Are the robots really better than humans at investing? With JPMorgan’s quant master Marko Kolanovic making the jaw-dropping claim only 90% of stock market volume is computer or algorithmic quant driven – exceeding the previous whisper number of 80% now involving systematic quant strategies – has the financial world reached a singularity point of no return? Or are the best computer-driven systematic programs actually driven by innovative humans?
Quant strategies expand
Regardless of the study – Barclays recently tagged systematically driven strategies at near $500 billion in assets under management, near the high end of the range – what is clear is that such strategies continue to rapidly expand in assets and strategy derivatives.
What started as primarily a systematically-driven trend following world, originally propelled in the 1980s but with roots dating back further, has morphed into a host of momentum, volatility, relative value, mean reversion and equity-based quant strategies. New strategy directions excitedly being watched include research into machine learning, artificial intelligence, and deep learning. But these strategies have yet to positive returns on a consistent basis nor reach their highly promoted potential.
In fact, the laugh among quantitatively-driven portfolio managers when many Silicon Valley tech firms said they would turn on its head quantitative investing was the claim they had already built systems that could “learn from market behavior, self-build a strategy, then pick the timing to execute that strategy and deliver positive performance.”
That was the general deep learning, artificial intelligence promise. Just like the show the Jetsons and Star Trek, technology hasn't reached that point, yet. What remains unclear is what will be uncovered in the process of getting from Point A to Point B.
Some of this work is currently underway.
Certain systematic hedge funds, large asset managers and investment banks in Boston, New York and Chicago are said to be looking at algorithms to identify a market environment and then adjust strategy. This is but one point of strategy deviation that makes the world of quant strategies seem so complex. But at the core there are building block insights that come from those with financial market knowledge.
Most of the Silicon Valley crowd didn’t much recognize market environment analysis any more than they didn’t recognize the role of connecting noncorrelated dots in building a trading algorithm. They have been good at marketing the dream, but their work needs to generate tangible results. That might not be how a Silicon Valley start-up works, but it is how financial markets operate.
Core insight into quant strategies: All formulas are if-then statements
At the little-discussed core of the strategy, certain practitioners – including those researching new strategies -- recognize that all formulas are built upon if-then Boolean logic. If-then logic has fundamental limitations into how information can be interpreted and decisions made. When building a market formula, success is found in identifying repeatable patterns that ideally have an economic justification. But many times finding these patterns requires an understanding of economics and market structure, which is best lubricated with a creative combination of noncorrelated dot connecting found in fundamental minds. Quants work best when their algorithms have a fundamental market understanding. At times, that’s what drives a quant strategies. Talk to many of the quant leaders, from AQR founder Cliff Asness to those at Renaissance Technologies and what they all have at their core is an understanding of markets and identifying repeatable patterns in general market behavior.
Silicon Valley’s promise of a learning machine building its own formulas without significant human minds behind it creating the core formulaic logic – and delivering positive returns – that was more than a stretch for some, including “Mr Momentum” Cliff Asness.
What Asness has voiced through white paper and news interviews is a key concept. When defending the value of discretionary stock picking human managers, he pointed to their ability to connect noncorrelated dots that a formula might not pick up. “I don’t think those patterns -- to the extent they find them, and I don’t think they’ll find anything because I think they’ll be finding patterns that don’t repeat,” Asness told Bloomberg. “I don’t think it will take over the need to have individuals look at companies.”
Artificial intelligence and deep learning may get there one day. But talk to those who build formulas and many will stress the human intellectual driving the quant strategy decisions. One of the best-known quant hedge funds, Two Sigma, has a keen interesting in watching the developments. They haven’t generated any positive returns from it to date, a source at the fund said, but they are watching closely to find strategy innovation.
Building an investing formula requires significant creativity in understanding how markets and how previously unknown or uncharacterized events might impact markets. The people behind quant formulas typically recognize how markets work inside the structure of a math formula, which is a conceptual knowledge derived from a degree of creativity that moves beyond the limitations of Boolean logic.
Barclays frames growing interest in quant strategies: "Strong case for portfolio inclusion"
A June 2017 Barclays Capital Solutions Hedge Fund Pulse report frames the key issues of the day, bolstered by a study of some of the largest fund managers and allocators.
The report touches on market cycle analysis, the difference between alpha and beta and in doing so points to a wide variety of information that shows how vast and nuanced it is. To understand the increasingly important investment sector, some fundamental managers start by looking through the lens they natively understand: Performance, market cycles, and beta correlation.
Fundamental portfolio managers look through the lens they best understand to understand quant strategies
In regards to performance, systematic and discretionary managers perform close to equal in the aggregate, with certain strategies generating noncorrelated return streams being one point of differentiation.
“Over the last decade, the performance disparity between corresponding systematic and discretionary strategies in the equity space has been small, although the difference appears to be larger in macro,” the report noted, correlating with other studies that show systematic CTA strategies generating close to the same – if slightly smaller – absolute returns, but doing so in different market environments. The result of Barclays analysis is that systematic strategies display a positive Sharpe ratio and the “correlation characteristics of systematic strategies make a strong case for portfolio inclusion.”
While touching on several key topics, the report did not discuss the Sharpe ratio or the differential in risk weighting between upside and downside deviation and how that is used in algorithmic formula development.
The world is just starting to awake to the systematic world. For those inside the industry and watching its development it isn't new. For others who might not be acquainted with what is a way of thinking as much as a way of investing, it all seems new. How could this systematic method of looking at the world be ignored for so long?
In part, it may be due to a math aversion.