Data-driven auction design for blockchain-based digital asset trading: A mixed method

This paper is motivated by a real project with a leading culture assets and equity swap organization in South China. Our preliminary survey assesses the potential of digital collectibles swap, with professionals aiming to boost liquidity and consumers driven by social attributes and meta-item diversity. In our setting, each agent owns a digital asset and wants another meta-item. However, the traditional Vickrey-Clarke-Groves (VCG) auction runs at a deficit. We thus introduce a novel mechanism that combines the VCG auction with limited supply and platform escrow concepts, called LSE-VCG auction. To improve the surplus of platform (auctioneer), we use limited supply to constrain the number of winners and platform escrow to increase market demand. The LSE-VCG auction and its externality-inclusive variant satisfy both truthful telling and participation rationality. If multilateral matching achieves maximal social welfare, then the substitute condition does not hold (impossibility theorem). We prove that the platform's surplus can be improved by limited supply in some conditions. Our experimental results show that the VCG auction solely with limited supply could reach greater social welfare, agents’ profits and ratio of swap relative to the sequential Vickrey auctions. Moreover, a mix of limited supply and platform escrow schemes can further improve the platform's profit and successful trading ratio. For the platform, truthful telling is a desirable strategy that brings high revenues, which promotes a transparent and beneficial auction environment. Besides, the impacts of externalities, auction timing, market size, value distribution and size of XOR bids are investigated. Furthermore, our auction mechanism is likely effective in addressing large-scale problems. Finally, we apply four effective machine learning methods to predict the limited supply number with partial information before the auction.