The Future of Investing: Humans + Machines

Next-generation investment firms will integrate well-trained humans with well-designed machines (M2X Capital LLC Newsletter, March 2019 - see www.M2Xcapital.com for PDF forms of all newsletters)

"…there is no way I would have won…you have to know how to wrestle and box . That’s just the truth." - Mike Tyson, Boxer, when asked about a hypothetical fight against Royce Gracie

On November 12, 1993, Royce Gracie won the 1st Ultimate Fighting Championship, beating his much larger competitors with ease. His fighting style, a grappling method known as Brazilian Jiu Jitsu (BJJ), confused the boxers and strikers. They would punch, he would duck, take them to the ground and beat them. For many years, the Gracie family and BJJ dominated the fight game.

Then something happened. The smart boxers and strikers adjusted, learning how to both defend and integrate BJJ. A few years later, these newly formidable fighters started to beat the grapplers.

Then something happened…again. The smart grapplers began to adjust, learning how to integrate boxing and striking.

Now, all that remains are mixed martial artists – fighters who integrate both striking and grappling. Prior categorizations have become distinctions without a difference.

Investors should pay attention. Human-driven firms, often categorized as fundamental investors where humans drive decision-making, are like the early boxers. Many are getting beaten badly by machine-driven firms, have not adjusted their approach and naively expect different outcomes. They keep punching, the competition keeps ducking and the pain will continue.

Machine-driven firms, so-called quants where computers drive the decision making, are like the confident, early grapplers. These firms should be vigilant, however, as the smart competition will adjust. Newly augmented fundamental investors will understand the often not-too-difficult math of the machines and develop strategies to both defend and attack.

The smart investment firms of the future, no matter whether human or machine-driven, will adjust like the smart fighters of the past. They will realize that each approach has strengths and weaknesses and set out to seamlessly integrate both. Just as today’s fight game contains only mixed martial artists, tomorrow’s stock market may become dominated by integrated investment firms – part human, part machine.

This begs a question: What might such a firm look like?

Imagine a Team that is Part Human & Part Machine

Quants were quick to realize how advances in technology could help them, both in reducing costs and developing unique trading strategies. Some forward-thinking fundamental investors have employed technology a bit, but it is often still quite primitive. Is there another, more radical, application of technology that could lead to a step change in efficiency and effectiveness?

Imagine an investment team composed of several human and several virtual teammates. The humans are well-trained in identifying and analyzing promising investments. The virtual teammates are well-designed amalgamations of computer code, data and math.

Picture Siri with a spreadsheet - bringing freshly analyzed data and thoughts on the portfolio to discuss with the team each morning. This is not a passive analytical tool, but an interactive teammate. Not a technology or quant overlay, but true process integration. Not a replacement of humans, but an augmentation of the team’s collective decision-making ability.

Sometimes the virtual teammates bring up points not considered which, upon reflection, all agree are based on solid logic and data. Other times, the humans explain other factors to consider such as a change in future regulation, industry structure or company positioning. Either way, the discussions are constructive, fact-based, and devoid of emotion.

Example: A Virtual Teammate in Risk Management

The starting point for creating a virtual teammate is the same as recruiting a human. First, define the responsibilities for the role. After the role is defined, the difference is more distinct: Virtual teammates are created from code, data and math whereas human teammates are recruited.

For example, I wanted a virtual teammate in risk management to analyze hundreds of datapoints in real-time, contextualize their importance mathematically and communicate them efficiently. One task of many was to track the price skew of put versus call options, a measure of forward risk perception. Below is a small part of the code I wrote which drives this virtual teammate to signal me when this risk measure deviates more than normal (note: calculations like this can now be done in the cloud without having to download reams of data):

=[coded redacted]

Virtual teammates are possible in all aspects of the investment process: idea generation, research, trading and risk management. Each can be built in a similar way, using technology, data and math to accomplish their tasks. The M2X virtual team is keeping track of well over 10,000 such data points, in real-time, every day, for the portfolio.

The Data Challenge

In addition to collecting data like put/call skew discussed above, the team must holistically understand what it means in combination with other data points and decide what actions should be taken, if any. Making a good decision is hard, however, given the dizzying amount of data combinations possible. A very simple example helps to illustrate:

1. Assume the virtual risk team above says that put/call skew is now +3 standard deviations
2. Your human analysts believe a stock has +20% upside over the next 12 months
3. Their scenario analysis highlights base, upside & downside cases of +25%, +30%, and -15%
4. The stock rallies 10% given a combination of macroeconomic, industry and company news
5. Your virtual team highlights several quantitative market signals foreshadowing headwinds
6. And, they indicate a certain factor (e.g., momentum) in the stock is +3 standard deviations
7. The stock is trading +0.7 standard deviations above its recent valuation range
8. The stock is trading -0.2 standard deviations below its market-relative valuation range
9. Short interest has grown from 3% to 6% in the past month
10. Implied volatility in the options market has surged materially

What do you do after the stock moves up (#4 above)? Keep it at full size? Sell it all? Sell a portion? Sell call options to monetize the change in volatility? Most focus on just a few data points, going with their “gut.” Others vaguely talk about pattern recognition but lack any real mathematical argument. The reason: it is nearly impossible for humans to analyze such quantities of data well.

For example, imagine a three-number briefcase lock where each number has 10 numerical options (0-9). This simple lock has 1,000 possible combinations. Four-, five-, and six-number locks have 10,000, 100,000 and 1,000,000 possible combinations, respectively. Now, think about the investing example above which had 10 simple data points, many of which could be broken down further in terms of distributions and standard deviations. A massive number of combinations are possible. Unable to analyze such quantities of information well, most human-driven firms rely on simple logic applied to a few data points. For example:

• Deere will earn $12/share mid-cycle; At a 15x multiple, there is 15% upside
• Berry Global will generate $675mm in free cash flow; An 8% yield means there is 30% upside

Maybe. Maybe not. But, for sure, there’s other information that can help inform your decisions. If you are a human-driven investment firm, why ignore it and let the quants get the benefit?

The Virtual Assistant to the Human Portfolio Manager

To overcome the data challenge, I created a virtual assistant to help aggregate the data, understand its meaning and decide on actions. Their intelligence is driven by coded sets of conditional probabilities that enable them to understand that no single data point means much, but certain combinations can signal a lot. Some simple lines from its much longer algorithm help demonstrate:

=[coded redacted]

=[coded redacted]

My virtual assistant helps augment decision-making in many areas:

• Timing & Sizing: Helps to better time entry and exit of positions; along the way, identifies where position size should be increased or decreased
• Research: Identifies where fundamentally-driven stock views are at risk given increased factor exposure (e.g., momentum) that might act differently in the coming trading regime
• Trading: Develops gameplans to use near-term volatility to trade a medium/long term fundamental view
• Structuring: Suggests options trades that more optimally monetize the fundamental view given changing market conditions
• Idea Generation: Highlights names on the watchlist that have the highest potential for mispricing and a higher probability of the mispricing closing sooner

My virtual assistant is evaluating well over 100,000 combinations of data every day, only communicating when signals show risk or present an opportunity. The volume of combinations can sound extraordinary but, given the automation, analyzing this information and making quality decisions takes less time than when far less data was evaluated in the past.

Do We Even Need Humans? Yes, We Do.

With such power from the virtual team, some may wonder if humans are even needed? There are some very good reasons to have humans on the team, including their ability to:

• Understand Complex Systems in Ways Machines Cannot: The stock market is a complex, adaptive, system where feedback often creates outcomes that are hard to predict. Well-trained humans can help understand this in ways that machines cannot. Both have something to add and working well together is better than either working alone.
• Recognize When the Future Might Not Look Like the Past: Machines typically inform their view of the future by what has happened in the past. This is often very helpful but, at times, can be misleading. For example, what if a new regulation or technological innovation changes industry structure? Past performance might not be indicative of future results.
• Identify Spurious Correlations: Increased computing power can detect relationships that were previously unknown, but useful. However, machines often find patterns that are spurious and not useful in predicting the future. Well-trained humans can help understand the difference (see AQR’s paper, The 7 Reasons Most Machine Learning Funds Fail).
• Generate Proprietary Data: Quantitative analysis is driven by data, the holy grail being datasets that are predictive and proprietary. Short of sending a satellite into space, your human team’s assessment of future trends might be one of your only proprietary datasets.

The key is to have a process that gets the benefits of well-trained, human judgment while mitigating its inherent limitations. This is where well-designed, virtual teammates can help – augmenting the collective decision-making ability of the team, not replacing its humans.

How the CIA Helped Spark the Idea

Over a year ago, I participated in research with IARPA (the Intelligence Advanced Research Projects Activity), a research group under the Office of the Director of National Intelligence that aims to help the CIA and FBI do their jobs better. As a decision science wonk with a graduate degree in the topic, I have studied the CIA’s approach to intelligence and was excited to participate (e.g., see Words of Estimative Probability, published 1964, declassified in 2007, for a classic paper).

Several times a week, IARPA would pose questions such as: What will inflation be in Somalia in 6 months? Or, how many terrorist attacks will occur in Europe from now until year end? Forecasts had to be quantified in a probability distribution and could be adjusted during the forecast period.

One night, while I was updating my Somalia inflation forecast at 1am, my mind started to wander (forecasting Somalia inflation late at night can do that to you!) The CIA, I thought, should develop computer-driven virtual agents that keep track of this type of data, in real-time. Integrating virtual agents with human agents could lead to much better intelligence, if done well. But I concluded the CIA likely had such technology (e.g., Palantir) and, given it was late, I went to bed.

The next morning, I started to brainstorm how some of these ideas could be applied to the investment process. First, I evaluated prior investments and trades to understand what could have been done better. Second, I brainstormed where technology, math, data and logic could be implemented with more rigor. Third, I explored and further educated myself as to how technologies such as various APIs, web scraping, and machine learning could help.

And the virtual teammates were born.

M2X = Man + Machine

To be clear, this is an enhancement in execution, not a radical change in approach. M2X’s approach is still focused on understanding the probabilities of potential security price movements. Get that right and you can make good decisions. If not, you are guessing no matter your confidence. While Wall Street is filled with many extremely confident guessers, M2X prefers to be a humble, probabilistic thinker – less exciting for television, but more likely to make you money.

To get probability right, analysis of future fundamentals – typically the most causal driver – is critical. But, especially in a market dominated by quants and computer-driven trading, there are many other factors at work (pun intended). The virtual team not only highlights a different perspective on future fundamentals, but also analyzes these other factors that drive security prices no matter the fundamentals. In short, the virtual team helps to understand probabilities better.

I could go on, but it’s late and my virtual team just sent an email of issues to discuss at 7am tomorrow. They are always on time and diligently prepared – so I should get some sleep.

***************************************

To learn more or keep in contact:

• Please reach out to schedule a call, [email protected]
• See www.M2Xcapital.com for more information
• Email here if you would like to be put on the mailing list https://www.m2xcapital.com/contact

Thank you for your interest,

Michael J. Molnar

M2X Capital LLC | Portfolio Manager and Managing Member

[email protected]

Michael Molnar Biography

Michael is the Portfolio Manager and Managing Member of M2X Capital LLC: a long/short, public equity investment firm. M2X’s investment process integrates well-trained human judgment with well-designed machine analysis to better identify, analyze, select and manage investments for clients.

Previous to M2X, Michael was a Founding Partner and Co-Managing Member of Lorem Ipsum Partners LLC, a long/short equity hedge fund. During his tenure, the firm had top-tier performance and grew to approximately $200 million in assets under management.

Prior to Lorem Ipsum Partners, he was a Founding Partner of Greentech Capital Advisors, an investment bank focused on serving clients in the sustainable infrastructure industry. He advised clients on M&A transactions, strategic joint ventures and private capital raises. He served on the Board of Directors and the Commitments Committee, helping the firm to grow nearly 10 times, raise two rounds of capital and expand to three offices around the world. In 2017 and 2018, the firm was ranked the #1 firm worldwide in Bloomberg’s clean energy transaction league tables.

Prior to Greentech Capital Advisors, Michael was the lead equity analyst for the U.S. alternative energy and coal sectors at Goldman Sachs (he was the first alternative energy analyst at Goldman Sachs). At Goldman, he also helped to start the Small and Mid-Cap Research Team and was a member of the Special Situations Research Team.

Prior to Goldman Sachs, Michael was a Visiting Research Fellow at Accenture’s Institute for High Performance Business, a company-sponsored think tank. His research focused on how companies' intangible assets and liabilities led to future value creation or destruction and was published in internal and external business journals. This was a form of environmental, social & governance (ESG) research before the term was widely used. He was also a manager in Accenture’s strategy consulting practice.

He started his career as an auditor with Arthur Andersen LLP.

Michael earned his MBA in finance at the University of Chicago, his MSc in decision science at the London School of Economics and his BS in accounting (major) and philosophy (minor) at Rutgers University. While in university, he was 1 of 100 people in the country selected for the Federal Bureau of Investigations (FBI) internship program where he assisted in white-collar crime investigations.

He is a CFA charterholder and has earned a CPA (inactive), CMA (inactive) and CFM (inactive). He has numerous publications to his credit including the book, Decoding the Energy Enigma: Improved Decision-Making on This Generation’s Most Pressing Issue (2016).

Outside of work, he enjoys yoga (completed yoga teacher training), stand-up comedy (both watching and a retired performer) and training in mixed martial arts.

He can be reached at [email protected].

Disclaimer

This report is confidential and intended only for the person to whom it has been delivered and may not be published, distributed or reproduced for any purpose without prior written consent from M2X Capital LLC. 

References to specific securities and issuers are for illustrative purposes only and are not intended to be, and should not be interpreted as, a recommendation to purchase or sell such securities. The contents hereof should not be construed as investment, legal or other advice. Where information provided in this document contains forward-looking information including estimates, projections and subjective judgment and analysis, no representation is made as to the accuracy of such estimates or projections or that such projections will be realized.

The views and information expressed herein are solely those of M2X Capital LLC as of the date of this letter and are subject to change without notice. This is not an offer or solicitation for the purchase of interests in any investment. Past performance may not be indicative of future results.

Copyright 2019 by M2X Capital LLC. All rights reserved.