A new study argued that Polymarket’s five-minute Bitcoin contract became a machine for wealth transfer. It moved money from retail bettors to a small band of manipulators, and it made Bitcoin’s spot price worse in the process.
The paper, “Settlement Manipulation in Prediction Markets” by David Dai, Ruizhe Jia, and Shihao Yu of Stanford and Singapore Management University, studied a product that did not exist before February 12, 2026.
On that date Polymarket launched a binary contract that paid $1 if Bitcoin closed a five-minute window above where it opened, and $0 otherwise. A fresh contract opened every five minutes around the clock.
Within months, Polymarket’s five- and fifteen-minute crypto up/down markets traded more than $4 billion and tripled the platform’s daily volume. The flaw in polymarket was when the contract settled against a Chainlink oracle that averaged Bitcoin’s price across major spot exchanges.
A trader who held the contract could buy or sell real Bitcoin in the closing seconds, drag that reference price across the strike, and win the bet.
The oracle’s blend of exchanges looked like a defense, because moving it seemed to require moving many venues at once. The authors showed it was not much of a defense. Binance, the largest crypto exchange, sat about two and a half basis points from the oracle and moved near one-for-one with it. It finished on the same side of the strike as the resolution about 85%of the time. A push that drove the Binance price a few basis points past the strike carried the outcome.
The pattern was in the Binance data. After the five-minute contract went live, net order flow in the final ten seconds before each close jumped about 50% above the pre-launch level. The spike was sharpest where a push mattered: in the 6% of cycles the market judged near-even, the jump was about 3.9 times the rest.
The reversal gave it away. Real information stays in a price; a manipulative push does not. Within ten seconds the price reverted, by about a quarter in the near-even cycles. The pushes clustered in thin hours, when a dollar of flow moved the price the most: 56% landed overnight and 44%on weekends.
Who won, who paid with these Polymarket bets
In near-even cycles, a push against the favored side flipped the winner 65% of the time, against 41% in normal trading. Even when one side held a 90-to-100% chance before the close, a push against it reversed the outcome 34% of the time, against 1% in cycles with no push. A bet the market treated as near-certain lost one time in three.
Because Polymarket settled on a public blockchain, the authors traced each wallet. Just 821 traders fit the manipulator profile, about one in three hundred of the 243,000 who traded the contract. They took $8.2 million in the pushed cycles and broke even in the rest. Of the losses, 93% fell on retail.
The authors ruled out hedging as the innocent explanation. A binary contract carried little exposure to hedge once one side was near-certain, yet those were the cycles a push flipped. And the trades arrived in one burst in the final fifty seconds, not as a position built over the window.
The remedy
The fix was the contract’s horizon. Manipulation was absent from the fifteen-minute contract, because a longer window took in more ordinary trading before the close and made a fixed push a weaker force. The stakes reached past crypto: Nasdaq and Cboe each filed with the SEC to list binary asset-price contracts on equity indices, which would carry the same risk onto larger markets.
