NFT Liquidation Shortfall Study and Protocol Risk

In this paper, we look to check the robustness of borrower and protocol protections within our collateralized lending offering.

A user deposits Non-Fungible Tokens (ERC-721) and/or Fungible ERC-20 tokens as collateral in order to borrow other ERC-20 tokens. The available Loan-to-Value ratios and various definitions can be found on our asset risk page.

The loan remains valid as long as the value of total collateral remains above the Liquidation Threshold as calculated by:

$$Liquidation \ Threshold_i \ * \sum Collateral \ in \ ETH \leq Floor_{TWAP}$$

Executive Summary

Our study shows that the existing mechanism for pricing and liquidating ERC-721-collateralized loans subjects the Protocol and Liquidity Providers to minor $Loss \ given \ Liquidation$. We have intentionally applied a smoothing factor to the Floor price in order to minimize the risk of unwanted loan liquidation—providing a better user experience to Borrowers. Notably, sharp market volatility in NFT prices and broader crypto tokens could force$Loss \ given \ Liquidation$ given Liquidation across a small minority NFT collections in our collateralized borrowing system.

Our conservative simulations show that our system is stable in ETH and USD terms with low risk of $Loss \ given \ Liquidation$​.

Liquidation and risks to the protocol and borrower

By definition ERC-721 tokens are non-fungible, they trade with lower liquidity relative to ERC-20’s. When used as collateral for loans in ERC-20 tokens, liquidation of these NFT’s could result with insufficient proceeds to cover the total loan value. Clearly we set low maximum Loan to Value ($LTV$) ratios to account for said risk.

In order to minimize the volatility of liquidation triggers and prevent undue liquidations, we likewise set a liquidation trigger threshold short of full $LTV$to also cover any potential shortfall in the process of liquidation.

To calculate maximum risk to the protocol we observe historical volatility in NFT collection floor prices and potential volatility as well as an additional margin for liquidation risk.

• Assume a user has borrowed the full LTV available given posted collateral valued at the Floor Price of the NFT collection. Here

$$MaxLTV = \sum_{\substack{i}} LTV_i \ * Collateral_i \ in \ ETH$$

For the purposes of this study we will assume the $Collateral$ is given by the value of the Non-Fungible Token in ETH terms and no further ERC20 tokens are provided as collateral. Further we assume that the $Price_j$ is equal to the $Floor_{Max}$ as measured by the Maximum $Floor_{TWAP}$ of the preceding 10 days.

$$Floor_{{TWAP}} = \frac{\sum_{\substack{j}}Floor_{j}T_{j}}{\sum_{\substack{j}}T_{j}}$$

$$MaxLTV = LTV_i \ * \ Floor_{TWAP}$$

• Measure the historical frequency with which that $Floor_{TWAP}$ indicates that the collateral has fallen below the minimum $Liquidation \ Threshold$ .

Liquidation occurs when the following conditions are met:

$$Floor_{TWAP} \leq \frac{MaxLTV}{Liquidation \ Threshold}$$

• Given the Liquidation trigger, observe the maximum $Loss \ Given \ Liquidation$ according to the Floor Price of the given NFT collection at the time.

$$Price_{Liquidation} = Floor_{Liquidation} *(1 - Liquidation \ Penalty)$$

$$Loss \ Given \ Liquidation = Price_{Liquidation} - MaxLTV$$

The Protocol will see shortfall if the $MaxLTV$ exceeds the total proceeds as noted by the $Price_{Liquidation}$.

We use the above assumptions and complement this with real NFT sales data to simulate the frequency with which the Protocol may see $Loss \ Given \ Liquidation$. Given the average monthly volume of a given NFT collection, we assume a new $MaxLTV$ loan is made with the frequency of an average sale.

For example, on-chain data shows CryptoPunks have sold 7,820 times through 2022-06-19, indicating that a sale occurs every 60 minutes. This is the frequency of the $MaxLTV$ loan in our model.

The table below summarizes data for all collections, while charts can be viewed for individual collections in the images which follow.

ETH Losses Given Liquidation

For this study we assume all NFT-collateralized borrowing is in ETH which itself is the quote currency for the NFT's themselves.

With the notable exception of Azuki, our model and liquidation mechanism provide protection against loan liquidation and $Loss \ Given \ Liquidation$.

The $Loss \ Given \ Liquidation$in Azuki will be the subject of further iteration on our Liquidation mechanism.​

Collection	Liquidation at Raw Floor Price	Liquidation at TWAP Floor Price	Total ETH Loss Given Liquidation
Azuki	1.7%	1.7%	-2,450
Bored Ape Yacht Club	0.0%	0.0%	0
CryptoPunks	0.0%	0.0%	0
Doodles	0.0%	0.0%	0
Mutant Ape Yacht Club	0.0%	0.0%	0
Bored Ape Kennel Club	0.0%	0.0%	0
VeeFriends	0.0%	0.0%	0
CloneX	0.0%	0.0%	0

USD Losses Given Liquidation

When we assume the Borrower uses their NFT's as collateral to borrow in USD terms the results show greater risk of Liquidation and a higher risk of $Loss \ Given \ Liquidation$​. With the clear exception of Azuki, however, our model and liquidation mechanism provides ample risk controls for our NFT lending Protocol.

Collections	Liquidation at Raw Floor Price	Liquidation at TWAP Floor Price	Total USD Loss Given Liquidation
Azuki	5.97%	5.60%	-$26,202,597
Bored Ape Yacht Club	0.01%	0.00%	$0
CryptoPunks	0.00%	0.00%	$0
Doodles	1.06%	0.87%	$0
Mutant Ape Yacht Club	0.52%	0.37%	$0
Bored Ape Kennel Club	0.02%	0.00%	$0
VeeFriends	0.00%	0.00%	$0
CloneX	0.66%	0.00%	$0

Study of Bored Ape Yacht Club

The Bored Ape Yacht Club has seen exceedingly rare liquidation triggers and ultimately none which would have, according to our assumptions, produced a $Loss \ given \ Liquidation$.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of Azuki

The Azuki collection is the largest historical loser in the selected collections given the notable fall in $Floor$ and the ETHUSD price through early May.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of Doodles

The Doodles collection would not have produced $Loss \ given \ Liquidation$despite notable volatility in its floor price and the price of ETHUSD.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of Mutant Ape Yacht Club

The Mutant Ape Yacht Club shows a modest risk profile for borrowers and the Protocol.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of CryptoPunks

According to our Liquidation model, the CryptoPunks collection would not have produced a loan Liquidation throughout the nearly-year-long price history in our sample.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of VeeFriends

The VeeFriends collection has historically shown itself stable as the basis for borrowing USD-denominated coins against the collection's NFT's.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of Bored Ape Kennel Club

Although more volatile than the Bored Ape Yacht Club collection, the Bored Ape Kennel Club has historically shown itself as stable collateral against USD-denominated borrowing.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$

Study of CloneX

According to on-chain data, the CloneX collection has historically shown itself as stable collateral against USD-denominated borrowing.

$\color{blue} Floor_{Raw}$ $\color{orange} Floor_{TWAP}$ $\color{purple} Liquidation \ Threshold$ $\color{red} MaxLTV$