Algorithmic trading demands high speed. Reducing latency times – the time it takes to transmit trades and receive market data so trading decisions and executions can be made ahead of the rest of the market has become an area of competitive advantage. However, its importance varies according to changes in market conditions and by the type of client and their trading strategy. As market infrastructures change and as liquidity fragments, what is meant by low latency also starts to change.
The hedge funds and proprietary traders with high frequency strategies often catch the attention. Their need to generate huge numbers of trades to make up for the very fine profits they make per trade not only stresses the models they use, but also puts a premium on extremely rapid data collection, fast algorithmic run times and near instant execution of a trade in the market. However, other more traditional traders and investors are also using some of the same technologies to improve their performance. This has certainly been the case in the US, and is now an accelerating trend in Europe as traders and investors prepare for the impact of the Markets in Financial Instruments Directive (MiFID) and the potential splintering of sources of liquidity.
According to Richard Balarkas, Credit Suisse’s head of advanced execution services sales: “A couple of years ago, if the average player was trading their model with 200 milliseconds of latency, and if you got yours down to 100 milliseconds, this was a distinct advantage. Credit Suisse is now operating at speeds well below 20 milliseconds, and in the US below 10 milliseconds.
“At these speeds, it is questionable whether further cuts will make any substantial difference. If the whole low latency area has become commoditised, then it is difficult to see how cutting latency times from 16 to 15 milliseconds is going to add much value, and it is questionable whether models that require that kind of speed are sustainable.
“Smart traders know there are easier ways of making money than being one millisecond faster than another trader running a similar model,” says Mr Balarkas.
Proximity services
However, many technology companies and consultants are talking up the value of so-called proximity services to cut latency times even more. In this environment, servers running a firm’s algorithms are located next to or even in an exchange building so that data can be pumped into them as quickly as possible and direct market access trades can be executed even faster.
Mr Balarkas questions whether it is the right time for such technologies in Europe. “MiFID may throw a spanner into the works because one of its key objectives is fragmentation of liquidity through encouraging competition between venues. The question is: where do you put the box to achieve proximity?”
However, one approach is to bring the liquidity venues to the infrastructure rather than the other way around. “As new venues develop, you can bring them to the infrastructure rather than the infrastructure going to the exchange,” says Mark Akass, chief technical officer for global financial services at BT. “As new exchanges open up or as new data sources emerge, we can host them in our infrastructure because we are with Project Boat [a trade data reporting service established by a consortium of banks to provide post-MiFID competition to the European stock exchanges]. This ensures the sources of liquidity will have very low latency times.”
Liquidity monitoring
The US has already been through this process and many firms now have to watch 26-27 potential sources of liquidity; the trick is to spot where the liquidity is and respond quickly, and not to be aligned with one venue, Mr Balarkas says. “This means you need incredibly fast, smart order routing because there might be five venues that have reasonable liquidity, but there may be queues on each. If the queue on one of these venues shrinks, you may need to cancel the order you have in one of the other queues to get onto the shorter one. We now use smart order routing software that employs fantastically fast, low-latency technology.”
Post MiFID and following the development of new pools of visible and dark liquidity, the pace of trading and the need for improved latency times will once again grow, as they have in the US, driven by the need for smart order routing capabilities. Credit Suisse’s Advanced Execution Services (AES) product suite has several advanced algorithms designed to hunt out liquidity where it exists (for example, Sniper) and other tactics designed to reduce the signals clients send out to the market (for example, Guerrilla).
This splintering of liquidity has led to a fundamental challenge. Forget speed of access to a single venue; coping with the volumes is becoming the biggest challenge. According to Mr Balarkas, the number of messages flying around the market is growing at such a pace that even capturing the volume of data is a challenge. Measuring message volumes by the one-minute peak number of messages per second is a challenge, says Mr Balarkas. “In 2001, the peak in the US was 763 messages per second. In 2004, it reached 25,000 messages per second. It has already reached 250,000 this year and the industry expects it to hit 575,000 by the end of the year.”
Europe lags behind US
This difference between the US and Europe is also evident to Tony McManus, director, algorithmic trading solutions, at Wombat Financial Software. “In general, we see that Europe is 18 months behind the US in its use of these technologies. However, in the past 12 months, the increase in volumes on the London Stock Exchange (LSE) has been driven by firms using high frequency strategies and advanced algorithms”
Europe will be catching up nevertheless, and will be experiencing the same kind of volume challenges as in the US. According to Danny Moore, chief operations officer of Wombat: “There is a view that MiFID will force the market to be more fragmented. My view is simpler, based on experience in the US: if a number of firms roll their technologies from the US to Europe, then the market will change anyway, and if you look at the rise in volumes on the LSE, this is because of very aggressive technologies now being available.
“As the technologies become available to trade across the market, this will drive the fragmentation itself, MiFID or not,” he says.
It is not just markets that are different, but customers, too. Yvonne Hansmann, head of execution sales, Europe, Middle East and Africa (EMEA) at Merrill Lynch, says: “You need to know your clients, their needs and requirements. Some clients require low latency, some want a combination of algorithmic trading and direct market access, and others want more handholding.
“Sales support is just one component, and understanding the benchmarks they work to, what their trading style is, how their desk set up is, and what their infrastructure is like, is also important. There is still a very large part of the market where the best method is to place an order over the telephone to the traders,” she says.
Adapting to change
Conditions in the market also change and this may also require a change in trading technique. “On a quiet day when the LSE is very liquid, then maybe 100% of trades in a given name can go through the algorithm,” says Ms Hansmann. “But the next day, there may be results coming out in the same stock or liquidity is difficult to find, and then maybe you should make use of the sales desk.
“Some brokers use a tyranny of quotas telling customers that they should be doing 30% to 50% of their trades with algorithms and direct market access. We believe that clients should be aware of all their options and pursuing such a formula is wrong,” she says.
Brian Schwieger, head of EMEA quantitative execution desk at Merrill Lynch, says: “The degree of importance of latency depends on the clients and their trading style. The high frequency traders vary substantially from traditional institutional clients.
“They have very short alphas and very small order sizes, while institutional clients have much longer alphas and much larger order sizes. Getting the right execution strategies for the large orders is a part of the consulting approach.”
The splintering of liquidity in Europe, as has occurred in the US, will make the use of technology more important to all customers as the hunt for trading opportunities becomes more complex.
“As you get 15 or more sources of liquidity it becomes impossible for a trader, or any human being, to monitor them all and this is where smart order routing technology comes in,” says Mr Schwieger. “This can monitor and interact with all the potential visible alternative execution venues and also probes the dark pools for liquidity.
“It means the role of the traditional broker is changing rapidly – the skilful traders of the future will be the ones that can exploit technology effectively.”
