Opinion: The Race is Too Fast, the Costs Too High

By Neil McGovern, Director of Marketing, Sybase

Today, unfiltered sponsored access times are about 250 milliseconds (0.25 seconds). If history is any indicator, they will continue dropping.

This might seem fast, but a major New York institution recently claimed that simple trades have been clocked at 1 microsecond (0.000001 seconds) in their latest system tests. Even if this is under perfect conditions, it can only be a matter of time before we are talking sub-microsecond for execution.

Given the coy nature of the participants, sub-microsecond trading is probably happening. So it comes as little surprise that there were two major issues at this year’s HiFreq Trade conference in London in February: Regulators’ reaction to last May’s Flash Crash; and the new measure for latency – nanoseconds.

There is little individual institutions can do about the regulators’ response to the Flash Crash. But keeping up with the lowest latency systems is evolving into a very expensive challenge.

Firms must decide to whether or not continue to the race. How many players can keep up when the cutting edge of low-latency trading starts measuring in nanoseconds? Are there other low-hanging opportunities where a high frequency trader can yield better returns than simply being the fastest?

Let’s put the awe-inspiring speeds involved into perspective. Light travels one foot (30 centimetres) in one nanosecond (one billionth of a second). This means that a 20-foot cable from an exchange system to a co-located trading server burns 20 nanoseconds. These are ideal numbers. The speed of light in a vacuum will always beat electrons in metal wires and photons in optic cables.

Little wonder that the latest high-frequency infrastructures are adopting 40 gigabit Ethernet and other arcane techniques to drive out a few more nanoseconds here and there. I’m sure they also want shorter cables.

So if the laws of physics are starting to hamper further decreases in speed – with the law of diminishing returns also raising its ugly head – where will the next technology battle in high frequency trading occur?

More Complex, and Smarter, but Still Fast

“Near-microsecond turnaround for simple trades” is the qualifier to blazing fast times bandied about Midtown New York bars. Very few situations qualify as microsecond decisions because most require some form of pre-trade analytics.

Field Programmable Gate Array (FPGA) chips are the solution of choice to balance speed with a semblance of thoughtful decision-making. But even programming risk checks into hardware is not ideal. Pre-programmed hardware runs in a set pattern to achieve the required latency, dramatically reducing the depth of thought available.

And speed is not the only important factor in trading – even in high frequency trading. A slower but smarter system may be able to find more lucrative opportunities. You will appreciate this in cases when the fast money competes in an expensive race for the same opportunity – and your system can think outside the black box.

A more thoughtful approach can pay dividends in cases where latencies are unavoidable. Complex trading strategies increasingly incorporate cross-asset class decisions, which often involve multiple trades occurring across different execution venues. You can productively use time delays resulting from that pesky speed-of-light latency issue.

Moreover, these complex strategies can incorporate a wider range of inputs beyond market data, such as news feeds. And the high frequency crowd tends to self-congregate around a small number of highly liquid, highly traded instruments; so the smarter algorithms that focus on less frequently traded instruments also face less high frequency competition.

It is my contention that markets will congeal into liquidity centres that are cross-asset class and regional in nature, where latency is an important, but not definitive factor in trading excellence. The ability to bring together and analyze large amounts of data of many different types from many sources will pay dividends.

There is an interesting vignette in Michael Lewis’s Liar’s Poker where one of Lewis’s fellow traders identifies an opportunity in tomato futures while everybody else is watching the emerging news from the Chernobyl disaster on CNN.

Computers are beating humans at chess, backgammon and even Jeopardy. A future in which computers are tomato options pickers, while mankind is simply picking tomatoes off vines is not out of the question.

To end on a positive note, I must give the last word to Dr. David Leinweber (author of Nerds on Wall Street: Math, Machines and Wired Markets, Wiley 2009) in a recent article: “If history repeats, we can expect HFT-like speed wars between computers reading everything from Twitter to the Shanghai Financial Times."

Now that will be fun to watch – if we can just figure out how to be nanosecond spectators.