The asset we are modeling the arrival rate of is the ES Mini Future which tracks the value of the S&P500 Index. ES Mini Futures are contracts which expire on the third Friday of every quarter – thus there are four contracts per year.
Data was collected for the following futures from a Bloomberg Terminal (1) ES December 2017 and (2) ES March 2018. This provided more than 120+ datasets for both day and night which would be plenty for the purposes of this study.
An example of what a typical dataset would look like is provided below:
Column 1 = Trade Timestamp
Column 2 = Trade Confirmation Message
Column 3 = Trade Price
Column 4 = Trade Quantity
For a given day, a dataset could contain anywhere from 15k to 500k data points, with nighttime datasets having significantly less datapoint due to less activity.
Classifying trades as large or small “correctly” was one of the most important aspects of this study. This comes into play when making the distinction of a large trade during the day vs. at night. For example a large trade during the day may need to be at least 15 in quantity whereas for night it may need to only be 10 in quantity. This however is subjective to us how we define it. An example of the trade distributions that compares small/large and day/night for a particular date is provided below: