Microwave Piggybacking?

Introduction

Suppose you want to trade a DAX stock on a London MTF. For your pricing you can use the Cboe & Aquis market data and the Xetra market data which you receive via a latency-competitive fiber (but not microwave) connection.

Someone else does have a microwave connection which is about 2 ms faster. This participant uses his faster Xetra data feed to price his orders on Cboe/Aquis. As long as the round-trip-latency between the participant and Cboe/Aquis is less than 2 ms, this signal reaches you before you receive the Xetra update yourself.

It is hence not unreasonable to wonder if - assuming one does not have a microwave connection - one needs the primary's market data at all.

Methodology

This article is an extension to last week's article which introduced and quantified the net information share of the different lit venues to the EBBO microprice. We now extend this concept by introducing delays for the top-of-book updates of the London-based or Frankfurt-based venues.

We then apply the same methodology to the shifted market data, i.e., (i) compute the EBBO microprice time series, (ii) compute the T+10 s EBBO microprice, (iii) decompose the microprice changes into signal and noise components, and (iv) compute the net information contribution for each venue. This procedure is repeated for a range of feed latencies.

Results

The resulting net information shares are shown in Fig. 1 below. Last week's result is represented by the values on the y-axis (x=0). Xetra contribues more than two thirds of the net information to the market. Towards the right (left), the Xetra (Cboe/Aquis) market data updates are increasingly delayed.

As expected, a venue's information share drops the more, the more it's data is delayed. For example, the information share of Cboe/Aquis via fiber for a server located in FR2 drops from one thirds to about 25% combined (left edge).

One the other hand, even with an uncompetitive fiber connection (>4.5 ms one-way latency), Xetra still contributes more than half of all information in London (right edge).

The value of piggybacking on someone else's microwave data is represented by the step change around the ±2.2 ms delays. It is noticeable but not big enough to forego the market data from the primary. While I have not done this analysis, my guess is that if one was only interested in the mid price without any volume information then the step would be much more pronounced. I leave this as an exercise to the reader.