TL;DR for operators

A liquidity provider should compare AMM performance with an alternative that could actually execute in the external market, including its bid-ask spread. Masaaki Fukasawa shows that, under continuous pure arbitrage and finite-variation inventory dynamics, an AMM and such an inventory-matched rebalancing strategy have identical liquidation wealth. Transaction-cost loss-versus-rebalancing (LVR) is therefore zero.1

That equality does not make liquidity provision costless. A full external inventory hedge still loses the bid-ask spread multiplied by cumulative inventory turnover.

For liquidity providers, the findings support separating performance benchmarking from hedge-cost budgeting. For protocol designers, they also show how retained-fee pool geometry can implement specified portfolio inventory rules. Neither application has been validated through empirical trading or deployed-contract tests in this paper.

The benchmark must pay the same market prices

Suppose a liquidity provider wants to know whether supplying assets to an automated market maker is financially preferable to adjusting those holdings through an ordinary exchange.

The comparison depends on what the alternative strategy would actually pay to trade. An external exchange normally quotes different prices for buying and selling. A strategy that repeatedly adjusts its inventory must absorb that difference.

Traditional loss-versus-rebalancing analysis compares an AMM with a strategy holding the same changing risky inventory but rebalancing at frictionless prices. That construction isolates a particular disadvantage of AMM liquidity provision, yet it may overstate the advantage available to someone trading in a market with transaction costs.

Fukasawa replaces that frictionless alternative with a self-financing strategy that buys at the external ask and sells at the external bid.

Transaction-cost LVR is then defined as:

$$ \operatorname{LVR}^{\mathrm{tc}}_t = V_t^R-V_t^{\mathrm{LP}} $$

Here, $V^R$ is the external rebalancing strategy’s liquidation wealth, while $V^{\mathrm{LP}}$ is the liquidity provider’s liquidation wealth.

The comparison requires the same total risky inventory and matching initial wealth. For pools that collect fees separately, total inventory includes both active reserves and separately held risky-asset fees. For retained-fee pools, fees remain in reserves. These arrangements can produce different inventory paths, but each can be compared with its own matched external strategy.

Theorem 3.2 establishes exact wealth equality for both fee mechanisms under continuous external prices, pure arbitrage, continuous finite-variation inventory, and well-defined reserve dynamics.

The theorem therefore changes the interpretation of LVR. Zero means the pool does not underperform this particular executable alternative. It does not mean the investor has earned a positive return, avoided impermanent loss, or matched a buy-and-hold portfolio.

Why arbitrage produces equal execution cash flows

The equality depends on where arbitrage transactions occur.

An arbitrageur trades between the pool and an external market only when the price difference covers both the pool’s trading fee and external execution costs. Together, these frictions establish a range of internal pool prices within which arbitrage is unprofitable.

Section 2.1.3 expresses this no-arbitrage band as:

$$ \gamma S_t^\ast \leq S_t \leq \frac{S_t^\ast}{\gamma}, \qquad \gamma=(1-\tau)(1-\lambda) $$

Here, $S_t^\ast$ is the external reference price, $S_t$ is the pool’s internal marginal price, $\tau$ is the pool fee, and $\lambda$ measures proportional external transaction costs.

The effective friction is $\tau+\lambda-\tau\lambda$. It determines arbitrage thresholds, not the fee revenue actually collected by the pool.

Under the paper’s pure-arbitrage conditions, inventory changes occur only when the pool reaches a boundary of this band. At those boundaries, the pool’s relevant marginal execution price equals the corresponding external bid or ask.

Consequently, when the liquidity provider’s risky inventory decreases, the pool and its external rebalancing benchmark receive the same marginal sale price. When inventory increases, they pay the same marginal purchase price.

Their cash-flow differences vanish, producing zero transaction-cost LVR.

The inventory condition matters. Finite variation means that cumulative absolute inventory changes remain finite over the interval considered. Positive trading friction supports reflected inventory dynamics in which adjustments occur at arbitrage boundaries rather than following the frictionless rebalancing path.

This also explains a less intuitive result: a positive external bid-ask spread can be sufficient for zero transaction-cost LVR even when the pool charges no fee.

The case where both frictions vanish is mathematically different. Without the positive no-arbitrage band, the familiar frictionless LVR mechanism can reappear through the interaction of reserve-curve curvature and inventory quadratic variation.

Thus, the zero result is an exact consequence of the specified execution mechanism, not a universal property of market making.

Hedging inventory still consumes wealth

A liquidity provider may accept the benchmark equality but still wish to eliminate exposure to changes in the risky asset’s price.

That requires a different strategy: holding an external position equal and opposite to the pool’s risky inventory.

If the pool acquires risky assets, the hedge must sell externally. If the pool releases risky assets, the hedge must buy externally. These transactions occur on the unfavorable sides of the external spread relative to the pool’s boundary executions.

Proposition 4.1 establishes:

$$ \mathrm{d}W_t = -(S_t^a-S_t^b) \,\mathrm{d}\operatorname{TV}(Q)_t $$

In this identity, $W$ is the net wealth of the pool combined with its full external hedge, $S^a-S^b$ is the external bid-ask spread, and $\operatorname{TV}(Q)$ measures cumulative absolute inventory turnover.

The result is precise: under the theorem’s conditions, hedged wealth declines by the prevailing spread integrated against turnover.

Higher inventory turnover increases the accumulated cost when spreads are positive. A wider spread makes each unit of hedging turnover more expensive.

There is no contradiction with zero LVR. The rebalancing benchmark reproduces the liquidity provider’s changing inventory. The hedge instead takes the opposite position to remove risky exposure. They perform different transactions and answer different financial questions.

For a liquidity provider considering a delta-neutral strategy, the relevant budget must therefore include spread-weighted hedge turnover separately from the LVR comparison.

Pool geometry can implement an investment policy

The paper’s next contribution extends beyond performance measurement.

An AMM’s trading function determines how its inventory responds to prices. Fukasawa investigates whether a designer can begin with a desired inventory policy and construct a pool that implements it through arbitrage.

Sections 5.1–5.2 consider a prescribed risky-inventory schedule $q(p)$ that decreases as the internal price $p$ rises.

Under specified differentiability, positivity, monotonicity, and integrability conditions, Proposition 5.1 constructs a convex reserve curve that reproduces the schedule:

$$ f(q(p)) = \eta+\int_0^p q(r)\,\mathrm{d}r-pq(p) $$

Proposition 5.2 then provides a retained-fee utility-indifference pool whose internal price depends only on risky inventory. Fees can accumulate in the numeraire reserve without changing the intended inventory-price relationship.

This is a constructive theoretical result: the designer specifies an admissible inventory rule, and the pool’s trading function encodes it.

Section 5.3 addresses a related but distinct investment problem: maintaining the risky asset’s portfolio weight within prescribed no-trade boundaries.

For a retained-fee geometric-mean pool, the paper derives its reserve-curve exponent and fee from the desired lower and upper risky-weight limits.

The boundaries must already have been determined by the investor’s optimization problem. They must also be sufficiently wide to accommodate the external transaction-cost band. Not every pair of desired boundaries is feasible, and an unchanged constant-product pool cannot represent every such pair.

Under these conditions, arbitrageurs can implement the portfolio’s rebalancing policy by trading against the pool when its boundaries are reached.

This suggests an investment interpretation of AMM design: a fixed trading function can encode an inventory policy, while arbitrage supplies the transactions that implement it.

It does not establish that the resulting strategy is optimal outside the assumed investment setting or that the pool can execute arbitrary time-dependent or path-dependent policies.

Continuous-price equality does not cover jump execution

The central theorem relies on continuous prices and inventory adjustments supported at arbitrage boundaries.

When an external price jumps, a pool can become mispriced beyond its no-arbitrage band. Arbitrage trades may then execute at average prices different from the boundary prices required for the wealth-equality argument.

Section 6 identifies this stale-price exposure as a reason the continuous-price result cannot be applied directly to markets with jumps.

The same section considers continuous liquidity-taker activity away from arbitrage boundaries. Under its specified no-arbitrage inequalities and finite-variation accounting, transaction-cost LVR becomes nonpositive: the pool can outperform its matched rebalancing strategy. This is an analytical extension, not evidence of a realized trading advantage in deployed markets.

Other operational costs remain outside the model, including gas expenditures, smart-contract risks, protocol fee deductions, and liquidity deposits or withdrawals. General retained-fee reserve dynamics also have local existence results unless additional conditions establish global behavior.

The paper supplies mathematical proofs and constructions, not empirical datasets, backtests, or production tests.

What liquidity providers and protocol designers can change

Cognaptus inference: The paper supports three separable decisions.

First, liquidity providers evaluating AMM returns should use an inventory-matched external comparison with realistic bid and ask execution prices. This prevents frictionless execution assumptions from being confused with achievable alternative performance.

Second, liquidity providers intending to hedge should estimate the spread-weighted turnover of their total risky inventory. A favorable LVR comparison cannot replace that cost calculation.

Third, protocol designers considering strategy-oriented pools can examine whether their intended inventory schedules or portfolio-weight boundaries satisfy the construction’s admissibility and feasibility requirements before translating them into trading functions and fee parameters.

These are implications for evaluation and design, not measured improvements in return on capital. Real execution, price discontinuities, contract costs, and strategy-specific risks require separate assessment.

The paper’s central distinction is therefore between relative execution efficiency and economic profitability. An AMM can match what an inventory-equivalent external trader could achieve after transaction costs, while hedging still incurs losses and the underlying investment remains exposed to market risk.

That distinction makes zero transaction-cost LVR a more carefully defined performance result—and a less easily misinterpreted one.


Cognaptus: Automate the Present, Incubate the Future.


  1. Masaaki Fukasawa (2026). Loss-versus-rebalancing with external proportional transaction costs. arXiv:2610.21057. https://arxiv.org/abs/2610.21057 ↩︎