The Winner's Curse

Picture a sealed jar of coins passed around a room, and everyone writes down a bid for it. Nobody knows the true amount inside; each person eyeballs it and guesses. Suppose everyone's guess is honest — on average, people neither over- nor under-estimate. You'd think the auction is fair. It isn't. The person who wins is, by definition, the one who guessed highest, and the highest of many honest guesses is almost always too high. Win the jar by bidding your estimate and you'll typically overpay. That's the winner's curse: in a contest for something of uncertain common value, winning is itself bad news about the thing you just won.

The key phrase is common value — the item is worth the same to everyone (a jar of coins, an oil field, a company's true worth), and people differ only in their noisy estimates of it. Contrast a private value auction (a painting you personally love) where no such curse applies. In common-value settings the trap is subtle because each individual estimate can be perfectly unbiased; the bias appears only when you condition on having won.

Worked example: the coin jar

A jar holds exactly $100 of coins. Five bidders each form an honest estimate, off by a random error spread evenly between −$20 and +$20. Any single estimate averages exactly $100 — unbiased, no problem. But the auction is won by the maximum of the five estimates, and the max of several draws is systematically high.

For errors spread evenly on $[-a, a]$ with $n$ bidders, the expected largest error is $a\,\frac{n-1}{n+1}$. With $a = 20$ and $n = 5$ that's

$$20 \times \frac{5 - 1}{5 + 1} = 20 \times \frac{4}{6} \approx 13.3 .$$

So the winning estimate averages about $113.30. Bid your honest estimate and you win the jar for roughly $113 — losing about $13 on something worth $100. And it gets worse with more bidders: at ten bidders the expected top error climbs toward $16. More competition, deeper curse.

The fix: shade your bid

The remedy is to bid as if you've already won, before you have. Ask: "if my bid turns out to be the highest, how far above the truth was I likely to be?" Then subtract that expected overshoot from your bid. In the jar example you'd shave roughly $13 off your estimate, bidding around $100 rather than $113, so that winning leaves you at break-even rather than underwater. Sophisticated bidders shade more when there are more competitors and when their own estimate is noisier.

Why traders care: getting filled is winning

The winner's curse is the same phenomenon as adverse selection in market making, dressed differently. When you post a bid and ask around your fair value, a fill is not a neutral event — the counterparty chose to trade with you rather than someone else, often because they know something. Your resting offer gets lifted precisely when the true value is above your ask; your bid gets hit precisely when value is below it. Just like the auction winner, the market maker who gets filled has, on average, been picked off by whoever knew more. Every fill carries the quiet question: why was this person happy to trade against my price?

In interviews the winner's curse shows up as the coin-jar or oil-field question, and the answer they want is the reasoning, not a number: name that the winner is a biased sample of estimates, so a rational bidder shades below their honest guess by the expected overshoot given they win. Then, for bonus points, connect it to why a market maker widens and skews against informed flow — the mechanics of that defense live in Order Book Mechanics and Auction Theory for Markets.