About a decade ago, if you said that the trading floors on Wall Street were a good place to go for some peace and quiet, you would get some very strange looks. But in five or 10 more years, that will probably be the case. Old-school trading pits have, of course, already been cleared out.
Christoher Aney, the ex-head of short-term markets and equity derivatives trading at RBS and formerly an executive director at Nomura and UBS, believes that electronic trading platforms and, in particular, artificial neural networks will make trading floors merely hum with the buzz of computer hardware as both buy-side and sell-side firms employ fewer human traders.
An artificial neural network (ANN, also known as a connectionist system) is a computing system with an information-processing structure along the same lines as the human brain’s architecture that enables a deep-learning process from observational data inputs.
“A trader’s job is to funnel a vast amount of information down to some very simple decisions: buy, sell and at what price,” Aney said. “Until recently, this was an incomprehensible job for most mortals and machines.
“That is why the pay rolls for good traders was, well, jaw-dropping,” he said. “But, now comes the artificial neural network that can do that job better – digesting more information faster and more accurately, if calibrated correctly.”
Artificial intelligence breakthrough
It is the same principle that researchers are using for driverless cars, IBM Watson and DeepMind Technologies’ AlphaGo that beat the world champion in Go, a Chinese board game that involves complex strategies.
You train the neural network on a massive amount of test data until you are satisfied that it makes the optimal predictions or decisions. But unlike program trading of old, a good neural network will keep learning and changing on the new information, Aney said.
“So, if it is working well, it will continuously update its parameters with new information and constantly re-train itself,” he said.
Artificial neural networks go from fantasy to reality
Applying the machine-learning capabilities of artificial neural networks to trading and other financial services functions is no longer fantasy.
For example, XTX Markets, a quantitative electronic market-maker launched by Deutsche Bank’s former head of quantitative trading research, Alexander Gerko, is now within the top 10 firms in the massive global foreign exchange market and brags that it is “one of the world’s fastest growing trading shops but doesn’t have any traders.”
Many global banks, high-frequency trading firms and hedge funds have set-up teams to create their own artificial neural networks, from Point72 Asset Management, Two Sigma, GQR and KCG to Morgan Stanley and J.P. Morgan. The race is on.
“These firms are not only competing with themselves for the top-talent, they are also competing with Google, Amazon, IBM and other tech firms,” Aney said. “So what is an old-school trader to do? Evolve into a trader/programmer or look forward to [living off of] a good 401(k).”
Photo credit: Henrik5000/GettyImages