Execution Cost and Quality: Measure It, Then Cut It

Execution cost is not one number and it does not scale in one direction. Lamport-denominated fees are flat per attempt, so they dominate small orders and vanish on large ones. Price impact is the reverse. Pool fees sit in the middle, scaling with notional and multiplying with every hop. A configuration tuned for one size is usually wrong for another, and knowing which line dominates at your size is most of the work.

This section starts with sizing, because size against pool depth determines the largest variable cost you control. It then covers the risk that a wide tolerance on a thin pool invites, enumerates the full cost stack line by line, and ends with measurement: how to compute realised slippage from your own transactions, how to count attempts rather than only fills, and how to compare two execution paths without the comparison quietly grading itself.

Measurement is last on purpose. Every improvement claim in trade execution is a claim about a distribution, and distributions require a fixed benchmark, a stated denominator and enough paired observations to separate a real effect from an unusually quiet hour. Without that discipline, cost optimisation is just preference with arithmetic attached.