context: product availability and prices during COVID — ECB

Make execution part of the strategy

This changes the questions worth asking. Which conditions tend to precede unfavorable fills? How long should a quote remain active? If a signal weakens while an order is resting, should the strategy withdraw it?

In Quant Fun’s Strategy Studio, you can ask the AI to assemble supported components for those decisions. Combine resting orders with order-age limits and cancellation conditions that react to adverse moves in a supported reference price. Inspect the logic on the canvas and change the assumptions you want to test.

The mechanics still matter. A cancellation can lose the race to a fill. An immediate-or-cancel order can receive only part of the requested quantity, or none. Quant Fun’s replay engine models execution delays, available depth, maker queue assumptions, partial fills, and fees so these details can be part of the experiment. A simulation remains a model, not proof of live execution.

When reviewing a strategy, ask what happened after both its fills and its missed opportunities. A promising signal needs to survive the selection of which trades actually happen. Building those questions into your strategy is where the research gets interesting.