AMM vs market maker: why a pool cannot answer back

Every few months someone argues that automated market makers will replace professional market makers, and someone else argues that pools are structurally doomed. Both sides usually reach for volume share. Volume share cannot settle it, because it counts where trades printed and says nothing about what the capital behind those trades earned.

The difference between a pool and a quoter is not a matter of sophistication. It is a matter of what each one is allowed to do when it is losing. That is worth understanding properly, whether you are a token team deciding how to fund your market or an LP deciding where to park.

What an AMM and a market maker have in common

Both do the same job. Both quote a price to buy and a price to sell. Both earn the difference between the two. Both are trying to be there when someone wants to trade, and to not be there when someone knows something they do not.

That last part is the entire game, and it is where the two designs separate.

The cost that never shows up as a fee

A pool quotes from a formula. The market moves. The formula does not know yet. An arbitrageur takes the stale price, and the pool has sold something slightly too cheap or bought something slightly too expensive.

This happens continuously, and it is not an accident or a sign of a badly run pool. It is what the design does. The name for the cost is LVR, loss versus rebalancing: the gap between what the pool actually earned and what the same inventory would have earned if it could have rebalanced at the market price instead of the formula's price.

LVR does not appear on any dashboard as a line item. It is netted silently against fee income. An LP looking at fees earned is looking at revenue, not profit, and the largest expense is the one they cannot see.

A quoter on an order book has exactly the same problem. It also gets picked off when it is slow. In market microstructure this is called adverse selection, and it is the same phenomenon wearing different clothes.

So far the two are even. The difference is what each one can do about it.

What a quoter can do that a pool cannot

A quoter can widen when it is uncertain. It can pull its quotes entirely when something is about to happen. It can quote thin on one side and thick on the other because it is already long the asset and wants to get flat. It can skip a token for an hour, or a day, and come back. It can look at who has been trading against it and price accordingly.

A pool can do none of this. Capital committed to a pool sits at every price inside its range, all the time, available to everyone, at the same price for everyone. It cannot distinguish an informed buyer from someone rebalancing a portfolio. It offers both the identical quote.

This is not a flaw in the design. It is the design, and it buys something real. Passive capital is cheap to supply, it requires no operator, it never sleeps, and it will quote a market at three in the morning that no desk would staff. For long-tail assets and correlated pairs, that is worth a great deal.

But it means a pool takes whatever price information gives it. A quoter answers back.

The comparison that actually settles it

Strip away the argument and both sides are claiming something about the same ratio. Neither usually writes it down.

For a pool, per dollar of pooled capital, per unit of time:

fees earned, minus LVR, minus gas

For a quoter, per dollar of capital at risk, per unit of time:

realised spread, minus adverse selection, minus inventory carry

Same shape. Different plumbing. Every term in the first line has a counterpart in the second, and the honest question is simply which arrangement returns more per dollar committed, for a given asset, over a given horizon.

Almost nobody publishes either number. That absence is the actual state of the argument, and it is not a hard gap to close. It requires measuring rather than asserting.

Where passive capital goes wrong in practice

There is a second failure mode that gets less attention than LVR, and it is more common for small assets: passive capital ends up sitting where nothing trades.

When we measured the onchain footprint of 10 tokens that had recently lost a centralized listing, five of them had no pool at all. The other five were spread across 12, 12, 13, 16 and 20 separate pools each. In the worst case, $125,362 of $141,237 in pooled capital, 89% of it, sat in pools that did no volume at all over 24 hours.

That was the worst of the five, not the typical one. What the group does show is how uneven the outcome is. Two of the five did no onchain volume at all over 24 hours, across every pool they had. Another turned over $24,610 in the same window. Fragmentation is not uniformly ruinous. It is unpredictable, and nobody holding the position can tell which case they are in without measuring it.

That capital is doing nothing for anybody. It is not earning fees, because no trades touch it. It is not providing usable liquidity, because a buyer arriving at one of those pools finds almost nothing there. It is fragmentation, and fragmentation is the pooled-liquidity version of an inventory problem: the money is in the wrong place and cannot move itself.

A quoter would have consolidated it. A pool cannot, because a pool has no view about where it should be.

That measurement comes from 10 tokens selected because they had just been delisted, read from GeckoTerminal on 2026-08-17. It is a deliberately unflattering sample and not a random one, so treat it as an illustration of the mechanism rather than a base rate for onchain liquidity generally.

What this does not cover

We have not published our own LVR measurements. The mechanism above is well established and checkable against public research, but we are not offering a number for what a given pool loses to arbitrage, because we have not measured it and will not assert it.

The comparison also assumes you are choosing where to put capital that is already committed to being liquidity. It says nothing about whether providing liquidity in a given asset is a good idea at all, which is a separate and usually more important question.

Finally, the correlated-pair argument for AMMs is real and this post does not dismiss it. Pairing assets that move together shrinks the price path a pool has to absorb, which shrinks LVR. It shrinks the term. It does not remove it, and it does not give the pool the ability to widen, pull, or skew.

If you are deciding how to fund your market

The practical version of all this: passive liquidity is cheap and always on, active liquidity is expensive and can react, and most tokens need some of both. What you should refuse is anyone who tells you which is better without reference to what a dollar earns in each place.

If you want the same rigour applied to your own book, we measure spread, depth by price band, and order impact for token teams at no cost, and we will show you the method rather than a dashboard.

Related reading: market making or volume: what a token listing actually needs.