Loss-Versus-Rebalancing (LVR)
LVR is a way of measuring exactly how much value an automated market maker loses to arbitrageurs compared to a hypothetical trader who rebalanced the same portfolio instantly and for free — it isolates the adverse-selection cost of providing liquidity from ordinary price movement.
Prerequisites: Delta-Neutral Liquidity Provision
An AMM pool holds two assets and prices them by a formula, not by watching the outside market. When the outside price moves, the pool's price doesn't move with it until someone trades against it — and the first trader to show up is almost always an arbitrageur, buying whatever the pool is now underpricing and selling whatever it's overpricing, pocketing the difference. LVR is the metric that quantifies exactly that leak: the value an arbitrageur extracts by trading against a pool that's stale relative to the true market price.
LVR measures the gap between what a liquidity provider's portfolio is worth and what it would be worth if it had been continuously and costlessly rebalanced to track the true market price instead. Every dollar of that gap is a dollar an arbitrageur pocketed by trading against the pool before it caught up.
Why it's a cleaner number than impermanent loss
Impermanent loss compares an LP position to simply holding the two assets — but that comparison mixes together ordinary price risk (which the LP knowingly took on) with the specific cost of being picked off by arbitrageurs. LVR isolates just the second piece: it assumes a benchmark portfolio that rebalances perfectly and instantly with the market, with no trading cost, so any gap left over is purely the toll arbitrageurs extract because the pool can only react to a price change after someone trades against it, never before.
Worked example
An ETH/USDC pool holds $1 million split 50/50 when ETH is $2,000. ETH's true market price jumps to $2,100 on outside exchanges. Before the pool's price adjusts, an arbitrageur buys ETH from the pool at the stale, cheaper price implied by the pool's formula — say the pool ends up selling ETH at an average effective price of $2,050 across the trade, $50 below the new market price of $2,100.
If the arbitrageur buys 5 ETH this way, they capture roughly 5 \times (2100 - 2050) = \250 of value that came directly out of the pool's assets — that \250 (scaled up across every such price move over a day, week, or month) accumulates into the pool's total LVR. A continuously-rebalancing benchmark portfolio, by contrast, would have already been holding fewer ETH by the time the price hit $2,100, and would not have handed that $250 to anyone.
What this means in practice
LVR gives LPs and protocol designers a rate — often expressed as an annualized percentage of pool value — that can be compared directly against the fee income the pool earns, making it possible to ask a concrete question: is this pool's fee tier high enough to compensate for the arbitrage toll it's paying, given the asset's volatility?
LVR is a theoretical benchmark, not literally the cash an LP loses relative to not providing liquidity at all — a pool can have meaningful LVR and still be profitable overall if trading fees from ordinary (non-arbitrage) volume exceed it. The useful comparison is fees earned versus LVR incurred, not LVR versus zero.
Related concepts
Practice in interviews
Further reading
- Milionis, Moallemi, Roughgarden & Zhang (2022), Automated Market Making and Loss-Versus-Rebalancing
- Loesch et al. (2021), Impermanent Loss in Uniswap v3