Market Dynamics vs. Statistics: Limit Order Book Example

TL;DR

This paper analyzes NASDAQ limit order book data, revealing that many attributes differ from traditional assumptions and proposing a shift from statistical to dynamical market analysis to better capture market behavior.

Contribution

It introduces a dynamical approach to limit order book analysis, emphasizing the importance of spike-based attributes and non-stationary market dynamics.

Findings

Market attributes with spikes contain most dynamic information

No stationary state exists in market behavior

Market dynamics involve fast excitation followed by slow relaxation

Abstract

Commonly used limit order book attributes are empirically considered based on NASDAQ ITCH data. It is shown that some of them have the properties drastically different from the ones assumed in many market dynamics study. Because of this difference we propose to make a transition from "Statistical" type of order book study (typical for academics) to "Dynamical" type of study (typical for market practitioners). Based on market data analysis we conclude, that most of market dynamics information is contained in attributes with spikes (e.g. executed trades flow $I=dv/dt$), there is no any "stationary case" on the market and typical market dynamics is a "fast excitation and then slow relaxation" type of behavior with a wide distribution of excitation frequencies and relaxation times. A computer code, providing full depth order book information and recently executed trades is available from authors [1].