The race for the fastest automated trading
With programmed trading exploiting every advance in computing, are we headed for an environment where trades happen so fast that they cannot be stopped, even when something goes wrong, for fear the problems would only get worse? People make emotional judgments when trading, and may overreact to surges in activity that trend downward - but they do have response limitations that dampen cycles to a certain degree. With the implementation of super-fast hardware enabling programmed trading at speeds never before possible, at which point do our financial systems become so interdependent, and laced with competing algorithms, that we are afraid to over-ride the programmatic decisions - or worse, shut down a system without considering the consequences to every other dependent system?
Two companies are pushing the upper boundaries of processing financial messages at speeds that are alarming, and far exceed human capacities. Solace and Exegy are purpose built systems that have implemented software into the firmware of the machines exploiting the use of Field Programmable Gate Arrays (FPGA).
There are benefits to financial institutions that adopt such technical advances, but I am not sure the consequences to the financial markets are completely understood, should these types of systems become common and more trading is activated by algorithms that exceed human reaction time. I suspect that while we may see periods of market hyperactivity, we will have equal and potentially more frequent unsettling downturns as the financial systems go through a period of technological advancement that disrupts balance and requires systems to adapt.
The network theory to completely understand the interdependent forces and the degree of transparency needed to regulate the systems are not available. This is a fertile area for research in information value flows and the network economies of systems, that may help determine a sensible regulation framework to avoid a collapse of systems that we simply can’t anticipate using the current macro models of finance and economics.
| Comments | Print |
Share | |
| Add comment | RSS |
March 7th, 2009 at 4:51 am
after working out a couple of policies, couldn’t an interconnected global network of trading systems run on a delay (ala radio and tv brodcasting) to anticipate the destabilizing effect of a trade and deny its execution? [it would have to operate in realtime rather than simulation b/c non-programmed trades would still occur randomly]
what would be the policy definition of a destabilizing trade?
what would be the policy for compensating the parties denied their trade?
if trading were totally programmed, would it be a ‘free’ market anymore?
August 17th, 2009 at 12:30 pm
Trading is a zero-sum game. As I understand it, algorithmic trading systems have two purposes. They seek to harvest pricing anomalies, and they try to send false pricing signals to induce favorable market movements.
But both of these are self-limiting. Obviously, taking bid/ask quotes that are out of line clears them.
The second effect is self-limiting, too. As algorithmic trading grows more pervasive false signals will fail - they’ll fail to affect market conditions, and they’re sure to lose money for the trader.
Mistakes and counterparty failures remain problems, but algorithmic trading offers the promise of uncovering them faster — or even keeping them out of the market.
September 19th, 2009 at 5:45 am
Great Article but Can you also explain us also where we could find more information about perceptrons and how to program them? what about a neural network with more layers, will it be more accurate? Is there any books on the subject (in english) you can recommend?
Thanks again for your help.
September 28th, 2009 at 8:34 pm
I am Rhea S. I have visited your website and I would like to congratulate you on building such a valuable online resource. I am sure your visitors find your site as useful as I did.
Have a great day.
Thanks and regards,
Rhea S.
October 13th, 2009 at 8:46 am
interesting