A $7.5 billion money manager with roots almost as old as quant-investing itself is going all-in with machine learning.
Millburn Ridgefield Corporation placed robots at the very heart of its open systematic strategies after a six-year experiment. Now, the New York firm is raising cash for a new computer-powered strategy that will actively trade ETFs and baskets of underlying securities.
Millburn is banking on artificial intelligence as it moves further away from its 1970s-era tradition in trend following, which typically uses futures contracts to surf the momentum of assets.
Co-chief executive Barry Goodman says statistical-learning programs scanning a broader set of data can figure out the nervous system connecting markets. That’s how the firm plans to beat the increasingly crowded world of quantitative investing.
-
“I admit I’m nervous about how our clients will react,” AQR's Cliff Asness wrote following the worst run in global equities in six years.
November 8 -
"It's going to become increasingly more important for financial professionals to understand the potential benefits of sustainable investing," an exec says.
December 10 -
Asset managers who flocked to safety in utilities and dividend stocks at the end of 2018 have seen their pain compounded by the latest risk-on rally.
November 8
“The machine-learning approaches in a broad sense allow us to adapt relatively quickly to environments where alpha gets arbitraged away, or where the structure of the markets themselves changes,” the Millburn executive said in an email.
Systematic traders of all stripes are investing in machines designed to improve over time without explicit human instruction in order to get ahead of the pack.
Millburn’s new equity fund will use machine learning to decipher signals from ETFs in order to make bets on the products and baskets of underlying securities such as members of the S&P 500 and MSCI World.
The algorithm, for example, might discover that momentum trades work best during seasonal shifts in volatility — something often buried in masses of data.
“Figuring this out is not trivial, and not something humans could do,” according to Goodman, who joined the company in 1982.
Millburn is the successor to a quant shop set up back in 1971, shortly before research into options pricing helped unleash an explosion of systematic trading. The firm’s managed-futures program launched in 1977, representing one of the world’s longest-running trend-following strategies.
It said its $615 million Multi-Markets Program, a long-short strategy using futures and currency instruments, has returned a net 47% over the past five years through July, during which it played with AI methods.
That trumps around 18% over the same period for a Societe Generale index tracking commodity trading advisors which are typically trend followers.
The average expense ratio among the leading 20 is nearly 40 basis points cheaper than what investors paid on average last year.
The AI buzzword encompasses a wide range of techniques, from interpreting text to mimicking neural networks in the brain. To skeptics, it all remains untested, complex and prone to human-like pitfalls such as detecting patterns in backtests that fail to materialize in the real world.
That’s not stopping a herd of money managers betting computers will uncover patterns undetected by the human eye.
JPMorgan Chase’s asset management arm is planning a strategy to invest in statistical-arbitrage hedge funds powered by machines that learn. Berkeley-based Voleon, which depends on the technology for trading, has seen assets in its hedge fund double to $5.1 billion in the year through June 1.
Millburn hasn’t given up on trend following entirely despite its pursuit of systematic strategies with a more macro and cross-asset tilt. Price momentum remains one of many trading signals used by the firm. But with more and more cash chasing the same quant strategies, Goodman reckons it’s time to shake things up.
“Positions began to look more similar across various trading firms and we had what from time-to-time could be pretty significant crowding,” he said. “This meant less diversification for investors and potentially higher volatility.”
