The Gateway to Algorithmic and Automated Trading

A Question of Speed - The race to zero still attracts many entrants

Published in Automated Trader Magazine Issue 33 Q2 2014

Regulatory fear, fierce competition, rising costs - a host of arguments have been made for why the race to zero is a race not worth entering. Yet, technology vendors are still reporting demand for anything shaving off a microsecond here or there. Anna Reitman talks to the competitors staying in this race about keeping ahead of the curve.

Andrew Upward

Andrew Upward

"The investment in technology behind latency reduction, whether FPGAs or microwave towers, has almost become commoditised."

History is littered with stories of great wealth amassed by being first to actionable information. But the markets have come a long way since the carrier pigeon and some of the fastest operators are claiming that low latency has become less important than ever.

That might not be true for professional market makers, said Matthew Haraburda, president of Chicago-based passive HFT firm XR Trading. "Automation brings a level of efficiency that makes it tough to compete. You are not going to get outsized return just using technology, but to be a professional market maker you need some of the best technology."

Some of the smart money, Haraburda said, is going to wireless networks, particularly as microwave and millimeter microwave technologies are increasingly deployed. "It is still early days in this space but it is also where there is still tremendous catching up to do."

Microwave channels have become common because an electro-magnetic wave is closest to the speed of light, an important feature if strategy is dependent on trading between major financial centres. Firms such as McKay Brothers and Perseus are well known providers of hardware, though costs for installation and bandwidth are considerable and there are security and atmospheric risks to consider.

As an example, interception is possible and can be used to trade on tertiary markets with little chance of detection - called the A-B-C trade, said Graeme Burnett, an HFT architecture expert and executive director at Alfa Capital Markets. On the point of improving resilience, trades are generally sent via a microwave and fibre simultaneously then arbitrated at the destination in case of disruptions.

Millimetre microwave covers shorter distances (typically five to 10 kilometres) but carry much more bandwidth. "This is useful for market data distribution over the trading route which makes traditional microwave still viable," he said.

Also grabbing headlines is the potential for laser technologies, and XR Trading's Haraburda points to service provider Anova Technologies as a firm in particular focused on becoming competitive in this space. This wouldn't give any speed advantage over microwave and millimeter microwave but rather greater bandwidth, Haraburda said.

Hans Pieterse, director and head of market structure for Optiver Europe, said that as an electronic market maker, being as fast as the competition will continue to be of primary concern for the firm.

Bradley Wood

Bradley Wood

"There is a sense in the market place that low latency seems a bit like yesterday's news."

"Optiver makes money by providing the tightest price across thousands of different products. It is crucial to address low latency to make the right trades profitably just to make sure that we can continue our business model of providing two-way prices," he said.

Revenue slide

Pieterse said that there is a major shift for HFT firms and electronic market makers amid an environment of tight spreads, low volatility, stagnant volumes and rising technology costs. "We expect a consolidation among electronic firms as more firms will be struggling to deal with the required investment to remain profitable."

There is no doubt that revenue and profits have gone down since the industry saw peak revenues. TABB Group estimates revenues of about $7.2 billion in 2009, versus $1.3 billion expected this year - a slight lift from the $1.1 billion in 2013.

Institutional broker Rosenblatt Securities estimates peak profits in HFT of $5 billion in 2009, which had dropped to $1 billion in 2012. At time of writing 2013 figures were being compiled but given the relatively stable volume and volatility trends over the year, the firm anticipates roughly the same figure.

Andrew Upward, market structure analyst for Rosenblatt, said that low latency is a necessary, but perhaps no longer sufficient, condition for profitability. "The investment in technology behind latency reduction, whether FPGAs or microwave towers, has almost become commoditised. For HFTs it is the cost of doing business," he said.

Meanwhile, part of the growth story for exchanges is about market data and connectivity upgrades as well as the associated fees. "Whether it is a new 10G or 40G bandwidth, or other technology - once something is introduced it does have to be adopted by market participants because they don't want to be competitively disadvantaged," Upward said.

Hans Pieterse

Hans Pieterse

"It is crucial to address low latency to make the right trades profitably..."

Cost-benefit analysis

Along with communication providers, firms are addressing latency bottlenecks across network and computing infrastructures. Scott Newham, co-founder of low latency hardware firm Metamako, said that the greatest speed gains are made by improving the network. Trading firms have tackled this by co-location, reducing network hops and rapid data delivery. "For the ultra-low latency traders, investment will continue to be on that kind of level of optimisation," he said.

The year-old firm's first product is an ultra-low latency network switch called MetaConnect, which occupies a rack unit in a switching cabinet. The 16-cable version is able to pass data in four nanoseconds, while a recently released 48-port version can accommodate larger installations and comes in at between four and six nanoseconds. In addition, MetaConnect is able to provide network tapping (port mirroring), delivering data simultaneously to multiple locations and serving as an alternative to multicast.

Financial industry clients for the device are generally trading firms, the majority of which are located in North America. However, Newham is looking to increase the amount of data centre customers as well and plans are in place to branch out into Southeast Asia and Europe. Low latency, he said, is becoming more of a norm even for non-HFT firms looking to accurately capture market data and timestamp, backtest models, or in general remove jitter and variability, for example.

HFT Revenues up a little in ‘14 due to increasing volumes and volatility

"There are of course prop trading firms on the bleeding edge willing to spend to get the edge, but in general most trading firms given a choice and reasonable expense will choose the low latency route," Newham said, adding that clients starting out with Metamako's network switch can see a beginning spend as low as some $10,000.

Measuring performance is an important step in identifying exactly where latency gains can be made, not to mention correlating the results to aim technology spend directly at profitability, said Burnett, speaking at the London Quant Club spring seminar in May. "We can buy faster, better technology, but are we wasting our time? That is a question not being asked very often," he said.

Burnett estimates a start-up budget for the first year of some £5 million ($8.4M), inclusive of trading capital, staff and infrastructure costs. "Profitable trading can be started in six months as long as the connectivity and infrastructure is in place," he said.

However, he too questions how much room there is to run for the ultra-low latency strategies. "The biggest threat at the moment to HFT firms is the existence of dark pools because large institutions are using them to trade between themselves. I question how much money there is to be made now. If you look at the profit of HFT firms like GSA or Jump Trading, it is in the hundreds of millions, but not billions of dollars," he said.

Sanjay Shah

Sanjay Shah

"HFT strategy these days does not only rely on speed, it is also about pre-trade processing."

Other needs for speed

Interest in purely ultra-low latency strategies is waning and it appears that the industry has reached a point where the cost of new technology exceeds the value it adds, said Bradley Wood, partner at capital markets consultancy firm GreySpark Partners. Meanwhile, research shows that the largest contributors to latency are not related to network connectivity but to firewalls and algorithms.

"There is a sense in the market place that low latency seems a bit like yesterday's news. Technology is so quick, everything is so optimised, the market is mature and arbitrage opportunities in low latency are few and far between," he said.

For hedge funds and private equity shops, interest in financial technology tends towards better servicing an existing need - liquidity hunting algorithms that can deal with fragmentation for example, or in addressing market structure changes brought about by regulations. This is particularly true for fixed income markets - not exactly an ultra-low latency environment though one where intelligence management and innovation is resulting in significant investment.

HFT was rage, however regulation, volumes, volatility, & EU taxes have dampened market. Little uptick in ‘14 from up volume & volatility

Wood said that the market structure shift - like liquidity fragmentation brought about by swap execution facilities - might shake up the low-latency market and provide more opportunities for fast money operators to exploit. But it is difficult to predict how technology will evolve even five to 10 years from now. For example, disruptive technologies like quantum computing could replace silicon chips which could trigger a flurry of investment.

In terms of the investment going to ultra-low latency today, Wood said there is some talk about FPGAs, or technology that will normalise market data rapidly so it is easier for an algorithm to consume. However, any investments in this area will be niche and likely piecemeal in the short to medium term, he said.

Scott Newham

Scott Newham

"...most trading firms given a choice and reasonable expense will choose the low latency route."

Although it took a little while, FPGA technology has certainly seems to have gained traction in the financial industry over the last couple of years. Now, FPGAs are resulting in smarter low latency trading algorithms and trading strategies to be used by a wider number of slightly smaller HFT companies, said Sanjay Shah, founder and CTO of Nanospeed Technologies.

"HFT strategy these days does not only rely on speed, it is also about pre-trade processing. With advances in very large FPGAs, firms can process far more data before the algo makes the trading decision," he said.

Nanospeed is getting more interest from clients - which tend to be algorithmic traders in tier 1 banks, prop trading firms and hedge funds - that are coping with pre-trade risk management controls. Shah attributes the bulk of this push to regulations such as MiFID II in Europe and the SEC's 15c3-5 in the US, or the client's own custom requirements. Examples of this would include implementing controls such as per-instrument throttling, order-to-instrument ratios, or stringent position limits for a particular asset class.

"We are seeing a shift in customer focus from the actual speed and latency to being able to process much more data in a given time before making the trading decisions, an area where FPGAs perform very well," he said.

Optiver's Hans Pieterse said that though firms are sometimes strong in a given area - pricing or client execution for example - the aim for top market makers will be to address all aspects of connectivity and computing. "If speed gains are not happening in wireless, then things will be happening in computer hardware. You are at big risk as a market maker if a competitor is faster so there is no one area of focus," he said.

And the length of time that a firm is potentially disadvantaged after missing out on a technology swing varies greatly. Whereas price feeds or better lines from a third party can be solved in a matter of days, early FPGA adoption could make a firm competitive for a long time, he said.

XR Trading's Haraburda said that the biggest gains in latency have largely been achieved but there will always be need for improvements.

"I am not going to be one that predicts technology is going to stop any time soon," he said.