Eleven seconds late costs six in seven orders

A study posted to Ethereum Research puts the two defences available to a liquidity provider on one scale: repricing sooner, and screening out informed flow. A break-even condition pins down the exact screening rate at which a slower maker matches a faster unscreened venue, which is the selection that buys back the extra staleness. The derivation needs only an integrable price innovation, so it holds under jumps, atoms and infinite variance.

Taken to four USDC-WETH pools benchmarked against Binance's ETH-USDT from June to August 2026, the one-second benchmark innovation turns out to be sharply non-Gaussian. In 73 percent of seconds the price does not change at all, and the excess kurtosis is 154. That shape is what makes the usual assumption misleading rather than merely imprecise.

Cutting the quote cadence from twelve seconds to one sheds 84.5 percent of the expected shock. So a twelve-second maker has to screen at least 84.5 percent of informed flow to match a one-second venue, far above the 71.1 percent a Gaussian model implies. The conclusion does not depend on pinning toxicity down: across the informed shares admitted by the tightest netting bound, the threshold stays between 84.5 and 85.2 percent, which the author renders as screening out roughly six in seven informed orders.

Two further results point the same way. In trending markets the stale pool is picked off at execution: with-trend fills obtained an effective spread 11.1 basis points lower than against-trend fills, with the same sign in all four pools. And at missed proposer slots, the discrepancy a stale pool must absorb grows with elapsed staleness at an elasticity of 0.45, close to the diffusion rate of one-half. Predictable drift, by contrast, is a centering problem: it widens a spot-centred quote, but the cost disappears once the quote is recentred on the expected reference.