NFTs have exploded in popularity in 2021 capturing the interest of crypto investors, collectors and a wide range of new users. The market was estimated to be worth approximately $41 billion at the end of last year. What’s even more astonishing is that this outstanding growth has been achieved relying on market infrastructure and trading tools that are still somewhat in their infancy. Even OpenSea, founded in late 2017 and today by far the largest NFTs marketplace, offers only basic trading capabilities, essentially limited to buying and selling at a set price or through an auction.
It’s fair to say we are still scratching the surface in the development of NFT financialization. The highly anticipated convergence of NFTs and DeFi mechanisms would allow NFT collectors to stake NFTs for rewards, as well as lend and borrow against the value of their NFTs. However, the ability to evaluate, price, and assign NFTs a fair market value, among other factors, appears to be the biggest barrier to unlocking the financialization of NFTs.
At the end of last year, we formed an Innovation Lab within ConsenSys, dedicated to researching emerging trends and developing forward-looking concepts and prototypes within the cryptospheres and metaverses. Recently, our team set eyes on the current state of the NFT market and its convergence with DeFi. We started “Project Crystal Ball,” an exploration of the problems and opportunities in this space, starting with what’s arguably a stepping stone for NFTs valuation and pricing: rarity and rankings.
Currently, the most common method used to baseline and value individual NFTs within a collection is by assessing and computing their “rarity.” Rarity refers to how uncommon an NFT is within a collection, and it has been a critical element to the hype and valuation around some of the most notable NFT drops.
Most collections are created by deploying generative art algorithms that compose preset inputs, the “traits” designed by the artist, to programmatically mint unique NFTs. Assessing the distribution and combination of traits within the collection allows one to determine the rarity of each NFT and rank them in respect to one another.
For collectors and investors, understanding rarity and how it affects the value of an NFT is critical before making a purchase or listing an item for sale. Assessing the rarity of an NFT might sound simple, but when dealing with large collections that contain thousands of individual items composed of multiple traits each, calculating and scoring the rarity of each NFT requires complex statistical calculations.
For this reason, accompanying the astonishing growth of NFTs over the past years, we have also witnessed the development of multiple methods and tools to calculate and compare the rarity of NFTs.
We see 4 approaches, of which the first 3 are incremental improvements on individual trait rarities and the last one based on the Jaccardian distance being more data intensive and based on a data science technique applied in many fields. To get a quick glimpse of these approaches, let’s take a collection of 1000 cats with 4 traits (Eyes,Ears, Nose and Face) and 4 possible trait values Blue, Green, Red or Brown for each of the traits. Let’s also take 3 cats — Cat A, Cat B and Cat C — with the traits shown for this illustration.
In the simple aggregation based on trait value rarity approach, the individual traits rarities would be applied to figure the overall rarity of the NFT. So, Cat A and Cat B will have the same overall rarity as shown below
Cat A rarity = 100/1000 * 200/1000 * 200/1000 * 200/1000 = 0.0008 Cat B rarity = 200/1000 * 200/1000 * 100/1000 * 200/1000 = 0.0008
In the trait-type weighted aggregation approach, the trait-types that have fewer possible trait-values carry a higher weight over trait-types that have more possible trait-values. In our example, Eyes are better than Nose because Eyes can have one of 3 possible values and Nose can have one of 5 possible values. Thus, Cat A’s Blue Eyes weigh more than Cat B’s Blue Nose.
A rarity score (higher the better) computed based on these weights as shown below would show Cat A is rarer than Cat B.
Cat A weighted rarity score = (1000/100) * 8.33 + (1000/200) * 5 + (1000/200)* 5 + (1000/200)* 6.25 = 164.55
Cat B weighted rarity score = (1000/200) * 8.33 + (1000/200) * 5 + (1000/100)* 5 + (1000/200)* 6.25 = 147.90
In the meta trait incorporated scoring approach, a meta-trait not originally noted by the artist is included in the rarity scoring. To illustrate this, let’s look at Cat C and its weighted rarity score.
Cat C weighted rarity score = (1000/700) * 8.33 + (1000/100) * 5 + (1000/100)* 5 + (1000/400)* 6.25 = 127.525
This puts Cat-C at the bottom of the list. Now, if we added a meta-trait into the scoring mix,
Now computing the weight attribution for this meta-trait,
Cat A weighted rarity score = (1000/100) * 8.33 + (1000/200) * 5 + (1000/200)* 5 + (1000/200)* 6.25 + (1000/800)* 8.33 = 174.96
Cat B weighted rarity score = (1000/200) * 8.33 + (1000/200) * 5 + (1000/100)* 5 + (1000/200)* 6.25 + (1000/800)* 8.33 = 158.31
Cat C weighted rarity score = (1000/700) * 8.33 + (1000/100) * 5 + (1000/100)* 5 + (1000/400)* 6.25 + (1000/1)*8.33 = 8457.53
The last one is a holistic approach scored based on Jaccardian distance of an NFT from each of the other NFTs in the collection, which is the method that NFTGo uses. NFTGo explains their method in more detail in this article.
Our Experiment | Project Crystal Ball
Building on this background research, we built the first phase of Project Crystal Ball with a focus on NFT ranking to test with a varied group of enthusiasts, amateur, and seasoned collectors to:
- Learn if collectors cared about NFT ranks and the different ranking methods used.
- Observe how collectors responded when the scores from two ranking methods were different.
- Tease collectors with a price estimate (more on this in our next blog about NFT Pricing) and see how they would use it.
- Know if collectors would care if we showed more elaborate details behind the scores (e.g. what traits mattered).
- See if collectors would explore similar and different NFTs if we showed them (this was a by product of the Jaccardian similarity or the Standout Rank as we called it)
Our Findings
As the result of a bunch of user interviews based on our prototype and follow up discussions, we could distill these key findings:
- Collectors spend most time picking a collection to ape into (vs picking an NFT within a collection). This is an opportunity space and there aren’t a lot of tools available for this.
- Ranks and Ranking methods within a collection don’t matter unless the specific NFT is at the top end of the rarity spectrum of the collection.
- Price estimates, though meant to be a teaser, turned out to be super important for collectors. They used it to make offers, evaluate potential upside, and often used it in conjunction with prevailing floor prices in the collection.
- Collectors look for traits that they connect with, and like the ability to navigate and explore similar and different NFTs based on their traits.
Next Steps
We are building on these findings as we progress to the next part of the experiment: NFT pricing. We are working with a wide variety of partners, collectors, and investors to frame this experiment and validate our assumptions and user journeys.
About the ConsenSys Innovation Lab
We recently established our Innovation Lab to continue exploring the forefront of the cryptospheres and metaverses.
As a space capsule wandering in space far away from its mothership, we are a small team of ConsenSys’ veteran builders, designers, and operators on a mission to explore, conceptualize, prototype, and evangelize new ideas that can further the adoption of Ethereum and other decentralization technologies. We aim to research, build, and test in public. We seek ideas, feedback, and collaboration from the communities we will meet through our explorations. Find out more at www.consensys.net/innovation-lab and if you are interested in participating in our experiments, or want to collaborate, drop us an email at [email protected]
