If IBM’s Watson can consume one million books per second, does that make it smarter than you? Is the possession of more data and information always an advantage? Big data and algorithms are transforming business and the world as we know it, including investing. Data is the lifeblood of quantitative trading strategies, which in recent years have exploded onto the Wall Street stage and are especially favored by hedge fund managers. With discretionary/fundamental trading now representing a minority of trading volume, algorithmic-based investing is not only prevalent but also increasingly nuanced. This growing march toward quant impacts both investment management and advisory services, and begs the question: go the way of the bot, or not?
‘No man is better than a machine. And no machine is better than a man with a machine,” said Paul Tudor Jones after laying off 15% of Tudor Investment Corporation employees to intensify focus on technology-driven trading approaches, a fairly strong endorsement of quantitative investing. The numbers tell the story: a recent JP Morgan report found that pure discretionary traders now only account for about 10% of trading volume in stocks, with passive/quantitative accounting for 60%. Barclays estimates the current AUM of quantitative funds to be $500 billion, while HFR puts that figure even higher, at above $900 billion.
Algorithmic-based strategies, also known as systematic, quantitative, quant, or black box strategies, have become ubiquitous in the hedge fund space and have expanded to include more esoteric strategies like momentum, volatility, relative value, mean reversion, in addition to the more common equity-based ones. These strategies rely on sifting and processing large amounts of data on powerful computers to uncover specific signals.
Big data is also the fuel for game-changing artificial intelligence (AI), the basis of algorithmic investing and a progenitor of machine learning and deep learning. AI is already an essential part of things we use every day—smart phones, cars, household appliances—and is behind some of the major advancements in medicine, the military and more. AI is more sophisticated than simple data analysis as it is not just created to compute, but to emulate human thought processes. AI makes decisions based on quantitative analysis but also can make ad hoc decisions based on changing data inputs; in other words, it learns judgement through experience.
What about machine learning and deep learning? Machine learning essentially teaches a machine how to do something by feeding it data and directing it to make predictions based on new data; it adapts and learns over time as the data accumulates (like Watson). Deep learning is a subset of AI that attempts to mimic the neurons in the brain in a more complex, non-linear process: it’s what enables Tesla’s self-driving car and Amazon’s voice-enabled smart speaker, Echo. The next frontier for deep learning is financial processing, and while some estimate its full use in asset management to be years away, a few firms are already pioneering this approach and even hiring deep learning teams.
While it may be easy to say, “She with the most computer chips wins,” the argument for fundamental investing rightfully contends that humans have the unique ability to connect dots that a program may not recognize. Tudor Jones’ man plus machine argument is valid in that investors are not likely to trust solely a computer or an algorithm. The value of nuance cannot be underestimated. “Sophia,” an anatomically realistic female robot that was interviewed by Charlie Rose on 60 Minutes, uses AI technology. She could carry on a full conversation yet had no answer to his question, “Have you been programmed?”
The rise of robo-advisors has also been a trend; enhancing the advisory experience with big data can provide much more information to give a boost to performance and the overall business. But entrusting an investment portfolio entirely to a system devoid of any human oversight has its obvious drawbacks. Robo-advisory also presents its share of compliance and regulatory challenges: SEC filings and requirements, risk assessments, potentially unsuitable investments to name a few.
Systematic strategies can’t operate in a total algorithmic vacuum. In addition to the math formulas required to find patterns and connect the dots in a quantitative strategy, a grasp of economics and market behavior—a suspiciously fundamental exercise—is required. Some quant funds even track the old-school yield-curve. Humans and machines are (ironically) somewhat codependent here. The role of behavioral bias in the face of the scientific method also can’t be ignored. This behavioral bias, after all, accounts for much of the divergence between value and price in stocks.
The bottom line: it’s about information, whether digital or analog. How efficiently that information can be processed is the question. Machine learning can truly only shine when there is an ample supply of good data, and an objective, strong computational ability. This isn’t always possible; the data set on some great long-term investments simply is not big enough. One will not mine millions of data points on a company that has only been publicly traded for one year. Even Google, Amazon and Apple don’t have decades of data which can be input. Any shortcomings of quantitative investing therefore may be attributable to the same concept behind its advantage: information.
Finally, can mass amounts of data alone give rise to the next great mind who might move civilization, or investing, forward? A Mozart, Picasso, Hawking…or Buffet or Soros? Ask Sophia that one.