Image credit: Shutterstock
A Stanford-led working paper said Polymarket’s five-minute Bitcoin prediction markets created conditions that allowed traders to influence settlement outcomes through spot-market activity.
The paper identified sharp increases in Binance order flow immediately before contracts closed, followed by rapid price reversals. Researchers said the pattern was consistent with temporary settlement pushes rather than new-information trading.
Binance Order Flow Rose 50% Near Settlement
The contracts pay $1 when Bitcoin finishes a five-minute window at or above its opening reference price and nothing when it finishes below it. Polymarket uses Chainlink Data Streams to determine the opening and closing prices.
Researchers found that Binance’s absolute Bitcoin spot order flow during the final 10 seconds increased by about 50% after Polymarket launched the five-minute product on February 12. The resulting price moves typically reversed within 10 seconds after settlement.
Paper Says Chainlink Oracle Was Not Compromised
The working paper does not allege that Chainlink’s oracle was compromised. It argues that concentrated trading on a large spot venue can affect the aggregated reference price because arbitrage keeps prices across exchanges closely aligned.
That means final-second spot trades could influence contracts that were close to settlement, even if the oracle system itself operated as designed.
821 Wallets Were Linked to $8.2M in Flagged-Cycle Gains
The researchers examined on-chain trading records from February 12 to April 8. They classified wallets based on profits earned during cycles showing the settlement-push pattern. The method identifies economic beneficiaries rather than directly proving intent.
A group of 821 wallets, about one in every 300 traders in the contract, earned a combined $8.2 million during the flagged cycles while roughly breaking even elsewhere. Retail accounts absorbed 93% of the associated losses, with estimated losses of $7.6 million across manipulated cycles.
High-Certainty Outcomes Flipped in One-Third of Flagged Cycles
Outcomes priced at more than 90% certainty were overturned during about one-third of the affected cycles. That compared with roughly 1% in normal cycles.
Researchers said late spot trades could change contracts that appeared almost settled, creating a concentrated incentive to push prices near the close.
Fifteen-Minute Contracts Showed Weaker Pattern
The same activity was largely absent from Polymarket’s 15-minute Bitcoin markets, despite those contracts using the same settlement system. Their near-close order-flow increase was much smaller and produced no detectable immediate price reversal.
The researchers proposed lengthening contract windows, averaging or randomizing settlement prices, limiting position sizes or increasing the cost of final-second spot trades. Polymarket had not publicly responded to the findings when the study was reported.
