A limit order sitting on the bid side of an order book waits to be filled. When a sell arrives and fills it, the maker has purchased at their chosen price. This feels like success — the order executed exactly where it was placed.
Albers et al. (arXiv:2502.18625), running a live trading experiment on Binance's Bitcoin perpetual, document a fundamental negative correlation: the higher the probability that a limit order gets filled, the lower the post-fill returns. Orders that fill easily tend to lose money. Orders that rarely fill tend to make money. The relationship is not incidental. It is structural.
The mechanism is the order book itself. A limit order at the front of the queue — close to the top of the book, in a large queue, at a favorable position — has high fill probability because aggressive orders hit it first. But those same aggressive orders are often the informed flow, the trades that move the market through the price. Getting filled means the market is moving through your level. If you are first in line, the thing hitting you is the future price, going the wrong way.
The viable maker strategy is therefore contrarian. When the order book is heavily bid, a profitable maker sells. When it is heavily offered, a profitable maker buys. This is the opposite of what the order book's pressure suggests. The paper models these as “Reversals” — situations where the touch price snaps back after the imbalance resolves. The maker profits not by agreeing with the visible pressure but by betting on its exhaustion.
The structural point is that execution itself is adversely informative. In a market with informed flow, the act of being filled is not neutral. It carries a signal: someone who knows more than you was willing to trade against your price. The easier it is to get filled, the more certain it is that the counterparty has information you lack. Fill probability is a direct measurement of adverse selection — the very thing the maker sought to avoid by placing a limit order rather than crossing the spread.
This inverts the usual framing of maker strategy. The question is not “will my order get filled?” but “what does getting filled tell me?” A fill at the front of a large queue, during heavy directional flow, tells you that someone with better information just hit your resting order. The fill is the bad news.