Advancing Liquidation Mechanisms: The Journey to Dynamic Spread Tiered Liquidation (DSTL) in Fathom Protocol

Maintaining the stability and solvency of protocols is crucial in the dynamic world of decentralized finance (DeFi). A key tool for this is liquidation, which manages under-collateralized loans. The main liquidation strategies are Auction Liquidations, as seen in MakerDAO, and Fixed Spread Liquidations, used by Aave. Both have advantages but present challenges concerning price discovery, incentive alignment, and responsiveness to fast-changing market conditions.

As a progressive lending platform, Fathom Protocol is at the forefront of exploring innovative solutions to improve the liquidation process. Our journey led us to conceptualize and develop the Dynamic Spread Tiered Liquidation (DSTL) model, aiming to blend the advantages of Fixed Spread Liquidation with novel features that mitigate its limitations.

Existing Liquidation Strategies

Auction Liquidations [1] [2] [6]

Auction liquidations are common in platforms like MakerDAO. When a loan’s collateral falls below a certain threshold, the protocol triggers a liquidation event, initiating an auction to sell the borrower’s collateral to cover the debt.

MakerDAO’s Liquidation 2.0 employs Dutch auctions to address some challenges of the earlier English auction model in Liquidation 1.2. This change streamlines the liquidation process and reduces price volatility risk for participants.

Working Mechanism

1. Trigger:

A liquidation is triggered when the Health Factor (a metric indicating the safety of a loan) of a loan falls below 1 due to a decline in collateral value.

2. Price Decline Over Time:

• Unlike English auctions where bidders raise the price, Dutch auctions start at a high price, decreasing over time until a liquidator accepts the price and executes the liquidation.
• The price is calculated based on the initial price and the time elapsed since the auction began.

3. Possible Instant Settlement: Liquidators can settle instantly at the current price, eliminating the need for a prolonged bidding process.

Example

Suppose a user has a loan secured by 10 ETH as collateral. The debt is 5000 DAI, and the liquidation ratio is 150%. If the price of ETH falls significantly, the collateral value might drop, making the loan under-collateralized and triggering a liquidation event.

Let us assume that at the time of liquidation, the initial price of ETH is 500 DAI. According to the auction’s rules, the auction might start at a higher price, say 600 DAI per ETH, to encourage early settlement. This price will decrease over time, following a predetermined curve, say linearly, over 6 hours.

1. Auction Start:

• The auction starts with the price of 1 ETH set at 600 DAI.
• The total collateral value at auction start is 10 ETH * 600 DAI = 6000 DAI.

2. Price Decline:

• As time progresses, the price of 1 ETH decreases. Say it drops to 550 DAI after 1 hour, 500 DAI after 2 hours, and so on.

3. Liquidator Participation:

• At any point during the auction, a liquidator can accept the current price and settle the auction.
• Suppose a liquidator decides to settle the auction 2 hours in when the price of 1 ETH has dropped to 500 DAI.
• The liquidator pays 5000 DAI to cover the debt and receives 10 ETH.

4. Auction Settlement:

• The auction is settled instantly, with the liquidator receiving the collateral at the agreed price of 500 DAI per ETH.

5. Protocol Recovery:

• The protocol has recovered the full debt of 5000 DAI while the liquidator has acquired 10 ETH at a total cost of 5000 DAI, potentially making a profit if the market price of ETH rebounds or stabilizes above 500 DAI.

Pros

• Price Discovery: The competitive bidding process allows for better price discovery, potentially achieving a fair market price for the collateral.
• Potential Value Recovery: In favorable market conditions, auction liquidations can recover a substantial portion or even the full value of the outstanding debt.

Cons

• Time-Consuming: The auction process can take several hours, which might be detrimental in rapidly falling markets.
• Potential Undervaluation: If the market is moving quickly, the preset price decline might result in the undervaluation of collateral, leading to less recovered value for the protocol.
• Complexity: The auction mechanics can be complex and may deter participation from some liquidators.

By transitioning to Dutch auctions in Liquidation 2.0, MakerDAO aimed to address the time and price volatility challenges associated with the previous English auction model. However, despite these improvements, the potential for undervaluation and less competitive bidding still needs to be improved. Moreover, the challenges have remained.

Fixed Spread Liquidations [3] [4] [6]

Platforms like Aave use a fixed-spread liquidation model for quicker and simpler liquidation processes than auction-based models.

Liquidators can immediately buy the collateral at a predetermined (fixed) discount, often known as the “liquidation spread.”

Working Mechanism

1. Trigger Point:

A loan is flagged for liquidation when the Health Factor, a measure of the loan’s safety, falls below 1 due to a drop in the collateral value or an increase in the borrowed amount.

2. Liquidation Call:

Once a loan is flagged, any user (typically liquidation bots) can call the liquidation function on the smart contract.

3. Fixed Discount:

The collateral is sold to the liquidator at a fixed discount rate, known as the liquidation spread, which is predetermined by the protocol. This spread compensates the liquidator for the risk and cost involved in the liquidation process.

4. Close Factor:

The model employs a fixed close factor, say 50%, which dictates the portion of the borrower’s position that the liquidator attempts to liquidate. This close factor is applied alongside the liquidation spread.

5. Debt Repayment:

The liquidator pays off a portion (determined by the close factor) or the entire outstanding debt of the borrower and receives the equivalent value in collateral at the discounted price.

6. Remaining Collateral:

The remaining collateral is returned to the borrower if the liquidation does not cover the entire debt.

Example

Let us consider a scenario where a user has taken a loan of 4000 USDC, backed by 10 ETH as collateral.

Assume the initial price of ETH is 500 USDC, and the total collateral value is 5000 USDC.

Assume a Liquidation Threshold of 80%. This threshold indicates the point at which the loan’s collateral is considered insufficient, prompting a liquidation event. In our case, the maximum borrowing capacity is 4000 USDC (5000 USDC * 80%), health factor is 1 (4000 USDC / 4000 USDC). This means any price drop for the user will trigger liquidation.

1. Market Decline is triggering liquidation:

If the price of ETH drops to 440 USDC, the total collateral value becomes 4400 USDC. The loan becomes eligible for liquidation as the maximum borrowing capacity now is 3520 USDC (4400 USDC * 80%), health factor is 0.88 (3520 USDC / 4000 USDC) < 1.

2. Liquidation Process:

A liquidator identifies this opportunity and decides to liquidate the position.

Assume the liquidation spread is 10% and the close factor is 50%. So, the liquidator can purchase the collateral at 400 USDC per ETH (440 USDC / (1+10%)).

3. Debt Repayment:

The liquidator pays off 50% of the 4000 USDC debt, i.e., 2000 USDC, to the protocol, adhering to the close factor.

In return, the liquidator receives ETH equivalent to 2000 USDC at 400 USDC per ETH, which equals 5 ETH (2000 USDC / 400 USDC).

4. Profit for Liquidator:

The liquidator’s profit comes from the difference between the market price of ETH and the discounted price they paid. If they sell the acquired ETH at the market price of 440 USDC per ETH, they profit 40 USDC per ETH for a total profit of 200 USDC (5 ETH * 40 USDC).

5. Post Liquidation Scenarios:

Scenario 1 (Incomplete Recovery): If the Health Factor remains below 1 post-liquidation, another liquidation event is triggered.

The remaining debt is 2000 USDC with 5 ETH left as collateral. Assuming the market price of ETH remains at 440 USDC, the total collateral value is 2200 USDC. The maximum borrowing capacity now is 1760 USDC (2200 USDC * 80%), health factor is 0.88 (1760 USDC / 2000 USDC) < 1.

It could not be needed if, for example, the close factor was 80% and the fixed spread was 5%. In that case, the remaining debt would be 800 USDC with about 2.36 ETH, with a total profit for the Liquidator of approximately 160.06 USDC. Assuming the market price of ETH at 440 USDC, the total collateral value is 1038.4 USDC. The maximum borrowing capacity is 830.72 USDC (1038.4 USDC * 80%), health factor is 1.038 (830.72 USDC / 800 USDC) > 1.

Scenario 2 (Overliquidation): If the market price of ETH quickly rebounds after the liquidation, the position might have been over-liquidated, causing a significant collateral loss for the borrower and potentially impacting market dynamics.

Pros

• Speed: Fixed spread liquidations are executed quickly, reducing the protocol’s exposure to falling asset prices.
• Simplicity: The model is simple and easy to understand, which could encourage more participants.

Cons

• Lack of Price Discovery: The fixed discount may not reflect the current market conditions, potentially leading to unfair liquidation values.
• Inadequate Incentives: The fixed spread might need to provide more incentives for liquidators during high volatility.
• Incomplete Recovery: If the fixed close factor results in only a partial liquidation and the Health Factor remains below 1, another liquidation is required. This could lead to multiple liquidation events for a single position, each incurring additional costs and potentially further eroding the borrower’s remaining collateral.
• Overliquidation: Conversely, the fixed close factor could cause over-liquidation, where more collateral is liquidated than necessary to return the position to a safe Health Factor. This results in a loss of collateral for the borrower and could also impact the market negatively. Moreover, over-liquidation may deter future liquidations as the potential profits for liquidators diminish.

Comparative Analysis of Dutch Auction Liquidation Model and Fixed Spread Liquidation Model

1. Complexity and Understandability

• Dutch Auction: This model can be relatively complex for participants due to its dynamic pricing mechanism that decreases over time until a liquidator accepts the current price.
• Fixed Spread: The model is simpler and easier to understand, employing fixed parameters like the liquidation spread and close factor.

2. Speed of Liquidation

• Dutch Auction: The auction’s duration can extend over several hours, which could be detrimental in highly volatile market conditions.
• Fixed Spread: Liquidations happen instantly once triggered, which can be advantageous to secure the protocol’s funds quickly during market downturns.

3. Efficiency

• Dutch Auction: Efficiency might be compromised if the market price rebounds quickly or if liquidation is delayed due to the auction process.
• Fixed Spread: This model tends to be more efficient as it allows instant liquidation at a known discount, making it quick and predictable.

4. Potential for Overliquidation

• Dutch Auction: Dynamic pricing could lead to over-liquidation if market conditions change rapidly during the auction process.
• Fixed Spread: The fixed close factor could cause over-liquidation or under-liquidation, depending on market conditions at the time of liquidation.

5. Profitability for Liquidators

• Dutch Auction: The profitability can vary significantly based on the time of liquidation and market conditions.
• Fixed Spread: Liquidators clearly understand the profit margin due to the fixed liquidation spread, making it more predictable.

6. Market Impact

• Dutch Auction: The extended auction process could have a more pronounced impact on market prices, especially in illiquid or thinly traded markets.
• Fixed Spread: Instant liquidations might have a lesser market impact and potentially stabilize or restore confidence more quickly in the protocol during adverse market conditions.

Favoring Fixed Spread Over Dutch Auction

The Fixed Spread Liquidation model appears more favorable considering the speed of liquidation, simplicity, and predictability. Acting swiftly to secure the protocol’s funds is crucial in a volatile market. The Fixed Spread model provides a straightforward and fast liquidation process, essential for ensuring the protocol’s solvency and reducing systemic risk during market downturns.

Introducing Dynamic Spread Tiered Liquidation (DSTL)

The need for Fixed Spread Liquidation improvement

The Fixed Spread Liquidation model has downsides like Lack of Price Discovery, Inadequate Incentives, Incomplete Recovery, and Overliquidation. This leads to the proposition of the Dynamic Spread Tiered Liquidation (DSTL) model, aiming to meld the benefits of the Fixed Spread Liquidation Model (FSLM) while introducing innovative mechanisms to address its shortcomings.

Core Components of DSTL

1. Dynamic Liquidation Spread [5]:

- Implement a market-responsive liquidation spread to ensure fair liquidation value and adequate incentive for liquidators. This feature enhances the adaptability of the liquidation process to market conditions, addressing the fixed discount (liquidation spread) issue in FSLM.

2. Dynamic Close Factor [5]:

- Allow the proportion of debt repayable in a single liquidation to vary based on market conditions and asset rankings, addressing the over-liquidation and incomplete recovery issues seen in the fixed spread model.

3. Batch Processing of Liquidations:

- Bundle multiple liquidations into a single transaction to optimize gas fees and enhance efficiency, thus addressing cost-related concerns.

4. Asset Bundling [5]:

- Create bundled assets or index funds to diversify risks and unlock more value from long-tail assets. This feature enhances protocol resilience by promoting asset diversification.

5. Tiered Partial Liquidation:

- Establish a tiered system wherein the amount of collateral liquidated increases incrementally if the Health Factor remains below 1 post-initial liquidation. For instance, the first liquidation may take 20% of the collateral, the second (if required) takes 25%, and so forth. This mechanism ensures a position returns to safety more rapidly during volatile market phases, addressing the issue of multiple liquidation events.

Conclusion

The DSTL model represents a significant stride towards creating more adaptive, efficient, and fair liquidation mechanisms within the DeFi ecosystem. By addressing the limitations inherent in existing liquidation models and introducing innovative features like tiered partial liquidation and asset bundling, DSTL sets the stage for a more resilient and user-friendly DeFi liquidation framework.

Fathom Protocol: Current State and The Road Ahead

Currently, Fathom Protocol has successfully implemented the batch processing feature of the DSTL model, significantly enhancing the efficiency and cost-effectiveness of liquidations on the platform. However, this is just the beginning.

The upcoming Fathom Protocol 2.0 upgrade is poised to integrate the remaining core components of the DSTL model. With the introduction of Dynamic Liquidation Spread, Dynamic Close Factor, Asset Bundling, and Tiered Partial Liquidation, we aim further to bolster the resilience and user-friendliness of our liquidation framework.

References

[1] Maker Foundation. 2022. Liquidation 2.0 Module.

[2] Maker Foundation. 2022. The Auctions of the Maker Protocol.

[3] Aave Protocol. 2023. Liquidations.

[4] Alpaca Finance. 2023. AUSD Liquidation.

[5] Liquidation systems overview (design, code, comparing) [part 1]. 2022. Stable Unit.

[6] K.Qin, L. Zhou, P. Gamito, P. Jovanovic, A. Gervais. 2021. An Empirical Study of DeFi Liquidations: Incentives, Risks, and Instabilities