Talk of Bitcoin price manipulation is nothing new, but this time the accusation isn't coming from anonymous forum posts — it's coming from academia. A new study by researchers at Stanford University, working alongside colleagues from the Singapore Management University, has identified a specific and troubling pattern inside one of Polymarket's most popular products: its five-minute Bitcoin "up or down" prediction markets.
How the five-minute Bitcoin markets work
The mechanics are simple. Every five minutes, Polymarket opens a new market asking whether Bitcoin's price will be above or below a target set at the moment the market opens. Traders bet "up" or "down," and whoever calls it right at settlement wins.
Because Bitcoin can barely be predicted with confidence over a five-minute window, these markets lean heavily into short-term speculation — closer to gambling than investing. That hasn't stopped them from becoming wildly popular: according to Dune Analytics data cited in the research, these five-minute markets account for more than a quarter of Polymarket's total trading volume.
Where the manipulation comes in
The core finding is that the manipulation didn't happen on Polymarket itself, but at the source feeding it. Polymarket relies on Chainlink's price oracle to determine the settlement price, and that oracle in turn averages prices across major exchanges — weighted heavily by trading volume. Binance, as the largest centralized exchange, carries outsized influence over that average.
Over roughly two months of trading data, the researchers found a repeating pattern: certain traders placed large spot orders on Binance in the final seconds before a five-minute market closed, nudging the exchange price just enough to flip the outcome of the Polymarket bet in their favor.
The study estimates that around 821 addresses profited from this strategy, pocketing a combined $8.2 million — with retail traders absorbing the bulk of the losses on the other side of those bets.
Manipulation, or just arbitrage?
This is where the story gets genuinely debatable. One view is that this is straightforward manipulation: traders are reaching into the underlying market to engineer the very outcome they're betting on, rather than simply predicting it. The other view is less damning — it frames this as a form of arbitrage, exploiting a brief pricing mismatch between the spot order book and the implied odds on the prediction market, not unlike patterns researchers have documented elsewhere in crypto markets involving insider-style trading advantages.
What isn't really up for debate is the mechanism itself: it's far easier to nudge a price in the last few seconds of a tight five-minute window than it is to move markets over longer stretches, where deeper liquidity and broader participation make similar tactics much harder to pull off without triggering a visible reaction.
