Automated Market Making and Loss-Versus-Rebalancing

Jason Milionis, Ciamac C. Moallemi, Tim Roughgarden, Anthony Lee ZhangarXiv 2022 · DOI · arXiv

Abstract

We consider the market microstructure of automated market makers (AMMs) from the perspective of liquidity providers (LPs). Our central contribution is a "Black-Scholes formula for AMMs." We identify the main adverse selection cost incurred by LPs, which we call "loss-versus-rebalancing" (LVR, pronounced "lever"). LVR captures costs incurred by AMM LPs due to stale prices that are picked off by better-informed arbitrageurs. We derive closed-form expressions for LVR applicable to all automated market makers. Our model is quantitatively realistic, matching actual LP returns empirically, and shows how CFMM protocols can be redesigned to reduce or eliminate LVR.

Introduction: The Problem with Impermanent Loss

Providing liquidity in decentralized finance (DeFi) is a high-stakes endeavor. While liquidity providers (LPs) earn trading fees, they are exposed to significant market risks. Historically, the industry has relied on "Impermanent Loss" (IL) as the primary metric for assessing these risks. IL compares the value of an LP's current portfolio to a "buy-and-hold" strategy of the original assets.

However, IL is a flawed benchmark for two reasons. First, it is path-independent; it only considers the price at the start and end of a period, ignoring the volatility that occurred in between. Second, it conflates general market risk with the specific costs of the AMM mechanism. To truly understand the profitability of being an LP, we must isolate the cost of the AMM's passive nature—specifically, the cost of providing "stale" quotes to the market. This paper introduces Loss-Versus-Rebalancing (LVR) to address this gap.

Defining Loss-Versus-Rebalancing (LVR)

LVR is the fundamental adverse selection cost incurred by LPs. Because AMMs are passive and do not incorporate external price feeds, they rely on arbitrageurs to move their internal price to match the global market. These arbitrageurs essentially "snipe" the LP, buying assets from the pool at a discount or selling them at a premium relative to the true market price.

To quantify this, the authors define a benchmark Rebalancing Strategy. This hypothetical strategy holds the exact same amount of risky assets as the AMM at any given time but executes its trades at fair market prices (e.g., on a high-volume centralized exchange like Binance). LVR is defined as the performance gap between this rebalancing strategy and the actual AMM LP position. Unlike IL, LVR is a running cost: every price movement that triggers an arbitrage trade results in a permanent, non-reversing loss for the LP.

The "Black-Scholes Formula" for AMMs

Using a continuous-time finance framework, the authors derive a closed-form expression for instantaneous LVR:

$$\ell(\sigma, P) = \frac{\sigma^2 P^2}{2} |x^{*'}(P)|$$

In this formula, $\sigma$ represents price volatility, $P$ is the current price, and $|x^{*'}(P)|$ represents the marginal liquidity (the slope of the AMM's demand curve).

This derivation leads to a powerful heuristic for the most common AMM type, the constant-product market maker (e.g., Uniswap v2). For these pools, the daily LVR is approximately 1/8th of the daily variance of the asset price. For instance, if an asset has a daily volatility of 5%, the pool loses roughly 3.125 basis points of its total value to arbitrageurs every single day.

Empirical Validation

The model was tested using a year of data (August 2021 – July 2022) from the Uniswap v2 WETH-USDC pair. By combining on-chain data from Dune Analytics with minute-level price data from the Binance API, the authors compared theoretical LVR predictions with actual market outcomes.

The results were striking:

  1. Model Fit: The theoretical LVR formula matched the empirical "hedged" returns of LPs almost perfectly.
  2. Risk Isolation: By "delta-hedging" an LP position (going long the LP and shorting the rebalancing strategy), the authors reduced the standard deviation of returns by a factor of 20. This demonstrates that the vast majority of LP return volatility is simply unhedged market exposure, while LVR represents the true cost of the AMM mechanism.

Implications for Mechanism Design

LVR provides a roadmap for the next generation of DeFi protocols. To remain competitive and sustainable for LPs, AMMs must find ways to mitigate this adverse selection. Potential solutions include:

  • High-Frequency Oracles: Updating internal prices faster to reduce the window for arbitrage.
  • MEV Capture: Auctioning off the right to be the first trader in a block and returning that value to LPs.
  • Dynamic Fees: Adjusting trading fees based on realized volatility to compensate LPs during turbulent periods.

Conclusion

Loss-Versus-Rebalancing shifts the paradigm of AMM analysis from path-independent snapshots to a continuous understanding of adverse selection. By isolating the cost of stale quotes, LVR allows LPs to treat their positions as a "pure bet" on whether trading fees will exceed the mathematical cost of volatility. As the DeFi ecosystem matures, LVR will likely become the standard metric for evaluating the efficiency and sustainability of liquidity provision.

Summary updated

Abstract

We consider the market microstructure of automated market makers (AMMs) from the perspective of liquidity providers (LPs). Our central contribution is a ``Black-Scholes formula for AMMs''. We identify the main adverse selection cost incurred by LPs, which we call ``loss-versus-rebalancing'' (LVR, pronounced ``lever''). LVR captures costs incurred by AMM LPs due to stale prices that are picked off by better informed arbitrageurs. We derive closed-form expressions for LVR applicable to all automated market makers. Our model is quantitatively realistic, matching actual LP returns empirically, and shows how CFMM protocols can be redesigned to reduce or eliminate LVR.
BibTeX
@misc{milionis2022automated,
  author = {Jason Milionis and Ciamac C. Moallemi and Tim Roughgarden and Anthony Lee Zhang},
  title = {{Automated Market Making and Loss-Versus-Rebalancing}},
  year = {2022},
  eprint = {2208.06046},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2208.06046},
}
Figure