A working paper from researchers at Stanford University and Singapore Management University has found evidence of coordinated manipulation inside Polymarket's five-minute $BTC prediction mark
A working paper from researchers at Stanford University and Singapore Management University has found evidence of coordinated manipulation inside Polymarket's five-minute $BTC prediction markets, raising fresh questions about the structural risks embedded in fast-settling crypto betting contracts.
The paper, titled "Settlement Manipulation in Prediction Markets" and co-authored by David Dai, Ruizhe Jia, and Shihao Yu, studied a product that launched on February 12, 2026. On that date, Polymarket introduced a binary contract that paid $1 if Bitcoin closed a five-minute window above where it opened, and $0 otherwise, with a fresh contract opening every five minutes around the clock.
Contracts settle using Chainlink price feeds at the end of each five-minute window, creating incentives to move the spot Bitcoin price just before expiry.Researchers found repeated bursts of one-sided trading on the Binance exchange that temporarily moved Bitcoin's price in the final seconds before bets closed, benefiting traders positioned in the same direction.During settlement periods exhibiting the most anomalous trading behavior, order volumes on Binance surged to roughly 3.9 times normal levels, with irregular patterns predominantly emerging during overnight hours and weekend periods when reduced liquidity creates opportunities for price influence.
Singapore Management University assistant professor Shihao Yu noted that "these contracts have a structural vulnerability" because they settle on a price that traders can move by trading the underlying asset itself. Despite Polymarket's reliance on multiple independent price oracles, contract settlements aligned with Binance pricing approximately 85% of the time throughout the research window.
Researchers calculated that wallets identified as probable manipulators accumulated approximately $8.2 million in profits across the study period. The paper estimates 821 suspected manipulators were responsible for the bulk of those gains, largely at the expense of retail participants.
The findings land at a sensitive moment for the prediction market industry, with combined monthly volume on Kalshi and Polymarket rising nearly fivefold in seven months, from under $5 billion in September 2025 to about $24 billion by April 2026.
The researchers found that manipulation was largely absent in fifteen-minute contracts, suggesting that lengthening the contract horizon removes the effect and provides the market-design remedy their model and evidence support.The Stanford and Singapore Management University findings suggest that changing the settlement window length and using price-averaging methods could meaningfully reduce manipulation risk.
Sources"Settlement Manipulation in Prediction Markets" — Working Paper (arXiv)Polymarket Bitcoin Bets Show Signs of Price Manipulation, Stanford Study Finds — BloombergTraders Took $8.2 Million From Polymarket's Five-Minute Bitcoin Bets, Study Found — Bitcoin Magazine
