Liquidations are the mechanisms taking place when someone’s collateral value decreases too much compared to their debt.

Below a certain ratio, people called liquidators can repay their debt on their behalf: this is a liquidation.

When they do so, liquidators get users remaining collateral at a discount compared to the market price in exchange.

This allows liquidators to make an instant profit (incentive), and the user “bad” debt to be cleared.

Lending markets and stablecoins like Maker’s DAI or Angle Borrowing module rely heavily on debt based positions.

Minted DAI or borrowed assets both represent a debt that is collateralized by some tokens, potentially very volatile.

These systems stay healthy thanks to their liquidation mechanisms: they ensure bad debt is erased when borrowers become under-collateralized.

However, giving incentives to liquidators take funds out of the lender+borrower system.

They also sell the recovered collateral, having a negative price impact on the market.

In the end, every debt protocol face a trade-off between giving a high enough discount to liquidators to make sure liquidations occur, and to preserve users’ capital.

Since the beginning of these debt-based protocols, different liquidation models have been tried out.

They all make different trade-offs depending on their use cases and the evolution of the sector.

Liquidations on protocols like Aave or Maker have important penalties for borrowers, redistributed between liquidators and protocols. This incentivizes borrowers to lower their LTV and avoid liquidations.

For example, borrowers lose at least a 13% penalty when being liquidated on Maker. From June 10th to 16th, that’s $6M+ taken out of borrowers pocket.

In Maker’s case, most of this amount goes to the protocol as surplus, to protect itself from bad debt.

This is not a bad thing, but still not ideal for the borrowers getting liquidated, losing more than 13% of their collateral upfront.

The other factor is the amount of debt to be repaid during liquidation: 50% on Aave, and 100% on Maker.

This eliminates unhealthy borrowers more quickly, but is less efficient for the system and has a bigger impact on the market.

It also has a bigger impact on the market as liquidators usually sell all their collateral profits for stablecoins.

Liquity has similar trade-offs, but kept liquidations as an internal part of the protocol through a new design.

They created a Stability Pool for LUSD, used to liquidate Liquity vaults. During liquidations, LUSD from depositors is exchanged for ETH at a discount.

This way, they profit from liquidations and their impact on the market is limited.

Other trade-offs are still being explored, where borrowers are not as penalized for liquidations but liquidators compete for smaller discounts for example.

Part 3.1: we are seeing new liquidation mechanisms with different trade-offs

For example, Angle, inspired by Euler, takes a lower and variable liquidation penalty, favoring borrowers.

The amount of debt to be repaid and the liquidation penalty both increase linearly while the position becomes less collateralized.

On the other side, liquidators face a trade-off: the longer they wait, the higher the discount, but the first one to liquidate takes all the profits, similar to a Dutch auction mechanism.

Another specificity in this design is that different discounts are given to different liquidators, depending on their holdings.

In Angle for example, a better discount can be given to liquidators with veANGLE.

In the end, the first to liquidate will be the one willing to accept the smallest discount while still making a profit.

This result in much lower discounts, as we have seen in the recent Euler liquidations: https://twitter.com/MacroMate8/status/1536370171387863046 https://dune.com/shippooordao/Euler-Finance-Dashboard

Thanks to this mechanism, borrowers keep more of their capital, which stays in the lender+borrower system.

In the case of very big positions, liquidators could accept even lower discounts as long as they remain profitable.

Another interesting trade-off would be to decrease the size of liquidations, but increase their quantity.

The health factor / LTV after liquidation would be lower, and a smaller drop in price would cause another liquidation.

For example, instead of 50% of the debt being liquidated once, 10% would be liquidated ~5 times.

In fact, depending on the size of the initial price drop, the number of liquidations wouldn’t have to be 5, but somewhere between 1 and 5.

On one side, this would lower the market impact and price slippage for liquidators.

On the other, it would increase the number of transactions needed to liquidate a similar amount of debt, potentially congesting the network.

The minimum debt size would also have to be increased to make sure that

liquidation profits always cover gas costs.

However, this condition would be much lower in more affordable networks, like rollups or sidechains.

In Angle as in Euler, the amount of debt to be repaid is directed by a parameter called the Target Health Factor, which can be modified by governance.

Pushed to the extreme, we could imagine this trade-off becoming almost a continuous liquidation process, with a much smaller portion of the debt being repaid each time.

Liquidations play a key role in DeFi. Without them, many important protocols couldn’t function properly.

That said, only a few models have been experimented with so far, and they still have a big impact on the market as seen recently.

With this new liquidation model, more funds are left for borrowers and liquidations should happen more dynamically, leaving less unnecessary funds from borrowers to liquidators.

This is far from the end in exploring everything that’s possible around liquidations.

Finding ways to reduce their impact on the market would make them much more efficient, while protecting it from liquidation cascades.

This could be done in Angle vaults by lowering their Target Health Factor, something possible on more affordable networks.

For now, maintaining a sizable healthy liquidation system on-chain relies heavily on having enough liquidity available, when everyone needs it the most.