Final-Second Rigging in Polymarket Bitcoin Bets, Stanford Study Finds
Stanford Study: Polymarket’s Bitcoin Bets Were Rigged in the Final 10 Seconds
A Stanford University study found that betting activity on Polymarket tied to Bitcoin outcomes was manipulated in the final moments of trading, with the researchers concluding the market was “rigged” in the last 10 seconds before resolution.
The study focuses on how late-stage trades can influence prediction markets—platforms where users place bets on real-world outcomes and prices move based on perceived probabilities. In this case, the researchers said the decisive activity occurred so close to the finish that it undermined the integrity of the market’s result.
Why it matters: Prediction markets are often cited as tools for aggregating information and forecasting events. If outcomes can be pushed over the line by last-second manipulation, it raises questions about whether these markets reliably reflect collective belief—or whether they can be steered by a small number of participants at critical moments.
The findings also highlight a broader issue in crypto-based markets: many systems are designed to be open and fast-moving, but that same openness can create opportunities for actors to exploit timing and settlement mechanics. When a market resolves based on a snapshot or a narrowly defined resolution window, “end-of-market” strategies can become disproportionately powerful.
Polymarket has become one of the most visible crypto-native prediction platforms, with markets spanning politics, economics, and major crypto events. The Stanford researchers’ conclusions suggest that even when a market is liquid and widely followed, short time windows near resolution can still present vulnerabilities.
The study adds to ongoing discussions about how prediction markets should structure resolution rules and guardrails—especially for events linked to highly liquid underlying assets like Bitcoin—so that the final outcome reflects the intended benchmark rather than last-second positioning.
