Growth rate of liquidity provider's wealth in G3Ms

✨

This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a landlord renting out a very special, high-tech apartment building. This building is called a Geometric Mean Market Maker (G3M). Instead of normal apartments, it holds two types of "currency" (let's call them Gold and Silver).

In this building, the rules are strict: no matter what happens, the landlord must always keep a specific ratio of Gold to Silver. If you have 60% Gold and 40% Silver, the building's smart contract forces the ratio to stay exactly 60/40. If someone trades and the ratio shifts, the building automatically rebalances itself.

This paper asks a simple but crucial question: Is being the landlord of this building a good way to make money in the long run?

Here is the breakdown of the paper's findings, explained through a story.

1. The Setup: The "No-Go" Zone

Imagine the price of Gold and Silver fluctuates wildly in the outside world (the "External Market"). Inside your building, the price is determined by how much Gold and Silver you have.

Because the outside market moves faster than your building can adjust, a gap often appears. Sometimes Gold is cheaper inside your building than outside; sometimes it's more expensive.

Enter the Arbitrageurs. These are like professional "price hunters."

If Gold is cheap inside your building, they buy it from you and sell it outside for a profit.
If Gold is expensive inside, they buy it outside and sell it to you.

The Paper's Big Insight: The authors realized that these price hunters are actually the engine that keeps your building profitable. Without them, your building would just sit there. With them, they constantly push the price back to the "fair" level, and in doing so, they pay you a toll.

2. The Toll Booth: Transaction Fees

Every time an arbitrageur trades with your building, they have to pay a small fee (a "toll").

The Bad News: When they trade, they take advantage of the price difference. This is called "adverse selection." It feels like they are stealing value from you.
The Good News: They also pay you a fee for every trade.

The paper treats this like a toll booth on a busy highway. Even if the cars (trades) are driving fast and causing some wear and tear (adverse selection), the tolls they pay (fees) add up. The question is: Do the tolls cover the wear and tear, and then some?

3. The "Mispricing" Meter

The authors invented a clever way to measure the chaos. They imagine a meter called the "Mispricing Process."

Think of this meter as a ball bouncing inside a box.
The box has walls at the top and bottom. These walls represent the No-Arbitrage Bounds.
As long as the price difference is small, the ball bounces freely inside the box.
But the moment the ball hits a wall (meaning the price difference is too big for the arbitrageurs to ignore), the wall pushes it back.

The Magic of the Walls:
When the ball hits the wall, it's because an arbitrageur just made a trade. That trade generates a fee for you. The authors used complex math (called "reflected diffusion") to prove that the more the ball hits the walls, the more money you make in fees.

4. The Surprising Result: Beating the "Set It and Forget It" Strategy

Usually, if you want to invest in Gold and Silver, you might just buy them and hold them (Buy and Hold), or you might rebalance them periodically (Constant Rebalanced Portfolio).

The paper proves something amazing: If you set the fee (the toll) just right, your building (G3M) will actually grow wealth faster than a standard investment portfolio.

How?

In a normal portfolio, you just sit there.
In your building, the constant "bouncing" of the price against the walls (volatility) generates a stream of fees.
The authors show that these fees act like a "volatility harvest." You are essentially getting paid to let the market swing back and forth.

5. The "Goldilocks" Fee

The paper does a lot of math to find the Perfect Fee.

Too Low: The arbitrageurs trade too much, and the fees don't cover the "wear and tear" (losses). You lose money.
Too High: The arbitrageurs stop trading because the toll is too expensive. The ball stops bouncing. You get no fees.
Just Right: There is a "sweet spot" (a specific fee percentage) where the bouncing is frequent enough to generate massive fees, but not so frequent that the losses eat them all up.

Interestingly, this sweet spot changes depending on how much Gold vs. Silver you hold. It's not a one-size-fits-all number.

The Bottom Line

This paper is a mathematical proof that Automated Market Makers (like Uniswap or Balancer) are not just passive storage for crypto assets; they are active, high-speed trading machines.

By carefully setting the transaction fees, a Liquidity Provider (the landlord) can turn the chaotic, noisy market swings into a steady, long-term profit stream that beats traditional, passive investing strategies. The "losses" from price swings are transformed into "gains" through the fee mechanism.

In short: The paper tells us that if you are smart about your fees, you can turn the chaos of the crypto market into a money-printing machine that outperforms a boring savings account.

1. Problem Statement

The paper investigates the long-term profitability of Liquidity Providers (LPs) in Geometric Mean Market Makers (G3Ms), a class of Automated Market Makers (AMMs) that includes protocols like Uniswap (Constant Product) and Balancer (Constant Weighted Geometric Mean).

The core problem is to determine whether the fee revenue generated by G3Ms can offset the losses caused by adverse selection (specifically, "Loss-Versus-Rebalancing" or LVR) due to continuous arbitrage. The authors aim to answer:

Can a G3M, through its fee mechanism, generate long-term wealth growth that matches or exceeds traditional passive investment strategies (like constant-rebalanced portfolios)?
How do transaction fees, market volatility, and portfolio weights interact to determine the optimal fee tier for maximizing LP returns?

2. Methodology

The authors develop a rigorous analytical framework based on stochastic reflected diffusion processes.

Key Assumptions

To isolate the arbitrage channel and create a tractable model, the paper adopts three simplifying assumptions:

No Noise Traders: All trades are executed by rational arbitrageurs. This represents a "worst-case" scenario for LPs, as there is no uninformed order flow to subsidize losses.
Frictionless External Reference Market: There exists an external market with infinite liquidity and zero transaction costs, serving as the true price reference.
Continuous Arbitrage: Arbitrageurs monitor and act on price discrepancies instantaneously. This is justified by modern blockchain infrastructure (e.g., Solana, Arbitrum) where block times are short enough to approximate continuous time.

Mathematical Framework

Mispricing Process ( $Z_t$ ): The authors define the mispricing as the log-ratio between the external market price ( $S_t$ ) and the G3M pool price ( $P_t$ ): $Z_t = \ln(S_t/P_t)$ .
No-Arbitrage Bounds: Due to transaction fees ( $\gamma$ ), arbitrage is only profitable when $Z_t$ exits the interval $[\ln \gamma, -\ln \gamma]$ .
Reflected Diffusion: The dynamics of $Z_t$ are modeled as a diffusion process confined to the no-arbitrage interval. When $Z_t$ hits the boundaries, it is "reflected" back into the interval by the action of arbitrageurs. This reflection is mathematically encoded by boundary local times ( $L_t$ and $U_t$ ), representing the cumulative volume of arbitrage trades.
Wealth Decomposition: The LP's wealth ( $V_t$ $V_{t}$ ) is decomposed into:
1. The growth of the underlying assets (market drift).
2. The accumulation of liquidity ( $\ell_t$ ) driven by fees collected during arbitrage (captured by $L_t$ and $U_t$ ).
3. A bounded error term related to the mispricing.

Analytical Tools

Feynman-Kac Formulas: Used to connect the expected growth rates to solutions of parabolic Partial Differential Equations (PDEs) with Neumann boundary conditions.
Sturm-Liouville Theory: Applied to the infinitesimal generator of the reflected diffusion to derive eigensystem expansions for time-homogeneous cases.
Mimicking Processes: Used to handle time-inhomogeneous and stochastic volatility cases by constructing a process with the same marginal distribution but simpler dynamics.

3. Key Contributions

The paper makes three primary contributions to the literature on DeFi and AMMs:

Explicit Growth Rate Formulas: The authors derive closed-form expressions for the long-term expected logarithmic growth rate of LP wealth under various market conditions (time-homogeneous, time-inhomogeneous, and stochastic volatility).
Optimal Fee Analysis: They characterize the optimal fee tier ( $\gamma^*$ ) that maximizes LP wealth growth. They reveal non-monotone phase transitions, showing that the optimal fee is not always at the extremes (0 or 1) but can lie in the interior of the interval $(0, 1)$ , depending on pool weights and market drift.
Connection to Stochastic Portfolio Theory (SPT): The paper demonstrates that G3Ms can function as superior on-chain index funds. It proves that under appropriate fee structures, G3Ms can outperform the excess growth rate of a constant-rebalanced portfolio (a standard benchmark in SPT), effectively transforming short-term arbitrage losses into long-term "volatility harvesting."

4. Key Results

The Growth Rate Formula

For a time-homogeneous market, the long-term expected logarithmic growth rate of LP wealth is given by:
$\lim_{T \to \infty} \frac{E[\ln V_T]}{T} = w\mu_X + (1-w)\mu_Y + \text{Fee Term}$
Where:

$w\mu_X + (1-w)\mu_Y$ is the growth rate of a constant-weight portfolio.
The Fee Term is a function of the fee parameter $\gamma$ , the pool weights $w$ , and the ergodic properties of the mispricing process (specifically the speed measure and boundary local times).

The Role of Fees

Dual Effect: Fees create a "no-arbitrage band" that prevents the pool price from deviating too far from the market price. While this reduces the frequency of trades (potentially lowering fee income), it also reduces the magnitude of adverse selection losses.
Volatility Harvesting: The paper shows that the fee mechanism allows LPs to capture the "excess growth rate" associated with volatility. In the limit as fees approach zero ( $\gamma \to 1$ ), the G3M growth rate converges to the SPT excess growth rate ( $w(1-w)\sigma^2/2$ ).
Optimal Fee Tier: Numerical analysis reveals that for certain weightings ( $w \neq 0.5$ ) and market drifts, the optimal fee is strictly interior. This implies that setting fees too high or too low can be suboptimal.

Comparison with Benchmarks

vs. Buy-and-Hold: G3Ms generally outperform buy-and-hold strategies in volatile markets due to the rebalancing effect.
vs. Constant Rebalanced Portfolio (CRP): The study shows that G3Ms can outperform a frictionless CRP. This is counter-intuitive because CRPs assume frictionless trading, whereas G3Ms have fees. The result suggests that the fee mechanism in a G3M effectively acts as a "volatility harvesting" engine that compensates for the friction, provided the fee tier is optimized.

5. Significance and Implications

Theoretical Validation: The paper provides a rigorous mathematical justification for G3Ms as viable infrastructure for on-chain index funds. It moves beyond heuristic arguments to provide explicit formulas for wealth growth.
Protocol Design: The findings offer actionable insights for protocol designers (e.g., Uniswap, Balancer). The identification of non-monotone optimal fees suggests that a "one-size-fits-all" fee structure may be inefficient. Protocols might benefit from dynamic fee tiers or specific fee structures tailored to asset weights and market volatility.
LP Strategy: For liquidity providers, the paper clarifies that LPing is not merely a passive income stream but an active strategy that harvests volatility. It highlights that even in a worst-case scenario (only adversarial arbitrage), LPs can achieve positive excess returns over traditional portfolios.
Future Research: The authors outline extensions to multi-asset pools ( $n > 2$ ) and the inclusion of noise traders, noting that the structural dynamics of reflected diffusions extend naturally, though the ergodic theory for higher dimensions remains a challenge.

In summary, this paper bridges the gap between stochastic portfolio theory and decentralized finance mechanics, proving that the interplay between transaction fees and continuous arbitrage can create a robust, long-term wealth generation mechanism for liquidity providers in G3Ms.