Latest / Masters in Business / The Intersection of Science and Finance with CFM's Jean-Philippe Bouchaud

Transcript

The Intersection of Science and Finance with CFM's Jean-Philippe Bouchaud · Masters in Business

0:02
Speaker 1: Bloomberg Audio Studios, Podcasts, radio news. This is Masters in
0:12
Speaker 1: Business with Barry Ritholt on Bloomberg Radio.
0:16
Speaker 2: This week on the podcast, yet another extra special guest.
0:21
Speaker 2: John Philip Buschow is chief Scientist, head of research, chairman
0:25
Speaker 2: and co founder at CFM. They're a quantitative trend following
0:31
Speaker 2: Hedge Fund. They run over twenty billion dollars in client money.
0:35
Speaker 2: They've been around for almost thirty five years, put together
0:38
Speaker 2: a very impressive track record. They also run a number
0:42
Speaker 2: of interesting academic research labs and things like that. Jean
0:48
Speaker 2: Philippe has published something like three hundred plus academic papers.
0:53
Speaker 2: They are deep into all the things that drive markets
0:56
Speaker 2: from a quantitative perspective. I thought this conversation was fascinating
1:00
Speaker 2: and I think you will also with no further ado,
1:04
Speaker 2: My interview of CFMS Jean Philippe the show. So what
1:09
Speaker 2: do people call you? JP? John Phillip Philippe, What what
1:13
Speaker 2: do you?
1:13
Speaker 3: Philippe? In France JP and Anglo Saxon Countries JP.
1:17
Speaker 2: All right, it seems a little informal, but I'll go
1:20
Speaker 2: with JP. So, so JP, let's start with your background
1:25
Speaker 2: PhD and theoretical physics from NS. You spent some years
1:29
Speaker 2: at very prestigious research in the institutions I mentioned Cavendish Labs.
1:35
Speaker 2: What was the original career plan.
1:38
Speaker 3: Yeah, I was planning to be a physicist, but then,
1:40
Speaker 3: you know, studying statistical physics, and we can go into
1:43
Speaker 3: that later if you wish. I realized that physics can
1:47
Speaker 3: offer much more than studying physics, and and then I,
1:51
Speaker 3: you know, was always fascinating, fascinated by numbers. I've always
1:57
Speaker 3: liked statistics and financial markets spit statistics every day, and
2:03
Speaker 3: I thought, you know, this is a very interesting complex system.
2:07
Speaker 3: There are crisis, crashes, jumps, this system seems to be
2:12
Speaker 3: driven by its own dynamics. Physicists have to do something
2:16
Speaker 3: about this, and so.
2:17
Speaker 2: The sounds very similar to chaos theory.
2:20
Speaker 3: Yeah, exactly, mean that was the high days of chaos theory.
2:23
Speaker 2: So I pulled some phrases from some of your papers.
2:29
Speaker 2: One was titled, and I'm going to mangle this disordered
2:32
Speaker 2: systems and complex phenomena which can be either physics or finance.
2:37
Speaker 2: That sounds like, but what are the dynamics of glassy
2:40
Speaker 2: systems and granular media?
2:42
Speaker 4: That sounds fascinating, Yeah, but you know it's all the
2:49
Speaker 4: problem is how to interacting elements give rise to something surprising.
2:55
Speaker 3: You know. Granuar matter is grains that you know, interract
3:00
Speaker 3: with one another, and then you have these strange phenomena
3:02
Speaker 3: called avalanches where you drop a grain on the slope
3:06
Speaker 3: and most of the time nothing happens, but sometimes there's
3:09
Speaker 3: a big landslide that takes all the grains down. And
3:13
Speaker 3: so this again is very reminiscent of financial markets, right,
3:18
Speaker 3: I mean, many things happen, nothing much follows, and then
3:22
Speaker 3: sometimes there's a crash, right, And so this was really
3:25
Speaker 3: intriguing for physicists like me.
3:27
Speaker 2: So I'm as you're talking, I'm just thinking of a
3:30
Speaker 2: concept in physics that really applies to markets, the three
3:35
Speaker 2: body problem. When you have those three gravitational masses interacting
3:42
Speaker 2: with each other, it's fairly unpredictable, which kind of seems
3:47
Speaker 2: like markets themselves.
3:49
Speaker 3: Yeah, I mean, there's there are two ways to be unpredictable.
3:52
Speaker 3: One is that the system is by itself unpredictable, that
3:55
Speaker 3: even with deterministics laws like the three body problem, you
3:59
Speaker 3: can't see much after a few seconds, days or weeks.
4:04
Speaker 3: But there are other kinds of unpredictability. When there's a
4:07
Speaker 3: true source of exognous noise that hits the system and
4:12
Speaker 3: you can't say anything. So that's the traditional way economists
4:15
Speaker 3: think about markets. They are kind of buffeted by things
4:18
Speaker 3: you can't predict because they come from outside. And then
4:21
Speaker 3: I think the physics hunch is that, hey, but there
4:24
Speaker 3: can be self generated shocks, self generated randomness that come
4:30
Speaker 3: from large assemblies of individuals, i e. Traders, you know,
4:36
Speaker 3: agents that trade and buy and sell to each other,
4:39
Speaker 3: and this can generate you know, intrinsic randomness that is
4:45
Speaker 3: not of the same kind as the three body problem,
4:48
Speaker 3: but really comes from the interaction of a huge number
4:51
Speaker 3: of elements.
4:53
Speaker 2: So I see the parallels between theoretical physics and finance.
4:59
Speaker 2: What I led you to begin shifting in the early
5:02
Speaker 2: nineties from studying theoretical physics to becoming fascinated by market
5:10
Speaker 2: microstructure and econophysics.
5:13
Speaker 3: Yeah, so, as I said initially, I've always been excited
5:17
Speaker 3: by data and trying to make sense of data. So
5:20
Speaker 3: there was something there anyway, But what really drove the
5:25
Speaker 3: transition was, in a sense, the nineteen eighty seven crash
5:30
Speaker 3: and the Black Shoal theory. So I didn't know anything
5:33
Speaker 3: about that, and then I wrote a paper on what
5:37
Speaker 3: I was working on, physics systems with large jumps, if
5:42
Speaker 3: you want large crashes that happened from time to time.
5:47
Speaker 3: And someone who was working in the banking industry called
5:51
Speaker 3: me and said, hey, it's really interesting because it resembles
5:54
Speaker 3: what happens in finance, and in particular what just happened
5:57
Speaker 3: the nineteen eighty seven crash. And there's this theory, the
6:00
Speaker 3: black steals theory, that is a theory that only works
6:04
Speaker 3: in a world where there are no crashes, where you know,
6:08
Speaker 3: all the motions are small and predict they're random, but
6:12
Speaker 3: they're kind of predictable even if they're random in some
6:15
Speaker 3: strange way. And I thought this is really weird. And
6:18
Speaker 3: this guy said, why don't you try to generalize black
6:21
Speaker 3: seals to a world where where there are crashes? And
6:24
Speaker 3: I thought, well, that's that's really interesting. So I read
6:27
Speaker 3: black Shoals and I thought it can't be right. They
6:29
Speaker 3: must be wrong, these guys. So I kind of redid
6:32
Speaker 3: everything myself and found something that looked more interesting than
6:36
Speaker 3: baccheors because it could be extended to non Gaussian statistics,
6:40
Speaker 3: as they are called non normal distribution Bell curves and
6:44
Speaker 3: so on, and so it looked to me interesting and
6:47
Speaker 3: I thought, okay, maybe we can do software out of
6:51
Speaker 3: that and commercialize it. And so I went and knocked
6:55
Speaker 3: on several doors and suddenly the door of Jean Pillard opened,
6:59
Speaker 3: and was someone who had founded actually CFM in ninety one,
7:05
Speaker 3: that was ninety four. And I started, you know, explaining
7:09
Speaker 3: what I had been doing and that I was interested
7:12
Speaker 3: in transferring ideas from physics to finance, and he said,
7:15
Speaker 3: why don't we create something together, and so at the
7:18
Speaker 3: time we created a company called Science and Finance and
7:22
Speaker 3: this was done, you know, in in two weeks. It
7:25
Speaker 3: was like amazing the way we met, and there was
7:30
Speaker 3: a fluid that was flowing between us immediately. And so
7:36
Speaker 3: CFM then merged with Science and Finance in nineteen ninety
7:40
Speaker 3: so it's now the same firm. But the idea he
7:43
Speaker 3: had at the time, he had this this small CTA
7:48
Speaker 3: trading firm, and he thought, I need to beef up
7:50
Speaker 3: my research team, and this guy seems to be interesting.
7:53
Speaker 3: So we just partnered and that's how it all started.
7:57
Speaker 2: And that CTA firm specialized in manage future Yeah, exactly much.
8:02
Speaker 2: Now I know most of the future straders. I know,
8:05
Speaker 2: they all seem to be trend followers. How do you
8:08
Speaker 2: think about applying quantitative research and theoretical physics to dealing
8:14
Speaker 2: with futures?
8:15
Speaker 3: Yeah, well that that was exactly Jean Pierre Haguilar's idea.
8:19
Speaker 3: He said, Okay, I'm doing trend following is good, but
8:22
Speaker 3: it's not you know, rocket science. Maybe we can do
8:25
Speaker 3: much better. And so he said, why don't we work
8:28
Speaker 3: on something more beefy than just trend following, And so
8:33
Speaker 3: you know that that started the whole thing. And the
8:35
Speaker 3: idea of the main idea is data. You know, physicists
8:39
Speaker 3: are good at looking at data and extracting structures, you know,
8:42
Speaker 3: looking at data and imagining that from that data you
8:46
Speaker 3: can build theories. You can identify what's important and what's not.
8:51
Speaker 3: And that's I think the way you know it all
8:54
Speaker 3: works in physics that you you you scrutinize data and
8:58
Speaker 3: then there's a flash and you think I can model that.
9:01
Speaker 3: And it's really the same process in finance, at least
9:04
Speaker 3: as far as we wear concern, and we are concerned now,
9:08
Speaker 3: it hasn't changed. It's the same process. Huh.
9:11
Speaker 2: Really really quite fascinating you keep your professorships at NS
9:16
Speaker 2: and you've maintained a foot in academia even as you're
9:20
Speaker 2: building and running an asset management firm. Tell us about that,
9:23
Speaker 2: you're still publishing papers. What keeps you interested in the
9:27
Speaker 2: academic side of finance.
9:29
Speaker 3: Well, first of all, it is me. You know, I
9:32
Speaker 3: feel I'm a researcher's researcher at heart, and I need
9:36
Speaker 3: to continue. It's like you know, people running the marathon.
9:40
Speaker 3: They are doing something else in life and then there's
9:43
Speaker 3: an urge to run the marathon. For me, there's an
9:45
Speaker 3: urge to understand what I'm doing, and understand also things
9:50
Speaker 3: that I'm not doing even now, you know, even physics problems,
9:54
Speaker 3: I can get excited about them, or trying new things
9:57
Speaker 3: like the mL revolution and how does mL work? Why
10:03
Speaker 3: do large language models work so well? Learn so well?
10:08
Speaker 3: I think it's fascinating. I want to understand. But there's
10:11
Speaker 3: another reason for doing this is that to attract talents,
10:16
Speaker 3: you need to identify them, you need to attract them,
10:20
Speaker 3: you need to be their professor at one point. And
10:23
Speaker 3: I think a lot of the success of CFM has
10:26
Speaker 3: been attracting talents, and I think part of that only
10:31
Speaker 3: part of that. Of course, it's a teamwork is due
10:33
Speaker 3: to the fact that I'm still very connected in academic
10:36
Speaker 3: circles and young students. They've listened to me giving talks, lecturing,
10:42
Speaker 3: They've read my papers, and so they feel let's go
10:47
Speaker 3: and work for that firm because it seems that they're
10:49
Speaker 3: really doing cool stuff.
10:50
Speaker 2: So is that the thinking behind establishing the research division
10:54
Speaker 2: at CFM. I mean, you run that as a full
10:57
Speaker 2: academic research department as oppose to you know, a lot
11:01
Speaker 2: of asset management shops. They have a couple of cfps
11:06
Speaker 2: and MBAs and their CFAs and they're working on their
11:10
Speaker 2: quantitative models. You guys seem like you've taken it to
11:13
Speaker 2: a whole different level.
11:14
Speaker 5: Yeah.
11:15
Speaker 3: I mean most, maybe even all, our researchers have a PhD.
11:21
Speaker 3: It doesn't mean that we're an academic lab. We're really
11:24
Speaker 3: working on concrete stuff. We're really there to, you know,
11:27
Speaker 3: make models that work, build portfolios that are robust, model risk,
11:32
Speaker 3: model execution, control costs. All these things are bread and
11:36
Speaker 3: butter for everyday work. But at the same time, we
11:40
Speaker 3: feel that when we find something that is beyond the
11:43
Speaker 3: kind of daily work and that can be published because
11:46
Speaker 3: it brings something to the academic debate or to do
11:49
Speaker 3: public debate. You know, why how do markets work? Why
11:53
Speaker 3: are there crashes? Are market sufficients? What about the economy?
11:58
Speaker 3: Do people understand invation? We need new theories to understand inflation,
12:03
Speaker 3: monetary policy and all these things. We believe it's our role.
12:08
Speaker 3: Also because we have access to so much data, and
12:11
Speaker 3: we're privileged, you know, academics they don't have access to
12:14
Speaker 3: so much data, and so we have to give back
12:17
Speaker 3: in a way. And the reason we're doing this, as
12:20
Speaker 3: I said, it's not only because we're driven to do that,
12:23
Speaker 3: but also because it creates an atmosphere where people are
12:28
Speaker 3: happy to work at CFM. I hope you know, I
12:31
Speaker 3: don't want to put words in their mouth.
12:34
Speaker 2: Well, you guys, open up a big, expanded, a big
12:37
Speaker 2: New York officer. You don't seem to be having much
12:39
Speaker 2: difficulty recruiting people there. What's the head count there? Now?
12:43
Speaker 3: We have one hundred and fifteen researchers and fifteen percent
12:47
Speaker 3: of them are in New York.
12:50
Speaker 2: What motivated expanding the New York office as much as
12:54
Speaker 2: you have?
12:54
Speaker 3: Well, first of all, a lot of our investors are
12:57
Speaker 3: in the US, so we need to be there interact
13:00
Speaker 3: with them, and we need to have a presence, if
13:05
Speaker 3: only for investor relation, but also because there's a lot
13:09
Speaker 3: of talents in the US that we want to grab
13:12
Speaker 3: and attract. There's a lot of data, a lot of brokers,
13:15
Speaker 3: so it makes a lot of sense. So we've been
13:17
Speaker 3: in New York for twenty years, and you know, it's
13:22
Speaker 3: obvious that you know it is a hub and we
13:25
Speaker 3: should expand there.
13:26
Speaker 2: So you mentioned earlier your co founder Jean Pierre Aguilar.
13:31
Speaker 2: He passed away in two thousand and nine. What was
13:34
Speaker 2: the impact on the firm. How did you guys manage
13:38
Speaker 2: around That's a big loss when you lose a founder.
13:41
Speaker 3: Yeah, it was a tragedy because he died in a
13:43
Speaker 3: glider accident. You know, we knew that he was gliding.
13:46
Speaker 3: We knew that gliding was dangerous, but in a sense,
13:51
Speaker 3: and it really means bad risk management, right, we never
13:55
Speaker 3: thought that he could crash. It never occurred to us,
13:59
Speaker 3: which was straight because these things happen, and so it
14:03
Speaker 3: was tragic because we were not prepared. And it was
14:06
Speaker 3: tragic because he was not only a friend, but he
14:09
Speaker 3: was you know, the public figure of CFM. He was
14:13
Speaker 3: not involved in constructing models. I mean quants in a way.
14:16
Speaker 3: What's great about quant investing is that you don't need
14:19
Speaker 3: star traders, you don't need you know, pms that know everything.
14:24
Speaker 3: It's a collective effort and so when someone disappears or
14:28
Speaker 3: resigns or dies, it's not a tragedy. But in the
14:31
Speaker 3: case of Jean Pierre, it was even that he was
14:33
Speaker 3: not really involved in the construction of models. He was
14:37
Speaker 3: just very inspiring, generous, and he was really great. Was
14:43
Speaker 3: he had a vision, you know when we met and
14:45
Speaker 3: he thought, Okay, with that guy, we can build something,
14:48
Speaker 3: build something great. It's amazing, you know, to think that
14:53
Speaker 3: he was so enthusiastic about creating what we created together,
14:57
Speaker 3: and so we owe him a lot. When he passed away,
15:01
Speaker 3: it was really difficult to well, you know, there were
15:05
Speaker 3: several issues. One is that he had fifty seven percent
15:09
Speaker 3: of the company, so we had to negotiate with the
15:13
Speaker 3: estate to get back control. That was pretty difficult, but
15:18
Speaker 3: we went through that. And also we needed to reassure
15:21
Speaker 3: our investors. You know, Jean Pierre was he seemed to
15:24
Speaker 3: be the public figure. He was a public figure, and
15:27
Speaker 3: it seemed to be the inspired inspiration behind everything, and
15:30
Speaker 3: so we had to we you know, communicate that we
15:33
Speaker 3: were at the helm and that we would navigate that
15:37
Speaker 3: and it worked and so it was it was very stressful,
15:42
Speaker 3: but it was very rewarding as well to go through
15:45
Speaker 3: that and.
15:46
Speaker 2: The firm carries on as his legacy.
15:49
Speaker 3: Yeah.
15:50
Speaker 2: Coming up, we continue our conversation with John Philip Schau,
15:54
Speaker 2: head of Research and Chief Scientists at CFM, talking about
15:58
Speaker 2: the growth of Capital Fund Management.
16:01
Speaker 6: I'm Barry Redults.
16:02
Speaker 2: You're listening to Masters in Business on Bloomberg Radio. I'm
16:16
Speaker 2: Barry Ridults. You're listening to Masters in Business on Bloomberg Radio.
16:21
Speaker 2: My extra special guest this week is Jean Philip the show.
16:24
Speaker 2: He is the head of Research, chairman and chief scientist
16:28
Speaker 2: at Capital Fund Management Hedge fund managing over twenty billion dollars,
16:33
Speaker 2: a quantitative shop specializing in managed futures and other quant
16:39
Speaker 2: type funds. So let's talk a little bit about the
16:45
Speaker 2: building of the fund. You meet you cofind found Science
16:49
Speaker 2: and Finance in nineteen ninety four with Jean Pierre Aguilar.
16:55
Speaker 2: What was the thought process did you think you were
16:58
Speaker 2: building a quant fund research shop? What was the original plan.
17:03
Speaker 6: A quand fund From day one.
17:05
Speaker 3: From day one a quand fund. But from day one
17:08
Speaker 3: we knew that we wanted to be strongly associated with academia.
17:13
Speaker 3: We knew that the only way to innovate, you know,
17:16
Speaker 3: again coming back to the fact that financial markets are
17:19
Speaker 3: complex systems, it's really difficult to beat the market. We
17:23
Speaker 3: know that everybody's trying to beat the market. If we
17:26
Speaker 3: want to have something else to say and not follow
17:29
Speaker 3: the crowd, we have to innovate, and innovating is hard.
17:32
Speaker 3: You have to spend time, you have to have new
17:35
Speaker 3: ideas that nobody else has, and so this means investing
17:39
Speaker 3: heavily in research. So the two are not contradictory. We
17:42
Speaker 3: really wanted to be a con fund. We had already.
17:46
Speaker 3: We knew already about Renaissance. We knew that these guys
17:49
Speaker 3: at Renaissance they were very close in spirit and in
17:52
Speaker 3: culture to what we were, and so we thought we
17:55
Speaker 3: were going to try to emulate them. Of course, they're
17:59
Speaker 3: so great that there's no way to emulate them, but anyway,
18:02
Speaker 3: this was the aim.
18:03
Speaker 2: They had a forty year head start on you guys.
18:06
Speaker 2: So yeah, But so it's funny you mentioned renaissance technologies.
18:10
Speaker 2: I think of when I'm doing my research for for
18:14
Speaker 2: CFM kind of reminded of D. E. Shaw and AQR
18:19
Speaker 2: and a few other little bit of Millennium, although they
18:22
Speaker 2: do so much of everything. The thought process behind being
18:29
Speaker 2: a QUND shop when there's so many other QUND shops.
18:32
Speaker 2: If we don't create our own models, if we don't
18:34
Speaker 2: create our own findings and innovations, we're just an also run.
18:38
Speaker 2: Is that ran? That is that the thought process behind it,
18:42
Speaker 2: like we have we have to do this otherwise because
18:45
Speaker 2: it's all of these other cound shop that I've mentioned,
18:50
Speaker 2: none of them are quite the academic lab that.
18:53
Speaker 3: They AQR is. I think that closest to us in
18:58
Speaker 3: that front. I think, you know, in the songs they
19:00
Speaker 3: took the initially the completely different term. They thought we
19:04
Speaker 3: have to be completely secretive about everything and be a
19:08
Speaker 3: kind of black hole where everything goes in but nothing
19:11
Speaker 3: goes out. And that was not our philosophy. We thought that,
19:15
Speaker 3: you know, life is too short as well. This is
19:17
Speaker 3: not only we want to make money for ourselves, for
19:20
Speaker 3: our investors. We want to excel, but not at any costs.
19:24
Speaker 3: We think that there's something else in life, that there's
19:27
Speaker 3: a legacy that we want to leave. And this legacy
19:30
Speaker 3: is intellectual as.
19:31
Speaker 2: Well, pursuing the truth as to what drives markets and
19:35
Speaker 2: what leads to alpha.
19:36
Speaker 6: And returns exactly.
19:38
Speaker 2: You know, every new discovery of alpha eventually gets arbitraged away.
19:43
Speaker 2: Is that is that the thinking?
19:44
Speaker 3: Yeah, not exactly, I mean transfollowing. You know, it's not
19:47
Speaker 3: arbitrage the way at all. And actually, if you think
19:50
Speaker 3: about it, it's very hard to arbitrage trend following. You know,
19:54
Speaker 3: if if people trend follow is going to lead to
19:58
Speaker 3: more trendfor.
19:59
Speaker 2: Right, it's not only a farm of French factor, but
20:02
Speaker 2: it takes on its own life.
20:04
Speaker 3: Right, Well, Fama doesn't like momentum but anyway.
20:06
Speaker 2: But isn't it part it's not part of the original
20:09
Speaker 2: three factor model, but wasn't it one of the later models?
20:14
Speaker 3: Reluctantly, I think he had to add it. But it's
20:17
Speaker 3: a disgrace for efficient market theory. So he doesn't like
20:20
Speaker 3: momentum at all.
20:22
Speaker 2: Listen, you know, if the math is there, it doesn't
20:25
Speaker 2: matter if you like it, if it works, if it's
20:28
Speaker 2: a valid factor, it's a valid factor.
20:30
Speaker 3: Agree, I agree, But you know that's again the physicists
20:34
Speaker 3: point of view. Experiments is about everything else. But sometimes
20:40
Speaker 3: when you talk to economists, they have a strange view
20:43
Speaker 3: that theorems and axioms are supersede any empirical observation. I
20:48
Speaker 3: was told that by an economist, and so there's a
20:52
Speaker 3: very strong difference in perception.
20:54
Speaker 2: I recently had Richard Thaylor and Alex Amus in the studio,
21:00
Speaker 2: and I was shocked to learn from them they still
21:04
Speaker 2: aren't teaching behavioral finance in economics course at a college level,
21:10
Speaker 2: which is kind of shocking. I think everything we've learned.
21:13
Speaker 2: So let's talk about another technology. There have been over
21:17
Speaker 2: the past decade, but especially the past few years, huge
21:21
Speaker 2: advances and artificial intelligence and machine learning, to say nothing
21:25
Speaker 2: about large language models. How are you guys thinking about
21:29
Speaker 2: real world investing driven by AI and what sort of
21:33
Speaker 2: opportunities does this open up?
21:35
Speaker 3: Well, you know, AI is really an advanced form of
21:39
Speaker 3: data analysis, and in a way, we've been doing machine
21:43
Speaker 3: learning forever. The thing is that techniques have evolved there.
21:48
Speaker 3: It's now much more efficient. There many more things that
21:52
Speaker 3: one can do, in particular reading text. For many years
21:55
Speaker 3: we were just using numbers, and actually for many years
21:58
Speaker 3: we were just using prices and volume and not anything else,
22:02
Speaker 3: and well and fundamental information about companies. But now there's
22:07
Speaker 3: so much data that you can use. There's new data
22:10
Speaker 3: set every day that we're presented by the data vendors,
22:13
Speaker 3: and so there's a need to handle that data to
22:16
Speaker 3: read sometimes huge data files. For example, if you think
22:19
Speaker 3: about microstructure high frequency data, you know there's events happening
22:26
Speaker 3: in the other book of major exchanges at the millisecond
22:30
Speaker 3: level or even faster. This generates a huge amount of
22:33
Speaker 3: information that has to be dealt with, analyzed, and machine
22:37
Speaker 3: learning helps you very much doing that, reading texts that
22:42
Speaker 3: no human would be able to read and extracting information
22:46
Speaker 3: statistical information from that text. So for us, it is
22:50
Speaker 3: I wouldn't say a revolution, but it's an acceleration of
22:54
Speaker 3: things that we were trying to do before, and obviously
22:57
Speaker 3: we're much in tune with that. We've actually created to
23:00
Speaker 3: the nomil ab at CFM to help transferring technology from
23:05
Speaker 3: what mL people are constructing and what researchers at CFM
23:09
Speaker 3: may be using, but also to try to understand how
23:13
Speaker 3: these things work right, because we're very uncomfortable with the
23:16
Speaker 3: idea of black boxes. Black box is something that can
23:21
Speaker 3: improve the research process, but when you think about implementing
23:26
Speaker 3: that in production and having models trading with these models,
23:30
Speaker 3: you really want to be sure that the machine has
23:32
Speaker 3: done something that makes sense, and so understanding what machine
23:37
Speaker 3: learning is actually doing? Why are these things working? To
23:41
Speaker 3: start with? You know, what is strange is that it
23:44
Speaker 3: works so well, but nobody understands why. When you're driving
23:48
Speaker 3: a car, the car works really well, but we know
23:50
Speaker 3: exactly why it works how it works machine learning, nobody
23:55
Speaker 3: really understands what's the magic, and I think it's a
23:59
Speaker 3: huge intellect challenge and we want to be part of that.
24:03
Speaker 2: How much of that is pattern recognition? Because when I
24:06
Speaker 2: think about the work, whenever I read about LMS, it's
24:10
Speaker 2: really just statistically what makes the most sense for the
24:14
Speaker 2: next letter, the next word, the next sentence. It's kind
24:17
Speaker 2: of hard to just think of crafting a document based
24:21
Speaker 2: on probabilities of the most likely word if you have
24:26
Speaker 2: these few words beginning. But apparently that's a big part
24:30
Speaker 2: of how they work. Or am I grossly overse.
24:33
Speaker 3: So why does it work? And can it work in
24:36
Speaker 3: finance too? You know? So is it because the language
24:40
Speaker 3: or images have such a strong structure that you know
24:45
Speaker 3: there's an internal logic to language, or to pictures, or
24:50
Speaker 3: to other things that the model is able to capture.
24:53
Speaker 3: And using these breathively simple ideas of statistical prediction of
24:59
Speaker 3: what's going to happen next is enough to generate meaningful sentences.
25:04
Speaker 3: But maybe maybe this part of that the structure of
25:08
Speaker 3: the of the data. Is it the case in finance too, well,
25:12
Speaker 3: he may.
25:12
Speaker 2: Be literally exactly where to go. If you're training an
25:15
Speaker 2: LLM on billions and billions of documents, pages, books, whatever,
25:19
Speaker 2: and it now has a giant data source of when
25:22
Speaker 2: you get these first few words or these first few sentences,
25:26
Speaker 2: here's what's most common, here's what's second most common. And
25:30
Speaker 2: here's a reference check for you to say, how does
25:33
Speaker 2: this compare grammatically structurally to the giant corpus we have.
25:40
Speaker 2: I can see the math behind that, because there's only
25:44
Speaker 2: so many trillions of combinations of letters where it sentences.
25:48
Speaker 2: But when you now apply it to markets, which seem
25:52
Speaker 2: to be so random tick to tick day to day,
25:57
Speaker 2: can you apply the same sort of logic to to
26:00
Speaker 2: invest there?
26:01
Speaker 3: Well, that are two problems. One is fundamental, are there
26:07
Speaker 3: structures that you can extract? And we believe that there are,
26:12
Speaker 3: because otherwise we wouldn't be there. I mean, you know,
26:14
Speaker 3: trend following is a structure. It's a pretty trivial one,
26:17
Speaker 3: but it is a structure. Now, many other types of
26:20
Speaker 3: structures in the data that we've extracted without using mL
26:25
Speaker 3: or using mL now or recovering with mL, or even
26:29
Speaker 3: more complicated one with emls. But the major difference between
26:34
Speaker 3: finance and languages or pictures is one the amount of data.
26:39
Speaker 3: Because in the end, you know, stock markets have only
26:42
Speaker 3: existed since I don't know, nineteen hundred or eighteen.
26:45
Speaker 6: Hundred year one small data set.
26:47
Speaker 3: Small said, except if you go to high frequency, as
26:50
Speaker 3: I said, you know, if you go to tick by
26:52
Speaker 3: tick or the book data, there's huge amounts of data
26:56
Speaker 3: and there you can think that there's more to but
27:03
Speaker 3: what it was I saying, yeah, so so, so there's
27:06
Speaker 3: the problem of the availability of data and the frequency
27:10
Speaker 3: at which you want to predict. So for high frequency,
27:12
Speaker 3: I think there's a lot of structure. For lower frequency,
27:15
Speaker 3: it's not clear yet that it is going to be useful,
27:19
Speaker 3: you know, used as a kind of technical model which
27:25
Speaker 3: only looks at prices without reading texts. For reading text
27:29
Speaker 3: we know that there's there's a lot of structure, which
27:31
Speaker 3: corresponds to to the structure of language. But having said
27:35
Speaker 3: everything you said, there's still something strange about lll m's
27:39
Speaker 3: or you know, generative AI is that with this process
27:42
Speaker 3: of constructing sentences that are statistic statistically valid, you can
27:47
Speaker 3: invent new things. And that's the thing that really is
27:51
Speaker 3: really strange, right, I mean, you can learn pictures, for example,
27:55
Speaker 3: you know this celebrity database where you have made the
27:59
Speaker 3: machine learn these these pictures and then you ask the
28:02
Speaker 3: machine to generate new ones and it does, and these
28:06
Speaker 3: are pictures that look exactly I mean that you look
28:09
Speaker 3: and you think it's a celebrity, but the celebrity doesn't exist.
28:13
Speaker 3: So there's something still weird about this that, as I said,
28:17
Speaker 3: nobody really understand.
28:19
Speaker 2: Really kind of interesting.
28:21
Speaker 3: And so continuing on that, what we are trying to
28:24
Speaker 3: do is to do the same thing with financial markets. So,
28:28
Speaker 3: as I said, one hundred years of data is not
28:30
Speaker 3: a lot, but maybe you can use these gen AI
28:33
Speaker 3: models to generate a millionaires of fictitious financial markets. That's interesting,
28:40
Speaker 3: very interesting.
28:41
Speaker 2: So let's talk a little bit about trend following and
28:43
Speaker 2: managed futures. It's had a few real standout years in particular,
28:48
Speaker 2: twenty twenty two real challenging year and manage futures at
28:56
Speaker 2: we're top of the asset quilt? What does that episode
29:01
Speaker 2: tell us about what strategies work, why they work? And
29:05
Speaker 2: the question I always find with trend following why did
29:09
Speaker 2: so few investors tend to stay with them? They all
29:13
Speaker 2: seem to get nervous and tap out right before things.
29:17
Speaker 2: It's almost when you see people giving up. It's just
29:21
Speaker 2: about when the turn occurs.
29:23
Speaker 6: I agree, So what is it?
29:25
Speaker 4: What?
29:25
Speaker 2: First of all, why was twenty twenty two such a
29:28
Speaker 2: standout year? I don't know just I mean, aside from
29:31
Speaker 2: the fact that we had big FED rate hikes and
29:35
Speaker 2: fixed income and equities both got sill act double digits
29:40
Speaker 2: kind of rare occurs the same year. I think you
29:44
Speaker 2: have to go back about forty to forty one years
29:46
Speaker 2: to see both of them down significantly. How do you
29:51
Speaker 2: think about how do you think about what environment leads
29:57
Speaker 2: to best results for trend following.
30:00
Speaker 3: It's very difficult to say, because otherwise we would have
30:02
Speaker 3: a meta model that albutrages and you know, increases the
30:06
Speaker 3: weight of trend following when it's going to work. Well,
30:09
Speaker 3: I think it's there probably is more research to do,
30:13
Speaker 3: and where we've been trying, we haven't found anything that's
30:16
Speaker 3: very convincing. But you know, beginning of twenty twenty six
30:19
Speaker 3: is also a very good period for trendsforward. Actually, since
30:22
Speaker 3: we wrote a paper in twenty fourteen called two hundred
30:25
Speaker 3: Years of trend Following, and we were reporting on the
30:28
Speaker 3: fact that since eighteen hundred, if you pay per trade,
30:32
Speaker 3: a very simple trend following strategy, you make money every
30:36
Speaker 3: decade with ups and downs. You know, there are years
30:39
Speaker 3: that are not so good years. But as you say,
30:42
Speaker 3: I mean, what is striking and about the very point
30:46
Speaker 3: you made about people getting out of trend following just
30:50
Speaker 3: before it gets back on is I think it's ingrained
30:54
Speaker 3: in people's behavior to chase performance. So if performance has
30:59
Speaker 3: been bad for a few years, everybody declares. And that
31:02
Speaker 3: was the case in twenty fourteen when we wrote our paper,
31:06
Speaker 3: transfollowing had been flat for the last five years, and
31:09
Speaker 3: people said, okay, well, trensfalling is dead now, and we
31:11
Speaker 3: were actually convinced that it was not the case. That
31:14
Speaker 3: you know, transfalling is such a strong behavioral bias that
31:19
Speaker 3: performance chasing is so ingrained in every one of us
31:23
Speaker 3: even rational, we can't help, and so we bet and
31:28
Speaker 3: it was confirmed that trenfollowing would come back. And since
31:31
Speaker 3: twenty twenty four is twenty fourteen, it's been very good.
31:34
Speaker 2: Actually overall, well, you've had a market that very much
31:38
Speaker 2: was trending, mostly in one direction. For I mean, you
31:41
Speaker 2: have Q four of twenty eighteen, and I think twenty
31:43
Speaker 2: sixteen was so so. But for the past fifteen years,
31:47
Speaker 2: the bias has been pretty much in one direction. If
31:50
Speaker 2: you're on the right side of that, you should do
31:52
Speaker 2: pretty well.
31:52
Speaker 3: Yeah, but I'm not speaking about being long. I mean
31:55
Speaker 3: I'm really speaking about you know.
31:56
Speaker 2: Different asset classes, different trends of your long and short.
32:00
Speaker 3: Yeah.
32:00
Speaker 2: Sure, so it doesn't matter as long as the trend
32:03
Speaker 2: is in place you want.
32:03
Speaker 6: To part exactly.
32:06
Speaker 2: And for people who are not familiar with managed futures,
32:10
Speaker 2: there's a decent amount of leverage used in that product,
32:13
Speaker 2: so you have to really manage the risk of the downside.
32:17
Speaker 2: But how do you think about, you know, the potential
32:19
Speaker 2: upside relative to the risk you're taking in a futures product.
32:23
Speaker 3: What do you mean exactly? I mean, I mean, we
32:27
Speaker 3: just think about risk we want to risk. Is a
32:30
Speaker 3: very complex object. Actually, you know, there's a volatility, but
32:33
Speaker 3: that's also correlation. If you deal with the portfolio of
32:35
Speaker 3: futures that has like one hundred and fifty futures, there's
32:39
Speaker 3: a very subtle correlation structure between all the assets that
32:44
Speaker 3: you have in your portfolio. So if you think about risk,
32:47
Speaker 3: you really have to think about how all these products
32:51
Speaker 3: interact with one another, talk to one another. And so
32:54
Speaker 3: it's not only a question of volatility that goes up
32:57
Speaker 3: and down that you have to control, but also a
32:59
Speaker 3: question of how these assets co move together or anti
33:03
Speaker 3: co move together. But the way we think about upside
33:06
Speaker 3: risk is the same as the way we think about
33:07
Speaker 3: downside risk is it's just a question of risk.
33:10
Speaker 2: So you mentioned one hundred and fifty different of assets.
33:14
Speaker 2: I'm assuming some of these are commodities with modities, bones, stockspawns,
33:19
Speaker 2: interest rates, right, But what else is in the full
33:22
Speaker 2: list of one hundred and fifty.
33:24
Speaker 3: Well, you know there's different the maturities that different countries,
33:28
Speaker 3: futures in China. I mean, if you count everything, it
33:33
Speaker 3: goes up to you know, one hundred. I don't have
33:35
Speaker 3: the exact number, but in the one hundred and fifties altogether.
33:39
Speaker 2: Has CFN been looking at prediction markets things like CALCI
33:43
Speaker 2: and polymarket.
33:44
Speaker 3: No, we haven't. They're not liquid enough for us.
33:47
Speaker 2: Really, you need size and they can't provide it exactly.
33:51
Speaker 2: Really really quite fascinating. Coming up, we continue our conversation
33:55
Speaker 2: with Jean Felipe Buschot, co founder and chief scientists at
34:00
Speaker 2: Capital Fund Management, talking about how market structures are changing today.
34:06
Speaker 6: I'm Barry Results.
34:07
Speaker 2: You're listening to Master's Business on Bloomberg Radio.
34:24
Speaker 6: I'm Bury Redults.
34:25
Speaker 2: You're listening to Masters in Business on Bloomberg Radio. My
34:29
Speaker 2: extra special guest this week, John Philip Buschow, chief scientist
34:33
Speaker 2: and chairman at CFM, a quantitative hedge fund managing over
34:37
Speaker 2: twenty billion dollars in assets. So let's just talk a
34:41
Speaker 2: little bit about risk management. I know that when you're
34:45
Speaker 2: dealing with leveraged or long short or futures, there's a
34:51
Speaker 2: very robust thought process around risk management. Tell us a
34:57
Speaker 2: little bit about how you think about out correlation and risk.
35:03
Speaker 3: Yeah, we have a discipline and systematic approach not only
35:06
Speaker 3: to alpha signals, to building prediction, but also to risk management.
35:10
Speaker 3: We have a pretty sufisticated tool to predict the velocity
35:15
Speaker 3: of tomorrow, the velocility of our portfolio tomorrow, and we're
35:19
Speaker 3: pretty good at that. So of course, you know, we
35:22
Speaker 3: know in financial markets are difficult beasts, and even if
35:25
Speaker 3: you have the best model of the world, you can
35:28
Speaker 3: still have an expected events that blow up your portfolio.
35:31
Speaker 3: That's that's something that we can't say will never happen.
35:36
Speaker 3: But in a way with you know, if you don't
35:38
Speaker 3: want to make take any risk, you shouldn't be in
35:41
Speaker 3: financial markets. You shouldn't be in that business. So we
35:43
Speaker 3: accept that there might be I don't know, a completely
35:47
Speaker 3: unexpected event that you know, breaks the whole financial markets
35:52
Speaker 3: everywhere in the world and you know everything is going
35:55
Speaker 3: to fail and there's nothing to do about that, So
35:58
Speaker 3: that can happen. But barring these extreme events, we think
36:03
Speaker 3: we're pretty good at predicting what's going to happen. And
36:06
Speaker 3: over the last thirty five years of the existence of CFM,
36:10
Speaker 3: we're actually anniversary is this year, we're celebrating our thirty
36:16
Speaker 3: fifth anniversary in June in Paris. Very proud of that. So,
36:22
Speaker 3: you know, it kind of resisted these thirty five years,
36:25
Speaker 3: although we've become much better with time. But having said that,
36:31
Speaker 3: there's always an element that you have to be ready
36:35
Speaker 3: to intervene, even if you're a contrap and so on
36:39
Speaker 3: several occasions in the past thirty five years, we decided
36:44
Speaker 3: that our risk model couldn't know about things that we
36:47
Speaker 3: humans knew, Like you know, I don't know the Brexit
36:51
Speaker 3: votes or in the let's.
36:53
Speaker 2: Talk about that because this year, just the past twelve months,
36:58
Speaker 2: between the tariffs and venezuel Ella and now the ongoing
37:02
Speaker 2: war and Iran. How does global market volatility around all
37:06
Speaker 2: these geopolitical events, How does a quad shop deal with that?
37:12
Speaker 2: What I'm hearing is the humans have to do what
37:15
Speaker 2: humans do and sometimes override the machines.
37:18
Speaker 6: Sometimes, yes, I mean when when.
37:20
Speaker 2: Unexpected geopol political events disrupt the world, our models just
37:28
Speaker 2: not built to really work their way through that.
37:32
Speaker 3: You're right, some events are okay, and like the war
37:37
Speaker 3: in Iran for the moment, is not, you know, something
37:40
Speaker 3: that our risk models are completely blind to. Right, It
37:44
Speaker 3: doesn't mean that they've predicted at all. It just means that,
37:47
Speaker 3: you know, we're comfortable with the risk that our model
37:50
Speaker 3: have predicted, and they've adapted sufficially fast toward to the events,
37:54
Speaker 3: so that we're comfortable with the risk level no human intervention.
37:58
Speaker 3: On the other hand, in some cases it is completely unexpected,
38:01
Speaker 3: like tariffs and Liberation Day, this created havoc, although strangely enough,
38:07
Speaker 3: Libration Day was announced. You know, everybody knew what it
38:11
Speaker 3: was going to be said, and still everybody.
38:14
Speaker 6: Was apprised of it.
38:15
Speaker 2: I don't believe despite I'm tariff man, it's the most
38:19
Speaker 2: beautiful in the dictionary. I think the one hundred, one
38:23
Speaker 2: hundred and fifty percent tariffs on specific countries later found
38:27
Speaker 2: to be completely unconstitutional, But at the time I think
38:31
Speaker 2: people were genuinely shocked by this. And then a week later,
38:37
Speaker 2: you know, the little bit of a taco trade where
38:40
Speaker 2: let's just put a pin in this for ninety days. Again,
38:44
Speaker 2: back to the volatility, how do you deal? I agree
38:47
Speaker 2: oil is trending upwards, and then you have a tweet
38:50
Speaker 2: the world the war's over, and then it resumes, and
38:53
Speaker 2: then you have a tweet I think we've got a deal,
38:56
Speaker 2: and then the other side says we're not even negotiating.
39:00
Speaker 2: I don't recall a period in history where the president
39:03
Speaker 2: of the United States just constantly disrupted the normal flow
39:08
Speaker 2: of market activity. How disruptive is this to a quant model.
39:14
Speaker 3: Well, as I said, the two cases seem to be
39:16
Speaker 3: pretty different. Liberation Day was really a surprise and we
39:20
Speaker 3: had to manually intervene. There was something in our models
39:23
Speaker 3: that was completely blind to these things, and we had
39:26
Speaker 3: to make a judge on the call. I think the
39:28
Speaker 3: idea really is that humans should use their best judgment
39:32
Speaker 3: in these cases and decide whether it's reasonable that the
39:36
Speaker 3: risk model knows something about what's going on or not.
39:40
Speaker 3: In some cases it does, in some cases it doesn't.
39:42
Speaker 3: I think the the tricky part is not to overreact,
39:47
Speaker 3: because you said you don't remember periods of the world
39:50
Speaker 3: where things like this happened. But you know, looking back,
39:54
Speaker 3: I've been in the markets for thirty five years and
39:56
Speaker 3: everything every year there seems to be something unexpect that happens.
40:00
Speaker 6: You know, just not every day, not every day, but
40:03
Speaker 6: every year.
40:03
Speaker 2: It seems like a lot.
40:04
Speaker 3: Right, But in a way, every day means that it
40:07
Speaker 3: becomes a new normal, I guess, and so it's not
40:11
Speaker 3: that bad.
40:11
Speaker 6: It's if it's every day.
40:13
Speaker 3: But really, you know, this idea that this time is
40:16
Speaker 3: different is something that's strange. If you look at the
40:19
Speaker 3: world and the history of financial markets. It's really being normal.
40:25
Speaker 3: That's not normal, and we've become used to that.
40:30
Speaker 2: So let's stay with the idea of modeling. You've been
40:34
Speaker 2: kind of skeptical of certain applications of deep learning in finance.
40:39
Speaker 2: There's overfitting. You know, no one's ever seen a bad
40:44
Speaker 2: back test because they all seem to work perfectly in
40:48
Speaker 2: the past. You have a lot of signal and noise issues.
40:52
Speaker 2: What are some of the problems with models that you're
40:56
Speaker 2: focusing on improving.
40:57
Speaker 3: Yeah, well, for example, this exactly what you just said.
41:01
Speaker 3: Can you have indicators that tell you whether your back
41:05
Speaker 3: test is overfitted or not? And for many years we
41:10
Speaker 3: struggled with that, and we use judgment again to say
41:14
Speaker 3: this is plausible, this is not plausible. We can believe
41:17
Speaker 3: that we kind of replace the trader that trades every
41:21
Speaker 3: day his signals or his beliefs to a higher level
41:28
Speaker 3: where we are traders of models. We know we kind
41:32
Speaker 3: of judge models. We say this model is good enough
41:35
Speaker 3: to go in production, this model is not convincing enough,
41:38
Speaker 3: but it would be great to have something more systematic.
41:41
Speaker 3: And over the years we've been struggling, and I think
41:44
Speaker 3: with some success, to have meta models that predict whether
41:49
Speaker 3: your back test is really fudged or if it's decent
41:54
Speaker 3: enough to go in production. So we're kind of industrial
41:58
Speaker 3: and industrializing this process of selecting models that will go
42:03
Speaker 3: into production. Does that make sense?
42:04
Speaker 2: Yeah, No, that makes perfect sense. You're quant, and we've
42:08
Speaker 2: seen some issues with a lot of quant shops in
42:12
Speaker 2: the US where crowding became a structural risk. You have
42:17
Speaker 2: all these system systematic strategies, and math is math, so
42:21
Speaker 2: essentially you end up with a crowded trade. We had
42:25
Speaker 2: what people called the quant quake way back when. How
42:29
Speaker 2: do you think about that? How do you manage that
42:32
Speaker 2: problem when you're constructing portfolio?
42:34
Speaker 3: Sure it's something that we you know, every day we
42:36
Speaker 3: think about this. We were in the quant quake in
42:39
Speaker 3: two thousand and oh seven, and actually we were fortunate
42:44
Speaker 3: enough to be out of the markets or to have
42:47
Speaker 3: the leveraged already two weeks before the worst day of
42:53
Speaker 3: the quont quake.
42:54
Speaker 2: Was that an external signal or one of your model
42:57
Speaker 2: signals or what like?
42:58
Speaker 3: Was it was something in the perfour months already in July.
43:01
Speaker 3: The one quick the really bad day happened maybe ninth
43:05
Speaker 3: of August, I don't remember exactly but.
43:07
Speaker 6: It was early August days of summer.
43:09
Speaker 3: Right, but starting like tenth of July already there was
43:14
Speaker 3: something really very strange in our portfolio. And we started
43:18
Speaker 3: reflecting on what was going on and decided someone was
43:21
Speaker 3: leveraging and you know, hitting us by shorting our lungs
43:27
Speaker 3: and buying our shorts, and you know, this process, this
43:30
Speaker 3: thought process of imagining that even if a fund that
43:35
Speaker 3: was like ten percent correlated with ours, not a lot,
43:38
Speaker 3: but ten percent and having every day a kind of
43:42
Speaker 3: systematic deleveraging policy, it would create exactly the kind of
43:46
Speaker 3: signals that we were seeing in our portfolio. So we thought, okay,
43:49
Speaker 3: this is maybe going to lead to a crash because
43:52
Speaker 3: people are going to you know, suffer suffering at one point,
43:56
Speaker 3: they're going to cascades and skate and so on, and
43:59
Speaker 3: so that was rationale for getting out. So in some
44:02
Speaker 3: cases you have you're lucky enough to have strong enough
44:06
Speaker 3: signals that tell you that your standard risk model is
44:09
Speaker 3: wrong and you should do something else.
44:13
Speaker 2: That's really that's really fascinating. So you mentioned almost thirty
44:17
Speaker 2: five years of doing this, What do you think that
44:21
Speaker 2: I'm gonna say that again? So you are now thirty
44:25
Speaker 2: five almost thirty five years into being a market quant
44:30
Speaker 2: what's the most important thing about markets that the mainstream
44:36
Speaker 2: funds still get wrong?
44:39
Speaker 3: Well, I think it's this question of what is price doing?
44:44
Speaker 3: What are moving prices? And a lot of people still
44:48
Speaker 3: believe that there's something like a fundamental value and that
44:51
Speaker 3: the price is really moving because fundamentals are moving whereas
44:57
Speaker 3: we believe, and this is going you know, this touches
45:00
Speaker 3: very recent academic papers that Gabex and coaches and to
45:05
Speaker 3: economists have put forward. They've called it the inefficient inelastic
45:09
Speaker 3: market hypothesis, and we've contributed to that debate as well.
45:14
Speaker 3: And the idea is really that markets are not driven
45:16
Speaker 3: by fundamentals, or at least they are. There are to
45:21
Speaker 3: some extent driven by fundamentals, but this is a small
45:24
Speaker 3: long term effect on short run. Short run meaning from
45:28
Speaker 3: one day to one year's pretty long already. It's really
45:31
Speaker 3: flows that matter. That is people buying or selling stuff,
45:36
Speaker 3: whatever the reason they buy or sell, is going to
45:39
Speaker 3: move prices. It's not going to move prices on a
45:41
Speaker 3: short timescale and then disappear. It's really going to leave
45:45
Speaker 3: a trace in markets. And this is really a fundamental
45:49
Speaker 3: change of point of view that I think that you
45:53
Speaker 3: know is going to percolate and convince more and more
45:58
Speaker 3: people looking for But having this change of tech is
46:03
Speaker 3: really important because in one case, what you need to
46:05
Speaker 3: do to make money is to predict fundamentals. In the
46:09
Speaker 3: other case, you need to predict what people are going
46:11
Speaker 3: to do. And so in a sense, crowding can be
46:14
Speaker 3: a good thing because if there's crowding, it's easier to
46:16
Speaker 3: predict what the crowd is going to do. And so
46:20
Speaker 3: if whatever the reason, people do things, they move prices
46:23
Speaker 3: and you're able to predict what people are going to
46:25
Speaker 3: do because you have, you know, behavioral models and structural
46:29
Speaker 3: models that tell you, you know, everything else being equal, people
46:33
Speaker 3: are more likely to do this and that then you
46:35
Speaker 3: can build models. And I think that's the reason why
46:38
Speaker 3: we've been successful is this change of philosophy. We were
46:42
Speaker 3: not kind of anchored to fundamentals. We're anchored to flows.
46:46
Speaker 2: So this sounds a little bit like the ben Gram line.
46:49
Speaker 2: I think it's Gram in the short run markets or
46:52
Speaker 2: voting machines and the long run their weighing machines. Is
46:55
Speaker 2: that is that the balance between flows and fundamentals.
47:00
Speaker 3: Yeah, I think it's an old idea. I mean, you know,
47:03
Speaker 3: Canes also said things like that. You know, but in
47:06
Speaker 3: the long run, we're all dead, right, Panes dead was
47:09
Speaker 3: saying that, So it's really a question of whether you're
47:12
Speaker 3: going to be solvable. I'm getting tired. I'll take that again.
47:25
Speaker 6: What's the word was that, Kip?
47:27
Speaker 2: By the way, Knes or Graham? I initially thought it
47:30
Speaker 2: was Canes, and then I checked myself.
47:32
Speaker 3: I'm not sure, but what Kines said is that you
47:36
Speaker 3: what was his quote? Markets can remain irrational longer than
47:43
Speaker 3: you can remain solid.
47:44
Speaker 2: So he never said that, but it's always attributed to him. Really,
47:48
Speaker 2: there's a great website called Quote Investigator. I see, and
47:52
Speaker 2: you give them a quote. Because people as a sort
47:55
Speaker 2: of authority, you know, appeal to authority, they'll put somebody
48:01
Speaker 2: sophisticated as the source. Einstein never said compounding is the most.
48:06
Speaker 5: Powerful force in the universe, but they always attributed to
48:09
Speaker 5: I didn't know that you would be I spend way
48:12
Speaker 5: too much time perusing quotes on the site, but you
48:15
Speaker 5: would be shocked who said markets are a voting machine
48:20
Speaker 5: in the short run, but a weighing machine in the
48:22
Speaker 5: long run.
48:23
Speaker 3: I don't think it's Canes by the way, No, I
48:25
Speaker 3: don't remember if.
48:26
Speaker 2: It was Canes or Graham, but I thought it was
48:28
Speaker 2: one or the other, and it'll tell me. Yeah, Benjamin Graham,
48:31
Speaker 2: But the first thing that came to mind was Canes
48:34
Speaker 2: because or Galbraith. They have so many favorite quotes from.
48:37
Speaker 3: Yeah, well he said many relevant things today. Yes, absolutely,
48:43
Speaker 3: but actually we have models that predict exactly that that
48:46
Speaker 3: you know, on the short run you can have trends
48:48
Speaker 3: and the irrational behavior, and on the long run it reverts
48:51
Speaker 3: back to fundamentals. But the long run from our estimate
48:54
Speaker 3: is like five ten years. I'm trying a long time scale.
48:58
Speaker 2: I'm trying to remember I'd have been French of Farm
49:01
Speaker 2: of French who said, you can't really tell if a
49:04
Speaker 2: manager is skilled innt you until you have twenty years
49:07
Speaker 2: of data, because it could just be good luck over
49:10
Speaker 2: five or ten years, which is kind of fascinating. I
49:13
Speaker 2: want to stay with the efficient market or the inelastic
49:18
Speaker 2: market hypothesis. I'm curious as to your thoughts on EMH.
49:24
Speaker 2: I've always thought markets were kind of sort of eventually efficient,
49:29
Speaker 2: but not very efficient in the short run. What's your
49:33
Speaker 2: criticism of EMH.
49:35
Speaker 3: Well, It really depends what you mean my efficient. If
49:39
Speaker 3: you mean that they're very close to unpredictable, then you're write.
49:43
Speaker 3: But I think it's a very dumb down version of EMH.
49:46
Speaker 3: The question is whether our prices reflect something from the
49:49
Speaker 3: mental that is, in principle noble, that reflects reality or
49:56
Speaker 3: not at all. And one I think smoking gun of
50:01
Speaker 3: that is do you have long term mean reversion? That is,
50:07
Speaker 3: can prices do random things in particular trending, which is
50:10
Speaker 3: really completely against the EMH. On the short run, that
50:14
Speaker 3: is from a week to six months, markets are trending
50:18
Speaker 3: over six months, one year, and then on the longer
50:21
Speaker 3: time scale they kind of hover around some long term
50:27
Speaker 3: trend and I think this is true, but this is
50:30
Speaker 3: really at odds with efficient market, which tells you that
50:33
Speaker 3: every day markets are around the correct price and there's
50:38
Speaker 3: no trend, no mean reversion.
50:40
Speaker 2: Ever, which is which you know that that's not exactly
50:45
Speaker 2: what your day to day experiences if you're in the markets.
50:48
Speaker 2: It seems it's not magic that the collective votes of
50:53
Speaker 2: all market participants don't magically bring you to the correct
50:58
Speaker 2: in quotes.
50:59
Speaker 3: But that's that's the assumption of efficient market, or actually
51:04
Speaker 3: not even an assumption. It's the argument that collectively, you
51:09
Speaker 3: know if you have. That's the difference between having rational
51:13
Speaker 3: investors that all take decisions based on noisy observation, but
51:18
Speaker 3: independent from one another. Then because they're independent, they realize
51:23
Speaker 3: the mean I mean, some overpriced, some underprice, and then
51:27
Speaker 3: it's a voting machine and the vote comes out right
51:30
Speaker 3: because there's enough investors and they're uncorrelated to one another.
51:34
Speaker 3: But the problem with markets is that it's not the
51:36
Speaker 3: way it works. The people are influenced by what other
51:40
Speaker 3: people are doing and what other people are saying. So
51:43
Speaker 3: instead of having independent guys doing random stuff, they're kind
51:47
Speaker 3: of one guy who's doing only one thing, which is
51:54
Speaker 3: a fictitious body that aggregates everybody in the same way.
51:59
Speaker 2: Let's jump to our favorite questions that we ask all
52:01
Speaker 2: of our guests, starting with tell us about your mentors
52:05
Speaker 2: who helped shape the direction of that career.
52:08
Speaker 3: That's an easy one. I have, you know, several mentors,
52:11
Speaker 3: but two of them are really close to my heart.
52:14
Speaker 3: One is but man that brought of course, you know,
52:17
Speaker 3: the fractal guy. I knew him personally, Oh really, yes,
52:21
Speaker 3: and he did a lot of things in physics as well, sure,
52:25
Speaker 3: and so he influenced a lot my wife. My wife
52:29
Speaker 3: was a physicist before turning a playwright now, and and
52:35
Speaker 3: she worked on fracture surfaces. The way you know, when
52:37
Speaker 3: you break a material, what emerges from from from the
52:42
Speaker 3: fracture is a kind of very rough landscape that is fractal.
52:46
Speaker 3: And Manda Brought had worked on that, and there was
52:48
Speaker 3: a lot of interaction and.
52:50
Speaker 2: The fractal at a molecular level or at.
52:52
Speaker 3: A yeah, well fractal from the kind of very fine
52:57
Speaker 3: structure to microscopic length skill. And so Manelrad also did
53:03
Speaker 3: his work on financial markets, and for me it was
53:06
Speaker 3: really a revelation. It was something very influencing and out
53:12
Speaker 3: of the dogma of Brownian statistics and Gaustian phenomena and
53:16
Speaker 3: so on, and so it was also very close to
53:18
Speaker 3: what I was doing myself in physics. So you know,
53:22
Speaker 3: it was clear that he influenced me enormously on that.
53:26
Speaker 2: Didn't he write a book on market crashes and how
53:30
Speaker 2: there's a fractal.
53:31
Speaker 3: Yeah, with the misbehavior of market that's right there, you go,
53:35
Speaker 3: I'm actually quoted in that book.
53:38
Speaker 2: Oh that's fascinating.
53:39
Speaker 3: And then Pierre Gilldzene who was a Nobel Price in
53:43
Speaker 3: physics of French physicists, who was so fantastic, and you
53:48
Speaker 3: know both these two, and and also Phil Anderson who
53:51
Speaker 3: was a Nobel Price in physics as well in the US.
53:54
Speaker 3: These three people, they convinced me that you shouldn't be
54:00
Speaker 3: stuck to your own field. You should broaden your scope,
54:04
Speaker 3: and what you learn from one field can be very
54:06
Speaker 3: useful understanding another field. The three of them, they're really
54:10
Speaker 3: kind of hovered around and not got tied to their
54:15
Speaker 3: specific initial field. And I think this creates well, at
54:20
Speaker 3: least for me, this created this The inhibited me in
54:26
Speaker 3: the sense that I thought, Okay, maybe I'm not legitimate
54:29
Speaker 3: to speak about finance because I'm a physicist, but you know,
54:33
Speaker 3: doesn't matter. If I have things that I strongly believe in,
54:37
Speaker 3: I should better say them and go to the end
54:40
Speaker 3: of them. So I think they were really influential in
54:42
Speaker 3: that way.
54:43
Speaker 2: So since we mentioned the misbehavior of markets, let's talk
54:47
Speaker 2: about some books. What are some of your favorites? What
54:49
Speaker 2: are you reading right now?
54:50
Speaker 6: Wow? So many?
54:51
Speaker 3: That's really a very broad question. So my last book
54:55
Speaker 3: is a book on John and Paul By in Leslie Johnny.
55:03
Speaker 3: It's a beautiful book. I really loved it.
55:05
Speaker 6: What's the name of it, John and Paul?
55:08
Speaker 2: Is it John and PAULA love story Memory?
55:10
Speaker 6: Correct?
55:11
Speaker 2: I'm a big Beatles fan, and.
55:14
Speaker 3: Oh you should read it.
55:15
Speaker 6: I'm very very I'm going to.
55:17
Speaker 2: Make a recommendation to you for a YouTube channel called
55:22
Speaker 2: you Can't Unhear This. They take apart Beatles songs in
55:26
Speaker 2: ways that just little things that were done in the
55:30
Speaker 2: recording process that in a million years you never would
55:33
Speaker 2: have noticed, and then once you hear it, you just
55:35
Speaker 2: can't unhear it. And if you're a Beatles fan, it's
55:39
Speaker 2: a rabbit hole. You'll you'll love this. What else besides
55:42
Speaker 2: John and Paul, Yeah.
55:44
Speaker 3: I'm reading someone that's something that I actually have read
55:46
Speaker 3: for years that Missus Dollaway Virginia or Wolf. Yeah, I'm
55:51
Speaker 3: read a great admirer of Virginia.
55:53
Speaker 2: Wolf really really interesting. Do you do much streaming TV
56:01
Speaker 2: podcast anything like that or we can skip.
56:03
Speaker 3: Both podcasts a bit, but they're kind of French podcast.
56:07
Speaker 2: So let's let's you know, people are always looking for stuff.
56:09
Speaker 2: I'm going to ask that, So what are you streaming today?
56:12
Speaker 2: What sort of podcasts are you listening to.
56:14
Speaker 3: Yeah, you know, in France, we were very fortunate we
56:17
Speaker 3: have something called Hoskulture. It's a radio where there's an
56:21
Speaker 3: enormal I mean, you could stay tuned all day if
56:24
Speaker 3: you want it. There's so many interesting things going on
56:27
Speaker 3: about everything cultural literature, but also you know, movies, politics,
56:33
Speaker 3: and so.
56:34
Speaker 2: We have NPR here.
56:35
Speaker 6: It's very similar.
56:36
Speaker 2: You can it's a rabbit hole you could fall down.
56:39
Speaker 3: Yeah, okay, so life of big peeple of major celebrities
56:44
Speaker 3: in culture, in in cinema and theater, all these things.
56:49
Speaker 3: So I'm really a big fan of Hostalture.
56:52
Speaker 2: And our final two questions, First, what sort of advice
56:57
Speaker 2: would you give to a recent college grad interest did
57:00
Speaker 2: in a career in either quantitative investing or theoretical physics.
57:06
Speaker 3: Well, study theoretical physics and study everything that's related to data,
57:12
Speaker 3: and pay attention to data. And think about something that
57:17
Speaker 3: you strongly believe in and that you feel has not
57:21
Speaker 3: been investigated, and it doesn't matter if it's big or small.
57:25
Speaker 3: Make the effort of building something you strongly believe in.
57:29
Speaker 2: Really good answer. And our final question, what do you
57:34
Speaker 2: know about the world of investing today? Might have been
57:36
Speaker 2: useful thirty five years or so ago when you were
57:39
Speaker 2: first getting started.
57:41
Speaker 3: That is extremely competitive, much more than we thought.
57:44
Speaker 2: Really wow, that's really fascinating. Jean Philip, thank you so
57:48
Speaker 2: much for being so generous with your time. We have
57:52
Speaker 2: been speaking with Jean Philippe Boucheau, cfm's co founder, chairman,
57:58
Speaker 2: and chief scientist. If you enjoy this conversation, well check
58:02
Speaker 2: out any of the six hundred plus we've done over
58:04
Speaker 2: the past twelve years. You could find those at Apple Podcasts, Spotify, YouTube, Bloomberg,
58:11
Speaker 2: wherever you find your favorite podcasts. I would be remiss
58:15
Speaker 2: if I didn't thank the crack team that helps me
58:17
Speaker 2: put these conversations together each week. Meredith Frank is my
58:21
Speaker 2: video producer. Anna Luke is my podcast producer. Sean Russo
58:26
Speaker 2: is that my head of research. I'm Barry Ritoults. You've
58:30
Speaker 2: been listening to Masters in Business on Bloomberg Radio.