As I am involving myself more in retail FX, I cannot stop being awed by a wave of home-based algorithmic traders, connecting from a range of tools, all the way from a humble Excel on DDE, to TradeStation, or even proprietary plug-ins being tweaked in long evenings.
There used to be a time when major revolutions in IT were assembled in garage by people everyone knows today.
Who knows whether the coming front wave of algorithms won’t be driven by some young individual trader, whose creativity will be spurred by a ‘nothing to lose’ factor unavailable to senior traders are well placed shops.
I was recently discussing with a large US hedge fund, which was looking for algorithmic specialists. One of the things which struck them is that the world of professional algo traders was roughly divided between flow traders, coming mostly from banks but not only, who have mastered the art of “earning the spread”, and sophisticated smaller entities traders, who, let us face it, are mostly glorified arbitragers (yes yes this still happens, would you believe it). If you meet an algorithmic trader more knowledgeable in the speed of light between New York and London, orthodromic routes on the floor of oceans, super-fast NIC cards and the virtues of fibre versus copper in a data centre than Poisson variables, he is one of those.
The quest for pure alpha remains elusive in FX, although I do know a few funds, which, with just 5 to 10 people, produce great alpha year after year. The difference with alpha-producing equity funds is that these are mostly not easily scalable. Which is also why they are not well-known? The owners often have sizable share of the funds, trade their own money very happily and are not looking for publicity. They may be low or high frequency, and, though interested in having up to date equipment, will general still speak in milliseconds when the above mentioned crowd is now talking micro if not nanoseconds.
And so, the brave new crowd of garage traders, limited by their wallets to a quest for alpha through MetaTrader, using MQL language, or TradeStation for the more sophisticated ones, who cannot compete with the super-fast sleuths using shiny 8 core engines and “fibre all the way” systems, have no choice but spend more time on their Matlab and R systems, checking correlations, learning systems and the like. Their volume is already huge. They won’t be able to scale, probably, but more because of a lack of capital than underperforming models.
Perhaps a good idea would be to organise structured, professional contests, so the best of them can faster become the star hedge funds of tomorrow. Better that than having nuclear physicists or rocket scientists ditching their socially useful career to make a few more bucks on checking whether bids are higher than offers.
Xavier's blog 