Every liquidity provider knows the pitch. Deposit two assets, earn trading fees, watch the displayed APY climb. What the dashboard rarely shows is what happens to the position when the market stops drifting and starts lurching.

In quiet weeks, impermanent loss is a rounding error. In violent weeks, it can outweigh every fee collected. The shift is not gradual, because it follows the shape of the pool's pricing curve, and that curve punishes large, fast moves disproportionately.

This article looks at why impermanent loss compounds in high volatility periods, and why the damage is structural rather than a matter of bad timing.

Key Takeaways

  • Impermanent loss scales roughly with the square of the price move for small to moderate moves, so a move twice as large costs about four times as much
  • Volatility drives arbitrage rebalancing, which transfers value out of the pool on every price update, not just at withdrawal
  • Fees scale with volume while the rebalancing cost scales with variance, so fee income often fails to keep pace in turbulent markets
  • Concentrated liquidity amplifies exposure, turning volatility spikes into one-sided positions that stop earning fees

The Common Misunderstanding

Most liquidity providers carry three beliefs. Impermanent loss only exists on paper. It disappears if the price comes back. And trading fees will eventually cover it.

Each belief contains something true. At the moment of withdrawal, the gap between the pool position and simply holding the same two assets depends only on the final price ratio. If the price returns to where it started, that particular gap closes.

But the beliefs are incomplete in a way that matters most during volatile periods. They treat the loss as a single number measured at the end, when the pool is giving up value through every price update along the way. They also treat fees as a steady stream, when fee income and losses arrive on different schedules and respond differently to volatility.

The hidden cost behind headline yield farming APY is exactly this mismatch: a trailing fee number sitting next to a loss that is rarely displayed.

One observation a week on liquidity, flow, and structure. 4 minutes. No price calls.

Subscribe →

What Actually Happens

The pool sells what rises and buys what falls

A constant-product pool does not hold a fixed basket. It keeps the value of its two sides balanced. As one asset rises in price, the pool sells some of it. As it falls, the pool buys more. LP exposure is therefore always tilted toward the asset that is losing ground relative to the other.

The pool cannot see external markets. It only learns the new price when someone trades against it. That someone is almost always an arbitrageur, who takes the underpriced asset from the pool until its price matches other venues. The profit from that trade comes out of the pool.

The loss is nonlinear

The shortfall versus holding follows a curve, not a line. For small moves it grows roughly with the square of the price change. For large moves it flattens in percentage terms but keeps climbing.

Price change of the volatile asset Pool value vs. holding
+25% or -20% about -0.6%
+100% or -50% about -5.7%
+200% or -67% about -13.4%
+400% or -80% about -25.5%

The figures are symmetric: a doubling and a halving produce the same shortfall. Calm markets sit in the first row. Stress sits in the last two.

Volatility is charged along the path

The table describes an endpoint. Researchers studying automated market makers describe a path-dependent cousin called loss versus rebalancing, or LVR. It measures what the pool gives up to arbitrageurs on each update, compared with a trader who simply rebalanced at market prices.

For a constant-product pool, that cost scales with variance, the square of volatility. As approximate figures, expressed as a share of pool value per year:

  • 40% annualized volatility: about 2%
  • 80% annualized volatility: about 8%
  • 120% annualized volatility: about 18%

Doubling volatility quadruples the drag. This is the core of why the cost compounds. Realized volatility tends to catch traders off guard, and an LP pays the realized figure, not the one assumed at deposit. The gap between realized and implied volatility matters here because the pool settles against what actually happened.

A violent round trip shows the effect. Price falls sharply and then recovers. The endpoint view says nothing was lost, since the ratio is back where it started. The path view says the pool traded with arbitrageurs on the way down and again on the way up. Those trades are completed and cannot be undone.

Fees respond differently

Fees are earned on volume. Volume does tend to rise with volatility, which is why displayed APYs often spike in turbulent weeks. But the volume that arrives during sharp moves is disproportionately arbitrage and informed flow, the same flow that creates the loss.

As a rough benchmark, a pool charging a 0.3% fee needs annual volume of about 27 times its liquidity just to offset an 8% LVR drag. That is around 7% of the pool turning over every day. Some pools clear that bar. Many do not, especially long-tail pairs.

The asymmetry is the point. Fees scale with volume. LVR scales with variance. When volatility doubles, the drag quadruples, and volume rarely quadruples with it.

Concentrated liquidity raises the stakes

Concentrated liquidity positions provide capital only within a chosen price range. While price stays inside the range, fee income per dollar is multiplied. Exposure is multiplied by the same factor. A tight range can mean an order of magnitude more sensitivity than a full-range position.

When volatility pushes price out of the range, the position converts entirely into the asset that is losing ground and stops earning fees. The LP holds one side of the pair, having sold the other on the way, with no fee income to show for the move.

Example from Crypto Markets

The numbers below are illustrative, not historical data. Consider an LP who deposits $10,000 into a 50/50 pool of a volatile token and USDC: $5,000 of each.

A quiet-ish week. The token drifts up 25%. Holding the two assets would be worth $11,250. The pool position is worth about $11,180. The gap is roughly $70, or 0.6%. Ordinary weekly fees can plausibly absorb a gap that small, and the LP never notices it.

A stress week. A leveraged altcoin is hit by forced selling, and the token falls 80% in a few days. Liquidation cascades push price down faster than ordinary selling would, and external venues reprice first. Arbitrageurs sell the falling token into the pool and pull out USDC. The pool absorbs the token all the way down.

Holding would now be worth $1,000 plus $5,000 in USDC, or $6,000. The pool position is worth about $4,472. The gap is roughly $1,528, or 25.5%.

To earn that back at a 0.3% fee, the position's share of the pool would need to see about $500,000 of trading volume. Meanwhile the pool's displayed APY has spiked, because volume surged during the crash. The yield figure looks best at the moment the loss is largest.

Now suppose the token bounces 50% from the lows, ending 70% below the original price. The gap does not close. The pool is worth about $5,477 against $6,500 for holding, a shortfall of roughly 16%. A partial recovery only partly repairs the damage, and the trades made during the crash were already settled with arbitrageurs.

What Traders Can Learn

The useful insight is about how the structure behaves, not about what to do with it.

Displayed APY is a trailing number. It rises with volume, and volume rises with volatility. So the headline figure tends to peak when the cost is peaking. Fresh capital often arrives at exactly that point, attracted by a yield that mostly reflects the turbulence it is about to absorb.

The honest benchmark is holding the same assets, not zero. An LP position that is up in dollar terms can still be well behind a simple hold. Measuring against an empty wallet hides the cost that this article describes.

A pool is a passive position with an active mechanism inside it. The case for holding through volatility rather than trying to time it rests on a position that carries no embedded trading cost. A liquidity position does. It is continuously trading against better-informed counterparties, so it behaves less like a hold and more like a standing offer to be picked off when prices move.

Pair structure sets the exposure. Pools of assets that move together, or of two stable assets, keep the price ratio close to one, where the curve is nearly flat. Pools pairing a volatile token with a stable asset sit on the steep part of the curve whenever the market turns.

Volatility clusters. Large moves tend to follow large moves. A single violent session is rarely isolated, which means the compounding is not a one-off event but a regime that can persist while the position stays open.

FAQ

Is impermanent loss really impermanent?

Only if the price ratio returns to its starting level before withdrawal. In that case the endpoint gap versus holding closes. The value given up to arbitrageurs along the way is not recovered, which is why the name understates the cost in volatile markets.

Can trading fees offset impermanent loss?

Sometimes, but not reliably. Fees scale with trading volume, while the rebalancing cost scales with variance. In calm, high-volume pools fees can comfortably exceed the loss. In sharp moves, the loss can arrive within days while fees accrue slowly.

Do stablecoin pools have impermanent loss?

Yes, but it is usually small, because the price ratio stays near one and the curve is almost flat there. The exception is a depeg. If one stablecoin loses its peg, the pool accumulates the weaker asset exactly as a volatile pair would.

What is loss versus rebalancing (LVR)?

LVR measures what a pool loses to arbitrageurs on every price update compared with a strategy that rebalanced at market prices. Unlike impermanent loss, it accumulates along the price path rather than depending only on the final ratio. For constant-product pools it scales with the square of volatility.

Related Concepts

Conclusion

Impermanent loss is not a fixed fee that grows in step with the market. Its cost rises faster than the price moves behind it, and the arbitrage that rebalances a pool charges more for every unit of variance. Fee income, tied to volume, only partly keeps up.

That is why liquidity provision looks most profitable on a dashboard just as its economics are most strained. Understanding the shape of that curve explains what the yield number leaves out. Fees are earned slowly; volatility is charged instantly.