Polymarket has confirmed that traders exploited a settlement flaw in its short-dated crypto prediction markets to extract millions of dollars from other participants, and the platform is now overhauling how those markets resolve. A peer-reviewed study from researchers at Stanford and Singapore Management University, released in July, found that 821 accounts engaged in what it called likely manipulated settlement activity across two months of five-minute Bitcoin contracts, draining a combined $8.2 million, with 93% of the losses in manipulated windows falling on retail traders.
The mechanism was simple once identified. Polymarket's five-minute, 15-minute and four-hour up/down markets resolved on a single price snapshot taken at the exact moment of expiry. Traders who built a position ahead of that moment could then fire off unusually large orders on Binance in the final seconds before settlement, temporarily distorting the reference price just long enough to flip the market's outcome before it snapped back. Researchers described spotting "unusually large orders on Binance in the final seconds before settlement, followed by rapid price reversals," adding that "a bet the market treated as near-certain was overturned one time in three" during the windows they studied.
Warnings ignored for months
The vulnerability wasn't a surprise to close observers of the platform. An onchain analyst going by Axis Robotics flagged manipulation in Polymarket's five-minute Bitcoin markets as early as May 11, pointing to "precise reversals in the last few seconds." Ten days later, a researcher known as Variance Lover published a more detailed breakdown showing traders accumulating positions and then moving Binance prices during the settlement window itself. Polymarket developer Josh Stevens acknowledged the platform was investigating, but no structural fix arrived until this week.
Moving to time-weighted pricing
On August 7, Polymarket said it would replace single-price snapshots with a time-weighted average price, or TWAP, sourced through Chainlink Data Streams. Five-minute markets will use a 30-second TWAP window, while 15-minute and four-hour markets move to a 60-second window. "To protect market integrity in our crypto up/down markets, we're updating how these markets resolve," the platform said in its announcement. Averaging price data over a window rather than reading one instant makes a last-second price push far more expensive to sustain, since an attacker now has to move the market for tens of seconds rather than a single tick.
Related: Bitwise CIO Says Hyperliquid Is Rewriting How DeFi Tokens Get Valued
Polymarket is also injecting $1 million in liquidity rewards into the affected markets through August, an incentive aimed at deepening order books so that large trades move prices less in the first place. The fix effectively brings Polymarket in line with regulated rival Kalshi, which settles its own short-dated markets against CF Benchmarks price indexes averaged over a 60-second moving window. A Kalshi developer known as IcoBeast.eth has said the manipulation pattern documented on Polymarket doesn't work against that design, since brief price distortions are "significantly harder and more expensive" to sustain against an averaged benchmark.
A recurring problem for fast markets
The episode is a reminder that ultra-short-duration prediction markets inherit the same settlement risk that plagued early derivatives products before benchmark providers moved to volume- or time-weighted references specifically to blunt last-second manipulation. Polymarket's founder and CEO, Shayne Coplan, has pushed the platform's crypto markets as a high-frequency complement to its slower-moving political and cultural markets, but the five-minute contracts' reliance on a single external price feed made them a comparatively soft target. With Chainlink's TWAP feeds now live and the incentive program running through the month, Polymarket is betting that deeper liquidity and averaged pricing will make the next five-second trick a lot less profitable.
