Evaluation of NFT Collections for Borrowing

The Parallel protocol provides the go-to interface for users to realize the true value of their NFT holdings. And a key goal is to list all NFT collections that meet our evaluation criteria for inclusion and minimize risks to the Protocol. We have taken a thoughtful and methodical approach to building out our list of supported collections.

We derive our selection from both quantitative (80%) and qualitative (20%) data points which produce a weighted score. A score closer to 100 shows a collection more likely to hold value and show lower risk to Parallel borrowers, lenders, and the protocol.

Collections	All-Time Volume	30-day Volume	Market Cap USD	Composite Score
Bored Ape Yacht Club	$2,303,309,946.82	$73,918,237.69	$1,144,204,000.00	100
CryptoPunks	$2,294,432,482.70	$41,210,434.22	$888,277,250.00	90.75
Mutant Ape Yacht Club	$1,580,986,602.26	$32,448,465.94	$397,189,300.00	85.5
Otherdeed	$987,249,603.22	$36,250,380.38	$536,814,250.00	80.5
CloneX	$703,290,051.83	$19,863,946.56	$284,190,300.00	71.25
Doodles	$508,750,781.44	$15,370,610.57	$159,455,550.00	68.25
Azuki	$778,441,939.14	$11,594,426.14	$144,305,450.00	59.75
Moonbirds	$560,172,379.59	$16,352,665.64	$234,695,450.00	57.75
Meebits	$507,154,137.37	$2,919,629.19	$170,892,300.00	55.0

Evaluation of an NFT Collection for Inclusion

Our data-driven methodology selects and ranks collections which add maximum value to our protocol and users, while limiting the Loss Given liquidation risks to the protocol. It’s likewise useful to view Aave’s methodology for risk factors in their evaluation of adding a specific token to their lending protocol. A key difference of course is that Parallel focuses more on NFT's than ERC-20's, and we have built a quantitative + qualitative scoring mechanism to rank NFT collections for inclusion.

Quantitative Analysis

Our ultimate goal is to limit the Parallel Protocol to the top collections which represent the most value to end users and limit risks to the Protocol. To focus exactly on value, we start with those NFT collections in the top 100 of USD transaction volume all-time. To control for more recent market conditions we likewise check those against the top 100 of the past 30 days. We collected on-chain transaction data via the Cryptoslam.io API as of June 30, 2022. And we evaluated each of these collections across the following criteria:

1. Liquidity Risk (25%):

   • USD Sales Volume - All time; weight = 15%

   • USD Sales Volume - 30 days; weight = 10%

2. Market Capitalization and Value to Parallel Protocol (25%):

   • Market Capitalization of Collection; weight = 20%

   • Number of Token Holders; weight = 5%

3. Volatility Risk (20%):

   • Volatility in the Floor Price of the collection given the metrics below; weight = 20%

     • Daily Value at Risk ($VaR$)

     • Daily $Expected \ Shortfall$​ or Conditional Value at Risk ($cVaR$)

Qualitative Analysis

In addition to quantitative metrics, we looked at project details that provide insight into the quality of the collections. We outline our criteria below:

1. Security Risk* - 20%

   1. Smart Contract risk - credible audits, collection age; weight = 10%

   2. Pricing oracle risk - deficiencies/vulnerabilities in marketplace listings; weight = 7.5%

   3. Centralization risk - use of centralized services, contract owner abilities; weight = 2.5%

2. Counterparty/Collection Risk* (5%):

   • Collection’s Founders + team - track record, investors; weight = 5%

3. Utility (2.5%):

   • Use cases; weight = 1.25%

   • Roadmap; weight = 1.25%

4. Community (2.5%):

   • Twitter followers and engagement; weight = 1.25%

   • Number of Daily Traders; weight = 0.625%

   • NFT owners who have not sold their first collection; weight = 0.625%

*Though nominally a minority share of a collection's score, any material deficiencies can disqualify a collection from inclusion as collateral on Parallel.

Liquidity Risk Analysis of NFT Collections

We look at data available to see lifetime transaction volume across all NFT collections and check that against transactions in the past 30 days. We then rank NFT collections on a scale of 1-10:

Collection	All Sales USD	30-day Sales	Liquidity Score	Market Cap USD	Holders	Supply	Value Score
Bored Ape Yacht Club	$2,293,928,507	$89,950,981	10	$1,144,204,000.00	6,430	10,000	10
CryptoPunks	$2,291,865,136	$42,103,210	10	$888,277,250.00	3,535	9,999	9
Mutant Ape Yacht Club	$1,576,149,744	$38,369,409	9	$397,189,300.00	13,105	19,423	8
Otherdeed	$982,387,078	$43,008,865	8	$536,814,250.00	34,326	100,000	8
Azuki	$776,763,294	$13,657,311	6	$144,305,450.00	5,151	10,000	6
CloneX	$700,155,515	$20,609,359	7	$284,190,300.00	9,527	19,312	7
Moonbirds	$555,819,800	$15,746,362	6	$234,695,450.00	6,602	10,000	6
Meebits	$506,787,165	$3,336,001	3	$170,892,300.00	6,598	20,000	6
Doodles	$505,809,831	$14,836,850	7	$159,455,550.00	5,169	10,000	6

Volatility Risk in NFT Collection Prices

Sharp declines in the Floor price of a given NFT collection could cause any loans collateralized by the same NFT’s to be liquidated. And just as critically, any especially drastic declines could also force Protocol losses if the liquidation value of collateral falls below the loan value. We discuss this at length in our liquidation shortfall research paper.

For the purposes of NFT Collection inclusion we want to model how much volatility puts us beyond our Liquidation Threshold and worse puts is at loss given liquidation. An individual user can borrow a $MaxLTV$ loan against $i$ collateral as defined by:

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

For the purposes of this study we assume the NFT represents all collateral posted—valued at the NFT $Floor$ price at $t_0$.

$$MaxLTV = Floor_{t_0} \cdot LTV$$

​The $MaxLTV$​loan is then in good standing as long as the Collateral remains above the $Liquidation \ Threshold$​ given by

$$Liquidation \ Threshold = \frac{ {Floor_t \cdot Liquidation}}{Total \ Borrow}$$

​Here $Floor_t$ is simply $Floor_{t_0} \cdot (1 - \Delta_{Floor})$ where $\Delta_{Floor}$ is the change in $Floor$ which would force a liquidation. If we hold $Total \ Borrow$ constant and solve for $\Delta_{Floor}$ we get the max change in price before liquidation is

$$\Delta_{Floor} = 1 - \frac{LTV}{Liquidation}$$

​For example an LTV of 0.3 and Liquidation ratio of 0.7 would tell us a $MaxLTV$ loan will be liquidated with a 57.1% decline in $Floor$ price with $Total \ Borrow$ held constant.

Value at Risk (VaR) and Expected Shortfall (cVaR) as key Guidelines for Volatility Risk

We model the Volatility Risk of an NFT Collection’s Floor price as part of our methodology. To do so, we take a concept commonly used in financial portfolio analysis in Value at Risk (VaR): the point estimate of a single-period loss of a given probability. Typically $VaR$ is used to say a portfolio or asset has an X% shortfall risk within a confidence interval—we will use 95%.

$$VaR_{\alpha} = (1-\alpha) \ quantile \ of \ r$$

​For a given $\alpha$ (e.g. 95%) we will calculate the returns at the $1-\alpha$ quantile (i.e. bottom 5th percentile) of returns. Or in other words, we'll look at the worst 5th percentile single-day loss in the NFT collection's floor price. For our model we will assume the $VaR_{95\%}$ represents a potential liquidation threshold trigger for a given NFT-collateralized loan.

Here we would like to go a step further, however, as $VaR_\alpha$ is simply a point estimate and does not account for the full tail risk—i.e. the expected loss below the $1-\alpha \ quantile$. We will assume the NFT-collateralized loan will be liquidated at $VaR_{95\%}$and thus model any $Expected \ Shortfall$ in liquidation with $Conditional \ Value \ at \ Risk \ (cVaR)$.

$cVaR_{\alpha}$measures the average loss below the $1-\alpha \ quantile$: the mean returns $r$ below $VaR_\alpha$:

$$cVaR_{\alpha} = E[r | r<VaR_{\alpha}]$$

​If we assume our $VaR_{95\%}$is the level at which the loan may be liquidated, the $cVaR_{95\%}$ then represents a view on potential Loss given Liquidation as discussed in our separate paper. In this we likewise have a potential marker for more-conservative liquidation thresholds for different levels of risk.

For our holistic NFT collection ranking we use top Blue-chip collections as a benchmark and produce a rating on relative downside risk as shown in the table below:

Row Labels	Daily VaR (95%)	Daily cVaR (VaR of 95%)	Rating
Bored Ape Yacht Club	-11.94	-15.46	10
VeeFriends	-12.45	-17.64	8
Mutant Ape Yacht Club	-13.34	-19.91	8
Meebits	-14.07	-19.29	7
CloneX	-15.29	-23.09	6
Doodles	-15.5	-23.49	6
CryptoPunks	-15.5	-23.49	6
Otherdeed	-17.54	-24.58	5
Azuki	-22.22	-36.37	3

For those NFT collections included in the protocol, the $1 - cVaR_{95\%}$ represents the maximum Liquidation Threshold we would implement for that collection.

Qualitative Analysis on NFT Collections

These measures are by definition more subjective, we want to remain cognizant of non-quantifiable risks and avoid Collections which NFT collectors do not deem trustworthy. Below are the ratings for our top selections:

Collections	Qualitative Score
Bored Ape Yacht Club	10
CryptoPunks	9.25
Mutant Ape Yacht Club	10
Otherdeed	10
CloneX	8.25
Doodles	9.25
Azuki	8.75
Moonbirds	8.75
Meebits	9

Para Space Protocol v1: NFT Collections

Based on research and scoring methodology outlined above, we’ve selected the following NFT Collections to include in the Para Space Protocol:

• Bored Ape Yacht Club

• CryptoPunks

• Mutant Ape Yacht Club

• Doodles

• Otherdeed

• Azuki

• CloneX

• Moonbirds

• Meebits

Each of these collections shows promise in terms of user value, utility, and ultimately relatively low risk to the Protocol.

Collection Collateral Limits

Once we approve a collection as collateral for borrowing we must also be careful of excess concentration risk--if the protocol owns too many NFT's from a given collection it could itself drive excessive volatility and liquidity risk. To that end we will target a maximum 1% of a given NFT collection's total supply as eligible collateral on Para Space. We see these as "Supply Cap" for NFT collections as listed on our Asset Risk page.

We may revisit and make this threshold more conservative or flexible with an eye on market and protocol risks.

Next Steps

As we continue to build the protocol, we will look at other metrics to add into our models and adjust as needed for collections that we add or update. This includes:

• Collateral bands:

  • ‘Blue-chip band’: Original collections are allowed higher LTV and Liquidation thresholds against their Blue-chip NFT collection tokens.

  • Secondary collections as part of ‘Higher-risk’ band with lower max LTV and lower Liquidation thresholds.

  • De-listing threshold for NFT collections which have fallen outside of our Risk parameters.

• Refined estimates of Floor prices and depth of collection liquidity

  • Better Floor estimates: There exist a number of competing data providers which propose improved Floor price and transaction data with true-value modeling.

  • Individual appraisal: Several of the same data providers likewise provide estimates for individual token values which would allow us the ability to provide competitive loans for higher-value individual tokens within a given NFT collection.

Appendix and Supporting Data

Below you will find collection-by-collection statistics on Floor price volatility as well as the distribution of Floor price returns.

Bored Ape Yacht Club

Bored Ape Yacht Club Historical Floor

BAYC Returns Distribution

Mutant Ape Yacht Club

Mutant Ape Yacht Club Floor returns

MAYC Distribution of Daily Returns

CryptoPunks

CryptoPunks Floor Volatility

Distribution of CryptoPunks Floor Returns

Otherdeeds

Otherdeeds Floor Volatility

Otherdeeds distribution of Daily Returns

CloneX

CloneX Floor Volatility

CloneX Distribution of Daily Returns on Floor Price

Doodles

Doodles Floor Price Volatility

Doodles distribution of daily returns in floor price

Azuki

Azuki floor price volatility

Azuki distribution of daily returns on Floor price.

Meebits

Meebits Floor Price Volatility

Meebits Distribution of Daily Floor Price Returns