Liquidity vs. Rebalancing (LVR) in AMMs

Updated 22 February 2026

Liquidity Versus Rebalancing (LVR) is a metric that measures the cumulative cost suffered by passive liquidity providers from stale pricing and delayed rebalancing in automated market makers.
The formulation rigorously isolates adverse selection costs from market risk, enabling precise calibration using mathematical models like the Itô process for CFMMs.
Extensions of LVR incorporate realistic market frictions, dynamic fee adjustments, and hedging strategies to minimize rebalancing slippage and optimize protocol performance.

Liquidity Versus Rebalancing (LVR) is a pivotal concept in the theory and empirical study of automated market makers (AMMs), formalizing the adverse selection cost systematically borne by passive liquidity providers (LPs) due to sluggish pool pricing and arbitrageur activity. It measures, both in theory and in practice, the cumulative value leakage from LPs to arbitrageurs, isolating this from market risk and broader “impermanent loss” while providing a rigorous basis for protocol optimization and comparative mechanism analysis (Milionis et al., 2022).

1. Formal Definition and Economic Interpretation

Loss-versus-rebalancing is defined as the shortfall between the realized mark-to-market value of a passive LP’s AMM position and the value that would have accrued had the LP continuously and frictionlessly rebalanced their inventory to match the AMM’s target portfolio at the true (external) market price. For a two-asset constant-function market maker (CFMM) with external market price $P_t$ , and pool reserves $(x_t, y_t)$ , the standard LVR at time $t$ is: $\mathrm{LVR}_t = R_t - V_t = \int_0^t x^*(P_s) dP_s - V(P_t),$ where $x^*(P)$ is the delta-optimal risky-asset holding, $V(P) = \min \{ P x + y \,\mid\, f(x, y) = L\}$ is the pool mark-to-market, and $R_t$ is the value of the frictionless rebalancer (Milionis et al., 2022).

Economically, LVR measures the cumulative cost of “stale quote arbitrage.” Every time the pool’s internal price lags the external reference, an arbitrageur can extract profits by trading at a slightly suboptimal AMM rate, at LP expense. LVR captures this loss, after neutralizing the first-order price exposure, making it the canonical adverse-selection metric for passive on-chain market making (Milionis et al., 2022); cf. for precise potential-based block accounting (Zhang et al., 2024).

2. Mathematical Formulation, Derivation, and Benchmarking

Assuming the external price $P_t$ follows a (risk-neutral) geometric Brownian motion with local volatility $\sigma_t$ , the core Itô formulation yields for any smooth CFMM with value function $V(P)$ : $d\,\mathrm{LVR}_t = -\frac{1}{2} V''(P_t) \sigma_t^2 P_t^2\,dt.$ For constant-product pools ( $f(x, y) = x y = L$ ), $V(P) = 2\sqrt{L P}$ , hence $V''(P) = -\frac{1}{4} L P^{-3/2}$ and the instantaneous per-dollar LVR simplifies to $\ell(\sigma, P)/V(P) = \sigma^2/8$ (Milionis et al., 2022). Integrating over time and normalizing by value, this closed-form “Black-Scholes–type” result allows precise calibration against empirical LP P&L.

Table: Closed-form instantaneous LVR rate for common CFMMs

CFMM Type	$\ell(\sigma, P)$	Per-dollar rate ( $\ell/V$ )
Constant-product (xy = L)	$(L \sigma^2 \sqrt{P})/4$	$\sigma^2 / 8$
Geometric-mean ( $x^\alpha y^{1-\alpha}$ )	$(\sigma^2 / 2) \alpha (1-\alpha) V(P)$	$\alpha (1-\alpha) \sigma^2 / 2$

This is valid under passive LPs, frictionless arbitrage, and continuous strategies. LVR is thus operationally and empirically distinguishable from impermanent loss (IL): while IL conflates slippage and directional risk, LVR strictly isolates rebalancing slippage (Alexander et al., 2024). For finite fees, it reflects only the “impermanent loss premium” not compensated by trading fees (Fukasawa et al., 2023).

3. Extensions: Bounded Liquidity, Block-Time, and Market Structure Effects

The canonical LVR model assumes a single AMM with frictionless external prices. This framework has been extended to various realistic frictions:

Bounded Liquidity: If the arbitrage counterparty (e.g., reference CEX) has quadratic trading cost, LVR is explicitly reduced as a function of relative depth: $\ell(\sigma, Q) = \frac{\sigma^2 Q^2}{2} \left(1 - \frac{|x^{*\prime}(Q)|}{|\tilde{x}^{*\prime}(Q)|}\right) |x^{*\prime}(Q)|,$ where $|x^{*\prime}|$ is the AMM’s marginal slope, and $|\tilde{x}^{*\prime}|$ its reference market analog (Schlegel, 2 Jul 2025). For two CPMMs with liquidity ratio $r$ , $\ell \sim (1-1/r)$ factor.
Stochastic Block Times: Under fixed intra-block volatility $\sigma_b$ and spread $\gamma$ , the per-block LVR admits the closed form: $\mathrm{LVR}_{\text{per block}} \approx \sigma_b^2 \left[2 + 1.7164\, \frac{\gamma}{\sigma_b} \right],$ with the coefficient $1.7164$ derived from ladder-height theory and Riemann zeta constants. Notably, deterministic block intervals attain the minimal asymptotic LVR among all admissible block-time distributions (Nezlobin et al., 8 May 2025).
Batch Auctions, Pool Networks: Defensive rebalancing between pools (“direct transfers” rather than trades) can be posed as a convex optimization, maximizing aggregate log-liquidity while preserving Pareto optimality and eliminating arbitrage (Devorsetz et al., 26 Jan 2026).

4. Empirical Calibration, Hedging, and Mitigation Mechanisms

Empirical studies on Uniswap v2/v3 confirm that LVR often exceeds the net fee revenues, especially on low-fee, high-volatility, or tightly concentrated pools. For WETH–USDC v3, LPs typically earn fees covering only 75–90% of LVR (i.e., negative net returns); in v2, higher fee tiers can reverse this (Fritsch et al., 2024). Delta-hedged replicating portfolios confirm that realized “fee minus LVR” aligns closely with observed LP returns, with residuals vanishing as rebalancing hedges become frequent (Milionis et al., 2022).

Protocol interventions designed to mitigate LVR, as formalized in contemporary research, include:

Dynamic Fee Adjustment: Adjusting AMM fees in response to realized volatility to offset instantaneous LVR (Campbell et al., 11 Aug 2025, Milionis et al., 2022).
Auction-based Arbitrage Right Sales: Selling arbitrage priority and rebating proceeds to LPs strictly curtails LVR, as in Diamond and RediSwap, with empirical LVR reduction factors of up to 99% observed in simulation (McMenamin et al., 2022, Zhang et al., 2024).
Oracle-Assisted Pricing: Directing AMMs to quote external-validated midprices can eliminate “stale” LVR at the cost of introducing oracle risk (Milionis et al., 2022).
Partially Active Reserves: By making only a fraction $\lambda$ of liquidity active per block, PA-AMMs can reduce expected LVR by a $1/(2-\lambda)$ factor, at a cost of increased tracking error (Ko, 10 Feb 2026).

5. LVR in Option-Theoretic, Asset-Management, and Dynamic Contexts

From a foundational perspective, LVR can be interpreted exactly as the time-decay (theta) rate of a portfolio of perpetual American continuous-installment (CI) options replicating the AMM position’s delta profile (Singh et al., 5 Aug 2025). This delivers actionable recipes for constructing liquidity bands with target LVR, calibrating using market-implied volatilities, and provides a bridge to fixed-for-floating fee swaps for quoting implied volatilities and correlations directly from realized AMM LVR (Bichuch et al., 27 Sep 2025).

For dynamic strategies, the LVR metric is extended to pay-as-you-rebalance portfolio management. Deep reinforcement learning approaches optimizing bracket width and rebalancing intervals on Uniswap v3 explicitly optimize “fee minus LVR minus gas,” dominating conventional heuristics, especially when capital constraints or gas costs are significant (Zhang et al., 2023). In portfolio management with significant transaction costs (gas, slippage), rebalancing strategies optimize the trade-off between tracking error and LVR, incorporating uncertainty and volatility as first-class inputs (Kashyap, 2024).

Table: LVR mitigation mechanisms in research AMMs

Mechanism / Protocol	Core Principle	Effect on LVR
Dynamic fee scheduling	Increase fee in volatile regimes	LVR matched by fees (Campbell et al., 11 Aug 2025)
Auctioned arbitrage rights	Rebate MEV/LVR to LPs	LVR $\downarrow$ to $(1-\beta)$ (McMenamin et al., 2022)
PA-AMMs	Activate fraction $\lambda$ only	LVR $\downarrow 1/(2-\lambda)$ (Ko, 10 Feb 2026)
Oracle pricing	External price lookup	LVR $\downarrow$ zero (oracle risk) (Milionis et al., 2022)
Option-theoretic band	CI-put equivalence, band design	Target “flat” LVR (Singh et al., 5 Aug 2025)

6. Comparative Analysis, Critiques, and Extensions

LVR provides a high-fidelity measure for slippage relative to a theoretically perfect, zero-cost, continuous rebalancing benchmark. However, practical evaluation must acknowledge the unrealistic assumptions of zero external trading fees, infinite liquidity, and instantaneous execution. The “Rebalancing-versus-Rebalancing” (RVR) framework refines the LVR metric by benchmarking AMM performance against a realistic centralized exchange rebalancer (with trading commissions, spreads, gas, and market impact), showing that for most parameter configurations, AMM rebalancing is often competitive—even superior—for medium-sized portfolios and frequent rebalancing strategies, except at the lowest available CEX fee tiers (Willetts et al., 2024).

Extensions to multi-asset pools, temporally dynamic market maker weights (TFMMs), and stochastic rebalancing policies demand further care, as direct LVR generalizations can collapse to degenerate or trivial metrics; multidimensional cases require matching weight-vectors rather than token inventories.

7. Practical Implications and Protocol Design Guidance

LVR findings have had direct impact on AMM and DeFi protocol design. For protocol designers:

LVR is the appropriate risk-neutral cost metric for passive AMM market making, isolating adverse selection from directional risk (Milionis et al., 2022).
No AMM can consistently offer positive expected PnL to passive LPs unless fees at least match LVR, as verified by both theory and large-scale empirical measurement (Fritsch et al., 2024).
Batching mechanisms, rapid block production, and in-protocol MEV capture are essential to minimize LVR and thus provide sustainable LP returns (Zhang et al., 2024, McMenamin et al., 2022).
Dynamic, volatility-sensitive fee schedules, and designs that proactively rebalance or throttle liquidity exposure, are prioritized by the empirical and theoretical LVR calculus (Campbell et al., 11 Aug 2025, Ko, 10 Feb 2026).