Finding the undervalued NFT

3 min readSep 19, 2021
Rinascita #1 by Stefano Contiero. Image Source: Art Blocks

The non-fungible token (NFT) market has been surprising us with not only new creative digital arts but also sky-rocketing prices. Everyone wants to find the next Crypto Punks, Chromie Squiggle or Fidenza before the price goes out of reach for most of us. Although the fair value of an NFT is difficult to evaluate and there is no set in stone method to price them, it is possible to compare NFTs with similar features and characteristics and find out the undervalued ones.

Ledger Status, the host of UpOnly TV has asked some interesting questions about seven Art Blocks collections they’re eyeing on. Using the Flipside Crypto data and these seven collections as an example, this article demonstrates a simple way to identify undervalued NFT collections using Decision Tree Regression.

The main goal is to compare the following 7 collections and assess which collection is undervalued.

  1. Rinascita by Stefano Contiero
  2. Alien Clock by Shvembldr
  3. Return by Aaron Penne
  4. Sentience by pxlq
  5. Eccentrics by Radix
  6. Obicera by Alexis André
  7. Ecumenopolis by Joshua Bagley

First, the prices of these 7 collections are predicted from a Decision Tree Regression. The prediction of the price is based on all Art Blocks collections with different features and prices they were sold for in the past. Based on all historic information prior to August 1st 2021, the Decision Tree Regression is able to give a tree path for each of the 7 collections and predict what the fair price will be. Since my previous work (here) shows this model gives a very high accuracy of 99.7%, the predicted price can be also used to rank order the fair value of each collection.

The predicted price for each of the 7 collections is of course not the exact price the collection will be sold for (otherwise I wouldn’t be sitting here writing this article), but it’s a good indication of where the value of the collection ranks among each other using all other collections’ price history and features. The full tree as trained from my previous work here is used to predict the price of these 7 collections.

Decision tree regression to predict August sale price

Since the predicted price is the average price of all collections with similar features (i.e. past sale price, curation status, aspect ratio etc.), if an actual recent sale price is lower than the this average predicted price which is considered the fair price, then the collection is deemed as undervalued.

Predicted vs. actual sale price in USD in August

Using the Decision Tree Regression and its predicted price of collections with similar features, the following 2 collections are undervalued:

  • Eccentrics by Radix: similar collections were sold for USD 22k on average in August while it was only sold for USD 13k.
  • Sentience by pxlq: similar collections were sold for USD 22k on average in August while it was sold for USD 16k.

Among the 7 collections the following 2 collections have a much lower expected value (USD 3k) than the other 5 (USD 22k):

  • Obicera by Alexis Andre
  • Rinascita by Stefano Contiero

The Decision Tree Regression model is only one of the many models you can use to rank order and compare fair values of NFTs at a collection level. In addition to the information related to the NFT itself, you can also incorporate data from the social media feeds i.e. Twitter, Discord, Telegram to see which collection is trending. There are definitely many more ways to evaluate and identify an undervalued NFT and take the opportunity early on by investing in the underdogs.




Specialise in NFT & DeFi analytics & modeling. Crypto and decentralisation enthusiast. You can DM me for questions or discussions on Twitter @elenahoolu