The primary project objective is to establish a robust and defensible methodology for calculating Non-Fungible Token (NFT) market capitalizations on an ecosystem basis.
NFTs have grown rapidly in the arts and collectibles space in the last 9 months. As far as we know, our work is the first calculation of the value of all NFTs within a specific NFT ecosystem to create a “market capitalization” for that ecosystem.
The secondary goal is to establish a robust and automated methodology for valuing individual NFTs with relevant confidence intervals
Having robust and accepted valuation tools in the NFT space is crucially important for artists, collectors and investment managers.
NFTs are already a multiple billion-dollar market. Our current valuation of the CryptoPunk ecosystem alone is $4.80B.
NFTs, however, are at a nascent stage. Our view is that they will grow to absorb a meaningful percentage of the art/collectible/community market and then extend to a whole array of non-fungible real world goods and services.
In other words, by the end of the decade, we believe NFTs will be a multi-trillion dollar market.
It is critical for this ecosystem for robust valuation methods to emerge and we hope to play an important role in this regard, particularly given the potential for composability between DeFi and NFTs.
NFTs and DeFi work using open protocols that can interact permissionlessly with each other (“composability”). This means more sophisticated, dynamic and capable systems can be built to finance NFT-based projects than with off-line projects.
It is critical for this ecosystem to emerge to have robust automated highly available valuations that can be fed through oracles into a variety of smart contracts.
The most obvious challenge is that they are non-fungible so valuation models have to take into account the general price level in an ecosystem and then adjust for the varying attributes of each NFT.
Calculations of fungible token market capitalizations is trivial by comparison and, even in that case, there are significant debates relating to token supply and wash trading that take place.
The NFT ecosystem valuation is at least an order of magnitude more difficult. We believe this can only be solved with machine learning. The goal of this project is to test various methodologies and determine which approaches produce the best results in this field.
There is a secondary challenge that relates to the large volatility swings in crypto-asset prices in general over time, which also complicates the modeling.
We plan to add other important NFT projects in the coming weeks focusing on those with meaningful market size and liquidity.
We hope in a few weeks to have a reasonably accurate market size for the NFT market as a whole.
Not yet! At this stage, we are comfortable that the overall market capitalization is broadly correct. The individual CryptoPunk level valuations should be considered as “beta” as this point. We plan to further fine-tune the model over time and determine its predictive value going forward.
We expect in a few months to be able to give stronger assertations about the model outputs, including confidence intervals of our predictions.
Very possibly. See above!
Absolutely not. Crypto-assets are extraordinarily volatile and art-oriented NFTs at this stage are probably more volatile than crypto-assets in general.
Anyone investing in NFT art should study the risks carefully and be aware that he or she may lose a large amount of his or her investment in ETH or USD terms.
We used all sales on the LarvaLabs.com website from January 1, 2021 to May 8, 2021.
We did not incorporate OpenSea sales of Wrapped CryptoPunks, but we are examining if we can add them into the model in the coming weeks.
We do not adjust for sales that are recorded publicly on LarvaLabs.com but may have been part of an off-line transaction where other assets or cash were exchanged off-line. In other words, we took all sales as reported on the LarvaLabs.com website
The current CryptoPunk valuation model is a weighted average of the output of certain Linear Learning, XGBoost and Tweedie machine learning models.
The models are trained using the attributes of CryptoPunks and the sales values from LarvaLabs.com
We consider our current model to be an acceptable initial model but believe there is significant room for improvement. Once we have coalesced on a more stable/definitive model, we will release more details about the methodology.
We have not at this point. This is a long-standing problem in illiquid markets and will require further research. Generically, the solution will have to be managed through confidence intervals and ‘margin for error’ in NFT valuations, as opposed to hoping it can be eliminated altogether.
The CryptoPunk valuation is updated daily taking into account the sales of the last 24 hours. We aim to provide an ongoing estimate of the market capitalization of an NFT ecosystem.
We will review if over time we will re-calculate on a faster cycle. Our assessment is that right now there is not sufficient liquidity to make material differences in the model on a sub-day basis.
The base currency for the initial model is ETH as we believe this is the unit of account in the CryptoPunk community. We plan to also test models denominated in USD terms to see if USD prices have higher or lower predictive value.
ETH-USD conversions on the website are real-time based on prices from CryptoCompare.com.
The algorithm can only take into account sales that have happened. Once the Christie’s and Sotheby’s sales have occurred, they will be entered to our database and the values recalculated.
The team is supervised by Dr. Dmitry Apraksin (faculty member and head of IT at the University of Nicosia) and Antonis Polemitis (@polemitis), Director of IFF and CEO of the University of Nicosia.
The team would like to also thank Professor Spyros Makridakis (@spyrosmakrid) and Artemis Semenoglou from the Makridakis Open Forecasting Center for their advice and Andreas Kitsios for his graphic assistance.
The University of Nicosia since 2013 is considered by many to be the leading university in the field of crypto-assets and blockchain. We believe we have the largest faculty, admin and researcher community in the world focused on crypto-assets, are involved in a wide range of research consortia, have over 40,000 student/alums community across academic and certificate programs, and advise the European Commission as the academic lead of the EU Blockchain Observatory and Forum.