A small group of traders may have extracted $8.2 million from Polymarket’s five-minute Bitcoin prediction markets by exploiting the platform’s settlement mechanism rather than correctly predicting Bitcoin’s price, according to a new academic study.
- 821 traders extract $8.2 million from Polymarket five-minute Bitcoin contracts by exploiting price settlement vulnerabilities.
- Researchers from Stanford University find retail participants absorb 93 percent of losses caused by last-second spot trades on Binance.
- Market manipulation risks increase during low-liquidity weekends when Chainlink oracles reference single-point prices instead of time-weighted averages.
Researchers from Stanford University and Singapore Management University found evidence that 821 traders repeatedly placed concentrated spot Bitcoin trades in the final seconds before contracts settled, briefly moving the price enough to determine the outcome of prediction market bets before the market quickly reversed.
The findings, published in the working paper “Settlement Manipulation in Prediction Markets,” offer one of the clearest examinations to date of how short-duration prediction markets can be vulnerable to manipulation when contracts rely on a narrow snapshot of an underlying asset’s price. The researchers concluded that the issue stems largely from market design and could be mitigated through changes to settlement rules rather than broader regulation.
Researchers Found Repeated Last-Second Price Pushes
The study examined Polymarket’s five-minute Bitcoin up-or-down contracts, introduced in February 2026. Each contract pays out based on whether Bitcoin finishes above or below a specified price after five minutes.
Settlement relied on a Chainlink oracle that referenced a near-instantaneous Bitcoin spot price, primarily from Binance.
Have a development worth tracking?
Share product launches, funding announcements, partnerships, research findings and market developments with The Grey Terminal's readership.
→ Submit a Press ReleaseAccording to the researchers, suspected manipulators first established positions in the prediction market before executing large directional trades on Binance during the final seconds before settlement. Those trades temporarily moved Bitcoin’s price enough to flip the outcome of the contract.
After settlement, Bitcoin’s price typically returned to its earlier level.
The pattern appeared most often in contracts that were still close to the strike price near expiration, where even a small movement could determine whether a contract paid out.
Retail Traders Bore Most of the Losses
The researchers analyzed trading activity from roughly 243,000 participants and identified 821 accounts that displayed trading patterns consistent with settlement manipulation.
According to the paper, those traders collectively earned an estimated $8.2 million from manipulated settlement cycles while showing little or no abnormal profits outside those events.
The study estimated that retail participants absorbed about 93% of losses associated with those manipulated contracts.
Researchers also found the behavior was more common during weekends and overnight trading hours, when lower market liquidity made short-term price movements easier to influence.
Longer Contracts Showed Far Less Manipulation
The paper found substantially weaker evidence of manipulation in Polymarket’s 15-minute Bitcoin contracts.
Researchers said the longer settlement window reduced the effectiveness of last-second price pushes because natural market activity had more time to offset temporary distortions before contracts expired.
The findings suggest that extending contract duration alone could significantly reduce opportunities for settlement manipulation.
Study Calls for Changes to Settlement Design
Rather than focusing on new regulation, the researchers argued that changes to prediction market design could address many of the vulnerabilities they identified.
Among their recommendations were replacing single-point settlement prices with time-weighted average prices, incorporating pricing from multiple exchanges instead of a single venue, and extending contract durations beyond five minutes.
The authors argued those changes would increase the cost of attempting to influence settlement prices while making temporary spot-market distortions less likely to determine contract outcomes.
The study has not alleged criminal wrongdoing by any identified trader, nor does it claim that every profitable trade resulted from manipulation. Instead, the researchers concluded that the observed trading patterns were statistically consistent with repeated attempts to influence settlement prices under the platform’s existing design.
Activate Terminal Layer
Structural analysis of the systems, pressures, and stakeholders behind this story.
