Impermanent loss, explained properly

It is not a fee. Nothing is deducted from your account. Nobody takes it from you. It is the difference between two outcomes, and it only becomes real when you walk away — which is exactly why so many people discover it too late.

The one-paragraph version

You deposit two tokens into a pool. One of them goes up more than the other. The pool automatically sells some of the winner and buys more of the loser, over and over, as the price moves. When you withdraw you hold less of the token that went up and more of the token that did not. Impermanent loss is the value of that difference — what you would have had if you had simply held the two tokens, minus what the pool actually hands back.

Why a pool does this at all

An automated market maker has no opinion about price. It holds a fixed mathematical relationship between the two token balances and lets traders move along it. When the market price of a token rises, traders buy that token out of the pool until its pool price matches the market. The pool has no choice in the matter — it is the counterparty to every one of those trades, always taking the losing side of the move.

That is not a bug. It is the service being sold. Traders get liquidity; the pool gets fees for standing there. The entire question of whether providing liquidity is worthwhile comes down to whether those fees exceed the cost of always being on the wrong side.

The actual numbers

For a standard full-range position, the loss depends only on how much the two tokens moved relative to each other. Not on direction, not on volatility along the way — just the ratio between where they started and where they ended.

Read that last row carefully, because it is the one that catches people. If a token 10×s against its pair, you do not lose 42.5% of your money — your position is still worth far more than when you started. What you lost is 42.5% relative to having just held the two tokens. You made money and still badly underperformed doing nothing. That distinction is the entire subject.

Why "impermanent" is a misleading word

The name comes from a real property: if prices return exactly to the ratio they had when you deposited, the loss vanishes completely. Nothing is permanently destroyed while you remain in the pool.

In practice this offers less comfort than it sounds. Prices return to old ratios when a move was noise. When the market has genuinely repriced one asset against another — a protocol grew, a token launched, a narrative broke — that ratio is not coming back, and waiting for it costs you every day you spend waiting. The loss is impermanent in the way that a paper loss on a stock is impermanent.

Concentrated liquidity sharpens both edges

Uniswap v3 introduced ranges, and nearly every major DEX has copied the idea since. Instead of spreading your deposit across every conceivable price, you concentrate it into a band. Inside the band you earn dramatically more fees per dollar deposited. Outside it you earn nothing.

The catch is what "outside" means. When price leaves your band, your position does not just stop earning — it has already converted entirely into the token that fell. You are fully exposed to the loser and hold none of the winner. A narrow range is a leveraged bet that price stays put, in both directions.

This is why the single most useful figure for a concentrated position is not APR but the share of days it actually spent in range. A pool advertising 80% APR that was in range 30% of the time did not pay 80%.

Re-centring is not a free fix

The obvious response is to move the range whenever price escapes it. That does keep you earning, and it costs on both sides: gas for every adjustment, and the loss on the token that ran away, which you crystallise at the worst possible moment by selling it to reopen. Chasing a trending market can quietly bleed more than sitting out of range would have.

When fees actually win

Providing liquidity beats holding under conditions that are genuinely identifiable in advance:

• Correlated pairs. Two tokens that move together barely change ratio, so there is little loss to cover. Stablecoin pairs and liquid staking pairs are the clearest cases.
• Range-bound markets. Sideways chop with heavy volume is close to ideal: fees accumulate while the ratio ends near where it began.
• Volume relative to liquidity. Fees are shared out by pool share. A busy pool with modest total liquidity pays each provider far more than a quiet pool holding a fortune.
• Short horizons in a known regime. The longer you stay, the more chances the ratio has to move against you.

And it reliably loses when one token strongly outperforms the other. If you have real conviction about a token, holding it outright will usually beat pooling it against something else — that conviction is precisely what the pool will trade away on your behalf.

Working it out for a real position

Every figure above is a general case. What matters for a decision is what a specific pool did over a specific period with a specific range — fees earned against loss incurred, on real volume and real prices.

Nothing here is financial advice. Providing liquidity carries a real risk of loss, including risks this page does not cover, such as smart contract failure and token-specific risk.