DeFi : Modeling and Forecasting Trading Volume on Uniswap v3 Liquidity Pools

We consider two major venues that exchange BTC-ETH on the blockchain, namely the Uniswap v3 WBTC-WETH liquidity pools with proportional transaction costs of 5 and 30 basis points respectively. A multivariate linear regression model with quadratic terms is introduced to predict the logarithm of incoming trading volume on either pool, where the possible prediction horizons are measured in number of blocks mined on the blockchain. We design and test features that relate to the latest activity on the mentioned pools, but we also allow spillover effects from a sample of one-hop liquidity pools in the network of venues that exchange a common asset. Furthermore, we augment the space of possible predictors with statistics computed from the BTC-ETH trading activity on the Binance crypto centralised exchange.Our results indicate that both the latest activity on our two main pools of interest and their neighbours, and the most recent evolution of Binance trading, are necessary information to explain a significant proportion of the variance in the upcoming trading volume. Thus, the drivers of liquidity takers' activity span from the broader decentralised ecosystem to centralised crypto exchanges