The Big Lebowski in Finance

Meet the New Boss. Same as the Old Boss

My Recent Podcast

My most recent guest on Brave New World was computational biologist Rich Bonneau, who is a professor of Computer Science and Biology at NYU. Rich inhabits the intersection of biology and machine learning, which is on fire these days. Rich is also co-founder of Prescient Design, which was acquired last year by Genentech, so he straddles the worlds of industry and academia.  That also means he’s got to make his research actually work, which is always exciting.

I had a great conversation with Rich about the role of machine learning in genomics and drug discovery. As pharma companies morph into technology companies, the old way, of trial and error in the messy analog world, is being replaced by machines and AI. So, tune into my conversation with Rich to learn more about the emerging brave new world of Pharma.

The Big Lebowski in Financial Markets

The Ides of March seems to have been particularly unkind to financial markets over the last few years. COVID drove the fear index of the market, the VIX, to a record high of 85 on March 16, 2020, as the S&P500 hit a low a week later. And here we are again, with the collapse of Silicon Valley Bank. As usual, the gurus are opining about the future, but none predicted it. Except for this article titled “Should You Sell Your Stocks in March And Go Away? Maybe” which was prescient, but just didn’t have the conviction to say “Yes.”

Why are all the gurus so poor at prediction? And why do people keep listening to them?

Being within blocks of Wall Street early in the early 80s, I became engaged in financial markets and investing. I loved the biotech pioneers that were emerging at the time, like Cetus and Genentech.  (Cetus invented the now widely used PCR test.) Investing was fun as long as you were winning. Then, on Black Monday in October 1987, I lost 30% of my meager savings. That was a painful lesson. Clearly I needed to learn more. I started tuning into shows like the Nightly Business Report, Louis Rukeyser’s Wall Street week, and other financial news shows. I read Barron’s and the Wall Street week diligently. I followed Initial Public Offerings.

I followed the advice of the various gurus of the time. I still didn’t do very well, but I kept trying.

It took me years of trial and error to realize that even the most knowledgeable people on Wall Street were highly inconsistent in terms of performance. Knowledge and smarts had little to do with predictive accuracy. Portfolio managers who had done well recently were paraded on Wall Street Week, but many of their recommendations were worthless. You were better off investing in the S&P 500 index, which represented the strongest companies in America.

In the mid-90s, a friend introduced me to Kevin Parker, a highly successful Morgan Stanley trader who had made a fortune shorting the Nikkei at its peak. He went on vacation immediately after that trade so he wouldn’t do something stupid, like trying to make another killer trade believing he was a true genius. His thinking was more along Clint Eastwood’s famous line, “a man has gotta know his limitations.”

In fact, Kevin was a big believer in technology, and that the future lay in data and machine learning. I met him over lunch in May 1994 when I was writing my book on machine learning methods. I told Kevin that I had been working on a “genetic algorithm” that could automatically discover “pockets of predictability,” or situations when  the probability of winning was high. That resonated with Kevin, who believed that markets were mostly quite efficient, but provided occasional opportunities during times of excessive panic or euphoria. That had been his Nikkei trade. He had seen the optimistic bubble develop in Japan and had the conviction to act. After the lunch, Kevin hired me to lead AI research at Morgan Stanley with an offer I couldn’t refuse.

That launched my trading career using machine learning for prediction. Data was becoming more easily available. I learned that a lot of professionals on Wall street had strong beliefs about markets. Some believed in fundamentals. Others believed in price patterns. Some believed in following analysts. Some believed in Fibonacci retracements, as if there was something magical about that series (years later I learned from a Math Fields Medal winner that the Fibonacci sequence was actually discovered in India in the 7^th century to count the combinations of beats and half-beats that could be packed into a specified number of beats in Indian classical music!)

I realized that rarely were these beliefs supported by real data or the application of the scientific method.  People somehow became authorities, some by force of personality, others by sheer luck, and people just believed them. At Morgan Stanley, a pundit made ten bold predictions every year, roughly half of which turned out to be wrong.

We now have new pundits, much like the old ones. It reminds me of the ending line of a song by The Who called “Won’t Get Fooled Again” that ends with “Meet the new boss. Same as the old boss.”

The painful lesson for me after ten years was that I couldn’t trust anyone with my money, including myself. My partnership with Kevin took me into the world of algorithms for investing, but over the last thirty years, I’ve applied machine learning algorithms to dozens of domains, including sports, health, real estate, insurance, supply-chains, and more. These experiences led me to propose a risk-based model for when we should trust AI systems with decision-making and when we shouldn’t, which I summarized in a Harvard Business Review article in 2016.

Humans vs Algorithms During Market Turmoil

And here we are again, with financial markets in turmoil. People are worried about contagion, that this could get a lot worse.

It could. But it is worth comparing the fear index, the VIX, with the last two major crises. During COVID, the VIX hit a high of 85. During the great financial crisis in 2008, the VIX similarly peaked in the high 80s. If you know some basic statistics, this level of the VIX implies that the market expects a one standard deviation daily move in the S&P500 of roughly 5%. That’s huge. In normal times, it is under 1 percent.

The VIX today is in the mid to high twenties. So, even as the media fans the flames to white heat, the data thus far are telling us not to over-react. But if we have over-reacted, the market currently presents opportunities for both algorithms and humans.

Interestingly, on the morning of the SVB failure, a student from my Systematic Investing class at NYU Stern messaged me that our recent session on “over-reactions” actually described the current state of financial markets very well. My course, which is based on my experience operating a machine-learning-based hedge fund on Wall Street, describes how to exploit mispricing opportunities systematically using algorithms. A challenge, for example, is how to recognize over-reactions. An algorithm can do that. In our class assignment, for example, we used a simple volatility-based formula to categorize over-reactions and implement a simple counter-trend system.

To make things interesting, my student sent me results from applying the strategy to JP Morgan. Her strategy had an Information Ratio (IR) of 2.7 from March 1993 until now. An IR is a risk-adjusted measure of performance, where the average return is divided by the risk taken to achieve it. An IR of 1 is very good. As a point of comparison, buying and holding JP Morgan during the same period would have had an IR of 0.4. For Bank of America, the same counter-trend strategy had an IR of 1.3, whereas buying and holding the stock had an IR of 0.25. The larger lesson here is that market turmoil often causes mispricing that may be quite far removed from the source of the turmoil.

The broader takeaway is that even a simple over-reaction algorithm like the one above that I used for my class assignment can work across the market because investors often over-react. The trick is to exploit the pocket of predictability that occurs from mispricing. But an algorithm can suffer in situations like the great financial crisis, when extreme price moves keep going in the same direction until the market finds a new equilibrium. That’s the next question: will there be significant contagion, or is the damage limited to SVB’s local ecosystem?

The current level of the VIX is telling us that it doesn’t expect contagion. If that turns out to be correct, it implies that there’s a lot of mispricing that’s occurred already. So, how should one exploit this mispricing without falling prey to all the opinions out there?

For this, I would suggest following a consistent process driven by data, that captures the larger picture that goes beyond the numbers. As my Stern colleague Aswath Damodaran revealed in our conversation on longer-term investing, an essential condition for successful investing is that the story match the numbers. This is the guiding principle for discretionary human investors at the moment: Identify the pockets in markets where the story is better than the numbers and where it is worse. That will reveal the opportunities. Then have the courage to act and wait for the returns to materialize. But don’t bet the farm.