The Intersection of Science and Finance with CFM's Jean-Philippe Bouchaud

28:52Machine learning, nobody really understands what's the magic. And I think it's a huge intellectual challenge, and we want to be part of that. How much of that is pattern recognition? Because when I think about the work, whenever I read about LLMs, it's really just statistically what makes the most sense for the next letter, the next word, next sentence. It's kind of hard to just think of crafting a document based on probabilities of the most likely word if you have these few words beginning. But apparently that's a big part of how they work or am I grossly oversimplified? Yeah. So why does it work?

29:34And can it work in finance too? So is it because the language or images have such a strong structure that there's an internal logic to language or to pictures or to other things that the model is able to capture and using these relatively simple ideas of statistical prediction of what's going to happen next is enough to generate meaningful sentences? But maybe, maybe there's part of that, the structure of the data. Is it the case in finance too? Maybe, maybe not. Literally exactly where I want to go. If you're training an LLM on billions and billions of documents, pages, books, whatever, and it now has a giant data source of when you get these first few words or these first few sentences, here's what's most common, here's what's second most common, And here's a reference check for you to say, how does this compare grammatically, structurally to the giant corpus we have?

30:39I can see the math behind that because there's only so many trillions of combinations of letters, words, sentences. But when you now apply it to markets, which seem to be so random tick to tick day to day, can you apply the same sort of logic to invest in? Well, there are two problems. One is fundamental. Are there structures that you can extract? And we believe that there are because otherwise we wouldn't be there. I mean, trend following is a structure. It's a pretty trivial one, but it is a structure. Now, many other types of structures in the data that we've extracted without using ML or using ML now or recovering with ML or even more complicated one with MLs.

31:29But the major difference between finance and languages or pictures is, one, the amount of data. Because in the end, stock markets have only existed since, I don't know, 1900 or 1800 if you want. Small data set. Small data set. Except if you go to high frequency, as I said, if you go tick by tick, all the book data, there's huge amounts of data. And there you can think that there's more to do.

32:01But what was I saying? Yeah, so there's the problem of the availability of data and the frequency at which you want to predict. So for high frequency, I think there's a lot of structure. For lower frequency, it's not clear yet that it is going to be used as a kind of technical model which only looks at prices without reading text. For reading text, we know that there's a lot of structure which corresponds to the structure of language. But having said everything you said, there's still something strange about LLMs or generative AI is that with this process of constructing sentences that are statistically valid, you can invent new things.

32:48And that's the thing that is really strange, right? I mean, you can learn pictures. For example, you know, this celebrity database where you make the machine learn these pictures. And then you ask the machine to generate new ones. And it does. And these are pictures that look exactly, I mean, that you look and you think it's a celebrity, but the celebrity doesn't exist. So there's something still weird about this that, as I said, nobody really understands. Really kind of interesting. And so continuing on that, what we are trying to do is to do the same thing with financial markets. So as I said, 100 years of data is not a lot, but maybe you can use these Gen AI models to generate a million years of fictitious financial markets.

33:37That's interesting. Very interesting. So let's talk a little bit about trend following and managed futures. It's had a few real standout years, in particular 2022, real challenging year. and Managed Futures were top of the asset quilt. What does that episode tell us about what strategies work, why they work? And the question I always find with trend following, why do so few investors tend to stay with them? They all seem to get nervous and tap out right before things. It's almost when you see people giving up, it's just about when the turn occurs. I agree. So what is it? First of all, why was 2022 such a standout year?

34:27I don't know. Just, I mean, aside from the fact that we had big Fed rate hikes and fixed income and equities, both got shellac double digits, kind of rare occurs the same year. I think you have to go about 40, 41 years to see both of them down significantly. How do you think about what environment leads to the best results for trend following? It's very difficult to say because otherwise we would have a meta model that arbitrages and increases the weight of trend following when it's going to work well. I think there probably is more research to do than we've been trying. We haven't found anything that's very convincing.

35:16But, you know, beginning of 2026 is also a very good period for trend following. Actually, since we wrote a paper in 2014 called 200 Years of Trend Following. And we were reporting on the fact that since 1800, if you paper trade a very simple trend following strategy, you make money every decade. With ups and downs, you know, there are years that are not so good. But as you say, I mean, what is striking and about the very point you made about people getting out of trend following just before it gets back on is I think it's ingrained in people's behavior to chase performance. So if performance has been bad for a few years, everybody declares.

36:00And that was the case in 2014. When we wrote our paper, trend following had been flat for the last five years. and people said, okay, well, trend following is dead now. And we were absolutely convinced that it was not the case. That trend following is such a strong behavioral bias that performance chasing is so ingrained in every one of us, even rational, we can't help. And so we bet and it was confirmed that trend following would come back. And since 2014, it's been very good actually overall. Well, you had a market that very much was trending mostly in one direction for, I mean, you have Q4 of 2018, and I think 2016 was so-so.

36:43But for the past 15 years, the bias has been pretty much in one direction. If you're on the right side of that, you should do pretty well. Yeah, but I'm not speaking about being long. I mean, I'm really speaking about, you know. Different asset classes and trends. You're long and short? Yeah, sure. So it doesn't matter. As long as the trend is in place, you want to participate in it. Up or down.

37:33a portfolio of futures that has like 150 futures, there's a very subtle correlation structure between all the assets that you have in your portfolio. So if you think about risk, you really have to think about how all these products interact with one another, talk to one another. And so it's not only a question of volatility that goes up and down that you have to control, but also a question of how these assets co-move together or anti-co-move together. But the way we think about upside risk is the same as the way we think about downside risk. It's just a question of risk. So you mentioned 150 different assets.

38:13I'm assuming some of these are commodities, oil, gold, stocks, bonds, interest rates. But what else is in the full list of 150? Well, there's different maturities, different countries, futures in China. I mean, if you count everything, it goes up to 100. I don't have the exact number, but in the 150s altogether. Has CFN been looking at prediction markets, things like Calci and Polymarket? No, we haven't. They're not liquid enough for us. Really? You need size and they can't provide it. Exactly. Really, really quite fascinating. Coming up, we continue our conversation with Jean-Philippe Bouchot, co-founder and chief scientist at Capital Fund Management, talking about how market structures are changing today.

39:05I'm Barry Ritholtz. You're listening to Masters in Business on Bloomberg Radio.

39:38We'll see you next time. and just 0-5 grams net carbs and 0 grams sugar. Hero Bread bakes with heart-healthy olive oil and delivers the soft, fluffy, flavorful experience you love. Breakfast burritos, schmear-loaded bagels, real mac and cheese. Hero Bread bakes loaves, bagels, and tortillas that don't taste or feel like cardboard. Noodles that don't fall apart in hearty sauces. Plus, limited-edition small batch bakes like the 2-grams net carb Hero Croissant or 1-gram net carb Hero Cheddar Biscuit, handmade in a Sonoma-based French bakery. Shop now on Hero.co. Use code IHEART for 10 % off. That's Hero.co.

40:13Per serving, not a low-calorie food. Some products contain Allulose. See nutrition info on Hero.co for sodium and sugar content. Every business has an ambition. PayPal Open is the platform designed to help you grow into yours. With access to business loans so you can expand and hundreds of millions of PayPal customers worldwide. Your customers can pay all the ways they want today with PayPal, Venmo, Pay Later, and all major cards. so you can focus on the future. When you need a partner trusted by millions, there's one platform for all business. PayPal Open. Grow today at paypalopen.com. Loan subject to approval in available locations.

40:49Discover a spectacular island destination with crystal blue seas, endless sunshine, and the cool Bahamian breeze. Bahamar, located in Nassau, Bahamas, offers your choice of three luxury hotels, the richly refined Rosewood, the playfully hip SLS, and the stylishly modern Grand Hyatt. With over 45 restaurants, bars, and lounges, Baja Mar serves up delicious dining from world-renowned chefs like Daniel Bouloud and Marcus Samuelson, nightlife venues like the new Jean-Baptiste Jazz Club, and the Caribbean's most luxurious casino. At Baja Mar, you'll find every pleasure under the sun and one-of-a-kind experiences for the entire family, like Baja Bay, our 15-acre lush tropical water park, interactive wildlife experiences, including our daily flamingo parade, world-class golf, tennis, spa, and so much more.

41:40Visit Bahamar.com today and discover a vacation destination where memories are made for a lifetime. Bahamar, life spectacular.

41:52I'm Barry Ritholtz. You're listening to Masters in Business on Bloomberg Radio. My extra special guest this week, Jean-Philippe Bouchot, Chief Scientist and Chairman at CFM, a quantitative hedge fund managing over$20 billion in assets. So let's just talk a little bit about risk management. I know that when you're dealing with leveraged or long-short or futures, there's a very robust thought process around risk management. Tell us a little bit about how you think about correlation and risk. Yeah, we have a disciplined and systematic approach, not only to alpha signals, to building prediction, but also to risk management.

42:38We have a pretty sophisticated tool to predict the volatility of tomorrow, the volatility of our portfolio tomorrow. And we're pretty good at that. So, of course, we know financial markets are difficult beasts. and even if you have the best model in the world, you can still have unexpected events that blow up your portfolio. That's something that we can't say will never happen. But in a way, if you don't want to take any risk, you shouldn't be in financial markets. You shouldn't be in that business. So we accept that there might be, I don't know, a completely unexpected event that breaks the whole financial markets everywhere in the world and everything is going to fail and there's nothing to do about that.

43:25So that can happen. But barring these extreme events, we think we're pretty good at predicting what's going to happen. And over the last 35 years of the existence of CFM, our anniversary is this year. We're celebrating our 35th anniversary in June in Paris. Very proud of that. so you know it kind of resisted these 35 years although we've become much better with time but having said that there's always an element that you have to be ready to intervene even if you're a quant shop and so on several occasions in the past 35 years we decided that our risk model couldn't know about things that we humans knew, like, you know, I don't know, the Brexit votes or and in these cases.

44:21Let's talk about that because this year, just the past 12 months, between the tariffs and Venezuela and now the ongoing war in Iran, how does global market volatility around all these geopolitical events, how does a quant shop deal with that? What I'm hearing is the humans have to do what humans do and sometimes override the machines sometimes yes i mean when when unexpected geopolitical events disrupt the world are models just not built to really work their way through that you're right some events are okay and like the war in iran for the moment is not you know something that our risk models are completely blind to.

45:11It doesn't mean that they've predicted at all. It just means that we're comfortable with the risk that our model have predicted, and they've adapted sufficiently fast to the events so that we're comfortable with the risk level, no human intervention. On the other hand, in some cases, it's completely unexpected, like tariffs and Liberation Day. This created havoc. Although, strangely enough, Liberation Day was announced. You know, everybody knew what was going to be said. And still, everybody was surprised. I don't think they knew the depth of it. I don't believe, despite I'm tariff man, it's the most beautiful word in the dictionary, I think the 100%, 150 % tariffs on specific countries later found to be completely unconstitutional.

45:57But at the time, I think people were genuinely shocked by this. and then a week later you know the little bit of a taco trade where let's just put a pin in this for 90 days again back to the volatility how do you deal i agree oil is trending upwards and then you have a tweet the world the war is over and then it resumes and then you have a tweet i think we've got a deal and then the other side says we're not even negotiating i don't recall a period in history where the President of the United States just constantly disrupted the normal flow of market activity. How disruptive is this to a quant model?

46:42Well, as I said, the two cases seem to be pretty different. Liberation Day was really a surprise and we had to manually intervene. There was something in our models that was completely blind to these things and we had to make a judgment call. I think the idea really is that humans should use their best judgment in these cases and decide whether it's reasonable that the risk model knows something about what's going on or not. In some cases, it does. In some cases, it doesn't. I think the tricky part is not to overreact. Because you said you don't remember periods of the world where things like this happened.

47:19But looking back, I've been in the markets for 35 years and everything, every year there seems to be something unexpected that happens. Just not every day. Not every day, but every year. Every day seems like a lot. Right. But in a way, every day means that it becomes a new normal. I guess. And so it's not that bad if it's every day. But really, this idea that this time is different is something that's strange. If you look at the world and the history of financial markets, it's really being normal that's not normal. And we've become used to that. So let's stay with the idea of modeling. You've been kind of skeptical of certain applications of deep learning in finance.

48:07There's overfitting. you know no one's ever seen a bad back test because they all seem to work perfectly in the past you have a lot of signal and noise issues what are some of the problems with models that you're focusing on improving yeah well for example this exactly what you just said can you have indicators that tell you whether your back test is overfitted or not and for for many years we We struggled with that and we used judgment again to say this is plausible, this is not plausible. We can believe that. We kind of replaced the trader that trades every day his signals or his beliefs to a higher level where we are traders of models.

48:59We kind of judge the models. We say this model is good enough to go in production. This model is not convincing enough. But it would be great to have something more systematic. And over the years, we've been struggling, and I think with some success, to have meta models that predict whether your backtest is really fudged or if it's decent enough to go in production. So we're kind of industrializing this process of selecting models that will go into production. Does that make sense? Yeah, no, that makes perfect sense. You're a quant. We've seen some issues with a lot of quant shops in the U.S. where crowding became a structural risk.

49:44You have all these systematic strategies and math is math. So essentially, you end up with a crowded trade. We had what people called the quant quake way back when. How do you think about that? How do you manage that problem when you're constructing portfolios? Sure. It's something that every day we think about this. We were in the quant quake in 2007. And actually, we were fortunate enough to be out of the markets or to have deleveraged already two weeks before the worst day of the quant quake. Was that an external signal or one of your model signals? It was something in the performance already in July.

50:29The quant quake, the really bad day happened maybe 9th of August. I don't remember exactly, but early August. Yeah, it was the dog days of summer, for sure. But starting like 10th of July already, there was something really very strange in our portfolio. And we started reflecting on what was going on and decided someone was deleveraging and hitting us by shorting our longs and buying our shorts. and this thought process of imagining that even if a fund that was like 10 % correlated with ours, not a lot, but 10%, and having every day a kind of systematic deleveraging policy, it would create exactly the kind of signals that we were seeing in our portfolio.

51:16So we thought, okay, this may be going to lead to a crash because people are going to suffer, suffer, and at one point they're going to… It cascades. Cascade and so on. And so that was the rationale for getting out. So in some cases, you're lucky enough to have strong enough signals that tell you that your standard risk model is wrong and you should do something else. That's really fascinating. So you mentioned almost 35 years of doing this. What do you think that – I'm going to say that again. So you are now almost 35 years into being a market quant. What's the most important thing about markets that the mainstream funds still get wrong?

52:07Well, I think it's this question of what is price doing? What are moving prices? and a lot of people still believe that there's something like a fundamental value and that the price is really moving because fundamentals are moving. Whereas we believe, and this touches very recent academic papers that Gabex and Coetion, two economists have put forward, they've called it the inelastic market hypothesis and we've contributed to that debate as well and the idea is really that markets are not driven by fundamentals or at least they are to some extent driven by fundamentals but this is a small long-term effect.

52:54On short run, short run meaning from one day to one year, which is pretty long already, it really flows that matter. That is people buying or selling stuff. Whatever the reason they buy or sell, it's going to move prices. It's not going to move prices on a short timescale and then disappear. It's really going to leave a trace in markets. And this is really a fundamental change of point of view that I think that is going to percolate and convince more and more people looking forward. But having this change of tack is really important because in one case, what you need to do to make money is to predict fundamentals.

53:36In the other case, you need to predict what people are going to do. And so, in a sense, crowding can be a good thing because if there's crowding, it's easier to predict what the crowd is going to do. And so, if whatever the reason people do things, they move prices and you're able to predict what people are going to do because you have behavioral models and structural models that tell you everything else being equal, people are more likely to do this and that, then you can build models. And I think that's the reason why we've been successful is this change of philosophy. We're not kind of anchored to fundamentals.

54:13We're anchored to flows. So this sounds a little bit like the Ben Graham line. I think it's Graham. In the short run, markets are voting machines. In the long run, they're weighing machines. Is that the balance between flows and fundamentals? Yeah, I think it's an old idea. I mean, you know, it's the Keynes also said things like that, you know. But in the long run, we're all dead, right? Keynes was saying that. So it's really a question of whether you're going to be solvable.

54:48I'm getting tired. I'll take that again. What's the word? Was that, by the way, Keynes or Graham? I initially thought it was Keynes, and then I checked myself. I'm not sure. But what Keynes said is that, what was his quote? Markets can remain irrational longer than you can remain solvent. So he never said that, but it's always attributed to him. Really? There's a great website called Quote Investigator. I see. And you give them a quote because people, as a sort of authority, you know, appeal to authority, they'll put somebody sophisticated as the source. Einstein never said compounding is the most powerful force in the universe, but they always attribute it.

55:37I didn't know that one. And you would be – I spend way too much time perusing quotes on the site, but you would be shocked. Who said markets are a voting machine in the short run but a weighing machine in the long run? I don't think it's Keynes, by the way. No, I don't remember if it was Keynes or Graham, but I thought it was one or the other. And it'll tell me, yeah, Benjamin Graham. But the first thing that came to mind was Keynes or Galbraith. They have so many favorite quotes from both. He said many relevant things even today. Yes, absolutely. But actually, we have models that predict exactly that, that on the short run, you can have trends and irrational behavior.

56:18And on the long run, it reverts back to fundamentals. But the long run, from our estimate, is like 5, 10 years. It's a very long time scale. I'm trying to remember. It might have been French, if I'm a French, who said you can't really tell if a manager is skilled until you have 20 years of data because it could just be good luck over 5 or 10 years, which is kind of fascinating. I want to stay with the efficient market or the inelastic market hypothesis. I'm curious as to your thoughts on EMH. I've always thought markets were kind of, sort of, eventually efficient, but not very efficient in the short run.

57:00What's your criticism of EMH? Well, it really depends what you mean by efficient. If you mean that they're very close to unpredictable, then you're right. But I think it's a very dumbed-down version of EMH. The question is whether prices reflect something from their mental that is in principle knowable, that reflects reality, or not at all. And one, I think, smoking gun of that is do you have long-term mean reversion? That is, can prices do random things in particular trending, which is really completely against EMH. On the short run, that is from a week to six months, markets are trending over six months, one year.

57:48And then on the longer timescale, they kind of hover around some long-term trend. And I think this is true. But this is really at odds with efficient market, which tells you that every day markets are around the correct price. Right. And there's no trend, no mean reversion ever. Which is, you know, that's not exactly what your day-to-day experience is if you're in the markets. It seems it's not magic. The collective votes of all market participants don't magically bring you to the correct, in quotes, price. But that's the assumption of efficient market. Or actually, not even an assumption. It's the argument that collectively, you know, if you have – that's the difference between having rational investors that all take decisions based on noisy observation but independent from one another.

58:48then because they're independent they realize the mean i mean some overpriced some underpriced and then it's a voting machine and the vote comes out right because there's enough investors and they're uncoilitated to one another but the problem of markets is that it's not the way it works the people are influenced by what other people are doing and what other people are saying So instead of having independent guys doing random stuff, they're kind of one guy who's doing only one thing, which is a fictitious body that aggregates everybody in the same way. Let's jump to our favorite questions that we ask all of our guests, starting with, tell us about your mentors who helped shape the direction of your career.

59:36That's an easy one. I have, you know, several mentors, but two of them are really close to my heart. One is Benoit Mandelbrot. Of course. You know, the fractal guy. I knew him personally. Oh, really? Yes. And he did a lot of things in physics as well. Sure. And so he influenced a lot. My wife, my wife was a physicist before turning a playwright now. and she works on fracture surfaces, the way when you break a material, what emerges from the fracture is a kind of very rough landscape that is fractal. And Mandelbrot had worked on that and there was a lot of interaction. Fractal at a molecular level or at a larger level?

1:00:21Well, fractal from the kind of very fine structure to macroscopic length scales. And so Mandelbrot also did his work on financial markets. And for me, it was really a revelation. It was something very influencing and out of the dogma of Brownian statistics and Gaussian phenomena and so on. And so it was also very close to what I was doing myself in physics. So, you know, it was clear that he influenced me enormously on that. didn't he write a book on market crashes and how there's a fractal nature within those the misbehavior of markets I'm actually quoted in that book oh get out, that's fascinating and then Pierre-Gilles Dejeune who was a Nobel Prize in physics a French physicist who was so fantastic and you know both these two and also Phil Anderson who was a Nobel Prize in physics as well in the US These three people, they convinced me that you shouldn't be stuck to your own field.

1:01:30You should broaden your scope. And what you learn from one field can be very useful in understanding another field. The three of them, they've really kind of hovered around and not got tied to their specific initial field. And I think this creates, well, at least for me, this de-inhibited me in the sense that I thought, okay, maybe I'm not legitimate to speak about finance because I'm a physicist. But, you know, it doesn't matter. If I have things that I strongly believe in, I should better say them and go to the end of them. So I think they were really influential in that way. Since we mentioned the misbehavior of markets, let's talk about some books.

1:02:16What are some of your favorites? What are you reading right now? Wow, so many. That's really a very broad question. So my last book is a book on John and Paul by Ian Leslie, John Lennon and Paul McCartney. It's a beautiful book. I really loved it. What's the name of it? John and Paul. Is it John and Paul, A Love Story? Am I remembering it correctly? I'm a big Beatles fan. And then that's in my queue. Oh, you should read it. Very emotional. I'm going to make a recommendation to you for a YouTube channel called You Can't Unhear This. They take apart Beatles songs in ways that just little things that were done in the recording process that in a million years you never would have noticed.

1:03:02And then once you hear it, you just can't unhear it. And if you're a Beatles fan, it's a rabbit hole. You'll love this. What else besides John and Paul? I'm reading something that I should have read for years, Mrs. Dalloway, Virginia Woolf. I'm a really great admirer of Virginia Woolf. Really, really interesting. Do you do much streaming, TV, podcasts, anything like that? Or we can skip over that. Podcasts a bit, but they're kind of French podcasts. So people are always looking for stuff. I'm going to ask that. So what are you streaming today? What sort of podcasts are you listening to? You know, in France, we are very fortunate.

1:03:45We have something called France Culture. It's a radio where there's an enormous – I mean, you could stay tuned all day if you wanted. There's so many interesting things going on about everything cultural literature, but also, you know, movies, politics. And so – We have NPR here. It's very similar. You can – it's a rabbit hole. You could fall down. Yeah. Yeah, okay, so life of major celebrities in culture, in cinema, theater, all these things. So I'm really a big fan of France Culture. And our final two questions. First, what sort of advice would you give to a recent college grad interested in a career in either quantitative investing or theoretical physics?

1:04:34Well, study theoretical physics. and study everything that's related to data. Pay attention to data and think about something that you strongly believe in and that you feel has not been investigated. And it doesn't matter if it's big or small. Make the effort of building something you strongly believe in. Really good answer. And our final question, what do you know about the world of investing today? They might have been useful 35 years or so ago when you were first getting started. That is extremely competitive, much more than we thought. Really? Yeah. Wow, that's really fascinating. Jean-Philippe, thank you so much for being so generous with your time.

1:05:19We have been speaking with Jean-Philippe Bouchot, CFM's co-founder, chairman, and chief scientist. If you enjoy this conversation, well, check out any of the 600 plus we've done over the past 12 years. You could find those at Apple Podcasts, Spotify, YouTube, Bloomberg, wherever you find your favorite podcasts. I would be remiss if I didn't thank the crack team that helps me put these conversations together each week. Meredith Frank is my video producer. Anna Luke is my podcast producer. Sean Russo is my head of research. I'm Barry Ritholtz. You've been listening to Masters in Business on Bloomberg Radio.

1:06:26from property and liability to cyber and regulatory challenges. At that level, managing risk becomes an ongoing discipline. At the Hartford, the focus is on helping businesses manage risk before it turns into something more disruptive. And when losses do happen, that work is paired with insurance coverage shaped by years of underwriting, risk engineering, and claims experience. Learn more at thehartford.com slash risk mitigation. Policies provided by Hartford Fire Insurance Company and its property and casualty affiliates, Hartford, Connecticut. Discover a spectacular island destination with crystal blue seas, endless sunshine, and the cool Bahamian breeze.

1:07:05Baha Mar, located in Nassau, Bahamas, offers your choice of three luxury hotels, over 45 fine dining and nightlife venues, John Batiste's all-new jazz club, the Caribbean's most luxurious casino, and one-of-a-kind experiences for the entire family, like our 15-acre tropical water park, wildlife sanctuary, world-class golf course, and so much more. Visit Bahamar.com today. Go on, get a little out there, into the big heart of Nevada, where you can go off-road and off the map, on two lakes or on horseback. Dip into hot springs and dive into deserts. Climb a mountain or make your best effort. See thousands of stars in some of the darkest skies.

1:07:46Stake out haunted hotels. Can you make it to sunrise? There's always something new to see because we've got plenty of space to just be. Plan your trip at TravelNevada.com.