Polymarket's August Rule Change Cut Bitcoin (BTC) Last-10-Second Trading to 3.02%

Last Ten Seconds Go Quiet

New research into Polymarket's five-minute contracts, which pay out on the Bitcoin price over each five-minute window, suggests an August change to settlement rules has taken the heat out of trading that used to bunch in the final ten seconds on Binance. The prediction market platform moved the reference price for its crypto up-and-down markets from a single snapshot to an average on August 7. Until then, the payoff keyed off the price at one instant, which made a single second decisive. From August 14, both the opening and the closing price of every five-minute contract has been struck from a 60-second time-weighted average price computed by Chainlink, a construction known as TWAP. A study by researcher Guangzhe Wang examined one-second trading data from Binance covering July 1, 2025 through September 4, 2026, a sample of 82,752 five-minute intervals in which only the five-minute contracts settled. The window closes about three weeks after the averaged pricing went live. In each interval, the share of volume traded in the last ten seconds ran 3.05% before the contracts existed, 5.68% during the snapshot-settlement period, and 3.02% after the 60-second average took effect. The snapshot figure sits nearly double the pre-contract baseline; the post-change reading lands back on it.

The mechanics explain why the design mattered. Under snapshot settlement, the payout turned on the exact price printed at settlement time, so a large order in the final seconds could tip the outcome while leaving the rest of the window untouched. The averaged design spreads the reference across a full minute, which raises the cost of any late push, because impact now has to persist for sixty seconds to change the payoff. A time-weighted average gives every second in the window equal weight, so no single print decides the payout.

The Pattern the Averaging Rule Addressed

The August change did not arrive unannounced. Work published on June 30 by Dai and coauthors had already documented what the snapshot design invited: after the five-minute contracts launched, spot trading in the underlying asset spiked at settlement times and then reversed shortly afterward. That earlier study estimated 821 wallets had taken in roughly $8.2 million from the pattern, returns that put individual accounts in whale territory. Wang's follow-up reads the tape after the rule change and reports what he did not find: in the three weeks after the 60-second average took effect, trading did not migrate to earlier portions of the window, and the reversal pattern the June work described did not reappear.

Both studies carry limits their authors state plainly. The newer numbers do not check the current profit-and-loss of the wallets behind the earlier gains, and the finding is no proof that every manipulation method has disappeared. Under the averaging method, the effect of pushing price depends on how much influence a trader can exert across the whole measurement period rather than in one tick, so the rule changes the economics of the attempt instead of blocking it outright. For Polymarket, whose payouts depend on how the reference price is struck, the readings go straight to settlement integrity. Spot-order-flow gauges such as the Coinbase Bitcoin (BTC) premium are the lens traders use for exactly this kind of underlying pressure at settlement boundaries. Whether the strategy stayed profitable after costs from trading the underlying asset and from execution is a question the data does not yet answer, and it is the verification still pending.

What the Averages Do Not Settle

The number to watch is 3.02% against 3.05%. When the last-ten-second share of volume returns to the pre-contract baseline, the specific pattern the snapshot design rewarded has, on this evidence, stopped paying. The logic is structural: a TWAP payout makes a late push profitable only if the price it forces holds for a full minute, which multiplies the cost of the attempt. What the studies leave open is the money question, and it is undated: the current standing of the 821 wallets is not on record, and after-cost profitability has not been measured. Large traders keep adapting, as Bitcoin whales on Hyperliquid turning net short for the first time since March shows, so whether late-window volume reappears in another form is the open question, with no follow-up study date announced.