The Future Belongs to the Quantamental

I recently finished the new Michael Lewis book (The Undoing Project) which explores the friendship between two ground breaking Israeli psychologists, Daniel Kahneman and Amos Tversky. Their studies showed the ways in which the human mind makes errors in judgements and forms biases particularly in uncertain complex situations. Their early works have revolutionized big data studies and have pushed forward advances in evidence based disciplines like medicine and finance. Their research reviews the limitations of the human mind to process massive amounts of information. Thus a human mind augmented with computational processing can be more powerful and accurate at predictions than a solo human mind if programmed correctly. The future of complex decision making and predictions is rapidly evolving thanks to data availability, the speed of computing power, and machine learning. As we know from our recent presidential election and Brexit -- polling data, surveys, and algorithms can be wrong but the capabilities are always improving. Statistical models are at best approximations and at worst incredibly misleading. In financial markets in particular we see the power of machine learning, and statistical processing of both fundamental & macro data leading to serious alpha at hedge funds like Renaissance Technologies where 30% annual returns at the Medallion Fund are the norm. 

“The best way to predict the future is to create it.” - Abraham Lincoln

One of my favorite movies that explores the future of prediction is Minority Report. It is loosely based on a Philip Dick short story set in 2054, where "PreCrime", a specialized police department in DC, apprehends criminals based on foreknowledge of crimes committed provided by three psychics called "precogs". The film's central theme is the question of free will versus determinism. It principally examines whether free will can exist if the future is set and known in advance. Is a society safer by jailing high probability future criminals or letting crimes occur. It is an exceptional film on many levels and good abstraction of a living neural network sharing subconscious visions to create a closed form prediction. Did person X commit a crime on Y? Sometimes we need to make necessary approximations to create simple models based on linear regressions but imagine all of the inputs required to establish motive, intent, and certainty of an action like murder? Maybe my 4 year old son will live to see this method of crime prevention. 

Now, fifteen years after Minority Report the headlines are plastered with optimism surrounding prediction algorithms with applications across industries including finance, science, and agriculture. This optimism is driven by significant advances in AI and machine learning, our understanding of neural networks and the volumes of big data collected from new sources. Neural networks are essentially a computational approach, with the goal of modeling problem solving in the same way that the human brain would. Modern neural network projects typically work with a few thousand to a few million neural units (measures of computation connectivity) and millions of connections, which is still several orders of magnitude less complex than the human brain and closer to the computing power of a worm. Kevin Kelly (co-founder of Wired) has suggested that processing power will be measured in AI units much the same way combustion engines are measured in Horsepower. The power of a horse is more concrete than an AI unit (perhaps for now). It is hard for us to grasp this new level of computation power so our minds need to make attribution to something we know (an availability heuristic). Noted inventor and futurist, Ray Kurzweil, tells us that by 2023, that computational abilities reaching 10^16 calculations (10 petaFLOPS) per second—roughly the equivalent of one human brain’s processing power—will cost $1,000.

In light of all the advances in computing power breakthroughs are coming in financial markets and professional investors are eager to fund the new investment geniuses entering the field from physics and statistical backgrounds. Even Ray Kurzweil himself, of course, is emerging as an investment manager with his own fund! With all of the talk of successful algorithmic centric funds it would almost seem that the path to riches is paved with smart people testing various mundane processes, applying machine learning, and boom….PROFIT. If it were that simple markets would be entirely efficient and the machines would already have taken over Wall Street. Perhaps some day, but not yet. Technology is a lever to let the user do more with less, but it is not yet omnipotent. Investment practitioners around the world are constantly in pursuit of alpha generation and this process starts with first principles of situational definitions, the testing of a thesis under various conditions, and complex implementations. It is hard to imagine that without rigorous process an investor can create repeatable strategies.

As the size and complexity of the markets have increased so too have the studies of the market participants. The field of behavioral trading and investment psychology in particular has undergone a significant evolution through the last few economic cycles. In addition to the early works of Kahneman and Taversky another expert, Brett Steenbarger Ph.D. reminds us that "The greatest big picture mistakes a trader (investor) can make are ‘failure to adapt’ & ‘failure to invent’ and they lose probabilistic edge." It seems that many hedge funds have recently lost their edge and struggled to beat their benchmarks net of fees in the past few years. Cambridge Associates only recommending 250 of the world’s 11,000 hedge funds. What institutional investors really want is to invest in funds with better returns. Better returns have been delivered by funds with multi-disciplinary and repeatable processes. In sum, hedge funds need to continuously adapt and invent.

Much like the 3 precogs with complementary and sometimes divergent visions in Minority Report it is important for investment teams to mix their different perspectives on the same situation. This blend of brains analyzing an investment as a “quantamental” (quant and fundamental) team versus a solo fundamental oriented PM shows a better systematic review of investment opportunities with higher likelihood of positive outcomes. The last few decades, hedge funds in particular were primarily relying on catalyst focused trading strategies and reacted to expected events like earnings announcements and unexpected events like changes in Fed policy or OPEC quota. The future of trading is going to see the best returns from the investors that understand more than just one simple correlation based on the reaction to new information and its impact on a stock. Trades that come out of a quantamental team are more sophisticated and can more ably be categorized as the “reaction to the reaction”. The “reaction to the reaction” is born out of a more complex systematic understanding of data interdependencies over the medium term. This methodology will push the alpha frontier by creating persistent uncorrelated returns synthesized from the most detailed and granular data sets tied to financial statements. By studying previous catalysts, understanding all of the key drivers of a stock both from a company & economic view, observing how a stock is repriced historically investors that have deep data infrastructures tied to this ontology and taking advantage of multiple unique data sources will repeatedly beat those that don’t. The next generation of great investors will be forged by this cross functional multi-disciplined deep data approach. This approach will show an evolved workflow that will not live in one terminal built many decades back but rather 3 dimensional data sets with time series history and all catalyst & drivers mapped, processed via significant AI units, and high confidence prediction capabilities per unit of risk.

As I meet asset managers regarding their data needs I hear requests like “Can you create the Alexa for Finance?” What everyone ultimately wants is an easy button with pre-configured data mappings and correlations for securities along with a simple interface. There is no question that this will happen it is more a question of when. This magic investing device starts with assembling the right data inputs, writing the relevant algorithms, and creating an intuitive workflow. The magic of data delivery is focusing on aggregating the right data and delivering it in the right formats. The incumbent data providers have taken decades to hone their craft and there is a clear oligopoly for basic financial data, company guidance, research, and news bundled into a seat. That is the traditional way and with the surplus of data it is now often a bit hard to search and organize data to make investment decisions under this workflow paradigm. When data is not optimally organized for computation investors hit a data limit described well by, John Naisbitt, who explains that "we are drowning in information but starved for knowledge." There is clear white space in filling the gap in the market for deep structured and cleaned data delivered in an optimal way for the sophisticated managers that seek to apply statistical methods to a blend of fundamental and macroeconomic strategies.

The investment best practices blueprint continues to evolve as demanded by an incredibly competitive industry. The Darwinian challenge for active investing comes down to finding, growing, and sustaining alpha while balancing risk. This discipline is incredibly challenging and data is both an opportunity and threat. Considering all of the evolution happening every day, hedge funds may need to consider being open minded to adapting their methods and stay on top of dynamic industry improvements to workflow. Investment returns for the past few years have tended to favor the “quantamental” approach and the LP dollar inflows have followed based on these returns. Quantamental success is driven by the breadth and depth of data available, the low price of computing power, and the sophistication of the market leading asset managers. These Quantamental managers have optimized for understanding intrinsic value and have a centralized data strategy with clear data ontologies for their investments including their historical drivers and interdependencies across industries and catalysts. Welcome to the future of predictions applied to investing!