Markouts: Measuring Post-Trade Drift

A market maker's bid at 99.99 gets hit — someone sold to them. Against the mid of 100.00 at the moment of the fill, that looks like a half-cent win: bought at 99.99, "worth" 100.00. But look at the mid five seconds later: it's 99.96. The price didn't stay put after the fill — it kept dropping, and the seller who hit the bid knew something. The fill that looked like a half-cent win is actually a two-and-a-half-cent loss, and the only way to see that is to check the price after the trade, not at it. That check is a markout.

The definition

For a fill at price $p$, side $\text{sign} \in \{+1, -1\}$ (buy = +1, sell = −1), and a horizon $\tau$ (say, 1 second, 10 seconds, 60 seconds):

$\text{markout}(\tau) = \text{sign} \times \big(m_{t+\tau} - p\big)$

In words: for a buy, markout is positive if the price rose after you bought (good — you bought before it went up); for a sell, markout is positive if the price fell after you sold. A market maker's fills are, by construction, the other side of whatever the taker wanted — so a taker's good markout is the maker's bad markout on the identical trade. Markouts are computed at several horizons because the informativeness of a fill often isn't visible in the first tick but shows up over the next few seconds as the informed trader's information percolates into the price.

Worked example

A market maker's bid at 99.99 is hit for 500 shares. The mid at the time of the fill was 100.00.

Horizon	Mid	Markout (sell-side math applies to the taker; maker is opposite)
At fill	100.00	maker "looks" +0.005 vs mid
+1s	99.98	maker markout: 99.99 − 99.98 = +0.010
+10s	99.96	maker markout: 99.99 − 99.96 = −0.030
+60s	99.95	maker markout: 99.99 − 99.95 = −0.040

The maker bought at 99.99. At 1 second the mid had only drifted to 99.98, so being long at 99.99 versus a 99.98 mid is still technically a loss of a penny relative to that mid, but small. By 10 seconds the mid is at 99.96 — the maker is marked down four cents from the naive "spread capture" story and three cents worse than even the 1-second check suggested. This is the signature of a genuinely informed seller: the price keeps moving the same direction well past the trade, and the damage grows with horizon rather than bouncing back.

Reading a markout curve

Aggregate markouts across thousands of fills and plot the average against horizon. A market maker earning genuine, uninformed spread sees markouts that are flat or even improve slightly with horizon — the counterparty had no information, so the price doesn't systematically drift. A market maker being adversely selected sees markouts that worsen steadily with horizon, and the shape tells you about the type of counterparty: markouts that stabilise after a second or two suggest fast, latency-sensitive informed flow; markouts that keep worsening out to a minute or more suggest slower, fundamentals-driven informed flow.

Where it drives decisions

• Quote sizing. Symbols or price levels with consistently bad markouts get quoted smaller or wider — see Skewing Quotes To Manage Inventory for the live response and P&L Attribution For A Market Maker for how markout losses get formally split out of total P&L.
• Counterparty and venue selection. On venues that let you see or infer counterparty flags, markouts by counterparty type reveal who is systematically informed.
• Model validation. A fair-value model (like The Microprice) that predicts short-term price moves should show better markouts when the maker quotes around it than around the raw mid — markouts are the natural way to validate that the model is actually working.

In interviews

If asked how a market maker knows if it's being picked off, "markouts" is the one-word answer, but explain the mechanism: check the price after the fill, not at it, because the whole point of adverse selection is that the damage isn't visible until later.