Impermanent loss is the most discussed and least precisely defined term in automated market making. It is not a loss in the ordinary sense, it is not usually impermanent, and it is not measured against your entry price. It is a comparison with a specific alternative, and once you know which alternative, everything confusing about it resolves.
Depositing into an automated pool is not lending your assets to a borrower and it is not holding them in a wallet. It is placing them into a formula that will exchange one for the other with anybody who asks, at a price the formula computes from the current ratio of the two.
You are therefore standing on both sides of a market permanently, without choosing when to trade or at what price. Anybody who wants to buy the asset that is rising buys it from you, and anybody who wants to sell the asset that is falling sells it to you, continuously and without your involvement.
That is the whole position, and every property that follows is a consequence of it. It is much closer to being a market maker with no ability to withdraw quotes than to any form of deposit, and describing it as passive income obscures precisely the part that matters.
You are a market maker who cannot withdraw quotes. Everything called impermanent loss is a consequence of that one fact.
As the relative price of the two assets moves, the formula changes the proportions of what you hold. When one asset rises against the other, traders take that asset out of the pool and put the other one in, so your share ends up holding less of the one that rose and more of the one that fell.
This is not a malfunction, it is the mechanism working. The pool has to end up with less of the scarce asset, because that is what makes its price rise inside the pool, and your holdings are the pool. You have been sold into strength and bought into weakness, automatically.
The result is a portfolio that rebalances continuously towards whatever is falling. Over a period where prices return to where they started, that behaviour costs nothing and may earn fees. Over a period with a sustained move, it produces less of the appreciating asset than you would have held by doing nothing.
The number people quote is the difference between the value of your pool share and the value of the same two assets if you had simply held them in the proportions you deposited. It is a comparison with holding, not with your entry price, and not with any other strategy.
That distinction settles most arguments about it. A pool position can be worth more than when you deposited and still show a large impermanent loss, because both quantities rose and holding would have risen more. It can also show none while losing money, if both assets fell together.
So the term is measuring an opportunity cost against one specific alternative. It is a useful number and it is not a profit and loss statement, and treating it as one is the source of most of the confusion around whether a position went well.
Impermanent describes the case where prices return to their starting ratio, at which point the gap closes and the comparison shows nothing. That case is real and it is the origin of the word.
It is also the exception rather than the rule. If prices do not return, the gap is realised when you withdraw, and at that point it is permanent in every ordinary sense. A term that describes the reversible case has been attached to a quantity that is usually not reversed.
Some practitioners use divergence loss instead, which is more accurate because it names the cause: the loss is a function of how far the two prices diverged, in either direction, and it grows with the size of the divergence rather than with its sign.
Providers are not doing this for nothing. Every trade against the pool pays a fee that accrues to the providers in proportion to their share, and that stream is the compensation for the position described above.
Whether the compensation exceeds the cost is an empirical question that depends on how much trading happens relative to how much prices diverge. High activity with little net movement is the favourable case; little activity with a large sustained move is the unfavourable one.
The honest statement is that neither is predictable in advance, and anybody offering a rule about which pools are worth providing to is making a claim the mechanism does not support. What can be done is measuring both quantities after the fact, which almost nobody does.
One further asymmetry is worth stating plainly. The fee stream depends on trading activity you neither control nor can forecast, and the divergence effect depends on price movement you also do not control. A provider is exposed to two independent uncertainties at the same time, and the position is routinely described as though only the first existed. Naming both is not pessimism, it is the minimum required to describe what has actually been entered into.
Newer designs let a provider specify a price range within which their capital is active, instead of spreading it across every possible price. Inside the range, the same capital provides much more depth and therefore earns a much larger share of the fees.
The trade is that the divergence effect is concentrated too. A narrow range earns more while the price stays inside it and converts entirely into the falling asset when the price leaves, at which point the position stops earning anything until the price returns.
This turns a passive position into an active one requiring management, which is a completely different commitment. It also makes the outcome far more sensitive to a decision the provider makes, which is worth being explicit about before treating it as an improvement.
A concentrated range earns more while the price stays inside it, and stops earning entirely when it leaves. That is not a passive position any more.
The divergence loss grows with the square of the price ratio change rather than linearly, which means it is negligible for small movements and grows quickly for large ones. A pair that moves modestly relative to each other produces almost nothing; a pair where one asset multiplies produces a great deal.
The worst case is therefore a pool holding one asset that moves violently against another that does not. Providers in such pools frequently find that they participated in very little of the move, which is exactly what the mechanism guarantees rather than an accident.
This is why the composition of the pair matters more than the headline fee rate. Two pools with the same activity and the same fee behave completely differently depending on how correlated their two assets are, and correlation is observable before providing.
Pools pairing two assets designed to hold the same value have very little divergence by construction, which removes most of the effect described above. That is a genuine difference and it is why such pools are frequently presented as the low-risk version.
What it does not remove is everything else. If one of the two assets stops holding its value, the mechanism will convert the pool almost entirely into that asset, because traders will sell it into the pool all the way down. The provider ends up holding the one that failed.
So a stable pair exchanges a continuous small effect for a rare large one, which is a different risk profile rather than a smaller risk. Which is preferable depends on the specific assets involved, and it depends far more on them than on anything about the pool.
Four things, all observable. The historical divergence between the two assets, which is the input the effect depends on. The activity in the pool relative to its size, which determines the fee stream. Whether the position is a full range or a concentrated one, which determines whether it needs management.
The fourth is the contract itself: whether it has been examined, whether it can be upgraded, and by whom. Providing liquidity means depositing into a contract, and every consideration that applies to any contract applies here regardless of how the economics look.
None of these predict the outcome and they are not meant to. They describe the position, which is the part that is knowable, and the difference between a provider who checked them and one who did not is not luck but information.
Everything above describes the mechanism. It says nothing about whether providing liquidity to any particular pool is a good idea, because that depends on future trading activity and future price divergence, neither of which is knowable.
It also says nothing about the assets themselves. A well-understood pool position in two assets that both go to nothing is a well-understood way to lose money, and no amount of clarity about the mechanism protects against the underlying.
What the mechanism does allow is measuring afterwards. Fees earned on one side, divergence effect on the other, and the comparison against having held. Running that once on a real position teaches more than any explanation, including this one.
Against simply holding the same two assets in the proportions you deposited. Not against your entry price and not against any other strategy. A pool position can be worth more than at deposit and still show a large impermanent loss, because holding would have risen more.
Because the formula must end up with less of the scarce asset for its price to rise inside the pool, and your share is the pool. Traders take out whatever is rising and put in whatever is falling, so you are sold into strength and bought into weakness continuously, without being asked.
Only if prices return to their starting ratio, which is the exception rather than the rule. If they do not, the gap is realised when you withdraw and is permanent in every ordinary sense. Divergence loss is the more accurate term because it names the cause.
Every trade against the pool pays a fee that accrues to providers in proportion to their share, and that stream is the compensation for the position. Whether it exceeds the divergence effect depends on how much trading happens relative to how much prices diverge, and neither is predictable in advance.
Capital is active only within a chosen price band, so inside it the same capital provides much more depth and earns a much larger share of fees. When the price leaves the band the position converts entirely into one asset and stops earning, which makes it an active position requiring management.
It grows with the square of the change in price ratio, so it is negligible for small moves and grows quickly for large ones. The worst case is a pool holding one asset that moves violently against another that does not, where providers participate in very little of the move.
They have very little divergence by construction, which removes most of the effect. What it does not remove is the case where one of the two stops holding its value: the mechanism will convert the pool almost entirely into that asset, so the provider ends up holding the one that failed.
The historical divergence between the two assets, the pool's activity relative to its size, whether the position is full range or concentrated, and the contract itself: whether it has been examined, whether it can be upgraded, and by whom. None predicts the outcome; all describe the position.