Sumo wrestlers on the bubble win suspiciously often
A thousand-year-old sport turned out to be a near-perfect natural experiment in how people actually respond to incentives.
In professional sumo, a wrestler needs eight wins out of fifteen bouts in a tournament to keep his rank, and rank determines income for the rest of the year. Steven Levitt, later of Freakonomics, and economist Mark Duggan noticed something in the win records: wrestlers entering their final match with exactly seven wins, needing one more to stay safe, won that match at a rate far above their normal odds, especially against opponents who already had eight wins and nothing left to gain.
The pattern wasn't subtle. On a tournament's final day, the data suggested roughly half of these "bubble" matches were effectively decided in advance. And the giveaway wasn't just the inflated win rate, it was what happened afterward. Wrestlers who won on the bubble were unusually likely to lose the next time they faced that same opponent, as if a favor had been quietly repaid. When Japanese media scrutiny of sumo spiked, the suspicious pattern shrank.
Nobody needed a confession. The incentive structure told the story on its own, match by match, for anyone willing to actually count.