Sunday, June 24

Man Group CEO on how machines are better at alpha, and not such black boxes at all

LinkedIn Google+ Pinterest Tumblr +

When it comes to generating alpha, computers are winning some and gaining elsewhere.

“You can use technology to improve almost any process,” said Man Group CEO Luke Ellis, speaking at a roundtable organized by Opalesque.

In terms of alpha generation, Ellis broke down that process into three basic steps: gross alpha, or an idea that isn’t currently reflected in the price; trading into and out of that idea; and building a portfolio of ideas.

“When it comes to trading in and out of positions, I would tell you that computers are infinitely better at it today than humans are in the futures, FX, and equities markets,” Ellis said.

“We use a significant amount of machine learning to do that execution,” he added.

In futures and FX, Ellis said it was debatable whether humans still have an edge in coming up with alpha-generating ideas.

For individual stocks and in credit as well, however, Ellis said it was clear, at least to him, that there are humans who seem to have an edge.

Humans a bust in portfolio construction

When it comes to portfolio construction, Ellis said he thinks humans “are pretty bad compared to machines”.

“The bit everybody focuses on is can you use machine learning to generate the pure alpha piece? And here too I can confirm that we can and we have done,” Ellis said.

“We have been running pure deep-learned strategies, which do not include a human designing what it’s supposed to do, for years with client capital,” he added.

That doesn’t mean those deep-learned strategies are perfect, but neither are humans, particularly when it comes to emotional biases.

Ellis also noted that the “black box” characterization is one he has always hated.

“Whether it’s a traditional algorithm, so a natural process where you know exactly what it’s going to do next, whatever happens, or a machine-learned process where you have to build in a set diagnostics so that you know what it’s going to do next, the reality is that it will tell you exactly what it’s going to do next if you tell it what the market is going to do next,” he said.

“To me, those are very transparent boxes compared to humans,” he added.

Artificial intelligence is expensive

Also speaking at the roundtable was Areski Iberrakene, chief investment officer of Arqaam Capital UK and the Arqaam Global Macro Fund.

He said that artificial intelligence is still “so much about the data” in terms of availability and quality.

It’s only been in the last 10 years that Iberrakene has seen data being gathered in a “robust” manner, an activity his team spent “a great deal of time” on before the fund launched.

And more than exchange-traded futures, that data gathering included option volatility and OTC prices as well, among other data points. One thing’s for sure: it’s expensive.

“We must take great care in being careful, and insightful, and corrective with this technology, especially when applying it cross-asset and in global markets,” said Iberrakene.

“Should we not be, then it is something which can cost multiple millions of dollars a year – to have the proper datasets to begin with, the algos, and neural networks or whatever machine learning or supervised learning algorithm you want to do,” he added.

Active management still human

Matthew Riley, regional sales manager at derivatives exchange Eurex, said to the roundtable that the use of artificial intelligence and machine learning as a successful, truly active management process is still in the future.

“I think now we still have mostly people telling a machine what to do and that machine is then executing the process more efficiently and faster than a human can do,” he said.

“These two types of applications often get lumped together, but these are two quite different applications,” added Riley.

Share.

About Author

Leave A Reply