Researchers from Stanford University have uncovered indications that traders may be manipulating the most popular Bitcoin prediction markets on the Polymarket platform by briefly influencing the cryptocurrency's price, with the outcome of the bets directly tied to that price.
This working paper, co-authored by Stanford scholars and a researcher from Singapore Management University, examined data from approximately two months of five-minute Bitcoin betting cycles on Polymarket. The study identified multiple instances of concentrated, one-sided trading on the Binance exchange, where Bitcoin's price was briefly pushed up or down just seconds before a betting settlement deadline, allowing traders with corresponding positions to profit.
These manipulative activities frequently occurred during periods where small, temporary fluctuations in Bitcoin's price could determine the profit or loss of a bet. The researchers have defined this pattern as short-term manipulation through spot price "pump and dump."
Prediction markets have traditionally been used for forecasting events like elections or sports, where participants have little ability to influence the outcome. However, the researchers point out that betting products linked to financial assets are highly susceptible to manipulation, as traders can directly trade the underlying asset that determines the settlement, thereby influencing the bet's result.
One of the paper's authors, Singapore Management University Assistant Professor Yu Shihao, stated on LinkedIn: "These contracts have a structural flaw. Their settlement price relies entirely on the market price of the underlying asset, and traders themselves can move that price by buying and selling it."
As major exchanges increasingly launch prediction market products tied to financial assets, the implications of this research may extend beyond just cryptocurrencies and the Polymarket platform. The Chicago Board Options Exchange (Cboe) has already listed several equity index prediction products, and Nasdaq is also applying to list similar contracts.
A Polymarket spokesperson responded, stating the platform uses multiple independent price oracles to aggregate market data, ensuring price fairness. The company plans to adjust the settlement mechanism for some markets next year, moving away from using a single instantaneous price at a specific time to using an average price over a longer period, which it says will further safeguard market integrity.
The research team did observe anomalous Bitcoin trading during the settlement windows of Polymarket's short-term Bitcoin bets, but the paper could not definitively prove that these trades originated from Polymarket users holding corresponding betting positions who would profit from the short-term price moves. Elton Sheshidura, Head of Research at crypto data analytics firm Allium, noted that while the study cannot directly prove traders' intent to manipulate, the evidence collected strongly aligns with the characteristics of such behavior.
Sheshidura stated: "This anomalous trading pattern is real. The hardest question to answer is whether the traders moving the price on Binance are the same people as the wallet addresses cashing out on Polymarket."
Polymarket's five-minute Bitcoin bets do not rely on a single exchange's price for settlement but instead use a composite price from multiple oracles. Despite this, during the study period, the contract settlement outcome moved in the same direction as Binance's short-term price 85% of the time, meaning Binance's price was sufficient to sway the final settlement. The researchers inferred manipulation based on trading patterns, as they could not directly access traders' subjective motives.
A Binance representative responded, stating that as a leading global cryptocurrency exchange, the platform serves a price discovery function. "We have robust market surveillance, order flow monitoring, and anti-manipulation risk control systems in place, but we cannot control the oracle data sources and settlement rules designed by external platforms themselves."
All of Polymarket's transaction records are publicly available on the blockchain, making the platform a frequent subject of academic research. The paper's authors noted that the structural flaw identified is not unique to Polymarket; any prediction market linked to the price of a tradable asset carries similar risks.
The research compared trading data before and after the launch of Polymarket's rolling five-minute Bitcoin bets in February of this year, analyzing approximately 16,000 betting contracts over a two-month period.
The data showed that during settlement periods where the profit/loss outcome was closely contested, the volume of market orders on Binance was 3.9 times higher than during ordinary settlement periods.
The researchers identified the top 10% of settlement windows with the most pronounced anomalous trading as high-probability manipulation periods. These periods often occurred during nights and weekends when market liquidity is low, allowing small orders to easily move the price. Based on this, the researchers estimated that the suspected manipulative traders profited by approximately $8.2 million over the two months, primarily at the expense of ordinary retail and other trading users.
In contrast, almost no similar manipulative behavior was observed in 15-minute Bitcoin betting markets. The research team believes that a longer settlement cycle significantly increases the difficulty of price manipulation, effectively mitigating the risk posed by this type of vulnerability.
A spokesperson for Cboe Global Markets stated that the paper highlights the importance of product design and market regulation, but emphasized that its new prediction market contracts are fundamentally different in their core mechanisms: Cboe's products are linked to broad-based indices like the S&P 500, not a single underlying asset, and are subject to U.S. securities regulation throughout.
Comments