New Innovations in HFT
Wed, 27 Jun 2012 04:13:40 GMT
By Nick Deacon
The world of low-latency trading is an arms race. From month to month, week to week and day to day, everything about a firm’s trading systems, strategies and the algorithms that execute those strategies must be better than the competition’s—better, faster and more intelligent.
But sources of innovation can be difficult to come by. As soon as one firm discovers a new toehold, yielding a sliver of advantage, you can expect that discovery to blaze a path across the marketplace in days or even hours. In a very real sense, advanced trading is the embodiment of the Red Queen’s lament: It takes all the running you can do to keep in the same place.
Not surprisingly, the pursuit of new advantage frequently drives quants and traders to break convention, uprooting and reinventing trading methodologies. And as high-frequency trading (HFT) gets more and more commoditized and the low-hanging fruit disappears, these tactics are becoming increasingly unconventional.
Mixing Asset Classes
Among the most popular of these unconventional innovations is identifying correlations in the price fluctuations between different types of assets. A simple example would be where a particular set of movements in energy futures might foreshadow a correlated equity-sector index price change. It starts to get more interesting when multiple diverse indicators are sought to signal future price fluctuations.
Although looking for correlations between securities is hardly a new idea, a huge leap in the sophistication of pattern identification has made it possible to find very subtle correlations today—not just among equities but across commodities, fixed income, foreign exchange, a host of derivative products and more.
There are, of course, obvious correlations that every trader worth his salt knows backward and forward. But for every obvious correlation, potentially hundreds or thousands more can’t be detected by humans. The reason is simple: data volume.
To discover heretofore unknown patterns of correlation requires the accumulation and rigorous analysis of an enormous quantity of historic data. The more assets tracked, the bigger the data set. Prior to today’s massive cloud architectures, which facilitate Big Data analytics by distributing processing power across many commodity servers, the high cost of such results would have been impossible to recoup.
Low Cost, High Value
As the cost of data storage continues to fall, organizations of all types seem open to the prospect of supporting more data. Global data growth is currently projected at 40 percent per year, according to a May 2011 report by the McKinsey Global Institute, versus a mere 5 percent growth in global IT spending. Such statistics suggest that Big Data should be a rich source of innovation for years to come.
Our growing ability to analyze massive data sets is the single most galvanizing trend in advanced trading today. It is fueling other HFT techniques, including trading assets across multiple geographies. This strategy attempts to profit from tiny price fluctuations among different geographic exchanges for the exact same security.
Perhaps the most important Big Data advantage is being able to hypermonitor the performance of HFT algorithms in real time. Emergent technologies are making it possible for firms to automatically disable an algorithm or switch to an alternative as soon as the algorithm’s performance begins to degrade. In essence, real-time data is not only helping firms create more profitable HFT algorithms but it is helping them prevent losses as well.
Trading on Tweets?
The level of sophistication engendered by Big Data has only just begun—and there are perhaps even more fascinating opportunities on the horizon, most notably the utilization of social media information in trading decisions.
Today, a small number of firms are attempting to enrich traders’ decision inputs with some very untraditional data sources, particularly Twitter. These firms know that traders can’t possibly keep abreast of all the many relevant conversations and market insinuations that can be gleaned from the Twitterverse, but what, they ask, if computers could do that for them?
These firms are effectively accumulating and aggregating public sentiment about different securities to determine and predict a future movement in the price of those securities. What that requires is to “sentimentize” the data—that is, to develop a method of scoring strings of text to measure the sentiment of the message, and then make a trading decision based on the analysis.
The combination of competitive necessity, falling data storage costs and analytic breakthroughs is a perfect storm for trading innovation. What sounds like science fiction today will be encoded into tomorrow’s HFT algorithms. If it can be stored as data, it can be a potential source of advantage. Those firms willing to invest in new ideas can hope for some relief—and reward—in the advanced trading arms race.
Nick Deacon has been working on risk- and front-office-related projects across a number of tier-one firms in the capital markets sector over the last 18 years. Prior to joining SAP, he held various roles at the CEP startup and pioneer Aleri, EMEA sales and the oversight of all aspects of Aleri CEP technology development.