Rare Events Analysis of High-Frequency Equity Data

In this work we present a methodology to detect rare events which are defined as large price movements relative to the volume traded. We analyze the behavior of equity after the detection of these rare events. We provide methods to calibrate trading rules based on the detection of these events and illustrate for a particular trading rule. We apply the methodology to tick data for thousands of equities over a period of five days. In order to draw comprehensive conclusions, we group the equities into classes and calculate probabilities of price recovery after these rare events for each class. The methodology that we have developed is based on non-parametric statistics and makes no assumption about the distribution of the random variables in the study