Making sense of rule 605

Abstract

The paper analyzes monthly reports of two companies: one market maker, Virtu, and one broker-dealer, SoHo, in the period between January 2021 and March 2024. The predictive power of standard methods for the probability laws of stock orders- linear regression and LSTM- is poor, despite their probability laws being relatively stable. However, the month-to-month correlations are highly stable and predictable. In this paper, we study the distributions of the sixteen datasets resulting from the 26 stock-trading parameters reported by brokerages under regulatory requirements.