Find the Value of Nonfungible Tokens Automatically and in Real-Time with Machine Learning
Updated: Apr 28
Owners of nonfungible tokens (NFTs) often wonder what their assets are worth. Unlike fungible and liquid cryptocurrencies like Bitcoin or Ether, nonfungible tokens tend to be less liquid, much like physical collectibles, real estate, and art in the real world. As a result, the best estimate of an asset's value is often it's last sale price, even though the sale may have occurred several months ago under very different market conditions. Most nonfungible tokens are selling for the first time ever, so they don't have a rich price history. In such cases, buyers need to learn about the market, the mechanics of the application, find comparable assets, and form a valuation based on this information.
This process creates a huge friction which prevents new users from participating in decentralized application (dApp) economies. Existing users may be losing money because of imperfect information about the true value of the assets they own or wish to purchase.
These frictions and risks can be reduced by machine learning algorithms that automatically generate valuations for nonfungible tokens. These algorithms follow the same process that a human would in valuing a digital asset, but are capable of objectively scanning an entire market and using statistical methods to find similar assets and validate valuation predictions on data not seen before.
Note: Image shows the interface for our Sorare valuation app. Users can obtain valuations for a blockchain-based token simply by inputting its ID.
These algorithms look at historical transactions and correlate sale prices to potentially high-dimensional sets of token characteristics, taking into account the market conditions at the time of any prior sales. They then generate predictions for an input token based on its characteristics and the current market conditions.
Valuations are always uncertain, because the market may be changing in unexpected ways, or different users may assign different values to the same NFT. To reflect this uncertainty, we use multiple machine learning models trained on different perturbed versions of transaction histories, and generate distributions of valuations for a single NFT.
Try our valuation applications and learn more about how to use them:
Note on updates: Remember to check the "Last Update" date in the app notes. If it's too far in the past and the market's behavior changed significantly in the meantime then the estimates may not be reliable.