An analysis of over 60 million on-chain records and high-frequency spot market data has uncovered systematic manipulation on the Polymarket prediction platform. Researchers from Stanford University and Singapore Management University found that a group of 821 wallets (just 0.34% of all traders) earned $8.22 million by using anomalous activity in the final seconds before the settlement of five-minute Bitcoin contracts.
The scheme turned out to be cynically simple and effective. The BTC Up/Down 5-minute contract, launched on February 12, 2026, pays $1 if the Bitcoin price at the end of the window is not lower than the starting price, and $0 otherwise. The BTC/USD price feed from Chainlink is used to determine the outcome. Manipulators first opened a position, then, seconds before settlement, executed large spot trades, shifting the price in the desired direction. This short-term market pressure was recorded by the oracle, the contract settled in their favor, after which quotes partially returned to their original level.
The data confirms this scenario: directional order flow in the last 10 seconds of the window increased by 50% after these contracts appeared, and the price movement began to reverse within the next 10 seconds. In cycles with an equal probability outcome, the activity spike was 3.9 times stronger than normal, and the average spot trade volume in manipulated cycles reached $1.7 million compared to $68,000 in regular ones. 56% of such episodes occurred during night hours and 44% on weekends — periods of lower liquidity when less capital was needed to shift the price.
821 "Whales" vs. Retail Traders
The researchers classified 1,613 cycles as manipulated — the top 10% in terms of directional trade intensity in the final seconds. Wallets that participated in at least 5 such cycles and earned a total profit of at least $2,000 were categorized as likely manipulators. These 821 addresses earned $8.22 million in anomalous cycles and only $90,000 in others. Retail and other unclassified participants lost $7.61 million — approximately 93% of the manipulators' profits.
The most telling point: in cycles where Polymarket estimated the probability of one side winning at 90–100%, directional movement against the favorite led to its loss in 34.2% of manipulated cases, compared to just 1% in normal ones. This proves that manipulators deliberately "flipped" nearly settled outcomes.
Polymarket's Response and Market Lessons
Polymarket has already announced plans to change the settlement mechanism. In fifteen-minute contracts, the activity spike was significantly weaker, and no pronounced price reversal after settlement was observed. Increasing the time window is an obvious but effective step. Other protective measures include settlement based on the average price over the period, random selection of the fixing moment, and position size limits.
My expert opinion: This situation is a classic example of the "last mile" vulnerability in DeFi, where oracles become an attack point. Polymarket, like any platform using instantaneous price snapshots, must reconsider its contract architecture. Otherwise, the "wisdom of the crowd" risks turning into a "cash cow" for a handful of insiders with enough capital for short-term market pressure.