2.4 — LP Profitability

LP Profitability — Fees vs Impermanent Loss

A rational liquidity provider cares about one thing: net P&L. The LP earns fee income from every swap and loses to impermanent loss as the price moves. Profitability reduces to a race between these two forces.

$$\text{LP P\&L}=\text{Fee Income}-\text{Impermanent Loss}$$

This article builds a simple model of each component and derives the break-even condition.

Fee Income Model

Per-Swap Fees

Each swap through the pool pays a fee $\gamma$ (e.g., 0.30% for Uniswap V2, 0.25% for PumpSwap). If a swap has notional size $V_i$ (measured in the output token), the fee is $\gamma V_i$.

Over a period with $N$ swaps, total fees collected by the pool are:

$$F=\gamma \sum _{i=1}^N{V}_i=\gamma \cdot V_{\text{total}}$$

where $V_{\text{total}}$ is total trading volume through the pool.

Fee Yield for a Single LP

An LP owning fraction $w$ of the pool earns:

$$F_{\text{LP}}=w\cdot \gamma \cdot V_{\text{total}}$$

Expressed as a return on the LP’s capital $V_{\text{LP}}$:

$$\text{Fee yield}=\frac{F_{\text{LP}}}{V_{\text{LP}}}=\frac{w\cdot \gamma \cdot V_{\text{total}}}{w\cdot V_{\text{pool}}}=\frac{\gamma \cdot V_{\text{total}}}{V_{\text{pool}}}$$

The $w$ cancels — fee yield depends only on the volume-to-liquidity ratio $V_{\text{total}}/V_{\text{pool}}$, often called pool utilization or turnover.

Annualized Fee APR

If daily volume is $V_d$ and pool TVL is $L$:

$$\text{Fee APR}=\frac{365\cdot \gamma \cdot V_d}{L}$$

For a pool with $\gamma =0.003$, V_d = \10\text{M}$,$L = $50\text{M}$:

$$\text{Fee APR}=\frac{365\times 0.003\times 10}{50}=21.9\%$$

This is the gross yield before IL.

Impermanent Loss as a Cost Rate

From impermanent-loss, IL over a period where price moves by ratio $r=P_1/P_0$ is:

$$\text{IL}(r)=\frac{2\sqrt{r}}{r+1}-1$$

For small log-price moves ${\sigma }_{\Delta t}$ over a short interval $\Delta t$:

$$\text{IL}\approx -\frac{{\sigma }_{\Delta t}^2}{8}$$

where ${\sigma }_{\Delta t}^2$ is the realized variance of the log-price over that interval.

Over a longer horizon with annualized volatility $\sigma$:

$$\text{IL per year}\approx -\frac{{\sigma }^2}{8}$$

(This uses the property that log-price variance scales linearly with time under a diffusion model.)

Break-Even Condition

The LP breaks even when fee income equals IL:

$$\frac{\gamma \cdot V_d\cdot 365}{L}=\frac{{\sigma }^2}{8}$$

Solving for the maximum tolerable volatility:

$$\boxed{{\sigma }_{\text{break-even}}=\sqrt{\frac{8\cdot 365\cdot \gamma \cdot V_d}{L}}}$$

Or equivalently, the minimum volume needed to offset a given volatility:

$$V_{d,\text{min}}=\frac{{\sigma }^{2}L}{8\cdot 365\cdot \gamma }$$

Numerical Example

Pool parameters: $\gamma =0.003$, V_d = \10\text{M}$,$L = $50\text{M}$.

$${\sigma }_{\text{break-even}}=\sqrt{\frac{8\times 365\times 0.003\times 10}{50}}=\sqrt{1.752}\approx 1.32=132\%$$

This pool can tolerate annualized volatility up to 132% before IL exceeds fee income. That is quite high — ETH/USDC realized vol is typically 60—100%, so this pool would likely be profitable.

Now consider a thin pool: V_d = \100\text{K}$,$L = $5\text{M}$:

$${\sigma }_{\text{break-even}}=\sqrt{\frac{8\times 365\times 0.003\times 0.1}{5}}=\sqrt{0.01752}\approx 0.132=13.2\%$$

This pool can only tolerate 13.2% annual vol — far below typical crypto volatility. LPs here are almost certainly losing money.

The Profitability Quadrant

LP profitability maps onto a two-dimensional space:

	High volume / TVL	Low volume / TVL
Low volatility	Very profitable	Marginally profitable
High volatility	Depends on ratio	Unprofitable

The best LP positions are in high-turnover, low-volatility pools: stablecoin pairs (USDC/USDT), or major pairs during calm markets.

The worst are low-turnover, high-volatility pools: long-tail meme tokens with thin liquidity and violent price swings. These are exactly the tokens that attract retail LPs with headline APR numbers computed from a single day’s volume.

Dynamic Considerations

The simple model above assumes constant $V_d$, $L$, and $\sigma$. Reality is messier:

Volume chases volatility

When price moves sharply, arbitrage volume spikes. This creates a partial natural hedge: the same volatility that causes IL also generates fee income. The question is whether the hedge ratio is sufficient — empirically, for most pools, it is not. Arbitrage volume covers only a fraction of the IL.

LP capital is mobile

When a pool becomes unprofitable, LPs withdraw, shrinking $L$. This increases the fee yield ($V_d/L$ rises) and increases price impact (which may reduce arbitrage volume). The pool finds a new equilibrium with less liquidity and higher yield. This is analogous to market makers widening spreads in volatile conditions in TradFi — see market-microstructure.

Fee tiers segment the market

Uniswap V3 offers multiple fee tiers (1 bps, 5 bps, 30 bps, 100 bps). Pairs naturally sort by volatility: stablecoins at 1 bps, blue-chip pairs at 5—30 bps, long-tail at 100 bps. This is the AMM analogue of tick-size tiering in equity markets.

The Market-Maker Analogy

The LP’s P&L equation is structurally identical to a traditional market maker’s:

TradFi market maker	AMM liquidity provider
Earns the bid-ask spread	Earns swap fees
Loses to informed flow (adverse selection)	Loses to arbitrageurs (IL)
Profits when vol is low, flow is retail	Profits when vol is low, volume is high
Widens spreads in volatile markets	Fee tier selection; LP withdrawal

The key difference: a TradFi market maker can choose which orders to fill and dynamically adjust quotes. An AMM LP is passive — the pricing function is fixed, and the LP cannot discriminate between informed and uninformed flow. This structural disadvantage is why IL is such a persistent drag.

---

Companion notebook: notebooks/defi/04-lp-profitability.py — Altair heatmap of net LP return as a function of $(\sigma ,V_d/L)$; break-even frontier; historical backtest on Uniswap V2 ETH/USDC pool data.

Questions to sit with:

1. The break-even formula gives ${\sigma }_{\text{break-even}}\propto \sqrt{\gamma V_d/L}$. All else equal, doubling the fee rate increases tolerable volatility by only $\sqrt{2}\approx 41\%$. Why does the square root appear — what is the deeper reason?
2. If arbitrage volume partially offsets IL, could you model the net effect as a “fee-adjusted volatility” ${\sigma }_{\text{eff}}<\sigma$? What data would you need to estimate ${\sigma }_{\text{eff}}$?
3. An LP in a Uniswap V3 concentrated-liquidity position earns higher fees per dollar of capital but faces higher IL (the position is more leveraged). Does concentration change the break-even condition, or does it cancel out?