Creating an algorithm to automate trading used to take an army of PhDs in pure maths and months of sifting through databases.
But now automated trading models are being created in real time, with algorithm-generating software that allows funds to go from concept to trading within a few hours. The software builds on ‘complex event processing’ technology, enabling traders to define the parameters and signals that will create unique algorithms.
John Bates, founder and general manager of the Apama, the algorithmic trading system division of Progress Software, says the technology evolved from the recognition that traditional database systems, which are sequential, are inefficient at responding to streams of data that are being updated asynchronously. “Complex event processing technology is applicable to any field where you need to look through large amounts of complex data,” Bates says. “But capital markets turned out to be the most willing to pay to see it developed.” According to Bates, the potential applications of the algorithmic technology are limitless.
“From monitoring spreads between stocks to creating derivative products, anything you can trade manually you can encode,” he says. As the technology becomes more flexible and affordable, Bates says that funds managers who would have once simply relied on their brokers’ algorithmic programs are now implementing their own systems.
“The enquiry level from hedge funds has been tremendous,” he says. The growing number of hedge funds executing cross-border and multi-asset class strategies are required to monitor what is also a growing population of trading venues, making automation more important in the implementation of algorithmic strategies, Bates says.
Richard Massey, senior sales executive at Progress, says he sees a lot of interest in experimenting with different parameters for trading currency. “I think foreign exchange is about to become its own asset class,” he says, explaining that an algorithmic system makes it possible to play synthetic crosses, such as the pound/peso, which aren’t actually listed anywhere.
Bates estimates that globally 50 per cent of equities and 60 per cent of futures are currently being traded algorithmically. The proprietary trading desks of investment banks like JPMorgan and Deutsche have implemented the Apama software across various asset classes, and hedge funds are reportedly using the technology to test an infinite number of scenarios. Bates describes one client whose process he likens to genetic tuning. “[The manager has] 300 Apama engines on 300 different computers running thousands of variances. It is like genetics, experimenting with permutations to see which set of parameters turn out to be most profitable, and killing off the less profitable branches.”
The technology can also be applied in market monitoring and regulation. Some estimates put the level of insider trading around mergers and acquisitions at around 30 per cent, but the best that analysis that’s not in real time can do is pinpoint irregularities long after the event. “If you find there has been nefarious action after a market move, it’s too late,” Bates says. “Sure, the offender might be caught and punished eventually, but the market has already been artificially distorted.” Creating algorithms in real time has the potential to detect unusual movements before the event. The technology could also pick up other illegal activities that are currently difficult to monitor, such as brokers front running large clients’ orders or traders “painting the tape” – that is, trading small amounts of a specific security among themselves to create the illusion of high investor interest, in the hope of artificially inflating the price.
“We’re talking to regulators all around the world,” Bates says. “You can’t afford to wait for manpower to encode algorithms.”
