The winner's curse

A jar of coins is up for auction. It is worth exactly £100, but nobody knows that: each bidder makes an honest guess, as likely to be too low as too high. The highest bid wins. And the winner, time after time, has paid too much.

The rules, all of them. Every bidder's estimate is unbiased: equally likely to fall anywhere from the spread below the true value to the spread above it. Each bids their estimate minus a fixed amount (the "shade" slider; zero means bidding exactly what you think it is worth). The highest bid wins and pays what it bid.

The average of all the estimates is right. But the winner is not an average bidder: the winner is whoever made the largest mistake upwards. With N bidders the highest estimate overshoots, on average, by exactly spread × (N − 1) ÷ (N + 1): £6.67 with two bidders and a £20 spread, £16.36 with ten, £19.22 with fifty. More competition makes it worse. The cure is to bid as if you already knew you had won: shade your bid by that amount and the winner breaks even on average. The effect was first described in bidding for offshore oil leases in 1971. What it leaves out: bidders who value the thing differently, second-price auctions, and bidders who learn from each other's bids.