Abstract
Abstract
Passive liquidity providers on concentrated-liquidity automated market makers are ranked almost everywhere by fee APR, the ratio of fees earned to pool value. This paper shows that the metric orders pools close to the reverse of their realized net profitability. Using per-minute swap flows valued against a block-synchronised centralised-exchange reference price, I decompose passive returns into fees, realized loss-versus-rebalancing (LVR) and gas for 62 Uniswap v3 pools on Ethereum, Arbitrum and Base over 37 weeks of 2026, with 2,232 pool-weeks entering the ranking tests. Sorting pools weekly on headline fee APR and holding for one week returns − 9.39% APR, against − 5.72% for holding the whole panel and − 0.49% for sorting on realized LVR-net APR. The sign replicates on Arbitrum, where the top-minus-bottom spread is -4.67 pp (t = -3.66), and on Base, where it is -7.90 pp (t = -4.81); the pooled rank-space coefficient is -0.306 (t = -4.49, clustered by pool). Volatility is not the channel: adding realized volatility of the external pair leaves the fee-APR coefficient intact and volatility itself insignificant. I then test whether the spatial structure of liquidity improves on aggregate LVR, which was the hypothesis this study set out to confirm, and find that it does not. Two further measurements follow. Just-in-time liquidity captured 0.089% of gross fees on Ethereum, and gas is 2.13% of LVR on Ethereum against 0.14% and 0.10% on the two rollups.
JEL classification:
G12; G14; G23; D47; C58