Your dashboard is green. The yield counter ticks up every time you look, and for a while it feels like the smartest money you have ever parked. Then you withdraw, compare it with what simply holding would have done, and the floor drops out from under you.
Here is the example that should keep you up tonight. You put 1 ETH and 1,000 USDC into a pool. ETH wanders off, comes back to exactly where it started, and you take out exactly what you put in. Impermanent loss reports $0. Yet $1,500 went to arbitrageurs across the two trades, while your share of the trading fees came to about $6, which is 0.4% of what was taken from you.
Think about what that money was supposed to be. The emergency fund. The deposit you are saving for. The cushion that lets you sleep through a surprise repair bill. Pools are getting more sophisticated every month, and the bots trading against you are faster than they were last year.
So the urgency is about understanding this before you deposit another dollar. Liquidity provision takes anywhere from $100 to $10,000 of your capital, returns between $0 and roughly $900 a month, and pays on no schedule you can plan your life around.
Tonight, pick one pool and write down two figures side by side: the fees it paid last week, and how far its price moved in that same week. That small comparison could protect more of your money than any yield screenshot.
Here is the study I would want you to know before you deposit anything. A 2021 paper on impermanent loss in Uniswap v3, cited more than a hundred times since, analysed 17 pools representing 43% of the protocol's total value locked. Over the period studied, those liquidity providers collected $199.3 million in fees and suffered $260.1 million in impermanent loss. Net, they would have been better off by roughly $60.8 million if they had simply held their tokens and done nothing.
That is the combined result across nearly half the protocol, and nobody went hunting for a bad pool to get it. A separate analysis by Topaz Blue and the Bancor Protocol, widely reported at the time, found that around 49.5% of Uniswap v3 liquidity providers had negative returns once impermanent loss was counted.
Neither finding means you cannot profit from this. Both mean the default outcome is a loss, which is the opposite of how liquidity provision is usually sold to you, and it is why this guide spends most of its length on how the thing works and much less on the returns.
You might wonder why a page about a losing default exists at all. Liquidity provision is genuinely useful infrastructure, the fees are real, and there are setups where the arithmetic works. What makes it dangerous for you is that the interface shows a rising number the whole time while your position may be losing, so every signal you get feels positive. Almost nothing else on this site works like that. A failing shop shows you empty shelves and a failing freelance business shows you an empty calendar, while a failing liquidity position shows you a fee counter that keeps climbing. Understanding the mechanism is the only stand-in you have for feedback that tells you the truth.
The word "yield" does a lot of damage here. Providing liquidity works differently from lending, and the fee you earn works differently from interest.
When you deposit two assets into an automated market maker, you are posting a standing offer to swap either asset for the other at prices set by a formula. You will buy when someone wants to sell and sell when someone wants to buy, automatically, at whatever the curve says, whatever the wider market thinks the price is.
a16z crypto's analysis puts the result plainly: liquidity providers "suffer losses from adverse selection, which is part of the price of doing business as an LP. By virtue of offering to take either side (buy or sell) of a trade at a given price, every LP in an AMM runs the risk" of trading against someone who knows more than the pool does.
That someone has a name. When the market price of an asset moves, your pool's price is out of date until someone corrects it, and the person who corrects it is an arbitrageur who buys the underpriced side from you and sells it elsewhere. Their profit is your loss, and it happens on every meaningful price move, automatically, whether or not you are watching.
So here is the honest description of what you would be doing: running a market-making operation with no way to update your prices, and collecting a fee for the privilege. The one question is whether the fee covers what the arbitrageurs take. Everything else is detail. Would you still call it yield if you described it to a friend in those words?
Impermanent loss is the measure everyone learns first, and it is a poor tool. The reason is worth understanding, because it changes how you judge a pool.
Impermanent loss compares your position's value with simply holding the same tokens, and it depends only on the starting price and the ending price. a16z's write-up flags exactly why that should bother you: "The independence of impermanent loss on the price trajectory (other than its initial and final values) should strike you as fishy."
It should. Picture a price that shoots up and then falls back to where it began. Impermanent loss says you lost nothing, because the start and end prices match. Meanwhile arbitrageurs traded against your pool on the way up and again on the way down, taking value each time. The measure shows zero while your money was drained twice.
The worked example in the paper is short. Take a constant-product pool holding 1 ETH and 1,000 USDC, and suppose the market price of ETH jumps from $1,000 to $4,000. An arbitrageur buys 0.5 ETH from the pool. The reference strategy sells that same 0.5 ETH, at the real market price of $4,000. The difference between what the pool got and what the market would have paid is the LVR, and it is the arbitrageur's profit.
Two things make LVR the right tool for you. It depends on the path the price takes, so a round trip up and back down adds up as a cost instead of showing zero. And it builds up trade by trade, which matches how the loss is actually taken from you.
The practical test, in a16z's words, is whether "the fees collected exceeded the LVR suffered". That is the whole question. A pool advertising a high APR is telling you about one side of it.
The gap between the two measures is easiest to see with real numbers. Take a constant-product pool holding 1 ETH and 1,000 USDC at a 0.30% fee tier, and send the price up to $4,000 and back down to $1,000.
The pool always rebalances so the ratio of its holdings matches the new price, and arbitrageurs are the ones who push it there.
On the way up, your pool sells half an ETH at an average of $2,000 while the market is paying $4,000. On the way down, it buys that same half back at an average of $2,000 while the market is charging $1,000. It sold low and bought high, in that order, automatically, with your money.
Your position ends holding exactly what it started with: 1 ETH and 1,000 USDC. So:
Nothing unusual happened here. One round trip in a volatile pair, no strange event, no hack. The measure everyone is taught reported a perfect outcome while your position leaked value on both legs.
That is why an APR that counts fees and leaves out LVR measures the wrong quantity entirely. If your dashboard showed you $6 earned and $1,500 lost side by side, would you still leave the money in?
Once you see providing liquidity as selling to arbitrageurs, you can predict which setups make money and which lose it.
Three setups survive the analysis, and their yields are far less exciting than the ones being advertised.
The risks here are peg risk and contract risk, with very little price risk. When this goes wrong, it goes wrong suddenly: one asset loses its peg and your position turns almost entirely into the broken one. Correlated pairs swap a steady trickle of small losses for the occasional large one. The risk takes a different shape, and it is every bit as real.
What fails is the common case: supplying a volatile pair in a standard fee tier, passively, because the displayed APR looked good.
Stablecoin pairs earning small fees are the version a careful household can live with, though even these can lose the money you deposit. Put in only what would leave your emergency fund whole and your rent untouched if the contract failed tomorrow, and treat the modest return as extra. Work out what a total loss would leave at home before you supply anything.
Picture a small stablecoin position that quietly adds a little each month, enough to cover a streaming subscription or a dinner out, while the rest of your savings stays safe elsewhere. A pleasant extra you never lean on. That is the version worth having, and it starts with sizing the deposit around your family's safety.
You have seen how a position can end flat while $1,500 leaves the pool. Now for the numbers a protocol puts in front of you before any of that happens.
Pool APRs are the most misleading number in this space, and the reasons are mechanical. Nobody has to be lying for them to mislead you.
It looks backward and stretches a short window into a year. A figure taken from one day of unusually high volume and multiplied out to a year describes a year that will not happen.
It counts fees and leaves out LVR. This is the core problem. The advertised number is one side of the comparison that decides whether you make money.
It assumes your money stays in range. For a concentrated position, the quoted return applies only while the price sits inside your band. Out of range, you earn nothing and hold entirely the wrong asset.
Everyone else who sees it dilutes it. Fees are shared across all the liquidity in the pool. A high advertised return attracts money, and the return falls as that money arrives. You are buying into a number that your own deposit helps shrink.
Reward tokens are valued hopefully. An APR that includes rewards assumes you can sell the reward token at its quoted price, in your size, again and again. Thin trading in the reward token makes that assumption false exactly when you need it to be true.
A more useful check is to look at the fees actually paid over a meaningful period, next to how far the pair's price actually moved over the same period, and ask whether the first plausibly beat the cost implied by the second. When did you last see a pool page that showed you both?
The costs that never appear in the APR
Beyond adverse selection, several real costs are routinely left out of the sums.
Gas. Entering, exiting and any rebalancing all cost transaction fees. For concentrated positions, active management means repeated transactions, and on an expensive chain that alone can eat the whole return on a small position.
Rebalancing locks in your loss. Moving a concentrated range means closing at current prices and reopening, which turns a paper loss into a real one and usually creates a taxable event.
Smart contract risk. Your money sits in code. Established protocols have long track records, and new pools often have none. This risk shows up in no yield figure, and it is the one that takes your whole position at once.
Bridge and chain risk. Chasing higher yields onto newer chains adds bridge risk, and bridges have historically been where the largest losses in this industry happened.
Tax complexity. Depending on where you live, depositing into a pool may itself count as a disposal, fees may be income as they build up, and every rebalance may be another event. This is some of the heaviest record-keeping in crypto, and it is almost never mentioned next to an APR.
How AMMs came to exist
The design behind all of this was a deliberate trade, and knowing what it bought explains why the costs are built in. Picking a better pool cannot remove them.
Traditional markets run on order books. Market makers post prices to buy and sell, update them constantly as conditions change, and pull them when they sense danger. That works because a market maker can react. Their whole edge is how fast they change a price when the world moves.
An order book cannot run on a slow, expensive, public blockchain. Every price update is a transaction with a fee and a delay, and everyone can see your order before it goes through. Early attempts at on-chain order books were unusable for exactly that reason.
The automated market maker solved this by removing the quoting entirely. A formula sets the prices in place of a person, and the classic one is the constant product curve, `x * y = k`, which keeps the product of the two reserves fixed. Liquidity providers deposit and the formula quotes for them, forever, without anyone touching it.
That was a genuine breakthrough. It made open markets possible for any pair, with no middleman and no minimum size. It is also the source of every cost in this guide, because the very property that lets it work on-chain is that it cannot update a price. A market maker who cannot change prices when the market moves will, by design, be traded against by people who saw the move first.
Later versions have chipped at the edges. Concentrated liquidity let providers choose a price band and earn far more inside it, which made capital work harder and raised adverse selection in the same stroke. Analysis of the two designs found that v3 positions tracked LVR losses closely, while v2 positions did better largely because they saw more ordinary, uninformed trading from people who simply wanted to swap.
The useful lesson from that comparison: what limits your returns is who trades in your pool, far more than how efficiently your capital is spread. A pool that mostly serves people who want to swap is a good business, and a pool that mostly serves arbitrageurs is you paying them a subsidy.
Checking a pool before you deposit
That settles the theory. From here it is a decision you make with your own money, on one specific pool.
Six questions, in order. Most pools you look at will fail one of the first three.
How closely do the two assets move together? This decides LVR more than anything else. Two assets pegged to the same thing leave almost nothing to take. A volatile asset against a stablecoin leaves a great deal.
Does the fee tier match the volatility? The fee is your payment for adverse selection. A volatile pair in the lowest tier loses by design, and no amount of management fixes it.
What is the volume actually made of? Volume from people who want the other token pays you. Volume from arbitrageurs is a partial refund of what they took. A pool whose volume jumps exactly when the price moves is telling you which kind it has.
What would this position have returned over the last three months? Ignore the advertised APR here. Look at actual fees against actual price movement over a real stretch that included a fall. If the interface cannot show you performance against holding, treat that as your answer.
How much of the return depends on rewards? Value the reward token at what you could really sell it for in your size, and assume the reward rate falls. If the position only works at today's rewards, its expiry date is in somebody else's hands.
How risky is the contract, honestly? How long has this code held serious money, has it been audited, and would you accept the risk of losing everything for the fee alone, with no rewards attached?
A pool that passes all six is usually a dull one, and dull is what you are looking for.
Answer the last of those six questions at the kitchen table. Accepting total loss for the fee alone means the deposit has to be money your kids would never feel go missing, so school shoes and the holiday fund never ride on a pool. If that loss would reach your children, reduce the amount or skip the pool.
How it compares with your other options
This is worth placing in context, because the same money has other uses covered elsewhere on this site.
Against staking. Ethereum staking yields roughly 2% with no adverse selection, no impermanent loss and no smart contract risk beyond the protocol itself. That is the closest thing to a risk-free benchmark inside crypto, and you should judge any LP position against it, with zero as the wrong yardstick. A volatile pair yielding 8% while bleeding LVR is worse for you than 2% that keeps your holdings exactly as they were.
Against simply holding. This is the comparison the research makes and the dashboards avoid. Across 43% of Uniswap v3 TVL, holding beat providing liquidity by about $60.8 million. Any pool you enter should come with a clear reason it will do better than that average. What is yours?
Against lending. Lending an asset in a lending market earns interest without your position turning into whichever asset is falling. Lower ceiling, different risk, no adverse selection. If you want a return on an asset you plan to keep, lending usually fits better.
Against active market making. The professionals who do this profitably rebalance, hedge their directional exposure on another venue, and stop quoting when conditions turn bad. If you are not doing those three things, you are the person on the other side of their trades.
The honest summary: providing liquidity earns its place for correlated pairs, for genuinely rewarded positions with the reward counted properly, and for people running it as an active operation. For everything else, one of the alternatives above beats it.
Rookie Mistakes
Treating LP as yield. It is market making with stale prices. Every other decision follows from that, and this one wrong idea sits upstream of all the other mistakes.
Judging a pool by impermanent loss. IL ignores the price path and so understates what round trips cost you. Judge by whether fees plausibly beat LVR.
Chasing the highest APR. The highest advertised return usually comes from the most volatile pair, the newest contract or the most aggressive rewards. All three are payment for risk, and none of it is free money.
Supplying a volatile pair in a low fee tier. This loses by design. The fee was priced for a smaller amount of adverse selection than the one you are taking on.
Setting a narrow range and walking away. Concentrated liquidity out of range earns nothing and leaves you holding only the asset that fell. Narrow ranges need attention; if you cannot give it, do not set one.
Counting reward tokens at quoted prices. Value them at what you could really sell them for, in your size, after the reward rate falls. Many positions only make money under an assumption nobody says out loud.
Ignoring gas against position size. A small position on an expensive chain can spend its whole year's return on entry, exit and two rebalances.
Forgetting to record where you started. Without the token amounts and prices at deposit, you can never work out whether you beat holding. A surprising number of people never find out whether they made money at all.
Gotchas Worth Knowing
Your position turns into whichever asset is falling. This is the mechanism doing exactly what it was built to do. As one asset drops, the pool buys more of it with the other. If you would be uncomfortable ending up heavily weighted toward the loser, stay out of volatile pairs.
"Impermanent" is a marketing word. The loss becomes permanent the moment you withdraw, and prices have no obligation to return to where you came in.
Depositing may be a taxable disposal. In several countries, swapping tokens for pool shares counts as disposing of the original assets. People find this out afterwards.
Rewards fall by design. Nearly every incentive programme shrinks over time. Judge a position you enter on today's rewards by what tomorrow's will be.
Vampire attacks and pool migrations move the trading. A competing protocol offering better terms can drain the volume your fees depend on, quickly, leaving you in a pool full of liquidity and empty of trades.
Some pools make it hard to leave. Lock-ups, withdrawal queues and vesting on reward tokens all exist. Read the terms before you deposit, while leaving is still a choice.
Front-running and sandwich attacks can hit your entries. Large deposits and withdrawals can be exploited the same way large swaps can. How big your position is and how you enter matter more than people expect.
Concentrated liquidity, and why it is leverage
Concentrated liquidity deserves its own section, because it is the feature you will be steered toward most and the one most likely to lose you money quickly.
In the original design, your money was spread across every possible price from zero to infinity. Almost all of it sat at prices that would never happen, doing nothing. Concentrated liquidity lets you supply only within a band you choose, so the same money backs far more depth where trading actually happens.
The result is a multiplier, and it applies to both sides of your ledger.
Fees multiply. A narrow band around the current price can earn many times what the same money earns spread thin. This is the number the interfaces show you.
Adverse selection multiplies just as much. You are now on the other side of a much bigger share of the trades inside that band. Every arbitrage trade in your range hits your money harder. LVR grows with your fee income, because both come from the same trades.
Out of range, you earn nothing. When the price leaves your band, your position stops collecting fees completely and sits converted into whichever asset did worse. It sits idle and weighted the wrong way.
Rebalancing has a real cost. Following the price means closing and reopening: gas, a loss locked in, and in most countries a taxable event. Doing it often enough to stay in range on a volatile pair can cost more than the extra fees the narrow band earned.
Think of the width of your band as a bet on how much the price will move. A narrow band pays well if the price stays put and badly if it moves. A wide band earns less and needs less of your attention. Setting a narrow band and then not watching it gives you the worst of both: you take the concentrated adverse selection while the price is inside, then earn nothing once it leaves.
Here is the part that matters for your decision. If you cannot commit to watching and rebalancing, a wider range or the older unconcentrated design suits you better, even though the advertised return is lower. The advertised return on a narrow band assumes management you are not going to do. How many evenings a week would you really spend checking a range?
What it actually feels like
Living with a position feels nothing like the dashboard suggests.
You deposit, and straight away the interface shows fees building up. This is the pleasant part, and it lasts about a week. The number climbs steadily in small steps, and it feels like yield.
Then the price of the volatile asset moves a long way in one direction. Your position's value stops tracking what you expected, because the pool has been selling the rising asset all the way up. The fee counter is still climbing, so your dashboard still looks positive, and the two numbers are heading in opposite directions.
If you set a concentrated range, sooner or later the price leaves it. Fees stop entirely. You now hold one asset, and it is the one that did worse. Your choice is to rebalance, which locks in the loss and costs gas, or to wait for a turnaround that may never come.
Then comes the sum nobody enjoys: comparing your position's value today with what the same tokens would be worth if you had never deposited. That needs the starting snapshot most people never took, and for many positions in a volatile pair the answer is bad news, which the climbing fee counter had hidden the whole time.
On a stablecoin pair it really does feel different, and duller. Small fees build up, the position stays balanced, nothing dramatic happens, and the return is modest and real. The dullness is what you are paying for.
Dull fees still add up to something warm. Left to build for a year on a stablecoin pair you would hold anyway, and if they survive costs and losses, they might pay for a dinner out with your parents, on you. Size the position so a failed contract would still leave your rent and your emergency savings in place.
Who This Suits
I will be direct here, because with this much money at stake a mismatch is expensive.
It suits you if you already hold both assets in a correlated pair and plan to keep holding them. If you hold two stablecoins, or two assets that track the same underlying, and you would hold them anyway, supplying that pair earns a modest fee with little adverse selection by design. This is the version that works for an ordinary person, and it is the one advertised least.
It suits you if you will run it as an operation. Rebalancing on a schedule, hedging directional exposure on another venue, pulling out when volatility spikes, and tracking performance against holding instead of against what you deposited. That is a trading job, and it can pay well.
It suits you if you want to understand how markets work underneath. There is no faster way to learn how adverse selection works than to watch it happen to your own money in small amounts. Treat a small position as tuition with a partial refund, and it can genuinely teach you a lot.
It does not suit you if you want passive yield. The word yield suggests a lender's position, and this is a market maker's position. The default outcome for a passive provider in a volatile pair is a loss against holding, which is exactly what the research measured.
It does not suit you if ending up holding mostly the asset that fell would upset you. That outcome is the mechanism working correctly, and it happens every time.
It does not suit small positions on expensive chains. Entry, exit and two rebalances can swallow a whole year of returns before adverse selection is even counted.
Checking whether you are actually winning
Your money is in and the position is live. The question changes from which pool to pick to whether this one is paying you.
Most people never find out whether a position made money, because the dashboard answers a different question from the one that matters.
Record where you started. Token amounts, both prices, the date, the gas you paid. Without this you cannot work out the only number that counts. Do it before you deposit, because the interface will not keep it for you.
Measure against holding, and set your deposit value aside. The right comparison is what those exact token amounts would be worth today if you had never deposited. A position up 5% in dollars during a stretch when holding would have been up 15% is a losing position dressed as a winner.
Count fees after gas. Entry, exit and every rebalance. On a small position this often flips the result from a gain to a loss.
Track the reward token separately, at the price you really got. Use what you actually sold it for, and ignore what it was quoted at when it arrived.
Review on a fixed schedule, whatever the price is doing. Checking whenever the price moves means checking when you are most likely to act on emotion. A monthly review with the same four numbers each time is worth more than daily watching.
The reason to be strict about this is that the research finding, holding beating liquidity provision by $60.8 million across nearly half of Uniswap v3's TVL, is made up of thousands of individual positions whose owners mostly believed they were earning yield. The counter on their screens was going up the entire time. If you checked your own position against holding today, which side of that line do you think you would land on?
Every week you leave a position unchecked is a week it can quietly bleed. You might be fine. You might be funding someone else's trading desk. Open your position tonight, compare it against simply holding the same tokens, and write the answer down. If it hurts, you found out early, which is the cheapest way to find out.
Where This Goes Next
These are judgements about direction, without dates attached. The regulatory items are the ones most likely to move first.
AMM design is aiming at LVR directly. Research has shifted its focus from impermanent loss to adverse selection, and designs that capture arbitrage value and hand it back to liquidity providers, including dynamic fees and auction-based approaches, are an active area. If they work, passive LP economics improve at the root, and today's numbers understate the future.
Liquidity provision is turning professional. Concentrated liquidity rewarded active management, which favours firms with infrastructure. Expect the share of liquidity supplied by sophisticated operators to keep rising and the room for a passive individual like you to keep shrinking, just as happened with market making on traditional exchanges.
The stable and correlated corner stays workable. Low LVR there comes from the structure itself, whatever the market is doing, so pegged pairs remain the lasting corner open to ordinary people. Yields there are modest, and that is the honest version of this activity.
Reward-driven yields keep shrinking. Paying out tokens to buy liquidity is expensive, and protocols have been cutting back as their models mature. Returns that relied on rewards will keep falling toward the fee-minus-LVR baseline.
Who trades in your pool becomes the thing that matters. Because ordinary, uninformed trading is what actually pays liquidity providers, designs that separate everyday swappers from arbitrageurs are where the value is. Order flow auctions, batch settlement and routing deals that send retail trades to particular venues all aim at this. Your future returns as a liquidity provider will depend more on which trades reach your pool than on which pair you picked.
Better tools show the truth. Position analytics that report performance against holding, instead of against what you deposited, are becoming standard. That is bad news for the advertised-APR story and good news for anyone deciding honestly, and it is the most useful thing you can ask of any interface before you deposit.
As the analytics improve, the gap between providers who measure honestly and those who trust the counter will grow wider. The careful ones keep what they earn. The rest keep funding the arbitrageurs. Decide which side you want to be on before the next shiny APR catches your eye.
Regulation and tax: the part the yield calculator leaves out
Everything so far has been about what your position earns. What follows is about what you have to report on it.
Providing liquidity needs nobody's permission at the protocol level. Everything around the protocol is regulated, and the tax treatment in particular is unsettled in a way that should affect how much you put in.
Europe: the decentralisation exemption is narrower than it sounds
The Markets in Crypto-Assets Regulation applied fully from December 2024, and the transitional grandfathering under Article 143(3) ran only until 1 July 2026, or until a firm was granted or refused authorisation. That window has closed, so operating as a crypto-asset service provider in the EU now requires authorisation.
DeFi is often described as simply exempt. The real position is narrower. Recital 22 provides that where crypto-asset services are provided in a fully decentralised manner without any intermediary, they fall outside the regulation. Both halves of that phrase carry weight, and the exemption depends on there being no intermediary at all. A protocol calling itself decentralised settles nothing.
For you as an individual putting money into a pool, this mostly does not bite, because you are a user of the service and someone else is providing it. It does matter to you in three indirect ways.
The front end you use may itself be an intermediary. A hosted interface with an operator, a fee, and control over what you see looks very different from a protocol, and interfaces are where regulators have focused.
The venues you use to get in and out of positions do need authorisation to serve EU clients. ESMA keeps a register under Articles 109 and 110 listing authorised providers and, separately, non-compliant entities providing crypto-asset services, updated weekly. It is a free check before you send money anywhere.
And if you ever pool other people's money into your positions, you have left the exemption entirely. Managing money for others is the line, and stepping over it casually is the most common way a profitable personal position becomes an unlicensed financial service. Has a friend or relative ever asked you to "just put some in for them too"? That request is exactly where the line sits.
The tax question nobody can answer cleanly for you
Providing liquidity creates several events, and only some have clear treatment.
Rewards and fees are the clear part. Where you receive tokens as a reward, the general principle the IRS applies is that property received is income at fair market value when you gain dominion and control over it, which is the test set out in Revenue Ruling 2023-14 for validation rewards and applied on the same reasoning elsewhere. Rewards arriving continuously are a stream of small income events, and each one needs a value and a timestamp.
The deposit is the unclear part. Putting two assets into a pool and receiving an LP token in return may count as disposing of those assets, or it may not, and no clear general guidance settles it. If it is a disposal, you may have triggered a taxable gain the moment you entered a position you have not left. Practitioners disagree, and the answer varies by country. Take this question to an adviser who knows digital assets in your country before you deposit, while the answer can still change what you do.
Impermanent loss has no standing in tax law. It describes an economic outcome, and a dashboard showing it does not turn it into a deductible loss. Any relief generally waits for an actual disposal and follows your country's capital loss rules, including limits on what those losses can offset. Your position can be down in real terms while your tax bill reflects income you were treated as receiving along the way.
What that means for your numbers
Take any quoted APY and make three adjustments before you believe it.
Subtract impermanent loss under a realistic price gap between the two assets, and skip the flat case. A yield that only survives if both assets move together is really a bet on correlation.
Subtract gas for entry, exit and every reward claim, which on some chains is more than the rewards on a small position.
Then treat what is left as pre-tax income that arises as it builds up, with no waiting until you withdraw. That last point is the one that catches people out: you can owe tax for a year in which your position ended down. Could you pay that bill from your savings if the position itself had shrunk?
The honest summary: liquidity provision is a real strategy with real yield and unusually messy accounting. Size it as a speculative slice of your money, keep records of every reward as it arrives, and get local advice on the deposit question before it becomes history you have to explain.