Intraday Short-Wave Triggers

Introduction

In a previous post I had looked at the growing use of short-wave data transmissions from Chicago to Frankfurt. Today I want to focus on the intraday pattern. For example, one would expect little influence of the US markets on Europe while the former is closed and the latter is open.

And this is indeed what Fig. 1 below shows. The color coding represents the number of order book updates (i.e., not just trades) for the DAX40 future on Eurex (FDAX) vs. the time delay from a preceding trade in the S&P500 E-mini future on the CME (ES) and the time of day. Note that the time of day along the vertical axis is in US Eastern time to eliminate artefacts due to daylight savings time.

The left-hand panel only considers days without non-farm payroll (NFP) news events; the right-hand panel only includes days with NFP events.

Observations

• First, we observe the two distinct bands of reaction times. The first one starting at around 26.8 ms which corresponds to the reported latency of a low-bandwidth transmission using short-wave radio frequencies. The second one is around 35.8 ms and represents the latency when using microwave links from Chicago to the coast, then a dedicated trans-Atlantic fiber, and finally via microwave from the UK to Frankfurt.
• Second, the short-wave band is much wider than the very narrow later band. This is because there is much more variance when using short-wave radio stemming from changes in atmospheric conditions (altitude of the reflecting ionosphere and the number of earth/ionosphere reflection hops taken).
• Third, we observe that there is little reaction to ES trades on the FDAX before the US cash open. The cash open itself at 09:30 EST triggers a spike of reactions. The intensity then drops off markedly over time.
• Fourth, just like there were few reactions before the US cash open, there are also very few reactions after the European cash close.
• Fifth, the singular importance of isolated news events is evident from the intense peak at 08:30 EST which only lasts a few minutes. Note that there are also some traces in the left-hand panel at 08:30 EST which are due to other economic numbers being released at the same time such as jobless claims and inflation numbers.

Washington DC vs. Chicago as Origin

The NFP numbers are released in Washington DC. If there was a direct short-wave link from DC to Frankfurt then we would expect to see reactions approximately 6 ms earlier (since we only measure reactions to CME data) than for events that originate in Chicago. But this does not seem to be the case. Fig. 2 shows latency histogram cross section (denoted by the green rectangles in Fig. 1). There is no difference in the location of the peaks between reactions to market data from Chicago and the economic news from Washington DC.

At least some other economic numbers are distributed to data service providers ahead of time, then distributed to major trading centers internationally, and then released simultaneously at the end of the embargo period. This could also be the case here. The reactions shown would then not be reactions to the NFP numbers but to the market's reaction on the CME.

Or, alternatively, there is for now no such short-wave data link from DC. Possibly, 5 minutes once a month does not make this economical.

ZN/FGBL

For the sake of completeness, Fig. 3 shows the same data for ZN (10 yr treasury futures) triggering FGBL (German Bund future) reactions. Here the we see that the short-wave data is almost exclusively used around economic news events.

#eurex #lowlatency #hft #derivativestrading #microstructure #marketdata #a7 #fpga #algotrading