While the general public may only be gaining awareness of high-frequency trading’s existence with the recent launch of Flash Boys by Michael Lewis, there are still many who participate in the capital markets industry struggling to understand the real role of latency in electronic trading. What follows is a perspective about the future role of latency, based on the past several years of monitoring the world’s electronic trading networks.
In 2009, we presented a view to the marketplace that “speed is good, transparency is better.” At this time the electronic trading industry was in the middle of what seemed to be an endless latency arms race. At that time we felt that this arms race would inevitably come to a halt as electronic trading advantage is a function of relative latency, not absolute latency.
This is often hard for non-technical people to fully understand. It is important to realize that improving relative latency at a point in time requires a non-linear increase in technology spend. Therefore investment in latency reduction was always destined to end at some point, as the return on investment tends to zero faster than latency tends to zero. In fact it is sometimes worse, as specific reductions in latency do not necessarily improve trading performance.
The relationship is much more complex. This fact is sometimes forgotten and not fully understood by IT professionals supporting the trading business.
Everyone is fast
What has happened in the past five years is the mass adoption of high-speed IT systems for electronic trading. Everybody is fast today. To be more precise, everybody is "fast enough." Fast enough is often good enough for most organizations. The trading mantra has shifted from "I must be fast" to "I can't be slow." In essence, the industry has found its desired center of gravity, balancing the cost of achieving latency advantage versus the economic return from improved trading performance.
So, where do we go from here in a world where everyone just wants to be “fast enough”? Does this mean you do not have to worry about latency any more? Unfortunately, it is not so simple. While it no longer makes sense to spend an un-ending amount of money being the fastest in the market, you do have to make sure you are not slower than the market. The secret to getting this tradeoff right is analytics. To be precise, we are talking about real-time latency analytics.
Imagine if you were a professional athlete and your specialty was the 100-meter sprint. How would you prepare for the next big racing year? Let’s assume that on your day you can sprint as fast as most of the field and you want to optimize your win rate over the year. To do this, you need to have better intelligence than your competitor. You need to know:
- What were the winning times at the planned venues?
- Who won the races?
- How many athletes competed?
- What were there best times?
- What precise conditions existed for this race?
- How did the conditions impact the race outcome?
Based on these answers, you would devise a training plan and strategy. If you execute your plan and prepare well, you will know statistically your chances of winning each race. The critical observation is that the athlete is able to devise a more predictable plan and strategy for a successful season based on mathematical analysis of all key metrics associated with the race, the competitors, and the environment.
In electronic trading, thousands of races happen every minute. If you do not have a clear speed advantage, you can increase your likelihood of winning by having superior ability to monitor and analyze the IT systems that support execution of trading transactions. Basic questions include:
- How fast is price delivered (market data latency)?
- How fast can your algorithm respond to advertised price (tick to trade)?
- How fast is the order executed (order response time)?
- How fast is the order acknowledged by the market (order to tick)?
If you can answer these accurately then you have a good basic start for predictability of trade execution and outcome. But real competitive insight comes from much deeper analytics relating to speed and operations of the IT systems. Examples of these analytics cover more complex questions like:
- If I see a drop in fill rate, how do I know it is a latency problem?
- If my latency increases, should I withdraw from the market or stay in?
- If I reduce the latency of my IT systems, will it make a difference to fill rate?
- If I notice my order is delayed, how do I know if it is my issue or the exchange’s?
This is the emerging field of latency analytics for electronic trading. In the new world of “fast enough is good enough” latency becomes a real-time analytic that can be used to weight the trade execution decision. In effect, latency, as an analytic, becomes the beta coefficient of electronic trade execution. Latency analytics qualifies the level of uncertainty or risk for getting the price advertised. It tells you if latency is a factor in a missed fill. It explains what happened with an unexpected trade outcome, and it shows you what to do about it.
When viewed this way, latency analytics acknowledges the reality of IT systems. In other words, it is impossible to provide a perfectly consistent level of latency. Latency will always vary, just like the weather. As traffic levels change, so will latency. Trading strategies need to be able to cope with this reality.
If traders were able to receive a latency analytics feed showing the real-time state of the IT systems, then the algorithm would be able to take account of the latency variation in the IT systems to achieve better trading performance. This reality is not too far away, as people realize that analysis of latency is now more important than latency itself.Donal Byrne is the CEO of Corvil, the leading real-time analytics company for monitoring and safeguarding the performance of the world's electronic trading networks. As CEO of Corvil, Donal has been at the forefront of technology innovation and its application to financial ... View Full Bio