Market Maker Economics: Why Liquidity Leaves When You Need It

A market maker quotes a price to buy and a price to sell, earning the difference when both sides trade. The business is a race between spread income from traders with no particular view and losses to traders who know something. Every feature of market behaviour that frustrates other participants descends from that arithmetic.

The two counterparties

Uninformed flow — someone rebalancing, converting, or acting on no particular edge — is profitable to trade against. The maker buys at the bid, sells at the offer, and the spread is earned.

Informed flow is the opposite. Someone trading because they know the price is about to move takes the maker's quote precisely when it is wrong, and the maker is left holding a position that immediately loses.

The maker cannot distinguish them in advance. The spread is therefore priced so that the earnings from the first group exceed the losses to the second, in aggregate, over time. That single equation explains almost everything about how quoted markets behave.

Why spreads widen when it matters

In volatile conditions, the proportion of informed flow rises. People trade because something happened, and the maker's quote is more likely to be stale relative to information the other side already has.

The rational response is to widen the spread — charging more to compensate for being picked off more often — and to reduce quoted size, limiting how much can be taken at a wrong price.

Both happen automatically and simultaneously across every maker in a market, which is why depth evaporates rather than degrading gently. It is not coordination and it is not abandonment. It is the same calculation producing the same answer at the same moment.

The uncomfortable implication is that a market's liquidity is at its thinnest exactly when a liquidation engine most needs it to be thick, which is the structural reason cascades are possible at all.

Inventory, and the cost of holding it

A maker who buys must eventually sell, and while holding the position they are exposed to the market. Managing that inventory is most of the operational work.

In traditional markets, hedging is cheap and available around the clock. In crypto it is uneven: hedging instruments exist for major assets and thinly or not at all for the rest, which is why spreads on long-tail assets are wide regardless of trading interest. The maker cannot lay off the risk, so they charge for carrying it.

This also explains why liquidity does not follow attention. An asset with heavy retail interest and no hedging market will have poor depth, because the constraint is on the maker's ability to manage inventory rather than on demand for the asset.

Automated market makers face the same problem differently

A pool cannot widen its spread or pull its quotes. It provides the same curve regardless of conditions, which means it is picked off by informed traders continuously and has no defence.

That loss is the well-known cost of providing liquidity to a pool, and it is not an inefficiency to be engineered away — it is the same adverse selection that a human maker manages by quoting wider, borne instead as a persistent transfer.

Which is why fee income has to exceed it for pool provision to be worthwhile, and why concentrated liquidity, which allows repositioning, is best understood as giving pool capital some of the defensive ability a quoting maker has always had.

What this means in practice

• Expect depth to disappear during volatility and size positions on the assumption that it will.
• Treat displayed depth as an upper bound at all times and a poor one during stress.
• Understand that wide spreads on long-tail assets reflect hedging difficulty rather than lack of interest.
• For liquidation risk specifically, model exit liquidity at stressed levels rather than average ones.
• Recognise that both quoted and pooled liquidity face the same adverse selection, and only one of them can react.

None of this is a criticism of market makers. It is a description of a business that only functions because it is allowed to withdraw, and the withdrawal is the feature that keeps anyone willing to quote at all.