Clustering Uniswap v3 traders from their activity on multiple liquidity pools, via novel graph embeddings

Abstract

AbstractUniswap is a Constant Product Market Maker built around liquidity pools, where pairs of tokens are exchanged subject to a fee that is proportional to the size of transactions. At the time of writing, there exist more than 6000 pools associated with Uniswap v3, implying that empirical investigations on the full ecosystem can easily become computationally expensive. We propose a systematic workflow to extract a tractable sub-universe of liquidity pools, where the interconnection among such pools is maximised to capture broader dynamics within the ecosystem. The resultant set of 34 pools is then used to cluster market participants according to their liquidity consumption behaviour over such environments, for the time window January–June 2022. Introducing a novel approach, we proceed to represent each liquidity taker by a suitably constructed transaction graph. The graph is a fully connected network where nodes are the liquidity taker’s executed transactions on the 34 pools of reference, and edges contain weights encoding the time elapsed between any two transactions. We then extend the NLP-inspired graph2vec algorithm to the weighted undirected setting, and employ it to obtain an embedding of the set of graphs representing market participants. This embedding allows us to extract seven clusters of liquidity takers, with equivalent behavioural patterns that can be interpreted in terms of trading attributes, i.e. preference for exotic assets over stablecoins, frequency of activity, tolerance for higher trading fees.