I keep meeting people who think of an LP position as a savings account. You put two tokens into a pool, the pool pays you a slice of every swap that passes through, and the fees just pile up. The word people reach for is yield, and yield sounds like something you earn for waiting. That framing is what gets them into trouble, because an automated market maker LP is not a deposit. It is a position with an option-like payoff, and the fees are the premium you collect for taking the other side of a bet you did not know you were making.
Once you see it that way, a lot of confusing outcomes stop being confusing. The pool that pays a fat fee APR and still leaves you behind buy-and-hold. The stablecoin pair that behaves like a bond and the volatile pair that behaves like a landmine. Same mechanism, very different risk, and the difference is entirely about how much the price moves while you are in the pool.
The payoff looks like a short straddle
Here is the shape of it. When you provide liquidity to a constant-product pool, the pool automatically sells you more of whatever token is falling and sells off whatever token is rising. You are always leaning into the move. If the price stays roughly where it started, you keep your two tokens and pocket the fees, which is the best case. If the price runs hard in either direction, you end up holding more of the loser and less of the winner than if you had just held the two tokens in your wallet.
That gap between what your LP position is worth and what a simple hold would have been worth is the thing people call impermanent loss. I dislike the name because there is nothing impermanent about it once you withdraw, but the mechanic is real. And the payoff diagram of that gap, plotted against price, is a curve that peaks at the starting price and slopes down on both sides. That is the payoff of a short straddle. You have sold a call and a put at the same strike, you collected premium up front, and you lose money as the underlying travels away from strike in either direction.
So an LP is short volatility. You want quiet. You want price to chop sideways and swap volume to stay high, because that is fees rolling in with nothing eating them from the other side. What you do not want is a big directional move, and you really do not want a big move on low volume, because then you are paying the option cost without collecting much premium to offset it.
Arbitrageurs are your actual counterparty
This is the part that took me a while to internalize. The fees feel like they come from ordinary traders swapping in the pool, and some of them do. But the trades that actually cost you are the arbitrage trades, and those are the ones that make the pool track the outside market.
An AMM does not know the real price of anything. It only knows its own reserves and the curve. When the market price on a deep centralized venue moves, the pool is now mispriced, and someone steps in to buy the cheap side until the pool matches the world again. That someone is an arbitrageur, and every time they rebalance your pool for you, they take a little value out of it. They are not doing you a favor. They are picking off the stale price you are quoting, over and over, all day, on every move.
That is why the cleanest way to measure the cost is not impermanent loss versus holding. It is a quantity researchers named loss-versus-rebalancing, or LVR, usually said out loud as "lever." The idea is to compare your LP against a strategy that holds the same changing position but rebalances at the true market price instead of at the stale pool price. The difference is exactly the money the arbitrageurs extract from you because your quote is always a step behind. LVR is the honest cost of being the slow market maker, and it scales with how much the price moves and how volatile the asset is. More variance, more rebalancing, more leakage.
Pricing the trade before you enter
Because it is a short volatility position, you can price it like one. You are collecting a fee stream, and you are paying a volatility cost, and the only question worth asking before you deposit is whether the fees are bigger than the cost. Most people only look at the fee side, which is like selling insurance and never checking what you might have to pay out.
The rough shape of the LVR cost is that it grows with the square of volatility. Double the volatility of the pair and the cost of quoting a stale price is not double, it is closer to four times. That single fact explains most of what you see in the wild. Stablecoin pairs barely move, so their variance is tiny, so their LVR is tiny, and even a thin fee tier can come out ahead. A freshly launched token that swings violently has enormous variance, so the fee has to be huge to compensate, and usually it is not for long.
Here is the checklist I actually run before providing liquidity to a pool:
- Estimate the realized volatility of the pair over a window that matches how long you plan to stay in. Higher volatility means you need proportionally more fee income, and the relationship is steeper than linear.
- Estimate the fee income as fee tier times the volume that genuinely rebalances or trades through the pool, not headline volume that may be wash or routed elsewhere. Fees only help if they land in your pool.
- Compare the two as a rate. If the annualized fee return is not comfortably above your estimated LVR drag, you are selling volatility too cheap and you are better off just holding the two tokens.
- Check where price discovery actually happens for this asset. If a deep centralized book leads and your pool follows, arbitrageurs will rebalance you constantly, and your LVR is high. A pool that is itself the primary venue leaks far less.
- Watch the correlation between the two tokens. Two assets that move together have low relative volatility, so the pool barely rebalances. Two uncorrelated assets diverge often, and every divergence is a rebalance that costs you.
The most common failure mode I see is someone chasing a headline fee APR on a volatile new pair, sitting through one sharp trending move, and coming out well behind a plain hold while the fee counter still shows a big cheerful number. The fees were real. They were just smaller than the volatility being sold, and the arbitrageurs collected the difference.
What to do with this lens
None of this means providing liquidity is a bad trade. Selling volatility is a perfectly good business when you get paid enough for it, and insurers and options desks have run it profitably for a very long time. It only goes wrong when you do not know you are doing it and you forget to price the risk.
So treat every LP decision as an options question. Ask what strike you are effectively short, meaning the current price. Ask how much the underlying is likely to move over your holding period. Ask whether the fee premium covers that expected move with room to spare. If price discovery for the asset lives somewhere else, assume arbitrage will find you and quote yourself a higher required fee accordingly. Concentrated liquidity ranges do not escape this either, they just tighten your strikes and raise both the premium and the risk, so the same comparison applies inside the range.
When I am sizing this kind of thing I want the volatility number and the real trade flow in front of me before the fee APR, because the fee APR is the marketing and the volatility is the bill. Tools like Blockcircle help there, since being able to look at realized volatility, scorecards, and where flow is actually happening across venues is exactly the input this pricing decision needs. Get those two numbers side by side and the LP stops being a mystery yield and turns back into what it is, a short straddle you can choose to sell only when the premium is good.