Every trade needs somebody on the other side, and there are exactly two ways to arrange that. One matches you against another participant who chose their price. The other matches you against a formula that has no opinion at all. The difference determines your cost, your certainty, and which surprises are possible.
A market has to solve one problem: when you want to buy, somebody has to be willing to sell, at a price both of you accept. Order books solve it by collecting the intentions of participants who each chose a price and a size, and matching the ones that overlap. Automated pools solve it by holding a reserve of both assets and quoting a price derived mechanically from the ratio between them, with no human deciding anything at the moment of the trade.
Everything else follows from that. In a book, the price you get was chosen by somebody who wanted it, which means it reflects an opinion and can be withdrawn the instant that opinion changes. In a pool, the price you get is computed from a formula that cannot change its mind, cannot withdraw, and will quote at any hour regardless of conditions. Both properties are advantages in some circumstances and liabilities in others, and neither design is a better version of the other.
A book quotes what somebody is willing to do. A pool quotes what a formula says. The first can vanish and the second cannot, and both of those facts cost money in different situations.
Resting orders sit at prices, stacked outward from the best bid and the best offer. Your order consumes them starting at the best available price and working outward until it is filled. If your size fits within the first level, you pay that price; if it does not, you pay a weighted average across as many levels as it takes, and the average is worse the further you go.
Two consequences matter. The cost of a given size is not knowable from the quoted price alone, because it depends on what is stacked behind it, and that changes second by second. And the depth is supplied voluntarily, which means it can disappear precisely when conditions become uncertain, since the participants providing it are the ones most exposed to being wrong in those moments.
The most common design holds reserves of two assets and enforces a rule: the product of the two reserve quantities stays constant through every trade. Buying one asset removes it from the reserve and adds the other, and the price the formula quotes moves accordingly. The larger your trade relative to the pool, the further along the curve you travel and the worse your average price.
The property that makes this genuinely different is that the cost is computable in advance, exactly, from public information. The reserves are visible on-chain, the formula is fixed, so anybody can calculate what a trade of any size will cost before sending it. There is no equivalent certainty in a book, where the answer depends on resting orders that can be cancelled between your decision and your arrival.
The property that makes it worse in the ordinary case is that the price does not track anything external. A pool does not know what the asset trades for elsewhere; it only knows its own ratio. Alignment with the wider market happens because arbitrageurs profit from correcting the difference, which means the pool is continuously slightly wrong and continuously being repaired by people who are paid for the repair out of the pool itself.
Supplying assets to a pool earns a share of the trading fees, and the marketing usually stops there. The cost sits on the other side of the ledger and it has an awkward name: the value of what you deposited, measured against simply having held the two assets, falls whenever their relative price moves in either direction.
The mechanism is straightforward once stated. The formula automatically sells whichever asset is appreciating and buys whichever is depreciating, because that is what keeping the product constant requires. A provider therefore ends up holding more of the asset that fell and less of the one that rose, compared with someone who did nothing. The larger the price divergence, the larger the shortfall, and it is not a fee anybody charged: it is the arithmetic of the rule.
It is called impermanent because the gap closes if relative prices return to where they started, which is a real property and a misleading name. Nothing guarantees they return, and for a pair whose prices genuinely diverge over the holding period the loss is permanent in every sense that matters. The honest framing is that providing liquidity is a bet that fee income exceeds the divergence, which is a specific position with a specific risk rather than a yield.
Providing liquidity is not lending and it is not earning interest. It is a position that profits from trading activity and loses from price divergence, and both halves have to be measured.
Early pools spread reserves across every possible price from zero to infinity, which meant that almost all of the capital sat at prices that would never trade. Concentrated designs let a provider specify a price range and place their capital only within it, which makes the same capital produce far more depth where trading actually happens.
It is a substantial improvement in efficiency and it changed what providing liquidity is. A position with a range is no longer passive: it stops earning entirely when the price leaves the range, and the exposure it holds changes shape as the price moves within it. Providers now have to choose a range, monitor it, and move it, which turns a deposit into an active strategy with a cost of management that the fee income has to cover as well.
For a trader taking liquidity rather than providing it, the effect is simply that depth around the current price is much better than it used to be and depth far from it is much worse. Large orders behave differently as a result: the first part fills well and the remainder can hit a range boundary and become expensive abruptly, which is a discontinuity that books do not have.
Books are better where there are participants willing to quote, which means liquid assets during active hours. A professional making markets can price a spread far tighter than any formula, because they can incorporate what is happening on other venues, adjust for their inventory, and step back when they are uncertain. On a major pair in the middle of the day, a book will beat a pool on cost almost every time.
Pools are better where nobody would bother to quote. A newly created asset, a pair that trades a handful of times a day, an hour when no professional is watching: in all of these a book is empty and a pool is not, because the formula does not require anybody to be present. That is the actual innovation, and it explains why pools dominate exactly the segment of the market where books were never going to work.
The practical rule that follows is unglamorous. For anything with real volume, compare the cost of your size on both and take the cheaper, which will usually be the book. For anything thin or new, the pool is frequently the only place the trade exists at all, and the question becomes whether the computable cost is acceptable rather than which venue is better.
Trading in a pool means a blockchain transaction, which introduces two costs a book does not have. The network fee is charged whether the trade succeeds or fails, so a transaction that reverts because the price moved still costs money. And the delay between sending and executing means the price you calculated is not necessarily the price you get, which is why every interface asks for a slippage tolerance and why setting it too wide is a way to be filled much worse than expected.
There is also a category of cost that only exists where transactions are publicly visible before they execute. A pending trade sitting in a public queue can be seen, and a party can place their own transaction ahead of it and behind it to profit from the price movement it will cause. Mechanisms exist to reduce this and none eliminate it, and it is a form of execution cost with no equivalent on a venue where orders are not broadcast before matching.
Against that, pools have one cost that books carry and they do not: there is no counterparty holding your balance. A trade against a pool settles atomically from your own wallet, which removes the platform failure modes entirely. That is a genuine structural advantage and it is the reason the comparison is not simply about execution cost.
Before trading into a book, the question is depth: how much rests within a tolerable distance of the touch, and whether your size is a meaningful fraction of it. Before trading into a pool, the question is the reserve: what the pool holds, and what the formula says your size will cost, which most interfaces will compute and display as price impact before you confirm.
The second is the one traders skip, and it is the easier of the two because the answer is exact rather than an estimate. Entering the size you intend and reading the impact figure takes seconds and removes the most common expensive surprise in this venue type, which is discovering after the fact that a pool that looked adequate was a fraction of the size required.
For liquid assets during active hours, the book almost always, because professional market makers can quote tighter than any formula. For thin or newly created assets, the pool is frequently the only place the trade exists. Compare the cost of your actual size on both rather than assuming either.
Because the pool's reserves are public and its formula is fixed, so the price for any size is arithmetic. A book's cost depends on resting orders that other people can cancel between your decision and your arrival, so the quoted price is an estimate of what you will pay rather than a guarantee.
The shortfall between holding a liquidity position and simply holding the two assets. The formula automatically sells whichever asset is rising and buys whichever is falling, so a provider ends up with more of the loser and less of the winner. It grows with price divergence and is only impermanent if relative prices return to where they started.
It is a position, not a yield. You profit from trading activity through fees and lose from price divergence through the rebalancing the formula performs. Both halves have to be measured together, and a pair whose prices diverge substantially can produce a loss despite healthy fee income.
It lets providers place capital within a chosen price range instead of across every possible price, which produces far more depth where trading actually happens. For providers it turned a passive deposit into an active position that stops earning outside its range. For traders it means better depth near the price and a sharp discontinuity at range boundaries.
Because a blockchain transaction is charged whether it succeeds or reverts. If the price moved beyond your slippage tolerance between sending and execution, the trade correctly refuses to complete, and the network fee has already been consumed by the attempt.
Tight enough that a bad fill is refused, wide enough that ordinary movement does not cause a revert. Setting it very wide converts a failed trade into a badly filled one, which is the more expensive outcome, and on a thin pool it is an invitation for somebody to sandwich the transaction.
They fail differently. A pool settles from your own wallet, so there is no operator holding your balance and no insolvency risk, and the exposure moves to the correctness of the contract. A book on a platform introduces counterparty risk and removes contract risk. Which is preferable depends on which you are better placed to assess.