My fellow analysts and I have carefully examined a recent academic study that sheds light on a troubling practice on the Polymarket prediction platform. It turns out that five-minute binary contracts on Bitcoin price movement have become not just a tool for betting, but an effective mechanism for redistributing funds from retail traders to a narrow group of manipulators. Moreover, this activity had a noticeable distorting effect on the spot price of the leading cryptocurrency.
How the Scheme Worked
The essence of the contract was simple: it paid $1 if Bitcoin closed a five-minute window above the opening price, and $0 otherwise. A new contract was launched every five minutes. Within just a few months, the turnover on these five- and fifteen-minute contracts exceeded $4 billion, tripling the platform's daily trading volume.
The key vulnerability identified by researchers lay in the calculation mechanism. The contract was tied to a Chainlink oracle, which averaged the Bitcoin price across major spot exchanges. A trader holding the contract could, in the final seconds of the window, execute a large real Bitcoin trade on one of the exchanges. This action would shift the oracle's reference price in the desired direction, guaranteeing a win on the bet.
In practice, the protection built into averaging across multiple platforms proved illusory. The largest exchange, Binance, deviated from the oracle's final value by only 2.5 basis points and moved almost synchronously with it. It was through Binance that the pressure was applied. In 85% of cases where a manipulative "push" occurred, the price on Binance ended up on the same side of the threshold as the calculation result. Manipulators only needed to shift the price on Binance by a few points past the threshold, and the cycle's outcome was decided.
A characteristic "reversal" in price confirmed the artificial nature of these movements. Within ten seconds of the calculation, the price retraced about a quarter of the move. Notably, the majority of interventions occurred during "quiet" hours: 56% at night and 44% on weekends.
Numbers and Consequences
The scale of manipulation is impressive. In cycles close to equilibrium, a push against the favorite changed the winner in 65% of cases, compared to 41% in normal trading. Even when one side had a 90-100% chance of winning just before closing, a counterattack reversed the outcome in 34% of cases. A bet that the market considered almost certain lost one in three times.
The transparency of the blockchain allowed wallets to be tracked. Only 821 out of 243,000 trading participants fit the manipulator profile — roughly one in three hundred. In cycles with manipulative pushes, they earned $8.2 million, while breaking even in others. Meanwhile, 93% of all losses fell on retail traders. I reject the notion of harmless hedging: in hedging, a binary contract is nearly risk-free when one side is close to winning, yet it was precisely such cycles that manipulators flipped.
Interestingly, no manipulation was found on the fifteen-minute contract: the longer window allowed more regular trades to pass through, "diluting" a single push. In my view, the solution lies precisely in increasing the time window of the contracts.
My conclusion: This study is not just an academic curiosity. It demonstrates how manipulation can thrive even on decentralized platforms with public data, if the contract architecture fails to account for behavioral and market risks. Given that Nasdaq and Cboe have already filed applications with the SEC to launch similar binary contracts on stock indices, this case serves as a warning for the entire market. The problem could spread to much larger and regulated platforms, and it cannot be ignored.