.png?width=6912&height=3456&name=Untitled%20design%20(1).png)
Please see the important disclosures appearing here. PAST PERFORMANCE IS NOT NECESSARILY INDICATIVE OF FUTURE RESULTS. THE POTENTIAL FOR PROFIT IS ACCOMPANIED BY THE RISK OF LOSS.
***
Barry Goodman is co-CEO and Executive Director of Trading for Millburn Ridgefield Corporation. Mr. Goodman is a 40-year veteran of quantitative investing and has spoken frequently on disruption and the ripple effects being felt across commodity, equity, currency and fixed income sectors (and why this is likely to continue). In this interview, Mr. Goodman shares insights on how the firm evolved from one of the earliest practitioners of traditional trend-following into an innovator in the application of AI and machine learning—and how that engine worked when applied to China. This interview was originally published by Resonanz Capital, an independent investment advisor based in Frankfurt.
What has changed in quantitative investing since the firm’s inception and what has remained the same?
[Barry Goodman]: The core advantages of systematic investment strategies have remained constant: these approaches allow trading at speed and scale; can manage risk in real-time; preserve institutional knowledge so that IP is never forgotten or can walk out the door; and get closer to the goal of being truly objective and avoid the biases that we all have as humans. This isn’t to say that discretion doesn’t have a role or that the keys should be fully handed over. But we as a firm have been unwavering in our belief that quantitative, systematic approaches make a lot of sense.
What has certainly changed is the complexity and interconnectedness of the markets, and the exponential increase in the amount of data available to analyze to try to understand what is driving movements in those markets. New tools of analysis had to evolve.
What are the keys to developing quantitative investment strategies with the passage of time?
[BG]: Quant models today need to be shark-like—in a state of perpetual forward movement. By that, I mean that systematic approaches need to account for the passage of time by being in a constant state of adaptation. This doesn’t mean a complete reinvention, nor does it mean aggressively chasing short-term changes in the market. It just means building in change as a core element of the quantitative process and making sure you never assume that the past will be exactly like the future. Whereas that might have worked 40 years ago, the world is simply too connected and efficient now.
What is the best way to explain a data-driven systematic strategy to institutional investors?
[BG]: We start with the philosophy—why data matters, how data can get at the truth about drivers of a market, and why sophisticated methods are probably the only way to cut through the noise to get a real “signal” in the data.
However, we like to move quite quickly to practical matters, and we have found that meaningful transparency is one of the best ways to get investors comfortable with our quantitative approaches. We are adamant about the protection of our intellectual property, but we have developed a suite of tools that give investors and prospects legitimate insight into how the strategy works, under what conditions it has historically done well versus faced headwinds, and how it has the chance to benefit a portfolio. We approach every institutional investor as a partner rather than a client.
How does a quantitative investment manager balance trend following and short-term trading?
[BG]: I can’t speak for other managers, but for Millburn, these concepts live on a spectrum, so the idea of balance as some sort of externally imposed view doesn’t really exist here. The models are contextual, sometimes behaving like short-term traders (for example, in a very inexpensive-to-trade market where short-term opportunities have a higher likelihood of overcoming transaction costs), but other times behaving like the slowest of trend-following strategies. But the point is that we, as researchers or Investment Committee members, don’t predetermine that behavior—we source and select what we think are relevant data sources, relying on our experience and economic intuition to do that, but the models themselves determine how best to interpret that data, using what we believe to be sound, tested methods. The whole idea is that there are opportunities to profit from either long- or short-term market moves and no reason to limit oneself to one or the other, or to artificially pre-determine a particular mix.
What are the challenges of commodities-focused quantitative investments?
[BG]: We’ve been trading commodity markets since 1971. Specific knowledge in these commodity markets can be particularly important to understand what inputs might play a role in driving behavior. This is especially true in today’s hyper-communicative and data-rich environment where understanding things like the interconnectedness of global commodity markets like WTI and Brent Crude Oil or Copper and Gold, or the idiosyncratic behavior of Natural Gas, can be critical.
As these markets combine true hedging and speculation and are subject to some or all of the vagaries of weather, supply and demand, technological innovation, and substitution effects, they can be additionally difficult to understand in simple terms. Combine this with commodity markets’ historical correlation to inflation, now globally at some of the highest levels in decades.
Finally, commodity markets have historically been volatile. The markets can seem opaque compared with traditional stock and bond markets. And while new technologies can potentially put downward pressure on prices, implementing these technologies or even bringing new supply to market can be slow, and markets can be susceptible to disruptive shocks in the meantime. The importance of political policy and incentives adds another variable to the mix. And post-COVID themes like near-shoring are a result of the desire for countries to more closely guard their supply of critical commodities to feed their populations or keep their economies running.
What next-generation evolutions are on the horizon for global commodities strategies?
[BG]: The transition from “brown-to-green” in terms of the move into more sustainable energy is going to introduce meaningful change to the ecosystem of global commodity investing. We’re beginning to see new dynamics play out between traditional commodity markets, but also the amplification of non-traditional markets and, in some cases, the introduction of brand-new markets. Take markets like lithium and cobalt, for example, which are critically important in the manufacturing of solar, wind, battery, and other elements of the green transition. Carbon emissions is another such market that has been around for a while but continues to evolve. And of course, you can look to the ongoing globalization of the Chinese commodity markets, where more than a dozen markets, including markets that have no true international proxy, like Iron Ore and Peanut Kernel futures, can now be traded by Western players. So, the next generation of commodity markets is already upon us. What’s clear to us is that these markets will take time to feather into the ecosystem, and as we go, we need to be nimble and evolve our approaches.
How has the investment opportunity in China evolved with the current interest rate and geopolitical environment?
[BG]: Unlike the US and many other Western markets, China has been in an era of moderately high interest rates since we started applying our strategies there back in 2013. So, from the perspective of the local Chinese market, rates have been fairly settled. Globally, of course, the situation is different. In general, though, we’ve seen that higher interest rates can mean a higher correlation between equities and fixed income, which translates to more volatile performance of a traditional “balanced” equity/bond portfolio. This has investors searching for true diversification, and we believe commodities—specifically the unique markets found in China—are a very legitimate option.
Is there a risk or opportunity in China-based investments institutional investors might be missing?
[BG]: Sometimes thinking about investing in China might seem overwhelming. Certainly, when we approached the idea back more than ten years ago the environment in China was complicated. The so-called “gray areas” stacked on top of one another to the point where things almost turned black. Today, while some things have become much clearer, geopolitics has introduced another layer of uncertainty.
Another challenge is truly understanding the markets and recognizing their uniqueness. We think managers who approach China with the idea of a simple “copy-paste” of existing approaches are doomed to fail. Our China machine learning models are trained on Chinese data, so they can learn the specific behaviors and drivers of those local markets, and through the regular process of retraining or re-learning can evolve as the characteristics of these markets change.
So, it is understandable when some investors choose to make the decision just to avoid China altogether. But as I’ve said earlier, these challenges bring opportunity. And as we’ve seen consistently over the years, regardless of the political or geopolitical environment, the Chinese government seems committed to opening these markets.
After more than fifty years of trading commodities and after more than a decade of trading in China, we feel like Millburn has a good process in place to address these challenges. We’ve demonstrated that opportunity exists. And we are expanding the opportunity set beyond futures markets. As the search for true diversification becomes even more important, to us the real risk is throwing up one’s hands and choosing not to invest.
How is AI impacting quantitative investing?
[BG]: Back when we began researching using machine learning we used to caution investors about recognizing when people are using the term as a buzzword versus as a core strategy. Today the buzzword is AI, with ChatGPT and similar large language models dominating the collective mindshare. ChatGPT’s innovation, aside from some effective marketing, is probably best characterized as a technical innovation of scale, with the ability to use more and more data points to train and leverage increasingly higher-powered computers.
For us, AI, or machine learning as a subset of AI, includes concepts that we’ve been using for years. The difference with using AI in investing is the immense amount of noise. This demands practical knowledge of how to set the parameters of your AI models and, we believe, a relatively skeptical view of just how much the machine should be trusted to do. In short—the recognition that forecasting returns in, say, Brent Crude Oil futures is not the same thing as forecasting what movie you might want to watch next on Netflix.
However, AI is an incredibly powerful tool for processing and understanding data. As a data-driven approach, quantitative investing can’t help but be impacted by it.
How do you see quantitative investment evolving over the next decade?
[BG]: Data will continue to increase and diversify, with more and more so-called “alternative data” being made available, like data that quantify real-time pollution from factories in Beijing to try to form an opinion on demand for steel rebar ahead of government reports. Statistical and AI algorithms will also advance, which, combined with more available data, will require advances in computing power. And in a virtuous cycle, more powerful computing will make possible further advances on the algorithmic side.
But nobody wants to see investment models that hallucinate like we’ve seen with early versions of ChatGPT and other large language model applications. Risk, common sense, and genuine market experience will also play an important role in the continued evolution of quantitative investing.