What impermanent loss actually is
Impermanent loss (often shortened to IL) is the difference between the dollar value of assets you would have today if you had simply held them in your wallet, and the dollar value of your share of a liquidity pool after the same price move. It's not a fee taken by a protocol, and it's not a hack. It's a mechanical consequence of the way automated market makers like Uniswap, Curve, and Balancer rebalance your position as prices change.
To make that concrete, imagine you put $1,000 into a 50/50 ETH/USDC pool at a price of $2,000 per ETH. The pool mints you LP shares worth $500 of ETH and $500 of USDC. If ETH doubles to $4,000 and you had just held, you would have $1,500 in ETH and $500 in USDC, for $2,000 total. As an LP, the AMM has been selling your ETH for USDC the whole way up to keep the pool balanced. Your share is now worth less than $2,000. The gap between $2,000 and your actual pool value is the impermanent loss. It is called "impermanent" only because if ETH fell back to $2,000, the loss would close. In practice, prices rarely round-trip cleanly, and realized IL is simply the loss.
This is why IL matters for a depositor: it is a silent drag on returns that can outweigh years of fee income if the price moves against your assumption. The rest of this article gives you a checklist to size that drag against the fees the pool is actually paying before you commit capital.
The risk you should size first: IL versus fees
The honest framing is that the only question that matters is whether the fees you earn will exceed the impermanent loss your position will suffer. Both numbers are uncertain, and both deserve the same skepticism. Treating IL as a footnote is one of the most common ways retail LPs quietly underperform a simple buy-and-hold.
Three concrete failure modes show up over and over in post-mortems of bad LP decisions. First, the LP deposits into a low-volume volatile/stable pool right before the volatile asset rallies 3x to 5x, and the pool rebalances by selling the entire way up, so the position ends up mostly in the stablecoin. Second, the LP ignores that pool fees are dominated by a short-lived incentive program that ends in weeks, leaving the position collecting a fraction of the headline APR. Third, the LP holds through a correlated crash, where both assets drop together, and IL is amplified by the broader drawdown plus the rebalancing drag. None of these are exotic. They are the default outcomes for an LP who picks a pool by APR alone.
For sizing, the rule of thumb is conservative: if the trailing 30-day fee APR is, say, 15%, then any plausible price move that produces more than 15% IL is a structural loss for the position, regardless of how the pool is marketed. The rest of this article is the practical toolkit for putting numbers behind that statement.
The IL formula, intuitively, for 50/50 pools
For a constant-product 50/50 pool (Uniswap v2 style, SushiSwap, etc.), the IL for a given price change is a fixed curve. Let r be the ratio of the new price to the old price of one asset versus the other. The IL, as a fraction of the held position, is roughly 2*sqrt(r)/(1+r) minus 1. You don't need to memorize the formula, because the numbers themselves are what matter.
A few reference points make the intuition stick. A 1.25x move in either direction produces about 0.6% IL. A 1.5x move produces about 2.0%. A 2x move produces about 5.7%. A 3x move produces about 13.4%. A 5x move produces about 25.5%. A 10x move produces about 42.9%. Notice how the curve is gentle at small moves and brutal at large ones. This non-linearity is why a pool that looks safe over a quiet month can be devastated by a single large repricing event.
For a quick mental check, a useful shortcut is that a doubling of one asset against the other eats roughly 5 to 6 percent of pool value relative to holding. If a pool is advertising a fee APR that would take two or three years to accumulate 5%, and the asset has historically moved 2x or more within a single cycle, the math does not work. The pool can pay you for years and still leave you behind a holder.
Correlation between the paired assets
IL is driven by price divergence, not by price direction. Two assets that are highly correlated produce small IL even if both are very volatile, because they tend to move together. Two assets that are uncorrelated or negatively correlated produce large IL even if each one is "stable" in absolute terms, because they tend to drift apart.
This is why a wETH/wstETH pool (a wrapped, yield-bearing form of ETH on Ethereum) tends to have low IL: the two assets are almost the same thing. It is also why a USDC/USDT pool tends to have low IL: both are dollar-pegged and historically trade within a few basis points of each other. Conversely, a USDC/ETH pool has high IL because USDC stays near $1 and ETH can move 2x in a quarter. And a stETH/ETH pool, despite the names, had meaningful IL events during the 2022 depeg, when the two temporarily diverged by several percent.
The practical takeaway is to ask, before depositing: are these two assets likely to move together, or likely to drift apart? For correlated-asset pools, IL is rarely the binding constraint. For volatile/stable pools, IL is the binding constraint almost every time.
Volatile/volatile versus volatile/stable: the asymmetry
People often assume that pairing two volatile assets is more dangerous than pairing a volatile asset with a stablecoin. The opposite is usually true for IL, and it is one of the most counter-intuitive points in DeFi.
In a volatile/stable pool like ETH/USDC, every percent move in ETH is also a percent move in the price ratio, so the rebalancing is constant and one-sided. In a volatile/volatile pool like ETH/SOL, the two assets can move in the same direction, and the price ratio can stay flat even while each asset doubles. IL depends on the ratio, not on each asset's individual volatility. This is why concentrated-liquidity LPs on Uniswap v3 often prefer volatile/volatile ranges: IL only accrues on the moves that actually diverge.
The trade-off is that volatile/stable pools typically earn higher fee APR, because traders use them as on-ramps and off-ramps. Volatile/volatile pools usually earn less, because trader flow is thinner and the assets are more correlated in traders' mental models. So the "safer" pool from an IL standpoint often pays you less, and the higher-paying pool often costs you more in IL. This is the asymmetry to internalize before you choose.
How to read a backtest of historical IL
Most pool analytics pages (DeFi Llama, Revert Finance, Apy.vision, and the analytics tabs on Uniswap and Balancer) show historical IL for a pool over a chosen window. The right way to read them is not as a forecast, but as a stress test. You are asking: given how these two assets actually behaved, what would my IL have been if I had been in this pool?
Three things to look for. First, the maximum drawdown from IL alone, ignoring fees. If the worst historical IL window was 15%, and the pool paid 20% fee APR over the same period, you survived. If the worst window was 30% and fees were 10%, you would have lost money relative to holding, even before counting gas. Second, the time-to-recovery: how long did it take for cumulative fees to climb back above the worst IL drawdown? If a 30% IL drawdown took eight months to recover, you should expect that pattern to repeat. Third, the correlation regime: was the worst IL window a one-off depeg, or did it look like ordinary divergence between two assets? One-offs can be discounted. Ordinary divergence will happen again.
A common mistake is to backtest only the bull case. Always look at the worst historical window, not the best. The fee APR in the worst window is what you should use for your break-even calculation, because that is the closest thing to a realistic floor you will get.
A worked example: sizing IL against a 30-day fee APR
Suppose you are looking at an ETH/USDC 0.05% pool on a major AMM. Trailing 30-day fee APR is 18%. Volatility over the past quarter has been high; ETH has moved 1.4x to the upside and 0.75x to the downside within the window. Here is the calculation.
Step one, estimate a realistic price move. For ETH, a 1.5x move in either direction over a 90-day holding period is not a tail event. It is roughly one standard deviation given the past year's realized volatility. Step two, look up IL at a 1.5x move in a 50/50 pool: about 2.0%. At a 2x move, which is plausible but not central, IL rises to about 5.7%. At a 3x move, which would be a strong trend, IL is 13.4%. Step three, compare to fee APR. Over 90 days at 18% APR, fee income is about 4.5%. So the central scenario (1.5x) is fine: fees exceed IL. The plausible scenario (2x) is roughly break-even. The strong-trend scenario (3x) is a clear loss, because 13.4% IL is more than three times the 4.5% of fees earned in the same window.
Step four, adjust for things the headline APR hides. Is the 18% APR supported by real trading volume, or by temporary incentive emissions? If 60% of the APR is from a token incentive that ends next month, the sustainable fee APR is closer to 7%, which is 1.75% over 90 days. That changes the conclusion: even the central scenario becomes a loss, and the 2x scenario becomes a 4% net loss on top of the opportunity cost of not just holding ETH.
This is the framework. For any pool, list: realistic price move over your intended holding period, IL at that move from the table above, fee APR over the same window with incentive emissions stripped out, and a comparison of the two numbers. If the realistic move's IL is more than the fee APR, the pool is a structural loser for your horizon. If it is less, the pool is at least defensible, and the rest of the decision is about gas, smart-contract risk, and how confident you are in your price assumption.
Practical implications for the LP about to deposit
Three habits separate the LPs who come out ahead from those who don't. First, never pick a pool by APR alone. Compute the IL at a realistic price move, compare it to the fee APR net of incentives, and only deposit if the spread is comfortably positive. "Comfortably" means enough margin that an adverse move of one or two standard deviations still leaves fees ahead. Second, size the position so that a worst-case IL event does not impair your overall portfolio. Most experienced LPs cap any single pool at a small fraction of total crypto exposure, precisely because the tail of the IL distribution is fat. Third, decide your exit before you deposit. If the price ratio moves far enough that IL exceeds your pre-set threshold, exit. Waiting for fees to "catch up" is how LPs end up holding through the worst drawdown.
Two more practicalities. Watch for fee APRs dominated by emissions, and treat any pool where more than half of APR is a token reward as a short-term trade, not a position. And remember that concentrated liquidity (Uniswap v3, etc.) changes the math: IL is bounded by your chosen range, but you also earn fees only when the price stays inside that range, so the same checklist applies with a smaller capital base and more active management.
The honest summary is that IL is not exotic or mysterious. It is a deterministic function of price divergence, and it is fully knowable for any given move. The reason LPs still lose money is not ignorance of the formula, but unwillingness to do the boring arithmetic of comparing that formula to the fee APR actually being earned. The few minutes of math is the entire edge.
Stay ahead of DeFi pool risk with smarter signals
DeFi pool dynamics move fast and the news around them moves faster. Manually tracking fee APR shifts, incentive emissions, IL backtests, and pool migrations across a dozen chains is a losing game. Zippfeed surfaces DeFi headlines with sentiment scoring (bullish, neutral, or bearish) and an importance rating, so you can spot pool-level risk before it shows up in your position.
