MEV Topology on Stablecoin CLMM Pools: Searcher Behaviour, LP Loss Decomposition, and Net Yield Implications

The received wisdom on stablecoin concentrated liquidity positions is that narrow-range provision is capital-efficient and low-risk. The two assets do not diverge materially in price; impermanent loss, in the traditional sense, is negligible; fee revenue per unit of liquidity is high. This account is incomplete. It ignores the adversarial microstructure that operates within concentrated liquidity market maker (CLMM) pools, the systematic extraction of value from passive liquidity providers (PLPs) by well-capitalised searchers, and the implications of that extraction for the realised yield of institutional LP positions. This article develops a rigorous decomposition of LP value flow in CLMM stablecoin pools, examines the topology of the MEV strategies operating within them, and draws practical conclusions for institutions allocating capital to this yield layer.

In a CLMM pool, as implemented in Uniswap V3, Aerodrome, Orca, and Kamino, liquidity providers define a position (L, pl, pu), where L is the liquidity magnitude and [pl, pu] is the price range over which it is active. The pricing curve within each tick is derived from the constant product invariant x·y = L², where x and y are the virtual reserves. The pool price P(t) moves across ticks as swaps execute, with each tick boundary representing a discrete jump in active liquidity.

The total value of an LP position V(t) at time t can be decomposed into three components: fee income F(t), representing the pro-rata share of swap fees accrued while price remains inside the active range; rebalancing loss R(t), the difference between V(t) and the value of simply holding the initial token quantities; and adverse selection cost A(t), the component of R(t) attributable to informed order flow rather than noise trading. The seminal framework for this decomposition is Loss-Versus-Rebalancing (LVR), introduced by Milionis, Moallemi, Roughgarden, and Zhang (2022). LVR defines the LP's adverse selection cost as the difference between the LP's position value and the value of a benchmark portfolio that trades at market prices, formalising the observation that LP positions are economically equivalent to writing a short gamma option on the price process.

For a continuous-time model under geometric Brownian motion with volatility sigma, the instantaneous LVR rate is dLVR/dt = (sigma^2 / 2) · (x'(P) · P^2), where x'(P) is the slope of the LP's token X reserve as a function of pool price and the expression captures the convexity of the LP payoff curve. For a CLMM position concentrated over a narrow range delta = pu - pl, the concentrated liquidity amplifies the effective virtual reserve size, and therefore the LVR, by a factor proportional to 1 / sqrt(delta) relative to a full-range position. Importantly, Cartea, Donnelly, and Jaimungal (2024) formalise this as Predictable Loss (PL), demonstrating analytically that PL increases as the width of the liquidity range decreases, even holding fee income constant. The narrower the range, the higher the per-unit adverse selection cost.

The primary mechanism through which value exits stablecoin CLMM pools is CEX-DEX arbitrage. When the price of a stablecoin pair shifts on a reference venue, typically a centralised exchange with lower latency and deeper liquidity, a stale spread opens between the DEX pool price and the true market price. Informed arbitrageurs, operating either as independent searchers or as vertically integrated builder-searchers, close this spread by trading against the pool, capturing the difference as profit at the expense of passive LPs.

On Ethereum, this process operates almost entirely through MEV-Boost. Searchers do not broadcast to the public mempool; they submit signed bundles to block builders, who aggregate them into maximally profitable blocks. Wang et al. (2024) document that the top two builders capture over 90% of Ethereum block auctions, and that the search-to-builder bid ratio on profitable CEX-DEX arbitrage transactions averages roughly 90%, meaning searchers transfer most of the arbitrage profit to builders in competition for inclusion. This concentration matters directly to LPs: the efficiency of the arbitrage pipeline determines how quickly stale prices are corrected, which in turn determines how long passive LPs remain exposed to adverse selection on each block.

For stablecoin-to-stablecoin CLMM pools, the arbitrage opportunity is structurally different from volatile asset pairs. The magnitude of the price dislocation is bounded by the stability of the peg, typically within a few basis points for USDC/USDT pairs. However, the frequency of arbitrage opportunities is high because stablecoin pools carry significant volume, and even small spreads can be profitably exploited given the low gas cost of a targeted swap. The net effect on LPs is a continuous drip of small adverse selection costs rather than the large periodic dislocations characteristic of volatile asset pools. Empirically, Gudgeon et al. (2021) find that realised interest rates in DeFi lending protocols cluster around the kink of kinked rate models, a finding that generalises: concentration of activity at specific parameter values is characteristic of competitive MEV environments where efficient extraction erodes margins toward zero.

A second class of MEV activity directly targeting CLMM LPs is Just-In-Time (JIT) liquidity provision. A JIT LP detects a large pending swap in the mempool, or via private orderflow, mints a concentrated position in the exact tick range the swap will traverse immediately before execution, and burns that position immediately afterward. The result is a disproportionate share of fee capture from a single swap with minimal duration risk.

The formal model of JIT strategy is characterised in Llacer Trotti et al. (2025), the first transaction-level game-theoretic analysis of JIT provision in CLMMs. Their key results are material for institutional allocators. First, a JIT LP entering immediately before a swap, when the pool price equals the market price, can benefit from swap-induced price impact in addition to collecting fees. Second, the optimal JIT strategy is a non-linear optimisation over liquidity magnitude and tick range that maximises the combined fee and price-impact gain net of capital opportunity cost. Third, at high JIT budget allocations, JIT LPs erode passive LP fee income by up to 44% per trade, a meaningful extraction from the LPs who created the liquidity depth that made the pool usable.

For stablecoin pools specifically, Xiong et al. (IEEE 2023) find that they are less attractive for JIT attacks than volatile asset pools because the narrow range constraint limits price-impact gain, the high baseline liquidity reduces the marginal fee share a JIT LP can capture, and the attack requires substantial capital to dilute passive LP fee share meaningfully. The practical implication for institutional stablecoin LP positions is that JIT pressure is lower than in volatile pairs, but not absent, especially in pools with large individual swaps from institutional counterparties or MEV-aware aggregator routing where the entry conditions are reliably predictable.

Importantly, Capponi, Jia, and Zhu (2024) demonstrate a paradox: if JIT adoption increases among sophisticated LPs, the resulting cream-skimming of uninformed order flow can displace PLPs entirely, potentially leading to a liquidity freeze in passive pools. This second-order effect, where individually rational sophistication degrades aggregate market quality, is directly analogous to adverse selection spirals in traditional market microstructure.

The architectural feature of CLMMs most relevant to institutional LP yield is the discretisation of the price axis into ticks. Each tick represents a price level at which active liquidity changes discontinuously. Between adjacent ticks, the pool behaves as a constant product AMM with fixed reserves. As price crosses a tick boundary, a discrete quantity of liquidity enters or exits the pool, changing the effective depth for subsequent swaps.

For a passive LP allocating capital to a stablecoin CLMM pool, the decision variables are the fee tier, the price range [pl, pu], and the rebalancing frequency. These interact non-trivially. A narrower range amplifies capital efficiency because the effective liquidity L provided per dollar of capital scales as 1 / (sqrt(pu) - sqrt(pl)), meaning tighter ranges provide exponentially more depth. But it also amplifies LVR per unit of capital, as demonstrated by the predictable-loss result above. The optimal range width is therefore not "as narrow as possible"; it is the range that maximises fee income relative to adverse selection cost, and that optimum is asset-pair and regime dependent.

For USDC/USDT pools with a 0.01% fee tier, the relevant regime is quasi-peg with occasional micropeg deviations of 1 to 5 basis points driven by redemption queue constraints, CEX liquidity dynamics, and market stress. At baseline, a plus-or-minus 5 basis point range captures nearly all swap volume with maximum capital efficiency. During stress events, such as the USDC depeg following the SVB collapse in March 2023, positions with very tight ranges can fall out of range and effectively become single-asset holdings, eliminating fee income while retaining adverse selection exposure if the position is not actively unwound. That is the institutional edge: not identifying the highest displayed CLMM APY, but correctly sizing and actively managing the range-width tradeoff through normal and stressed conditions.

The realised yield of a stablecoin CLMM position is best expressed as Ynet = Fgross - LVRrealised - G, where Fgross is gross fee income, LVRrealised is the actual adverse selection cost, and G is gas plus rebalancing cost. For a USDC/USDT 0.01% pool on Ethereum mainnet, typical Fgross may run 3% to 5% annualised in high-volume conditions. LVR_realised depends on arbitrage-pipeline efficiency but remains bounded by peg stability. The annual extraction burden is therefore determined not by one dramatic event but by the cumulative frequency of small dislocations and the speed with which searchers monetise them.

The empirical finding of Milionis et al. (2022), that LVR and impermanent loss have identical expectations under Brownian motion but vastly different distributions, is directly useful for portfolio construction. LVR provides a more reliable forward-looking diagnostic for expected extraction costs than historical impermanent loss because it isolates the adverse-selection component from the broader market-risk component that IL conflates. For institutional positions managing to a target net yield, monitoring rolling LVR, rather than headline APY or superficially attractive fee rates, is the operationally correct diagnostic.

Stablecoin CLMM positions remain a viable yield layer, but only for allocators who understand the adversarial microstructure operating within them, model their net yield correctly, and maintain the operational infrastructure to monitor range exposure and rebalance or exit during stress. The difference between a well-managed stablecoin CLMM allocation and a poorly managed one is measured in basis points of LVR per month, and that distinction compounds materially over a portfolio cycle.

Stablecoin CLMM pools are not passive yield instruments. They are adversarial microstructures in which searchers with superior capital, latency, and block-building access systematically extract value from passive liquidity providers through CEX-DEX arbitrage and JIT liquidity attacks. The relevant literature, LVR, JIT game theory, predictable loss, and the broader MEV research stack, provides a rigorous framework for quantifying these extraction mechanisms and their impact on net LP yield. The operational implication is straightforward: stablecoin CLMM yield must be modelled as fee income minus LVR minus rebalancing cost, not as headline APY. Institutions that fail to make that distinction will systematically underperform their theoretical yield targets. Those that do can still access genuine value from the strategy, but only with a clear-eyed understanding of the structural forces working against them.