Whilst attending the Fix Protocol Nordics event, I heard a number of definitions from various industry members on HFT, and not one was the same.
So I was wondering whether it would be even possible to define HFT, as, also
at the same event, I heard a very worrying definition that has recently been used in
German High Frequency Trading Act. The definition of “high frequency trading” in the Act uses technology and market interaction as the guiding factors for whether trading activity is or is not HFT. In this case, the more worrying aspect is that the definition was particularly specific, therefore allowing for a number of loopholes to arise. The definition utilises the criteria of server location as well as the bandwidth size of networking pipes. So, for example, if you do not have a server co-located for trading on German markets, then you are not an HFT player; neither are you a player if you have bandwidth of less than 10 gigabits. I now predict a rise in demand for Zurich data centre space and multiple 1 gigabit network connections!
However, if you push this crazy definition aside, what is a reasonable definition for HFT? One definition proposed during the conference was that HFT is the strategy of short-term liquidity provisioning or market making. However, I believe this to be too focused, and HFT deserves a broader description to catch a greater range of activity.
To get to an appropriate definition, I believe that we should take a step back and look at the HFT investment strategy rather than which technology or specific behaviour patterns define it, so that no loopholes are created, and no ambiguity is left.
I view HFT to be a highly repetitive turnover of capital within a short timespan; the time period should be considered one day with very little or no positions being held overnight. Due to the way that markets have evolved, this could mean that HFT firms show the following attributes: the need for accurate low latency market data; their strategies hold very short-term positions; the profit margins per trade are very slight. However, the firm’s activity would not just be focused around these particular aspects. These attributes are ultimately a result of technological innovation, and therefore increase an HFT’s competitive edge. But they do not define HFT.
There are a number of differing types of strategies that could fall into the HFT category. I believe these could be broken into several categories:
- Modern day market making / Extreme short-term pricing inefficiency
- Market event prediction / Participant prediction
- Statistical Arbitrage
All of these strategies require accurate data, need to be latency sensitive in execution patterns, and will generally be traded in an automated manner. This is all true for highly liquid stocks. However, what if the strategy were transacted in an illiquid market? Well, due to the nature of the market, and the low trading volumes, firms do not necessarily need low latency infrastructure and therefore need neither co-location nor the infrastructure to handle high volumes of messages. However, these short term practices are practiced in the market and are considered HFT. But under some current definitions they would slip under the radar.
The next step would be to understand why people are trying to define HFT in Europe, when the guidelines that have created this need are written around automated trading. So what is
Automated Trading? The short answer: not HFT.
*Content originally published by Matt Coupe