Practicable optimization for portfolios that contain nonfungible tokens

Abstract

Non-fungible tokens (NFT) constitute a novel asset class that has the potential to diversify portfolios. Scant research supports that hypothesis at a collection level, yet it remains an open question how to leverage the potential in practice. Owing to their non-fungible nature, liquidity of the asset that leads to a mathematically optimal portfolio does not always exist. This letter introduces a practicable portfolio optimization strategy for NFTs based on machine learning, more specifically robust hierarchical risk parity. When applied to portfolios that contain high valued NFT collections, the latter's inclusion into the portfolio is shown to improve overall portfolio return.