SIGNATURE WORK
CONFERENCE & EXHIBITION 2023

Cryptopunk is an NFT art collection generated based on blockchain technology. Cryptopunk are algorithmically generated images, each image consisting of one or more of 87 traits at random. There are 10,000 Cryptopunk in total, and each image is unique. Most research analysis suggests that the price of different types of Cryptopunk varies widely due to the different rarity of Cryptopunk. Intuitively, scarce Cryptopunk will have a higher value. However, there are no studies analyzing to what extent the scarcity explains the price difference of Cryptopunk. In addition to this, there are currently many models in the industry that calculate rarity scores. However, there are no academic articles comparing different rarity score calculation models, and there lacks a study analyzing how the rarity score calculated from different models influences the price prediction of Cryptopunk. Therefore, in this paper, we want to answer the question that how different rarity score calculation models influence the price prediction of Cryptopunk. In order to figure out this question, we implement three common models to calculate the rarity score of Cryptopunk through code and use the Hedonic pricing model, random forest regression, and auto-machine learning to predict the price of Cryptopunk with the three rarity scores respectively. For each prediction method, we compare the value of the adjusted r-square and root mean square error (RMSE) to decide which rarity calculation model could better predict the Cryptopunk’s price.