Faster Blockchain Trading Can Leave a Thinner Market

Does making a blockchain exchange ten times faster make its market better? A preprint by Ruizhe Jia and Shihao Yu finds a split answer: prices can track an outside market more closely while the pool of capital available for trading becomes shallower.

Their study, Optimal Trading Speed on Blockchain, combines an economic model with evidence from Base’s Flashblocks upgrade. The central question is not simply how quickly transactions execute, but how that timing changes competition among traders and liquidity providers.

When an auction becomes a race

An automated market maker, or AMM, holds tokens in a shared pool and prices trades using programmed rules. Liquidity providers deposit the tokens and receive trading fees. But when prices move elsewhere, a pool’s quote can become stale. Arbitrageurs exploit the difference, earning money while bringing the pool price closer to the outside market. Providers whose capital remains deposited bear the corresponding stale-price loss.

Transactions wait for a blockchain execution batch. On Base, higher transaction-fee bids buy earlier positions within that batch. With longer intervals, competing arbitrage transactions are more likely to wait together, creating an auction for priority.

Shorter intervals make it more likely that one transaction executes before its rival arrives. Paying more cannot displace an already completed trade. The model therefore predicts lower priority bids and a larger share of arbitrage profits staying with arbitrageurs rather than the sequencer, the operator ordering transactions. Participants’ message speeds need not change for this shift to occur.

Leaving before the loss

Faster execution also changes liquidity provision. An active provider can monitor outside prices, withdraw when a profitable arbitrage opportunity appears, and return after prices realign.

With slow blocks, the withdrawal may wait alongside the arbitrage transaction and execute too late. Faster blocks offer more chances to withdraw in an earlier batch. The active provider can then collect fees while avoiding some losses that passive providers still bear—a strategy the authors call cream-skimming.

Under the model’s assumptions, liquidity reaches its greatest depth at an intermediate block speed. Initially, faster execution saves customer trades that would otherwise expire while waiting, supporting fee income and attracting capital. Eventually that benefit diminishes, while arbitrage executes more frequently. At sufficiently high speeds, passive capital retreats and total depth falls.

What changed on Base

Base publicly launched Flashblocks on July 16, 2025, introducing execution sub-batches every 200 milliseconds within its existing two-second blocks. The researchers compare changes on Base with changes on Arbitrum, whose execution arrangements remained unchanged during the sample.

They study Uniswap v3 pools trading wrapped ether against the USDC stablecoin, matching blockchain events with Binance prices. Their comparison covers May 16–September 15, excluding the July 7–16 rollout period. This difference-in-differences approach asks whether Base changed differently from the comparison chain, rather than attributing every before-and-after movement to the upgrade.

For arbitrage trades in the largest size decile, the estimated gas-price decline was 73% relative to Arbitrum. The result was not universal across trade sizes: the smallest half showed an increase instead.

Arbitrageurs’ estimated net return improved by 2.73 basis points per dollar traded relative to Arbitrum. One basis point is 0.01 percentage points. This measure compares execution prices with Binance prices ten seconds later and subtracts trading and blockchain fees; it is a short-horizon performance proxy. Lower blockchain fees outweighed a statistically insignificant decline in returns before those fees.

Better timing, less depth

The active-provider analysis follows one qualifying Base wallet and four Arbitrum wallets, selected using pre-upgrade behavior. The Base wallet’s rebalancing duration fell by about 66% relative to the controls, without a statistically significant increase in its gas-price bids. Its exposure to unfavorable trades declined more than the pool-wide change, and its measured per-dollar payoff improved.

Meanwhile, estimated pool depth fell approximately 19% relative to Arbitrum. For trades around one ether, the effective spread widened by 0.07 basis points. That spread measures execution against the pool’s immediately preceding quote and excludes the pool’s separate five-basis-point trading fee.

What remains unresolved

The evidence concerns one asset pair and upgrade, not every blockchain market. Causal interpretation depends on Base and Arbitrum providing a valid counterfactual comparison. Some provider outcomes lose statistical significance in a narrower event window, and unfavorable-flow avoidance already showed movement before launch.

Price tracking also requires caution. The appendix shows that the model’s calendar-time price gap and a post-block statistic can move in opposite directions. Connecting the empirical tracking result to theory therefore requires explicit sampling and timing rules.

The preprint does not identify the best real-world block interval or establish an overall welfare gain or loss. Its narrower lesson is that faster execution changes incentives: fresher prices and deeper liquidity are different achievements.