Lending pools, market makers and liquidations, in plain terms
Three mechanisms do most of the work in decentralised finance, and all three are simple enough to follow with arithmetic. Here is what each one actually does, with a worked example of the price impact everyone underestimates.
Not financial advice. This article is for informational purposes only.
The short version
Most of decentralised finance reduces to three mechanisms. A lending pool lets anyone deposit and anyone borrow against collateral, with a rate set by how much of the pool is in use. An automated market maker holds two assets and quotes a price from their ratio, which means every trade moves the price along a curve. A liquidation sells collateral when it stops covering a debt. None of them require a counterparty who agrees with you — and none of them can pause when things go wrong.
Decentralised finance is often introduced through its vocabulary, which is the wrong order. The vocabulary is intimidating and the mechanisms are not. There are essentially three, they are built from arithmetic you can follow, and understanding them explains most of what happens when this part of the market breaks.
Mechanism 1: the lending pool
Ordinary lending needs a lender, a borrower and terms. A lending pool removes the matching problem: depositors put assets into a shared contract, borrowers take assets out against collateral, and nobody negotiates with anybody.
The interest rate is set by a formula, not a market maker. The input is utilisation — the share of the pool currently borrowed. Low utilisation means a low rate, because idle capital needs tempting borrowers. High utilisation means a high rate, because the pool needs to attract deposits and discourage further borrowing. Most implementations add a kink: beyond some level, the rate climbs steeply to protect depositors’ ability to withdraw.
Two consequences follow directly and matter more than any headline yield. First, the rate you see is not a rate you are promised — it recalculates continuously as others deposit and borrow, so a quoted yield is a snapshot. Second, if utilisation approaches its ceiling, depositors cannot all withdraw at once, because the assets are lent out. The high rate is the system asking for help, not a reward for cleverness.
Borrowing is always over-collateralised: you post more value than you take. There is no credit assessment because there is no identity — the collateral is the entire underwriting.
Mechanism 2: the automated market maker
A conventional venue matches buyers with sellers in an order book. An automated market maker holds a reserve of two assets and quotes a price derived from the ratio between them, so it will always trade — no counterparty needed.
The most common rule keeps the product of the two reserves constant. Take a pool holding 100 of an asset and 200,000 units of a stable asset. The product is 20,000,000, and the implied price is 2,000. Now buy 10 units:
| Step | Value |
|---|---|
| Reserves before | 100 and 200,000 |
| Constant product | 20,000,000 |
| Implied price before | 2,000 |
| Asset reserve after buying 10 | 90 |
| Stable reserve must become | 20,000,000 ÷ 90 = 222,222.22 |
| You therefore pay | 22,222.22 |
| Your average price per unit | 2,222.22 |
| Price impact | 11.1% |
You wanted 10% of the pool and paid 11.1% above the starting price. Nothing went wrong; that is the curve. This is why slippage on a small pool is brutal and why pool depth matters far more than any advertised fee. The fee might be 0.3%; the price impact above is more than thirty times that.
It also explains why arbitrage is structural rather than parasitic. The pool has no idea what the asset is worth — it only knows its own ratio. When the outside price moves, the pool is stale until somebody trades against it, and that somebody takes the difference. Providing liquidity to a pool means systematically selling the asset that is rising and buying the one that is falling. That is the honest description of what is often labelled “impermanent loss”, and there is nothing impermanent about it if prices do not come back.
Mechanism 3: liquidation
Every over-collateralised position has a threshold at which the collateral no longer comfortably covers the debt. Cross it, and anyone may repay part of the debt in exchange for a discounted slice of the collateral. The discount is the incentive; that is the whole design.
This works well in ordinary conditions and is exactly where the system is fragile in bad ones, for three reasons that all arrive together:
- Price feeds. A contract cannot see market prices; it is told them. During violent moves feeds can lag, and a stale feed either liquidates positions that were fine or fails to liquidate ones that are not.
- Block space. Liquidations are transactions and compete for the same congested blocks as everything else. A crash is precisely when fees spike, and a liquidation that cannot be included does not happen.
- Buyers. The discount only helps if someone wants the collateral. In a sharp decline the people who would normally take the other side are managing their own positions.
When all three bind at once, the protocol ends up holding debt worth more than the collateral backing it. See liquidation for the mechanics, and how forced selling interacts with resting orders for why the cascade accelerates.
What “decentralised” does and does not remove
It genuinely removes some things. There is no application to be approved, no minimum balance, no counterparty who can decline you, and the rules are readable code that executes identically for everyone.
It does not remove risk — it substitutes one set for another:
| Removed | Introduced instead |
|---|---|
| Counterparty refusing you | Contract bugs, which execute perfectly as written |
| Discretionary account freezes | No recourse, ever, including for your own mistakes |
| Opaque internal risk models | Dependence on an external price feed |
| A firm that can pause trading | Nobody who can pause anything mid-crisis |
| Credit assessment | Mandatory over-collateralisation |
That last row on the left is worth sitting with. The ability to halt is usually described as a flaw of traditional finance, and sometimes it is. It is also a circuit breaker. A system that cannot stop will keep liquidating into a market with no buyers, exactly as instructed, until there is nothing left to liquidate.
The three questions worth asking of any protocol
Where do prices come from, and what happens if the feed is wrong or late? What is the collateral, and is it correlated with what has been borrowed against it — because correlated collateral means the crash and the liquidation arrive together? And who can change the rules, how quickly, and with whose approval?
None of this is a recommendation to use any of it. These are high-risk systems in which total loss is a realistic outcome, including through mechanisms that involve no error on your part. Our disclaimer sets out the full position.
Key takeaways
- Lending pool rates are formulas driven by utilisation, so a quoted yield is a snapshot, not a promise.
- High utilisation means depositors cannot all withdraw — the high rate is a warning, not a reward.
- A constant-product pool moves price along a curve: buying 10% of a pool cost 11.1% above the start price in the worked example.
- Price impact frequently dwarfs the trading fee, so pool depth matters far more than the advertised rate.
- Providing liquidity means systematically selling what rises and buying what falls. “Impermanent” is optimistic.
- Liquidations depend on price feeds, block space and willing buyers — all three strained at the same moment.
- Decentralisation removes discretion, including the discretion to stop. Nothing can pause mid-crisis.