Liquidity Pools vs Order Books: Crypto Trading Explained

Two ways to buy the same coin

Imagine walking into a market for mangoes. In one stall, there's a shopkeeper standing behind a crate, willing to sell at a price she recalculates after every single sale. In the next stall, there's a chalkboard where buyers and sellers pin up notes: "will buy at 40," "will sell at 42." Both stalls sell mangoes. The experience, however, is not the same.

Crypto has both kinds of stalls. A centralized exchange (CEX) runs an order book — a live, ranked list of buy and sell orders, matched by software that pairs the highest bid with the lowest offer. A decentralized exchange (DEX) using an automated market maker, or AMM, works differently: there's no list, no matching of two humans. Instead, a smart contract holds a pool of two tokens, and it quotes you a price according to a formula.

That formula matters more than it sounds. The most common one is the "constant product" rule, usually written as x × y = k. Here x and y are the amounts of each token sitting in the pool, and k is a number the pool tries to keep constant. If you buy token A out of the pool, x falls, so y must rise to keep k steady — and the price, which is just the ratio between the two reserves, drifts upward. No human repriced anything. The maths did.

Where the price comes from

In an order book, price is discovered. Buyers and sellers reveal what they're willing to pay, and the best overlapping pair becomes the last traded price. In an AMM, price is mechanical. It is simply the ratio of reserves, and it only moves when someone trades against the pool.

This creates an odd little dependency. An AMM has no idea what the rest of the world thinks a token is worth. So it relies on arbitrageurs — traders who spot when the pool's price has drifted away from the price on larger venues. If a pool is selling ETH cheaper than a big exchange is, arbitrageurs buy from the pool and sell elsewhere until the gap closes. That's not a flaw; it's the plumbing. The pool stays roughly honest because someone is always paid to correct it.

Slippage and price impact: the same trade, different bruises

Here's where your execution cost starts to diverge.

On an order book, if you buy more than the offers available at the best price, you eat into the next price level, then the next. That's slippage — the gap between the price you expected and the average price you got. Deep markets have shallow slippage: lots of orders stacked near the top means your trade barely registers.

On an AMM, the pool is a smooth curve rather than a ladder of orders. Every token you buy costs slightly more than the one before it, because you're shifting the reserve ratio as you go. This is price impact, and it's a function of two things: how big your trade is relative to the pool, and which formula the pool uses. A useful rough rule of thumb from market practitioners is that your price impact lands at roughly twice your order size as a share of the pool. Trade 1% of a pool, expect something near 2% impact.

There's a second cost layered on top. AMM pools charge a trading fee to whoever swaps — commonly 0.30% in the classic constant-product design — and that fee is paid to the people who supplied the tokens in the first place. Because the fee applies in both directions, it creates an effective spread of about 0.6% on the pair. Even the most liquid AMM trade carries that as a built-in headwind. On a large, busy centralized exchange, the equivalent fee might be a fraction of that. This is worth saying plainly: sometimes the best execution venue is not the decentralized one, and paying more for on-chain access is a choice with real value attached — self-custody, no account application, instant availability — but it is still a cost, not a free lunch.

Small comfort: you can reduce price impact by splitting a large order into several smaller ones over time, letting arbitrageurs reset the pool between each. You can also route through a DEX aggregator, which spreads your order across multiple pools at once. Both add gas fees and complexity.

Impermanent loss, explained without the drama

If you deposit tokens into a pool, you're a liquidity provider. You earn a share of trading fees. You also take on a risk with a slightly menacing name: impermanent loss.

It isn't a loss in the sense of money vanishing from your wallet. It's an opportunity cost. When the two tokens in your pool change in price relative to each other, arbitrageurs rebalance the pool, and you end up holding more of the token that fell and less of the token that rose. Compare that to simply holding both tokens in a wallet, and you'd have been better off holding. That difference is the impermanent loss.

It's called "impermanent" because if the price ratio returns to where you started, the gap closes. Withdraw while the ratio is skewed, and it becomes permanent.

How bad is it? A widely cited framing puts a 2x price move at roughly a 5.7% deficit versus holding, and a 5x move beyond 25%. Notice the shape: the pain grows faster than the price change. It's also asymmetric — a token falling hurts more than the same token rising helps.

But here's the part that often gets lost. Those numbers assume no fees. Once you include the fees paid by traders and arbitrageurs over time, the picture changes considerably. Research simulating a year of real on-chain trading on a major ETH/USDC pool found that liquidity providers actually came out ahead of a simple holding strategy as long as prices didn't drop more than about 75% or rise more than about 300% within the year. Price moves alone didn't guarantee losses; only permanent, unexpected and very large moves did.

The same research turned up something genuinely counterintuitive. Arbitrage activity — which sounds like it should be bad for liquidity providers — actually improved their returns. Why? Because arbitrageurs pay fees too. The rebalancing damage they cause stays roughly constant, while the fee income they generate compounds. Pools that welcome arbitrageurs, rather than trying to block them, tend to treat their liquidity providers better.

So the honest summary is: impermanent loss is real, it's structural, and it is frequently overstated by people quoting the no-fee number as if fees didn't exist. The net outcome is fees and incentives minus loss, and that subtraction has to be done honestly, pool by pool.

Execution risk, and the rules catching up

Order books and pools fail in different ways. An order book can thin out in seconds during stress, and off-chain order books on some venues reintroduce the very centralization decentralization was meant to avoid. AMMs can't run out of a token, but they can be sandwiched — a practice where someone spots your pending trade on a public blockchain and places orders around it to profit from the price move you're about to cause. That's a public-blockchain quirk, not something you can fully avoid with today's DEX designs.

Regulators are now working through these distinctions, and the landscape is genuinely in motion. The CFTC has opened a formal rulemaking process around retail crypto transactions and proposed a tailored registration category for exchanges handling them, while asking whether existing customer-fund segregation and bankruptcy protections actually fit these products — a question that is, as of the latest available material, still unresolved. The SEC has proposed its own framework and issued temporary, conditional relief to venues trading tokenized stocks using AMMs and liquidity pools, with conditions around volume limits, auditable contracts and coordinated trading halts. Meanwhile, industry groups have argued that AMMs should be regulated by the function they perform rather than the architecture they use, pointing to slippage, conflicts for liquidity providers, thin manipulation controls and unclear governance as features that would need addressing before these models touch regulated securities markets.

Two things follow from that. First, jurisdiction matters enormously — what's permitted, protected, or prohibited depends on where you are and which asset you're touching. Second, because much of this is being done through agency rulemaking and temporary exemptions rather than legislation, it can be revised by future commissions. Nothing here is settled.

One last note on cost, since it's easy to forget. Every AMM trade also pays the blockchain's own transaction fee, which can swing wildly depending on network congestion. On a quiet day that's pocket change. On a busy day, it's the mango stall charging you for the privilege of walking in.

What this means if you're the one trading

Neither design is better. They're differently shaped. Order books concentrate liquidity where people want to trade, which tends to give tight pricing on popular pairs. Pools spread liquidity across a continuous price curve, which means a token with no order book at all can still have a market — at the cost of that smooth curve working against you as your order grows.

So the practical work is comparison, not loyalty. Check depth before you trade. Compare the fee, the price impact, and the network cost against what a centralized venue would charge. If a pool is offering you a price several percent away from what other markets show, that gap is a signal, not a bargain. And if you're supplying liquidity rather than trading, model the fee income against realistic price divergence before you deposit — a stablecoin pair and a volatile pair are not the same decision wearing different tickers.

None of this is a recommendation about what to buy, sell or provide. It's simply the shape of the machinery. Understanding which stall you're standing in front of is most of the battle.