Stale Quote Sniping

A market maker posts a resting limit order — say, an offer to sell at $100.05 — based on its current read of fair value. A moment later, news breaks or a correlated instrument moves, and the true fair value of the asset jumps to $100.10. For a short window — however many microseconds it takes the market maker's own system to notice the move and cancel or reprice the stale offer — that $100.05 offer is now a mistake sitting in the book, priced below what the asset is actually worth. Anyone fast enough to see the triggering event, recognize the mispricing, and hit that offer before it's canceled captures the $0.05 gap essentially risk-free. This is stale quote sniping: trading against a resting order that hasn't yet caught up to new information.

Why this is structural, not just bad luck

Sniping isn't a rare exploit — it's a mechanical consequence of the fact that canceling a quote takes nonzero time, and that time is never zero for anyone. Every market maker's quotes are stale for some window after a relevant price move, and the only question is whether a faster participant can act inside it. This is the essence of the "latency arms race": market makers spend heavily shrinking their own stale window, while faster participants hunt the residual window that's left — sniping is an easier trading problem than forecasting the next move, since the triggering information is already public and the only edge is speed.

Because being sniped is a real and unavoidable cost, market makers price it in: quoted spreads are wider than they would be in a world without latency risk, specifically to compensate for the expected losses from being picked off on stale quotes during fast moves. Widening the spread doesn't eliminate sniping, but it makes the market maker's average economics work despite it.

Worked example: pricing the snipe risk into a spread

A market maker's quotes go stale, on average, for 80 microseconds after a triggering price move, and such moves happen roughly 200 times a day for this symbol, with an average size of $0.03 when they occur. If, historically, snipers succeed in hitting a stale quote about 40% of the time within that 80-microsecond window (the rest of the time the market maker cancels first), expected daily cost from sniping on one side of the book is:

$$200 \times 0.40 \times 0.03 = 2.40 ,$$

i.e. $2.40 per day, per unit of typical quoted size.

If the market maker trades that size roughly 2,000 times a day on that symbol from normal two-sided market making, this $2.40 in snipe losses needs to be recovered from the bid-ask spread earned on those 2,000 round trips — roughly $0.0012 per trade added to the spread just to break even on adverse selection, on top of whatever margin the market maker wants to actually earn.

What this means in practice

Stale quote sniping is the concrete mechanism behind "adverse selection" in market making — a specific race that happens every time a relevant price moves, not an abstract statistical concept. It explains why market makers invest heavily in fast cancel infrastructure, why spreads widen around news and correlated-market moves, and why some exchanges use speed bumps or batch auctions specifically to shrink the sniping window.