Self-organised criticality in high frequency finance: the case of flash crashes

Abstract

With the rise of computing and artificial intelligence, advanced modeling and forecasting has been applied to High Frequency markets. A crucial element of solid production modeling though relies on the investigation of data distributions and how they relate to modeling assumptions. In this work we investigate volume distributions during anomalous price events and show how their tail exponents < 2 indicate a diverging second moment of the distribution, i.e. variance. We then tie the dynamics of flash crashes to self-organised criticality. The findings are of great relevance for regulators and market makers as they advocate for rigorous heavy-tailed modeling of risk and changes in regulation to avoid simultaneous liquidity withdrawals and hard risk constraints which lead to synchronisation and critical events.