High Frequency Forecasting, (HFF).
Mon, 11 Feb 2013 14:14:07 GMT
(c) Christopher S Kirk
[A summary note intended as a prelim to a short series of relevant highlights]
Whilst the regulatory arguments for and against High Frequency Trading, (HFT) continue, a number of investment teams are already forging the next path forward and grasping the capability to advance above and beyond the provisions of HFT by several orders of magnitude.
This advance is more middleware than hardware, more spongeware than software...it is a group of dynamic machine learning algorithms that adapt at every change in data and live in a unresting world of constant flux.
Some years ago, the so-called Artificial Intelligence methodologies developed theorems which astounded the world and provided promise. Sadly, many of them never materialised into useful tools in the public domain yet their groundbreaking ideals provided the foundation for one of the most advanced sectors known, that of machine learning; the fundamentals and capacity which powerhouse the release of computers to be relatively free of human constraints and to learn and adapt to new horizons along the data-knowledge continuum.
These new-breed programs are free to learn, they do so at an unbelievable rate, not simply memorising events but examining (and acting upon) triggers that humans can only gasp at! A well-written system is flexible enough to develop itself and indeed within a modern algorithm, functions morph even whilst new data is being acquired in order to increase the rate of machine-awareness to real-time.
These new generation machine learning algorithms are in use today; some of the latest implementations have been developed to exploit multidisciplinary advances in electrical and mechanical engineering and in face, speech and hearing recognition and they operate as non-stop cycles of computational excellence in obscure space that are in a 'world of their own' whilst being connected actually and virtually to data-feeds.
As autonomous systems they exist in undefined states of constant learning and decision-making. These decisions when directed at the financial and capital markets can co-exist with high frequency execution systems and develop a controlled market. They do so by making forecasts not just milliseconds ahead (to seize arbitrage opportunities as in the current use of HFT systems), but to learn from trigger and event management and their resolution to evaluate risk, return, positioning and liquidity up to several seconds ahead. This foresight of vector prediction and consequential flow control enables not just a more immediate inter-asset transfer capability but also inter-market positioning to framework global money management.
On balance therefore, When HFT regulatory arguments are considered one might also maintain an awareness of new-generation, wholly dynamic computational systems which are as alive as humans yet unrestrained by temporal or spatial limitations; systems that are already providing new frameworks by developing a 'non-stop' world of High Frequency Forecasting.