Michael Jordaan stakes his retirement on machine learning at AI-powered fund
NMRQL Research has a unit trust and hedge fund, and plans to launch factor-based funds, but its cutting-edge technology is proving a hard sell
NMRQL Research — SA’s first investment manager powered by machine learning — plans to launch factor-based funds that it hopes will be easier for investors to understand.
The Michael Jordaan-backed fintech uses artificial intelligence (AI) and machine learning to drive research, analysis and stock selection, offering a hedge fund and unit trust fund.
Jordaan, the former CEO of First National Bank, has invested his retirement savings in the NMRQL unit trust.
"It is important to demonstrate to outside investors that I have proper skin in the game," he told Business Day.
But the sophistication of NMRQL’s technology, which removes the human element in investment decision-making almost entirely, has proved a hurdle to sales. NMRQL’s assets of R75m, of that, R25m in the hedge fund and R50m in the unit trust fund have not grown much since the beginning of 2018.
"It’s been a little bit of a hard sell," NMRQL Research CEO Thomas Schlebusch told Business Day.
"I think it’s conceptually quite difficult… if people don’t understand it, why would they buy into it?"
NMRQL is feeding more than 1,200 machine-learning algorithms about 19,000 different "time series" — data on anything from stock returns and bond yields to commodity prices over time. By learning from these different sets of data, the algorithm identifies patterns and predicts future market moves.
It is one of a number of applications of AI and machine learning that promise to disrupt dozens of industries and jobs.
NMRQL’s model adapts in real time as new information becomes available. An uncertainty metric provides feedback on the certainty of predictions.
The technology could be used to predict factor performance, said Schlebusch.
Factors help explain the drivers of returns across asset classes over time.
They generally relate either to macro-factors, such as economic growth, or investment styles, such as value investing. NMRQL could, for example, identify stocks that appear cheap but then continue to languish or decline further, Schlebusch said.
NMRQL’s unit trust fund, a multiasset, high-equity fund, is down 2.1% since its launch in October, relative to a 2.8% decline among the category average, said Schlebusch.