[Paper Review] Mapping the NFT revolution: market trends, trade networks and visual features

Matthieu Nadini, Laura Alessandretti+4 more|IT University Of Copenhagen (IT University of Copenhagen)|Jun 1, 2021

Art History and Market Analysis25 citations

TL;DR

This paper analyzes 6.1 million NFT trades across 4.7 million NFTs from Ethereum and WAX (2017–2021) to map market properties, trader and NFT networks, visual clustering, and price predictability.

ABSTRACT

Non Fungible Tokens (NFTs) are digital assets that represent objects like art, collectible, and in-game items. They are traded online, often with cryptocurrency, and are generally encoded within smart contracts on a blockchain. Public attention towards NFTs has exploded in 2021, when their market has experienced record sales, but little is known about the overall structure and evolution of its market. Here, we analyse data concerning 6.1 million trades of 4.7 million NFTs between June 23, 2017 and April 27, 2021, obtained primarily from Ethereum and WAX blockchains. First, we characterize statistical properties of the market. Second, we build the network of interactions, show that traders typically specialize on NFTs associated with similar objects and form tight clusters with other traders that exchange the same kind of objects. Third, we cluster objects associated to NFTs according to their visual features and show that collections contain visually homogeneous objects. Finally, we investigate the predictability of NFT sales using simple machine learning algorithms and find that sale history and, secondarily, visual features are good predictors for price. We anticipate that these findings will stimulate further research on NFT production, adoption, and trading in different contexts.

Motivation & Objective

• Characterize statistical properties and evolution of the NFT market over time.
• Build and analyze networks of NFT traders and NFT assets to uncover specialization and community structure.
• Cluster NFTs by visual features to assess intra-collection homogeneity.
• Evaluate the predictability of NFT sales and prices using regression and classification models.

Proposed method

• Analyze a large dataset of 6.1 million trades across 4.7 million NFTs from Ethereum and WAX (2017–2021).
• Classify NFTs into six categories (Art, Collectible, Games, Metaverse, Other, Utility) and study category-level market dynamics.
• Construct trader and NFT networks; measure node/edge-level statistics, assortativity, modularity, and strongly connected components.
• Extract visual features from NFT images using AlexNet, project with PCA, and assess intra-collection visual homogeneity.
• Fit linear regression models to predict primary and secondary sale prices using features from network centrality, collection history, and visual features; evaluate AdaBoost for sale/non-sale prediction.

Experimental results

Research questions

• RQ1 What are the statistical properties and evolution of NFT market activity and prices from 2017 to 2021?
• RQ2 How are NFT traders and NFT assets organized in interaction networks, and do traders specialize by collection?
• RQ3 Are NFT collections visually homogeneous, and can visual features distinguish collections or categories?
• RQ4 How well can NFT prices and sales be predicted using past sale history, network centrality, and visual features?

Key findings

• The NFT market shows category-driven shifts; Art dominates volume since mid-2020 with higher average prices, while Games and Collectible drive transaction counts.
• Trader activity is highly heterogeneous and highly concentrated; the top 10% of traders handle about 85% of transactions and trade most assets.
• Trader networks show strong specialization by collection and high modularity (Q = 0.613) indicating collection-based community structure; assortativity is near zero, suggesting no strong preference by neighbor strength.
• NFT networks exhibit clustering around collections; the NFT collection partition yields high modularity (Q = 0.80), and there exist two large strongly connected components corresponding to WAX and Ethereum assets.
• NFTs within the same collection are visually more similar (AlexNet-based cosine distance within collection is lower, μ = 0.59) than across collections (μ = 0.87); PCA shows visible intra-collection clustering by category.
• Visual and centrality features jointly improve secondary sale price predictions; median past sale price explains substantial variance, with R_adj^2 up to ~0.60 for a one-month horizon; collectibles and art show stronger predictability from features; price predictions degrade for longer horizons.