One Tiny Auction Rule Can Stop Bidding Wars From Going Haywire

Auctions are everywhere: the ads you scroll past online, government contracts, radio spectrum licenses, even buying a house. In a "first-price auction," the highest bidder wins and pays exactly what they bid. Economists want bidders to land on a stable, sensible strategy — a Nash equilibrium (a stable standoff where nobody gains by changing their bid). A new paper by researcher Benjamin Heymann asks a surprisingly simple question: what happens if two bidders bid the same amount?

The answer, it turns out, depends entirely on who you declare the winner. Heymann studied bidders who learn over time by watching each other — an approach called fictitious play (bidders guess what rivals will do based on past bids). With the usual rule — split the prize if there's a tie — the bidding never settles. Instead it circles endlessly, drifting far from any sensible outcome, like two people in a doorway each stepping the same way forever. Change one line of the rules so that tied bidders win nothing, and the bidding calms down and lands close to a fair, stable outcome.

Why should you care? Because auctions set prices on things you buy every day. Your social media ads, your flights, your electricity, and your mobile data are all priced by auction-like systems, sometimes worth billions of dollars. If the tie-breaking rule quietly nudges bidding into chaos, everyone pays a hidden tax. The finding also matters for AI: modern automated bidding agents learn exactly this way, watching rivals and adjusting. A rule that sends human bidders into circles will do the same to software, just much faster.

The honest catch: this is a deliberately tiny model — two bidders, a handful of values, a short list of allowed bids. It's a warning sign, not proof that real markets misbehave. The stable outcome Heymann finds is also only an "epsilon-equilibrium," meaning close enough to fair, not exactly perfect. Still, it makes a good point for anyone designing a marketplace: the boring rules nobody reads may matter most.