TL;DR
This paper develops a mathematical model to understand how trading fees and discrete block generation times on blockchains affect arbitrage profits in Automated Market Makers (AMMs). The key takeaway is that faster blockchains and lower fees significantly reduce arbitrageurs' profits, ultimately benefiting liquidity providers by mitigating their losses.
What Problem Does This Paper Address?
Automated Market Makers (AMMs) are a cornerstone of decentralized finance, enabling trading without traditional order books. However, liquidity providers (LPs) who supply assets to AMMs face a significant cost: adverse selection. This occurs when arbitrageurs, who have access to real-time market prices, exploit discrepancies between external market prices and the stale prices within the AMM. This exploitation results in losses for LPs, often quantified as "loss-versus-rebalancing" (LVR).
Previous models often simplified this by assuming continuous trading and no trading fees. This paper addresses the crucial gap of incorporating realistic frictions: trading fees (paid to LPs) and discrete block generation times (the times when arbitrageurs can actually execute trades on a blockchain). Without these, previous models might overestimate arbitrage profits and underestimate LP losses.
Methods
The authors build upon existing models of AMMs and arbitrage, introducing two key innovations:
- Discrete Arbitrageur Arrivals: Instead of assuming arbitrageurs can trade continuously, they model arbitrageur arrivals (and thus trade opportunities) as a Poisson process. This realistically captures the discrete nature of block generation on blockchains.
- Trading Fees: They incorporate proportional trading fees, paid by arbitrageurs to LPs, into the model. These fees act as a friction, reducing the profitability of arbitrage.
The core of their methodology involves analyzing a Markovian jump-diffusion process that describes the mispricing (difference between AMM and market prices). They derive closed-form expressions for the probability of profitable trades and, crucially, the instantaneous rate of arbitrage profit. They also analyze asymptotic regimes, particularly the "fast block regime" where block generation is very frequent.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Faster Blockchains = Lower Arbitrage Profits: The most significant insight is that increasing the block generation rate (i.e., making the blockchain faster) directly reduces arbitrage profits. This is because arbitrageurs have less time to exploit price differences before the next block is generated, and the probability of a profitable trade decreases.
- Fees as a "Rescaling of Time": Trading fees effectively "scale down" arbitrage profits. The impact of fees is proportional to the fraction of time where profitable arbitrage opportunities exist, which is reduced by the presence of fees.
- LP Losses are Real and Quantifiable: The paper reinforces that LVR is a fundamental cost for LPs. The model quantifies how this loss is split between arbitrageurs' profits and fees paid to LPs, and in the presence of gas fees, also to validators.
- Gas Fees as an Additional Friction: The introduction of fixed gas fees (transaction costs to include a trade in a block) further reduces arbitrageurs' profitability and shifts losses from LPs to validators.
Key Results
- Closed-form probability of trade: The paper derives a formula for
P_trade, the probability that an arbitrageur can make a profitable trade, which depends on the fee level, volatility, and block rate. - Fast block regime analysis: In the fast block regime (very frequent blocks), arbitrage profits scale with the square root of the mean interblock time ($\Delta t^{1/2}$), the cube of volatility ($\sigma^3$), and inversely with the fee ($\gamma^{-1}$).
- LVR Decomposition: The total loss to LPs (LVR) is shown to be split between arbitrage profits (ARB), trading fees (FEE), and gas fees (GAS). In the fast block regime with small fees, ARB + FEE + GAS = LVR.
- Empirical Validation: The paper references empirical work (Fritsch and Canidio [2024]) that supports their theoretical predictions regarding the relationship between block times and arbitrage profits.
Limitations and Future Work
- Diffusion Price Model: The model assumes asset prices follow a diffusion process (geometric Brownian motion). The authors acknowledge that real-world asset prices can exhibit jumps, which might lead to different arbitrage dynamics, especially at very short time scales. They suggest future work could explore jump processes.
- Constant Gas Fees: While the model incorporates gas fees, it assumes they are fixed per block. Real-world gas fees can be dynamic.
- Stylized AMMs: While the framework is general for locally smooth AMMs, specific results are often derived for constant product market makers (CPMMs). Extending to more complex AMM designs could be valuable.
- Multi-block MEV: The paper touches upon "multi-block MEV" (where an agent controls consecutive blocks to censor trades), suggesting this is an interesting area for future research, particularly for consensus mechanism design.
Summary updated