Impermanent loss is the money a liquidity provider gives up by putting two tokens into a pool instead of just holding them. The name is the first problem. It suggests the loss floats away on its own if you wait long enough. Most of the time it does not.
This guide starts from zero. What a pool is, why depositing into one can leave you poorer than the person who did nothing at all, and why a study from years ago still describes the trap better than most of the marketing does. The math turns out to be simple once you see where the tokens actually go.
A pool is two piles of tokens and one rule
Most decentralized exchanges do not match a buyer with a seller. They hold reserves. A pool for ETH and USDC keeps a pile of each, and one rule governs every trade that touches it: the two piles multiplied together have to stay the same number. Traders call this the constant product formula, written as x times y equals k.
When you add liquidity, you deposit both tokens in equal dollar value and receive a share of the pool. From then on you earn a cut of every trading fee the pool collects. That fee income is the entire reason anyone deposits.
These pools sit at the center of DeFi, and they set prices without asking anyone. There is no order book. There is no price oracle feeding them a figure from outside. The ratio between the two piles is the price, and trades move that ratio.
Why the pool hands you the losing side
Picture ETH at 2,000 dollars the moment you deposit. Then ETH climbs to 4,000 on every other exchange. For a brief window your pool still prices ETH near 2,000, which makes it the cheapest ETH in the market. Arbitrage traders notice within seconds. They buy ETH from your pool until its price lines up with everywhere else.
Every one of those trades pulls ETH out of the pool and pushes USDC in. When it settles, your share holds less ETH and more USDC than you started with. You end up heavy in the token that lagged and light in the token that ran. Set that against the person who kept the same ETH and USDC untouched in a wallet, and the gap between the two positions is impermanent loss.
The direction does not save you either. Had ETH dropped to 1,000 instead, arbitrage would have run the other way, feeding cheap USDC in and draining ETH out until your pool matched the market. You would finish holding more of the token that fell. The pool always rebalances toward whatever is underperforming, which is exactly the opposite of what a holder wants. It sells your winners and buys your losers on autopilot, a little at a time, with no say from you. That mechanical rebalancing is the whole source of the loss, and it runs whether the market goes up or down.
How big the gap gets, in one table
The loss depends only on how far the price moves, never on the direction. A token that doubles and a token that halves produce the identical figure.
| Price change of one token | Impermanent loss versus holding |
|---|---|
| 1.25x | 0.6% |
| 1.5x | 2.0% |
| 2x | 5.7% |
| 3x | 13.4% |
| 4x | 20.0% |
| 5x | 25.5% |
Read the 2x row. If one token doubles against the other, a provider finishes about 5.7 percent behind a plain holder, and that is before any fees are counted. At five times the gap passes a quarter of the position. A pair that barely moves, such as two dollar stablecoins, lives up near the top of the table where the loss rounds down to almost nothing.
The curve behind those numbers is IL equals two times the square root of the price ratio, divided by one plus that ratio, minus one. You will never need to run it by hand. What the formula reveals is worth keeping in mind, though. The loss accelerates. Doubling the price costs 5.7 percent, but a fivefold move does not cost five times that. It costs more than four times as much, because the gap widens faster the further the pair drifts apart. The table is the part that matters when you are standing at the counter deciding whether to deposit.
Impermanent is the wrong word for it
The word carries a promise inside it. Impermanent says the loss reverses if you are patient.
It can. If the price ratio drifts all the way back to exactly where you deposited, the gap closes and you walk away with the fees on top. That single possibility is the whole reason it gets called impermanent.
The problem is timing. The loss only stays on paper while your tokens sit inside the pool. The instant you withdraw with prices still apart, it turns real. It crystallizes, in the language of the people who model this for a living. Markets almost never wander back to a precise past ratio on the schedule of someone who wants their money out, so for most providers the loss is not impermanent in any useful sense. It is deferred until the day they exit.
Fees are meant to cover it, and often do not
Providers accept impermanent loss because trading fees are meant to more than pay it back. Whether that actually happens is a measurable question, and someone went and measured it.
In late 2021 the advisory firm Topaz Blue and the team behind Bancor studied Uniswap v3, then the largest venue of its kind. They tracked 17 pools that together held about 43 percent of the protocol's value, running from early May to late September. Those pools earned 199.3 million dollars in fees across the period. Impermanent loss over the same window reached 260.1 million. On the whole, providers came out roughly 60.8 million dollars worse off than if they had held their tokens and skipped the pools.
Inside that group, 49.5 percent of providers were sitting on negative returns once impermanent loss was counted against them. In more than 80 percent of the pools studied, the loss outran the fees. Only a short list, including WBTC paired with USDC, cleared a genuine profit.
The study is old by the clock crypto runs on, and fee dynamics have shifted since. The lesson has not moved. A tall advertised yield on a volatile pair is a headline number. What lands in your wallet is that yield minus a loss the headline forgets to print. It is the same quiet drag that funding rates put on margin traders, a cost that only shows itself when you add up the account at the end.
Ways providers limit the damage
No switch turns impermanent loss off inside a standard pool. Providers manage the exposure instead, and a few approaches come up again and again:
- Pairs that move together. Two stablecoins, or a token and its staked twin, barely diverge, so the loss stays tiny. The cost is thinner fees.
- Fee income large enough to swamp the gap. High volume pools can pay so much that even real impermanent loss still leaves the provider ahead, which is how the WBTC and USDC style pool came out positive in the study.
- Shorter exposure. Some providers only add liquidity around stretches they expect to stay calm, then pull out before a large move can build.
- Yield that does not ride on a trading pair. Passive income such as staking a proof of stake asset sidesteps the mechanism entirely, while carrying its own lockups and risks.
Frequently asked questions
Does impermanent loss happen with two stablecoins?
Barely, as long as both stay near a dollar. The ratio hardly moves, so the loss rounds toward zero and the fees keep flowing. A depeg is the exception, and a sharp one flips the pool against you in a hurry.
Is it a real loss if I never withdraw?
Not yet. While your tokens stay in the pool it lives on paper, sitting next to the holder you are being measured against. It only becomes real the moment you exit with prices still diverged from where you came in.
Can fees fully cancel it out?
They can, and in busy pools they regularly do. Whether they will for your position comes down to how far the pair moves against how much it trades. The Uniswap study above is the standing reminder that the answer, across a lot of pools, was no.
Before adding to any pool, set the advertised yield next to the table above and estimate how far the pair might travel while your money is locked inside. That comparison, rather than the yield on its own, is what tells a provider whether the pool is worth walking into.