NFT Pricing via Machine Learning

NFTs enter the horizon in 2017, when cryprokit-ties as a game token became popular. In 2021, it is the meta year of NFT eruption. Not only do the two famous auctions confirm the unique values of NFTs, but a lot of research has also analyzed the features of NFTs and their values. This paper proposes common factors to price the NFTs, inspired by classical asset pricing theory in finance. The common factors are total sale, buyer/seller, transfer and etc. Moreover, we incorporate machine learning methods including regularized regressions, trees, and neural networks. We use the representative cryptopunks as head pictures in social communities, NBA top shot as the real-world reflection in NFT world, autoglyphs, and bored ape yacht club. The empirical results greatly support our proposed factors on NFT pricing.