Abstract
Market by order (MBO) data - a detailed feed of individual trade instructionsfor a given stock on an exchange - is arguably one of the most granular sourcesof microstructure information. While limit order books (LOBs) are implicitlyderived from it, MBO data is largely neglected by current academic literaturewhich focuses primarily on LOB modelling. In this paper, we demonstrate theutility of MBO data for forecasting high-frequency price movements, providingan orthogonal source of information to LOB snapshots. We provide the firstpredictive analysis on MBO data by carefully introducing the data structure andpresenting a specific normalisation scheme to consider level information inorder books and to allow model training with multiple instruments. Throughforecasting experiments using deep neural networks, we show that whileMBO-driven and LOB-driven models individually provide similar performance,ensembles of the two can lead to improvements in forecasting accuracy --indicating that MBO data is additive to LOB-based features.