A new study by researchers at Columbia University has found that trading volumes on Polymarket, one of the world’s most popular prediction markets, have been significantly inflated by so-called “wash trading,” where users rapidly buy and sell the same contracts to boost apparent market activity.
The paper, titled “Network-Based Detection of Wash Trading,” identifies patterns of artificial activity across Polymarket transactions between 2022 and 2025.
Using network-analysis techniques, the researchers found that roughly 25% of all buy-and-sell volume during the past three years could be attributed to non-genuine trading behaviour.
According to the authors, wash trading was not constant but fluctuated sharply over time, peaking at nearly 60% of total activity in late 2024 before dropping to around 20% by October 2025.
Sports-related markets were found to have the highest share of artificial trading, accounting for almost 45% of detected volume, while election and political markets were less affected.
Understanding the Implications of Wash Trading
The study does not suggest that Polymarket itself engaged in or encouraged such activity. Instead, it highlights how the design of decentralized prediction markets, which allow users to trade event-based contracts using cryptocurrency, can make them susceptible to manipulation by users attempting to simulate liquidity or attract attention to specific markets.
Polymarket, which operates on the Polygon blockchain, has grown rapidly in recent years as interest in crowd-driven forecasting tools has surged. It enables users to bet on real-world outcomes ranging from election results to sports and economic indicators.
Researchers said their findings underline the need for more transparent monitoring systems within blockchain-based financial platforms. They argue that detecting and deterring wash trading is essential to maintaining market integrity, especially as prediction markets gain mainstream visibility and regulatory scrutiny.
Study Details and Availability
The Columbia team’s paper is currently available as a working draft on the Social Science Research Network (SSRN).