The Derivative - Sowing the Seeds of AI Strategies with Howard Siow of Taaffeite Capital Management

Episode Date: June 17, 2021

Artificial Intelligence is a hot topic in every walk of life, and perhaps no more so than in the world of hedge funds and alternative investments. Howard Siow, Founding Principal and CEO at Taaffeite Capital, joins The Derivative to discuss how AI is used in the hedge fund world, and how everyone wants AI to predict every move in the market, but that’s not actually what AI does well. Howard talks how AI can be used as an incremental change agent, chicken farming, being a venture capitalist, the effect of human biases, the dirty little secret holding AI back, trying to solve complex problems, being an engineer, launching Taaffeite, opportunities in China, and growing up down under in Australia. Chapters: 00:00-02:47=Intro 02:48-10:51=AI = Getting the small things right 10:52-22:25=The most complex system we have? 22:26-26:49=Launching Taaffeite 26:50-47:15=The Strategy 47:16-01:10:04=Is AI getting closer to solving the market problem? 01:10:05-01:16:08=Humble Adjustments 01:16:09-01:22:47=Favorites Learn more about Taaffeite Capital and their strategies here: http://taaffeitecm.com/ And last but not least, don't forget to subscribe to The Derivative. Follow our host Jeff Malec on Twitter and RCM on Twitter, LinkedIn, Facebook, and sign-up for our blog digest. Disclaimer: This podcast is provided for informational purposes only and should not be relied upon as legal, business, or tax advice. All opinions expressed by podcast participants are solely their own opinions and do not necessarily reflect the opinions of RCM Alternatives, their affiliates, or companies featured. Due to industry regulations, participants on this podcast are instructed not to make specific trade recommendations, nor reference past or potential profits. And listeners are reminded that managed futures, commodity trading, and other alternative investments are complex and carry a risk of substantial losses. As such, they are not suitable for all investors. For more information, visit www.rcmalternatives.com/disclaimer

Thanks for listening to The Derivative. This podcast is provided for informational purposes only and should not be relied upon as legal, business, investment, or tax advice. All opinions expressed by podcast participants are solely their own opinions and do not necessarily reflect the opinions of RCM Alternatives, their affiliates, or companies featured. Due to industry regulations, participants on this podcast are instructed not to make specific trade recommendations nor reference past or potential profits, and listeners are reminded that managed futures, commodity trading, and other alternative investments are complex and carry a risk

of substantial losses. As such, they are not suitable for all investors. Welcome to The Derivative by RCM Alternatives, where we dive into what makes alternative investments go, analyze the strategies of unique hedge fund managers, and chat with interesting guests from across the investment world. Well, I think it's a little bit like, you know, the way that we would look at it as a human being, right? There would be certain things that we can glean from the past, right? So if you look at, for example, people are now talking about the 1970s and what happened in the 1970s with inflation so people can look back and say look there are some macro things that we can learn and there are ways that we can use historical data to back test our strategies and to kind of predict what would be expected behavior right and we can test to see, you know, can our systems, you know, respond to a large data set, right? I mean, if even as markets evolve, you know, ideally, we would like to have 100 years of data, right, that we could run all our systems through and be able to backtest them through 100 years of data.

Now, we could say back in, you know, back in 1921, markets were completely dissimilar to what they are today, right? But we would still love to be able to get that 100 years of data set to be able to run our systems through. That would still be a good test for the robustness of our system. But, you know, so there'll be things that you could learn from that entire data set, but also there'll be things that you would incorrectly take away from that, like, you know, in'll be things that you would be incorrectly take away

from that, like, you know, in terms of how much the market microstructure has changed, right? And, you know, so it's about being able to think about those two Coming at you around 10.30 p.m. here in Chicago to get our guest during his day down under. We've got a packaging company CEO turned PwC and Bain consultant turned AI-focused hedge fund manager, all from growing up in the chicken farming business in Australia. It's Howard Seau, founder and CEO of Taffite Capital Management. Welcome, Howard. Thanks. Good to see you.

We were just saying it looks like winter there. That tree behind you looks like it's missing some leaves. It's a cold day in Melbourne right now, actually. We're in the middle of a lockdown, so I can't really complain. Oh, so you're not even really supposed to be outside uh yeah so uh it's uh i think it's the last day of a two-week lockdown because we had a bit of an outbreak here in the last couple of weeks you guys have been one of the stars though in terms of controlling at least melbourne has right yeah well i think we had like three cases or something

like that overnight and people are freaking out about three cases um uh we'll get through it and then you were born and raised there in melbourne yes i was born in melbourne um and yeah i've uh i lived in melbourne until i was maybe like 20 22 23 and then i've been kind of traveling abroad doing things yeah that's one of my only spots i haven't been to in australia been sydney over to perth and freemantle up at uh surfers paradise but haven't made it down there so it's on my list for sure yeah well uh you should come by jeff would be uh it'd be good to entertain you i think i'll come in our winter though in, in your summer.

That would be a little better. And tell me about this chicken farming stuff. That was fun to read in your bio there. So your dad was in the chicken farming business? Yeah. So my dad, I mean, I grew up on a chicken farm. My dad owned a poultry egg business in Melbourne. So, you know, I grew up there, you know, ever since I was like five, six years old, I was kind of helping out on the farm.

I was, you know, selling eggs by the roadside at one stage when I was a little kid. So, yeah, I was in the egg business kind of growing up. And then we kind of left the farm and i was around uh 13 14 years old we moved into the city as my dad was kind of expanding the business and decided it was better that i kind of education in the city for a while so kind of like from about 13 14 onwards i was kind of in the city but you know pre that i was a little farmer boy and when i think of chicken um i think you're farming the actual like meat,

but you were just the eggs or both? Yeah, so predominantly an egg business, but, you know, it's a similar business, right? You know, it's an egg chicken rather than a meat chicken. Right. So you can tell us which came first, chicken or the egg? It still stunts me till today. Awesome and then so then you're I'm always I got in the back of my head you

ever seen Silence of the Lambs? Not really but I know of the film. Yeah and then she grew up on a lamb farm and he gets into her head like do you hear the lambs at night when they're killing the lambs? i i worry that you never ate chicken again after dealing with so much chicken but hopefully you did so then you got off the farm and you start doing getting your education so what what did that look like you were at a university there in melbourne yeah so i went to uh i was kind of a melbourne boy so i went to melbourne grammar then i went to melbourne university i studied uh um uh engineering software engineering and uh finance and commerce finance um uh which is a double bachelor's degree at melbourne uni i'm not sure

if they still offer it till today but i think um at the time it was quite controversial they said why would you want to study finance as well at the same time as something like engineering, like software engineering? And it didn't make sense to a lot of people. But I think I was the first batch to graduate from that double degree. It's like, you know, a five-year degree. And so kind of the precursor to today at like Cal Berkeley and stuff, they have these financial engineering degrees, right?

So they've made it into one degree at a lot of schools now. Right. I mean, like back in my day, I mean, this is, I mean, I graduated in 2001. So I started from 1996 to 2001. Back in the day, I think we were the first graduate class of software engineering. So software engineering in 1996 was completely

was completely wild you know that someone could start like a specialist degree in software engineering so i actually started off my degree doing electrical engineering for two years and then the last three years you know they actually came out with the software engineering degree that i kind of pivoted to because I just enjoyed the whole computer science software component of engineering. And yeah, and for my commerce degree, I gravitated towards finance because I thought it was very interesting. But because it was a double degree, the university actually allowed us to do a lot of things. So it kind of a a degree that was kind of between economics and finance um and also we also did bits of you know accounting marketing um strategy and things like that so you know it almost turned into like a virtual mba at that stage it was a kind of a

major in finance um but yeah i mean it was a very it was it was kind of like the first year that they would kind of offer us at such a such a bizarre degree and everyone was wondering what the hell we were doing with it um it just so happened kind of making future hedge funders um which always amazes me many of the guests we've had many of the uh people i know in the business they engineering backgrounds a lot of the clients too a lot of high net worth and family office people who get interested in alternatives have like an engineering mindset, right? Of like, okay, this is the ultimate puzzle. Let me figure out the pieces and try and figure out how to put it together.

I mean, the thing that's interesting about engineering that I think a lot of people would talk about is that it's a very systems-based approach to life right so you learn about abstraction and you learn about kind of um putting things into you know abstracting them into boxes and breaking down very complex systems into simpler systems and think of everything as you know as a complex uh complex system that can be simplified into into you know different abstracted systems i think that that's very interesting, particularly in the world of finance or economics where these are immensely complex systems and it requires kind of an engineering mindset to be able to break it down. I think it's very much like a lot of problems that we have.

If you look at people that make breakthroughs, whether it be in, you know, in electric vehicles or in, you know, building a machine that can fly or, you know, a lot of problems. These aren't necessarily the smartest people on the planet. You know, they're not necessarily fields medalists or things like that. Right. Fields medalists and things like that, right? But it's about people that can work on very, you know, very complex problems and commit to them for long periods of time. And it's about getting a lot of small things right. And I think, you know, and I think it was, you know, third in my career in managing consulting that a lot of the solution

mindset is about, you is about assembling the right team and being able to think about the problem correctly and that the solutions were generally not a result of the most sophisticated mathematics or the most sophisticated AI, things like that, but it's about getting a lot of small things right and it strikes me like is it the ultimate hubris to think like that you can solve the market that it is is even a solvable complex problem right like it's a little bit if you had a machine or you have an engine or a drone right it it can't fly 76 percent of the time right it

has to fly 100 percent of the time the engine has to fire nearly 100 percent of the time but in the markets you maybe only need to be right 50.46 percent of the time or something so i can see both sides of it like one no way anyone could figure this out no matter the computing power um but then two you have people like jim simons and you have you have groups that are figuring it out so how do you how do you think about that of like it's it is that complex problem you do have the solutions mindset but do you believe there actually is one solution out there um i mean i think that i mean in, you know, people ask this all the time, right? Can I build a system that can predict every move of the market, right? Every time the market, you know, something moves or something happens, you know, can you predict it? I mean, I think

that that almost by definition is impossible, right? I mean, I can't predict what will happen in the next hour, pretty much anything, right? Like, you know, is it possible to predict September 11 as an example? I mean, I think that that's just fundamentally very difficult to predict, right? I think that what can be done, right, is building a robust system that's able to withstand a lot of shocks. But can you build a system that can predict anything?

I think that that's fundamentally fairly impossible. I mean, you'd be basically very rich if you could build a system that would predict what every participant in the market is going to do at any point in time. It's like Westworld. I think part of that is because there are humans involved, right, which is why the markets have, you know, 10, 20, even 30 sigma moves, right?

You know, moves that shouldn't happen in like, you know, 10 lifetimes of the universe and they happen every five years. Right. And that happens just because, you know, we have humans and, you know, these are immensely complex systems. And when certain stars align, as they do every now and then, you know, things will collapse. And, you know, I think one of the most humbling things that I remember at the start of the, you know, when I first started TCM back in 2013, was if you actually go back and have a look at how many strategies or how many funds have lasted more than 10 years, right?

It's not a lot, right? So back in 2013, there was actually not a lot of funds that actually existed prior to 2007, right? I think even today if you look at how many funds that have been around for more than 10 years it's very few and far between right and that's because you know because markets move in in such kind of random and such you know as i said 10 20 30 sigma moves building a system that is robust to you know pretty much any eventuality is very, very difficult.

And I think that that's kind of the kind of degree of humility that we also approach the market. Whilst we are what we call a quant macro system, so we do think about the market as a very large data set, we also superimpose on that what we call kind of fundamentals of the strategy. So things that, you know, we kind of had a look at what has blown up strategies over the last, say, 50 years. And we built that into a set of principles that we said

that the system could never do, right, so that we would give ourselves, you know, the biggest buffer possible um to ensure that any kind of huge tail event um or black swan event that would occur um we would give ourselves a better a better chance of of of that not being our downfall um to the system so we data led we built reset. And I can go through that a little bit later. Yeah, yeah, we'll touch on that later. But, yeah, I love it. It's like you're building the least crappy set of many models versus the one best model to solve all problems, right?

Right. So, yeah, we buried the lead right there. I love it. We jumped right into it. But back to your background quick. So you went into consulting and then you said, screw that, I'm going out for for myself give us the quick story there yeah so i mean management consulting i mean it's a wonderful experience i mean you get to travel like you meet you know

people and executives from all over the world and i think that that's um management consulting is a is a very expensive account right life's good right um uh but, you know, a few things, I mean, you know, I think the reality is of, you know, for managed consultants is, you know, you work on a lot of projects that are generally about, you know, three months long in general, give or take. And at the end of the day, what you produce is kind of like a PowerPoint. Right. It's a, it's a set of the day what you produce is kind of like a powerpoint right um it's it's a

it's a set of recommendations um uh to you know to a board or to some kind of committee right um but ultimately you are the person providing the recommendation and they're the decision makers right and i think after about seven eight years you look, I don't want to be the person producing PowerPoint. I mean, I want to be the decision maker, right? And, you know, I think that there's, you know, some people are very comfortable, you know, in the advisory role. But I think I wanted to be the person executing and doing more with that, right?

Yeah. What were some of your big clients there anyone exciting uh you know we can't really you know we just generally don't talk about you know clients but um i specialize in uh i did the automotive industry i did the uh uh i did the energy i did a lot of energy clients as well. So it was predominantly energy and automotive and some telco as well. And so somewhere in all there, you kind of taught yourself coding? No, I mean, like most of my coding background was from university. So, I mean, I did a five-year degree in software engineering. So most of that was from my university days.

But I was, I mean, I built, I was trading in the background. I was building algorithms, trading the market with my own capital, you know, since university days, right? So I wish you could change places with the people in university today, trading GameStop and stuff, or you think you would have blown out? Think you would have done well or done poorly? Oh, I mean, I think that that's it. I mean, I think that, I mean,

they have a very different mindset to something that I can kind of comprehend. I speak to people, for example, about crypto or these kind of various strategies that they have today. It's a little bit beyond me right now. You know, they're kind of like, for example that they have today it's a little bit beyond me right now um you know they're kind of like for example how they're doing how they're calculating their moves um i don't necessarily believe in it um but um you know but but i'm sure something good will come out of all of this by the way um i'm sure the next one some of that sounds on something that they're doing right now yeah let's hope so

they were just that data today the most option volume like ever on GameStop and that was a GameStop it was AMC I think like 10 billion call options were purchased I mean to be honest I mean I think the people that are really

making money out of this are the high frequency guys right I mean I know some honest, I mean, I think the people that are really making money at this are the high frequency guys, right? I mean, I know some high frequency guys in both, you know, both in the option space as well as the crypto space that are making a lot of money right now. And I've got a feeling that it's actually the high frequency guys that are, you know, the real money makers out of this whole, you know, GameStop, Reddit stuff. I mean, I think I'm not sure what the spreads are but i'm sure that they're making very handsome profit on it yeah i think we're going to get like a michael lewis book within the next 18 months right of like here was the true gamestop billionaire or something um definitely out there and then speaking of billionaires one of your former

fellow coders went on to be quite successful? What was that story? I mean, yeah. I mean, there's a gentleman called Cameron Adams that actually worked for me. And he was a designer. I mean, he's gone on to be a billionaire. He built a business called Canva. Canva. And what do they do?

They build, they have a platform for people to go in and and uh and design like invitations and things like that so it's an automated platform for them to build um you know for people to do like design um all online right you ever wonder like maybe that was easier if i'd gone that way and created something that uh right i could get a few million people to pay me 12.99 a month to fix. That could be another approach at the billion mark. Well, I mean, like, I mean, I did, you know, I did fund a number of startups in the technology space.

You know, I, you know, I thought that I would be a decent, you know, a kind of, you know, a decent investor when it comes to tech space. But I know it sounds very simple when you read the stories about, you know, people that first invested in Google or Facebook and things like that. But it's difficult. I mean, you know, building a tech startup is a lot harder.

I mean, there's a lot of very, very smart people doing it. And it takes, you know, I think it takes an enormous amount of luck as well as talent. I think, I mean, you know, I tried my hand several times investing in several startups and I do think it's a very difficult space. But sure, I mean mean you know i i i certainly do believe that um you know it's a great space for a lot of young entrepreneurs to get into

and i'm sure there'll be plenty more billionaires built in the tech space yeah we did a blog post once i think it was apple's stock price in the six years after its ipo or something and just kind of presented it like a hedge fund with no name or anything, you know, in the monthlies and the annuals, the drawdown, it was like negative 18% annual return, 62% drawdown. And I was like, would anyone invest in this? And the answer would have been no. Like, why would you touch that with a 10-foot pole?

But there were people who believed in that. Right then it was all about narrative. It wasn't about numbers at all. Just we believe they're going to get somewhere. So anyway, the interesting thing right now is because you know, so because of the blowout year that they had in the last, let's say in the last 12 months, there's an enormous amount of money that has gone into venture capital. And it's just one of those things when you do have that much money kind of,

you know, going into every opportunity out there, undoubtedly that money will generate winners, right? But, you know, the chance of picking those winners is difficult because, you know, you know, it's like everything else. It will have a boom and then a shakeout at some stage. A lot of reworks as well right right right so anyway so then you decided to launch uh tap fight so how do you think i want to get into the hedge fund space you said i want to be a decision, but that's a jump from there all the way to like decision maker

and managing other people's money. Yeah. So actually the process went, so I worked in managed consulting for about seven years. And then after that, I actually moved to China to build a packaging business, which is one of my dad's companies. And I was working there for about seven years as well,

from about 2007, 2008 through to about 2013. And, you know, the really interesting thing, I mean, I had been in China several times. I saw the opportunity. I mean, I wanted to spend a bit more time there. And so, yes, I was the CEO of MyPack for about seven years and I was expanding both the sales base. I was building the business in terms of the client base into Europe,

US, as well as Asia and Australia, as well as expanding production there. So it gave me an opportunity to actually build the business. I mean, by 2012, 2013, it was a $40, $50 million business in annual turnover. So, you know, we, you know, I mean, I really enjoyed that time in China. Where in China was it? It was in Guangzhou. So it was in south of China. Yeah.

So, yeah, I mean i thought that that was fascinating um and uh you know but you know after about seven eight years i mean i i i came home and you know for personal reasons and things like that got married and things like that and uh and decided to try my hand in uh in finance right yeah but it was that simple it was from packaging company to finance or were you yeah i mean it's a little bit more complicated i mean uh you know actually at the time i i moved back to to to melbourne you know for a girl right yeah um i mean it didn't work out but you know i moved back for a girl um there's always a girl. Yeah. But then I intimately met my wife and then, you know, and then I, and then I actually ended up settling down quite quickly.

So, yeah. So it was, you know, it was all that ends well, right? Yeah. Yeah. But were you reading like stories of all the, like, I'm forgetting his name now, but who, but who called the big short guys and all that? And you're like, OK, I got to look, the reality with China, right, is there is an enormous amount of opportunity in China, right?

It is a demanding place, you know, from I'm talking about everything in terms of work life, everything. I mean, people, you know, we were working until, you know, 1, 2 a.m. in the morning and then you're up at like 7 in the morning as well yeah um you know this is like seven days a week i mean so it does become very demanding um and don't forget you know i was traveling i mean i was on a plane i mean i was on an international plane probably you know five six seven times a month um and so uh yeah, I mean, it's just become very demanding after a while. And after doing that for like seven, eight years, you just feel incredibly. I mean,

I remember the time, I mean, I was dating a girl and she said, I've calculated something like in two years that I've been dating you that, you know, I've only seen you for less than a month, right? It just becomes very, very demanding from a work-life balance point of view and at some stage i think you just you decide that you know you know you you need to have some time out for a little while that's it yeah my wife and i joke she was a consultant i was the one sitting at home uh but before we got married we did the same calculation but we thought like oh that's why it worked. Like if she had to see me every day for the two years we were dating, she probably would have been sick of me.

But when she only had to see me like once a month, it worked out. So then you start Taffy in 2013. And you guys have gone on to win a bunch of awards. You've done great. So give us the elevator pitch on how the strategy works. We touched on it a little bit there in the beginning, but let's back up and, yeah, the elevator pitch on how it works. Right. So what TCM does is, so we invest in a portfolio of across global, all the, you know, the major global assets, right? So our portfolio is 100 to 200 positions, right? Across global equities, fixed income, commodities, and currencies. And this is across North America, Asia, and Europe as well, right?

And there's no, you know, there's no bias towards one asset class. So it's fairly well diversified across the four asset classes, as well as well diversified across the three major geographies as well. There's no one particular, you you know position that is more than or we say no more than you know five percent of the portfolio but generally speaking it's less than two percent of the portfolio um and so you know the idea is um uh having a very well you know diversified uh portfolio gives you effectively you know the best 100 to 200 um positions or the best 100 200 you know opportunities across across the globe right

um and it's a nice it's you know it's it's a it's a nice um kind of philosophically way you know to view the market that you're owning, you know, very high quality opportunities and continuing to rebalance into the best opportunities. I think a lot of us ideally, you know, would like to do this on a discretionary basis. You know, for example, we think right now, oh, right now would be, you know, a great time to own some Tesla stock

or some, you know, semiconductor stock or things like that, right, right now would be a great time to own some Tesla stock or some semiconductor stock or things like that, right, right now. And I'd also like to own gold and I'd like to own this to diversify and things like that. But being able to do this in real time is just very difficult to do because, you know, there's just too many instruments, right? I mean, we scan over 130,000 instruments to come up with this kind of optimal portfolio, right?

So I think that, you know, it's always something that, you know, we would like to do on a discretionary basis. I think everyone tries to do that, you know, tries to own a diversified portfolio that has great diversification. But trying to do this is a lot harder in practice than what we think it is, right? And it was a long process to actually get there. As I said, we started off in 2013 with this ideal that we could build something that would significantly outperform the S&P 500 and that would be something, you know, because a lot of things are happening in 2013, you've got to remember, right?

We were just recovered out of the financial crisis. We were actually just entering a new wave of digital exchanges, right? So a lot of the products, for example, that we trade today, we're only really trading since about, you know, we're only electronic trading since about 2000, know, we're only electronic trading since about 2000, say 2005 to 2008, a lot of them started, right? And the volumes didn't really pick up until maybe 2009, 2010, right? So we're talking about things like simple like Eurostox or E-mini or Hang Seng Index, things like that, or just electronic contacts?

I mean, no, I mean, if you look across, you know, say the 130,000 instruments that we trade, sorry, that we, you know, the investment universe that we scan, right, a lot of the data sets only begin from about, say, somewhere between 2005 to 2008, right? I mean, some of them do this, you know to about you know the late 1990s um or like the early 2000s but generally speaking i mean in terms of actually getting a very clean data set where the majority of trading was kind of done electronically it's kind of around 2005 and 2008

um and the main reason for that i mean like who you know how many of us you know were using a computer on a day-to-day basis you know pre-2000 i mean you know not trading off your phone or right yeah right right i mean we forget you know the iphone was only really launched in 2007 right i mean you know we had our blackberries pre-2007 but you know how many people kind of had blackberries pre-2007 and it wasn't a heck of a lot right um so this kind of wealth of data was kind of exploding into the market, right? So, you know, as a quant, I mean, as a quant system developer,

being able to, you know, have such a nice pure data set to learn from and to exploit was something that was quite new. And, you know, and a number of other things were happening i mean a lot of the spreads were coming down um a lot of the transaction costs were starting to come down as well that was opening up a lot of these systems um and yeah i mean you know it was really the idea that there was a large sea change happening in the world of finance and that we could still jump in now and be ahead of this curve, right? People were still only just starting to exploit this data

by using, you know, AI algorithms and a lot of machine learning kind of concepts and a lot of, you know, based on, you know, data science and information theory and things like that. So, you know, it was a fairly new space. But, you know, even so, um you know i think that a lot of the systems were still failing right i mean even if you look until today how many pure quant systems are really performing that well today right i mean so it's still a very very difficult system that even you know even you know seven eight years later that people are still kind of grappling with, right?

You know, we obviously feel that we kind of cracked that and we made, you know, several novel kind of advances in the field that, you know, haven't been done anywhere else, you know, for a few reasons. But, you know, it is still a very rapidly developing market. This confluence of using, you know, information theory and database concepts together with new information sources that are coming out. So lots to unpack there.

Let me start with, right, so I hear you say multi-asset, multi-sector, multi-geography, and my brain goes to, like, that's classic managed futures kind of global macro trend following, right? They have their across energies, growing on the ground commodities, stock indices, fixed income, currencies, yada, yada, yada. So how does it differ from that kind of approach, from that kind of diversification, right? I'm assuming that that's much longer timeframe, but tell us how that differs from kind of diversification right i'm assuming that that's much longer time frame but uh tell us how that differs from kind of a classic managed futures type global macro type strategy right so if you look at kind of yeah so i think that's right so if you look at the classic

global macro right so that's all built on uh harry maskwitz Maskowitz's market portfolio theory, right? Which is this general concept that if you build a diversified portfolio that is uncorrelated, right, you can improve the risk-adjusted return of that portfolio, right? But when Harry Maskowitz did it back in, I think, in the early 1950s or late 1940s, right, he obviously had a fairly limited data set. So he was just thinking about diversifying it with, you know, simple equity and bond, you know, index portfolio, right?

Right. You know, but what we have now is obviously a much larger data set that we can calculate that with. So that's one thing, right? We can think of uncorrelation in terms of, you know, a hugely multifaceted kind of, you know, portfolio that he couldn't even, you know, and then you couple that with the computing power that we have

to be able to calculate all these things in real time. We can calculate covariance and things like that in real time, right? And then probably the most game-changing factor is you combine that with the data science theories that we have today and the computing power that we have today, right? And that allows us to compute an enormous data set almost in real time, right? So what, you know, so our strategy kind of has two components. The first one is, you know, we analyze, you know,

data set of about 130,000 instruments, you know, tick data, but we aggregate it up to about but we upgraded up to about minute data. And we, from that, we create a probabilistic forecast around three things. We will calculate expected return, expected volatility and expected covariance. For each of those 130,000 instruments? Right, right. Okay. For the next day, the next hour, the next day, the next week, all of the above?

No, no. I mean, it's done for multiple timeframes out to about eight weeks, right? So, you know, it's done from about now out to about eight weeks over kind of, you know know across multiple time frames right yeah um so you know as an example for each price point in the future you'll calculate you know you'll assign a probability to each price point you know out in the future right um or to each you know um you know or you know to various covariance, right, between the instruments out into the future, right? So what you get is kind of this three-dimensional probability

density function, right, for each one of these instruments, right? And that's a function of not just that instrument's historical price and volume data, but it's looking at the relationship of that particular instrument, you instrument, price and volume, with all the other instruments in that data set. So it's a function of the other instruments in that data set. Yeah. Just I'll interrupt you.

So to me, it seems like we're saying a couple of things. One, like the traditional 75 market trend following portfolio, right? It's kind of restricted just by old habits or limited data sets, right? Like you're saying, hey, if I know those covariances in more real time and I can be more confident in the correlations between those, I could add, I could triple those markets. I could quadruple that data set, right? Without adding much more risk because I can predict the, right, if it becomes kind of one trade, I want to make sure I don't have more risk by adding more markets. So to me what I'm hearing is like the math has allowed you

to kind of add more markets with not much more risk, or that's the goal. That's right. That's right. And it's also about using these techniques. So the traditional sense is I use the historical data and I use the historical correlations to purely be a predictor of what the future correlations or the future returns will be, right?

Right. I only want one market out of the energy sector and one out of the grain sector and yada, yada. Right, right, right. So you're not just looking at the historical covariances or historical volatility and price data. You're trying to make forecasts of the future returns and the future volatility, not just based on the historical price,

but there's certain things. So, for example, you know, if you look at, you know, people in the technical, you know, in the technical realm, they can see, you know, even though volatility may be decreasing, that may or may not be a predictor of volatility could soon break out, right? And so it allows us to look at, you know, future volatility or future covariances in a much more sophisticated way to make more accurate

predictors of what they might be in the future right um and i think that that is quite kind of create do you view it as like it becomes like a single trade like here's the trade today with all these 100 instruments together that's going to produce this tomorrow or this over the next week and then it's dynamically changing each day of like we're going to produce this tomorrow or this over the next week and then it's dynamically changing each day of like we're going to add three here take away three here so it just is kind of this one single entity that has all the data running through it and predicting over the next 10 periods or whatever what what the next 10 are going to be that's right well you know it's a little bit difficult to really ascertain what the system

is is kind of thinking at any one point in time like yeah if i look at the portfolio right now um in some cases i can see that for example it's um it's kind of uh betting that you know some price is going to mean revert right and in some cases you can see that it's it's kind of betting that that you know some movement is very strong and it's kind of assuming that this kind of this momentum is going to continue upwards. So it's not, you know, it's not really clear, if I'm honest, you know, exactly why, just from eyeballing it, you know, if it's kind of more of a momentum-based strategy,

like you said before, if it's more kind of mean reversion. But, you know but it's just doing that to maximize tomorrow's risk-adjusted return right exactly exactly and you know i mean we obviously are looking at the portfolio you know every day all day every day and you know it is incredible mean, we are very impressed by the quality of the portfolio, right? One of the things that is scary, I mean, when we did start, the idea was, you know, it was a pure kind of data-driven system, if you like. And what was scary, I mean, sometimes a system could go, you know, 70%, 80% into one position, right?

Yeah, that's what I was just going to ask. Could it put on, like, every stock index in the world full long and that's the portfolio heading into tomorrow? Right, right. I mean, like, yeah, so exactly. I mean, at the start, it was actually getting quite scary. I mean, I remember, you know, you'd be going into some kind of, you know, event, right, some kind of release that was pending

and, you know, some election or something like that. And you could see the system was extremely long into one position. And it was actually quite high leverage as well. I mean, in some cases, like seven, eight times leverage. It was like 70% in one position. And it was scary. I mean, like, you know, we did get it right in some cases,

but you could see that it wouldn't take a lot for it to go wrong. And so after, you know, I guess the initial years, as I said, it was about, you know, paring it back. So what we learnt to do was, and it was, you know, through a process of, you know, maturity and maturing the system, was, as I said, we now build the system so there's no single position that's, as an example, more than, say, 2% of the whole portfolio, right?

And we build in these kind of these core fundamentals and we say that even though we trust the system to be making very good predictions of the market, we have kind of like five major kind of risk principles that we say the system can never do. So, for example, it can never invest in a position that's more than 5% of the portfolio. And generally speaking, it's never more than 2%

of the portfolio, right? Right. It can never have, it can it be biased in one particular asset class it always has to be well diversified among all four major asset classes right it can't have a bias towards one geography so it can't be all you know europe or something like that it has to be well diversified geographically um it also can only invest in highly liquid instruments so even though we scan a universe about about 130,000 instruments, our actual investable universe in terms of what, you know,

the instruments that it can invest in, generally speaking, I mean, this changes depending on liquidity day-to-day or month-to-month, is generally less than, you know, less than 4,000 or 5,000 instruments that we actually, that provide us with sufficient liquidity. In some cases, it's like it goes down to less than than a thousand instruments that we actually can invest in at any one point in time right um so we have liquidity constraints and then we also have leverage constraints you know the system isn't allowed to you know leverage beyond a certain point so you know to ensure that

you know that if there's some large left tail event that you know we're not we're also not you know highly impacted by that. And probably the last thing is the system needs to be neutral, right? And we say risk neutrality not in a traditional sense of risk neutrality in terms of your long or your shorts, but risk neutrality in terms of your, you know, your risk-off positions have to equal your risk-on positions on a volatility-adjusted basis.

So we kind of build these kind these risk tenets in place. So what we're saying is if you look at the major reasons that a lot of quant strategies run into issues, we want to take all of those off the table and then we want to use the best systems possible to then create the best portfolio after these principles are in place. Right. So like ones that have blown up have ignored one or more of those principles in effect.

Right. I mean, I think if you were to trace back as to why a number of strategies didn't make it, for example, to the 10-year mark or beyond is generally because of one of those things not going well. Right. And this just popped in my head, I have seven other questions here, but do you ever write the one thing you can't protect, protect against? There's like the death of a thousand cuts, right? Like, so maybe you've taken away the big losses,

but maybe it could lose many times in a row over weeks or months or right. Right. Right. I mean, you know, so maybe, right, that's preferable. Yeah, I'd rather have that than, than the other way. No, I mean, there is so the, you know, it's an it's a good point. I mean, right. So the, the I mean, everyone wants to know, I mean, what is the kind of the weakness, right? So the weakness is, right, like, like in any other strategy, right, is that all your risk-on positions and all your risk-off positions across, you know, across global markets and across all asset classes all sell off at the same time, right?

So if you were to design the event, that would be a real problem for you. You'd look for maybe a 30% to 40% move across all asset classes. So you'd want to see things like equities sell off and you'd also say you'd want to see equities sell off by 30%, 40%. You'd also want to see fixed income, like, you know, I'm talking about government debt debt sell off 30%, 40%. And you'd also want to see commodity sell off 30%. So you want to see precious metals, agri stocks. You'd want to see them sell off.

And you want to see safe haven currencies and currencies and everything like that all sell off 30%, 40% overnight all at the same time. That, you know. And when you're saying you would want to see, like if were trying to break it this is what you'd want i mean you're trying to break it right um and it would have to happen quite quickly because the system as i said everything's on a volatility adjusted basis say as volatility increases you know the system would naturally scale out of it right so it would need to the volatility would have to be also quite high so you'd need

this all to happen very quickly almost like a march of 2020 like many of those were in place there how do you define that risk on risk off? That seems pretty important, right? So is that the model defines it, the AI defines it, or you predefine it, right? Like I'm assuming you're saying risk on is equities, currencies, whatever, commodities, and risk off is long bonds, long gold. And then your derivatives, like you're saying, of agribusinesses and stuff like that. But so that's predefined or the model defines it? Yeah, so the model is calculating. So it's making three calculations for each instrument.

One is an expected return, the second is expected volatility, and the third is expected covariance, so the correlation between those instruments. So when we say risk on, risk um you know that's yeah number three was right you know the risk on would be exactly that you have certain you know you have things like equities and you know and certain commodities and currencies certain you know even certain bonds and things like that that are that generally move to a a more a risk-on kind of movement. And then you have the, you know, the assets that are correlated

to a risk-off. That obviously changes over time, right? Like you said, in March 2020, you kind of do have what we say happens kind of once or twice every year. You kind of have these systemic events where just everything sells off, right? You know, you just kind of have this rush to cash, right? It's kind of the end of the world.

It happens once or twice a year. Everyone's like, look, you know, I'm selling my gold. I'm selling my fixed income and I'm selling my equities and I'm selling everything, right? And everything just sells off. And, yeah, I mean, I think, you know, everyone just goes to cash, you know, for particular reasons. I mean, we will scale out to a certain extent because there's a jump in volatility, but it's, you know, we more or less will rebalance during those periods into things that are generally mispriced, very cheap, and that tend to kind of bounce there. So we don't sell off, we tend to kind of just rebalance down. And then when markets bounce back, we hope to bounce

back a lot faster than the market. And then, so say we had a bond driven equity sell-off, right? Rates rise, bonds are down and equities follow them down, which we haven't seen in 30 years, but right. That's a very real possibility. So eventually the models would recalibrate and say, bonds is no longer a risk off. Or that's a tricky proposition because even in that scenario, they might be risk off in a sharp downturn. But there could be months long or years long downturn in both prices where the model may say, all right, that third leg, that third pillar, the covariance prediction is saying that's no longer a risk off market. Right, exactly, the 1970s. I mean, it has happened, you know, for short stints over the last, say, 10 years.

But it's the first time that we've really seen a clean data set into what price moves might eventuate from a potentially high inflation scenario, right? You know, I am not convinced that inflation is going to be necessarily an issue in the short run. Certainly not for the next two to three years. I think that, you know, debt globally is just too high. Governments would just run out of money to continue stimulating the economy to create the inflation. I mean, I think that governments would love to create inflation if they can,

but whether they have deep enough pockets to continue that inflationary stimulus, I think that is arguable. You said Janet Yellen coming out and making a point about that. Jerome Powell also mentioned, you know, he questions whether this inflation rally really has legs or not. I think a number of other people have questioned the real possibility. I mean, it's always a risk.

I mean, you can never discount the risk of inflation. But, you know, it's certainly probably a little bit, it's not a worry that I think that, you know, maybe it's been exaggerated a little bit recently. But I think the interesting thing is exactly what you said, Jeff. There is the possibility that inflation-led downturn that could see, for example, bond yields rising and bonds falling at the same time as equity.

I mean, if there is very high inflation, it could cause quite a large rotation in equities and even a downturn in equities as well. You know, we speak a lot about this internally, you know, what, if anything, should be required in the strategy change. Actually, in the recent sell-off, you know, we did make some adjustments to the strategy. Now, we don't really get involved in, obviously, what the portfolio is choosing, but we can make certain decisions, right? So, for example, you know, I can't, you know,

without really talking too much about it, I mean, one of the things that we do is, so I spoke about, for example, there being four major asset classes, right? One of the ways that we can do it is, so as I said... What are those before quickly? Equities, bonds? So global equities, global fixed income,

commodities and currencies, right? Right. So one of the things that we do do, for example, in the commodity portfolio, at the moment, it actually does have quite free range within the commodity portfolio to allocate. But the commodity portfolio is actually quite diversified. There's obviously agris, there's, you know,

there's precious metals, there's base metals, there's, you know, there's, you know, there's a number of kind of subcategories underneath that kind of commodity category, right? So one of the things that we have done is we've kind of, you know, asked the system to have a kind of a minimum kind of weight in there towards, for example, precious metals, right? And what we did find by doing that is it actually does improve the inflation robustness of the strategy over the long run.

And we were only able really to achieve that by looking at the price moves that have happened over like the last six months. It's a little bit more sophisticated than that but you know what we do in a system that actually makes it more robust inflation is actually quite interesting as well. And that's a good point though like you just went through all that stuff like Yellen saying this, Powell saying this but really the model doesn't care what you think even though you right even though you've designed it you're not going in there and saying oh pal's wrong on this i'm going to code up some stuff to get you know more long bonds or more short currencies or something right so the model's

doing it what it does regardless of what you think right right right i mean look um you know we do have a certain, you know, background in terms of finance and, you know, we do look at the portfolio on a day-to-day basis and always want to make sure that it also always makes sense to us as well. And so if we feel that there is, you know, something that looks very imbalanced,

we'll obviously go back to the drawing board and say, you know, is there a sixth or seventh principle that we've missed out on here, right? And it's a little bit like if you were trying to design, you know, a self-driving plane, right? So a self-driving fighter jet, right? Would you just look at flight data, right? And, you know, and use that flight data to design that fighter jet autonomous software. I think it would be a combination. You'd also want to talk to very experienced fighter jet pilots, right? And you'd also want to get their input and say, well,

is what this jet doing here, does this also make sense, right? And so the idea is we also want to have that check and balance. I mean, it's one of the dirty secrets, right, of the quant industry, right? We actually don't have a little bit what we touched on before. We actually don't have that much data, right? We only really have about 10 to 15 years of really clean data to work from. So you have to look at that with a certain amount of humility

to say what we are building on is a prediction that the next 10 to 15 years will look fairly similar to the last 10 to 15 years. And, you know, the reality is that's probably not correct, right? I mean, the markets are going to evolve in quite a significant way that we need to, you know, really think about eventualities that could occur that the system is just not prepared for, right? Yeah, and I jotted that down before when you were talking

of data only back to 05-08, right? Like that's a huge hole, right? There's for sure things that have happened and patterns that have existed that aren't represented since since that 05-08 period. But so I like that. That's the human interaction. Right, right, right. I mean, you can get data older than that, but it wouldn't necessarily be very clean. You know, if your quant system is working from from for example tick data i mean the the liquidity and how much the markets have evolved you know since pre-2005 is significant i mean if you think about

back in the 1980s when people are using um you know hand plotting you know back to the you know i remember you know watching that special on like computer drones and portuda drones was like kind of getting someone's like chart each dot back in 1986. so we've had a few of them on the pot here tom basso was a couple of years ago and he used to do it by hand yeah right right right so when you think back at those days and you think about you know people were using these kind of technical patterns that were very profitable in those dates right now if your system was learning back from data that was in the 1980s where, you know, simple technical patterns were working, I just don't know how valuable that would be to a con system anyway, right?

Because a lot of those pricing patterns would really not be profitable, you know, in today's market, right? I mean, even if you look back, you know, the patterns that were profitable five years ago are not profitable today. I mean, how many high frequency strategies, you know, have, you know, even the last two to three years have stopped working, right? So, you know, the markets evolve, right? So to a certain extent, it's about lack of data. And, you know, the other part is just evolution of markets overall as well. And which, yeah, like that's a, how do you square that, right? Because it's some firms, some con firms are like, no, anything past the last three years or five years irrelevant, like you're saying, right?

I think of it as like, if you're an alien and you go back in time and you're, or you come from back in time to today and you're tracking a kid, right? And you're like, okay, where's that kid going to go after school you're like oh he always goes to the comic book store except here in 2021 where there is no comic book store he's going to look on his phone or he's going to do this so to your point of like even if you know these really old patterns they simply might not exist because of the modern market right right right well i think it's a little bit like you know the way that we would look at it as a human being right there would be certain things that we can glean from the past right so if you look at for example people are now talking about the 1970s and what happened in the 1970s with inflation so people can look back and say look there are some macro

things that we can learn and there are ways that we can use historical data to back test our strategies and to kind of predict what would be expected behavior right and we can test to see you know can our systems you know respond to a large data set right i mean if if you know even as markets evolve you know, we would like to have 100 years of data, right, that we could run all our systems through and be able to backtest them through 100 years of data. Now, we could say back in, you know, back in 1921, markets were completely dissimilar to what they are today, right? But we would still love to be able to get that 100 years of data set to be able to run our systems through. That would still be a good test for the robustness of our system, right?

But, you know, so there'll be things that you could learn from that entire data set, but also there'll be things that you would incorrectly take away from that, like, you know, in terms of how much the market microstructure has changed, right? And, you know, so it's about being able to think about those two things at the same time. But does AI, like, self-reinforce itself? So, right, if it learned, say you did have those 100 years,

would it start to teach itself of, like, the last five years are way more valuable in the model and just pick and choose or it's always going to take the whole data set. Yeah right I mean so definitely the more recent data is obviously a lot more valuable like even a lot the data over the last say you know two to say in the last eight weeks as an example is but it... But is the model dynamically adjusting how much of that data it's kind of, quote-unquote, using? Right, right, right. So it's continually updating it.

So if there is a very large move in the last eight weeks, we'll certainly be weighting that considerably than, you know, what happened, for example, five, ten years ago. Got it. And then I started thinking about, you know, the AI and the Go that beat the Go masters, the game. You've read that article, I'm sure. So, right, as I'm thinking about it in relation to this,

most quants, it seems, are making, if they were the go engineers for that AI, they were just teaching it on all the players' past performances, right? Which is kind of in the market. Look at all the past market performance. But the AI, when they actually, the one that beat these guys, ignored all that. they just taught it the rules and it came up with its own strategy so do you think that's ever viable in today's

world or you're that's you're saying if you just let the computer go and said make as much money as possible tomorrow right it's which it seems like what they did with go like here's the rules of the game just go win the game um do you think that's possible or that leads to weird places that leads to blowouts and whatnot i mean i think that that's a very idealistic way of viewing the market right that you know like i mean whilst you know go is i mean i personally don't play it but from what i understand you know it's a it's a understand, it's an NP-defined data set. So there are an infinite set of moves, and it is an extremely complicated game to beat a human being in. So if you think about the advances in AI recently, I mean, I think that that is very significant.

But that is very different to the market, right? The market has a far more complex set of moves that it can move in than even a go board, right? So being able to- And a design win, right? Like how do you win the game do you win next minute right right right right so i think it's um i mean i think i mean you know i think everyone that that reads that would think you know there must be a way to kind of apply this to the market um i am you know we personally can't see it but we we're not necessarily, you know, we're certainly not experts in that space. But, you know, gosh, I'd love someone to prove me wrong.

I mean, I'm sure someone is working on trying to, you know, use those techniques and use it to kind of beat the market. I wish them well. I mean, our system is, you know, is not about using the most sophisticated and the most cutting-edge AI. We think of this a lot more like an engineering problem. Like, so if you look at, you know, if you look at, say, an iPhone, right, now, you know, the latest, you know, say an iPhone, iPhone 10, right, is,

you know, is not that, you know, if you look fundamentally from, say, an iPhone 1 to an iPhone 10, right, each new iteration of the iPhone, right, was an engineering problem. Like, you know, it wasn't that large an upgrade from, say, 1 to 2 and 2 to 3 and so on, right? was an engineering problem. Like, you know, it wasn't that large an upgrade from, from say one to two and two to three and so on. Right. But if you were to pull out an iPhone one and compare that to an iPhone 10, I mean, that is a dramatically improved device. Right. And I think order of magnitude different. Yeah. Right. You can say the same thing for cars, right? Like if you look at the latest, you know, Mercedes-Benzbenz right and if you were to compare that to the same mercedes-benz like you know 30 40 years ago right it looks vastly different but if you look at the new release that they have on every model it's actually not that large as you know some kind of incremental engineering improvement between every release

right and that's pretty much the same thing when you're developing quant systems, right? You're looking at the system as a complex system, and you're looking at what other things that we can do to improve on that system on a day-to-day basis, right? It's not about getting, you know, a NASA scientist to work on a turbo boost to your Mercedes-Benz or something like that. It's not about coming up with some, you know, quantum mechanic system to your Mercedes-Benz or something like that. It's not about coming up with some quantum mechanic system inside your Mercedes-Benz. It's about making these incremental improvements over time. And you trust that as an engineering problem, as you make continuous improvements to that system, that the system improves over time time and a lot of the work and building systems it's

really about grunt work right it's about making a lot of small incremental changes over time that doesn't sell as well it's not as exciting except if you i mean i wish that we had some quant computer that could you know that could calculate every possible move in the market and solve it but i mean i i'm i'm actually i'm skeptical that that's that's even possible did you do you watch the series westworld um i do i do actually i think that latest season right they had the main the thing that could tell exactly what each person was going to do in the future and that they were taking them out i think we talked about this on the phone before the david brailsford the sir david brailsford is that his name the cycling guy in england um but he he wrote this whole paper on that their cycling

team coming into the london olympics and it was this whole concept of like no we're not just going to be like we need faster people we're going to the, you know, the caliper on the wheel a little bit better. And we're going to make this wheel design and we're going to make the handlebar a little bit. So all these incremental changes, the next thing you know, they were winning the Tour de France with Team Sky and they did well in the Olympics and all that. So we'll put that in the show notes. It was an interesting piece.

I can't remember the name of it right now. Yeah. I mean, I think, look, the interesting thing about finance i mean people will always want to know you know is there one system that kind of beats it all um i mean i'm not i'm not convinced that i mean the markets are so so sophisticated right there's certainly going to be you know a whole basket of players that that win right and there'll, a whole basket of players that win, right, and there'll be a large basket of players that lose, right. The idea is you really want to be just in the top quartile and be that over a long period of time.

I don't think that the goal necessarily needs to be, you know, number one every year, you know, number one every year, you know, to be the winner. It's not really a winner-take-all market. It's a market where if you aim to be in the top quartile, you know, every, you know, say every year for like Warren Buffett did for like, you know, I mean, he did it in like 60 years, but, you know, for a long period of time, we'll come out on top. And I think that that's more but, you know, for a long period of time, we'll come out on top.

And I think that that's more realistically, you know, what the goal is in building strategies, right? Right. Just land the plane every time. Don't crash, right? If you crash on your last landing, no one will remember the 300 ones before that. And then one last thought on the whole AI and the computer.

So I've always had this thought, right? If everyone's using AI, if you have essentially the same software and same hardware moving towards that where everyone has the same stuff, do you get to some utopia, or I don't know if it's a utopia or dystopia, where everyone's models are kind of coming up with the same stuff do you get to some utopia or i don't know if it's a utopia or dystopia where everyone's models are kind of coming up with the same answer right like you have these principles the next guy might not have any of the principles but do the machines eventually become smarter and smarter and smarter where they all are arriving at a very similar answer to the market problem um yeah look i mean people you know the whole field of kind of data science or you know information theory and and kind of machine learning and things like that i mean this is a huge field of statistics right um so you know there will always you know if you think of it

as kind of this immense field right um uh yeah i mean i think that there will always be people coming up with different ways to to kind of model the markets and to make money off the markets i don't think that there will be necessarily one model to kind of rule them all so i think the the likelihood of you're kind of one model but if it's like a set of 50 principles becomes right bubbles up the same way diversification and and this and that has today excuse me but yeah you you've answered that 17 different ways so i'll let you off the hook on Any other thoughts before we go into our favorites? I did have one more thing, unless you had something. So the 130,000 instruments, does that include indicators

like employment rates or cost of crude production in the North Sea? Does it include things like that? Or no, it's all just market price action. Yeah, so it's purely, we're only using the price and volume data for those instruments. And I guess the premise of, well, the premise of information theory is that all information in the market is actually reflected in the price and volume data across the markets, right? So if there is some kind of, you know, employment data or some kind of news event, that will all be reflected actually

in the price and volume data across that data set, right, across that global data set. And, in fact, it's actually a purer way to see that data, right, than actually looking at um then you know trying to look at you know for example employment or pmi numbers or things like that yeah so and that to your point of there's not one model some other quant shop sitting there looking at all of that data and saying i'm going to predict what that data is going to be and where the price is going to move once that data comes in so there's two ways right right in that right right right then you talked a little bit about

the improvements what was that just getting those principles in there or there was something else when you said there was kind of these improvements back two three years ago yeah so i think it's a number of things. So, you know, we made a fairly large pivot in strategy back in around 2018 that was more or less driven by trying to, you know, reducing the leverage of the strategy, reducing the concentration of portfolios that they said rather than having 70%, 80% in one position, we want to reduce that down to, say, 2%, having a lot more positions, say, about, you know, 100 to 200 positions in the portfolio.

So we wanted to, you know, and increase diversification across both asset classes as well geographically. So it was about, you know, really enforcing a lot of fundamental kind of financial principles into the strategy. And, you know, we tried various things that, you know, we just felt limited success. So we were trying to use, you know, stop losses at one stage. We ended up not using them.

You know, we were trying to reduce the downside volatility before kind of waving the right white flag and deciding that that wasn't the right way to go, you know, because, you know, various volatility in the market has increased in the last few years that just makes using, you know, stop losses quite destructive in value. So, you know, we had to make various adjustments to the strategy which is a humbling experience but i think it's important to do um and you know certainly what we what we have as i said

it's an engineering process and what we have now is you know it's a system that we're extremely satisfied with right um you know i have you know um over five million you know us dollars of my own wealth you know invest in the strategy um and i do it because you know i don't invest in any other strategy i'm outside of a few houses that i own it's pretty much all my liquid net wealth and i invest in that in the strategy because you know i know that this is a very high quality strategy now um but it did take time you know it does take time to kind of build these strategies is what we've learned. And the strategy is likely to continue to evolve

as the markets evolve. But we are very comfortable with where it stands right now. Right. And it seems like everything we've talked about is don't, you know, invest because you think the strategy as it stands today is the end all right like it's exactly what we've been saying there is no model that will solve everything but invest in the group that you think is going to you know have the good process and has the machinery

and the tools to keep having it evolve and keep having it be dynamic i mean i think that you know at the start it's a little bit ironic but at the start we had a very you know we're trying very very sophisticated things at the start and um i think the reality is right now we actually think of the markets a lot more simply than what we did at the start, ironically, right? We were trying very sophisticated things at the start that, you know, arguably, you know, actually built in a lot more tail risk. The way that we view markets right now is taking away a lot of that early complexity, but what we learned was that it actually makes it a lot more robust to external shocks. And, you know, and based on the models,

it's just a lot more robust a system. And then I want you to apply your principles to your own net worth because you have a lot of exposure then, right? You're saying like um the the concept of diversification you're like oh but that's twofold right you don't have any diversification in your investments except that inside that one investment is a lot of diversification well yeah i mean like you know it was built as a fully diversified portfolio, right? So if, for example, this was, say, a pure kind of commodity CTA that was only invested in, say, certain commodities,

then sure, I mean, I would probably want to have a separate kind of, for example, equity portfolio, things like that. But the idea around this is that it is so well diversified in just, you know, 100 to 200 opportunities that it doesn't require kind of an additional portfolio out there because, you know, it's already optimally kind of diversified. So favourite, we'll just go quick by here favorite australian athlete oh gosh got me that's a hard one i mean you know i you know you know i i'd probably go um

steve smith you know the captain of the oh no longer the captain but you know australian cricketer um you know i kind of grew up playing cricket so yeah probably steve smith love it all right then my next gonna be favorite melbourne sporting event don't they do an f1 race there and they have the aussie open what uh yeah you know if i had a favorite probably be the boxing day test match once again i'm kind of into my cricket i've never been to a cricket match. I highly recommend. Don't they last like days?

Yeah, so it kind of lasts up to five days. So, honestly speaking, if you're not really into cricket, you might find it kind of goes on a little bit. Five days? But still, I want to mark it off the list, right? I got to at least go for a few hours one day see what it's all about um favorite thing to do in singapore when you're back in singapore uh there's so many things maybe just going down to you, the beach club or something like that, you know, just spending time in, you know, the bars or just whatever. That's what you do in Singapore. It's a relatively small city, but there's lots of things to do.

And they open back up pretty much all the way? So, actually, they're in lockdown right now as well in Singapore. They're hoping to kind of open up in the next kind of week or two. So we'll see. I mean, you know, people are fairly optimistic. I mean, I think cases are down to about, I think, about 10 cases a day. So we're also expecting a war to open up.

I mean, for us, it doesn't really make that much of a difference because we kind of, you know, we're all sitting in front of our computer screens most of the day anyway. So it doesn't really have that much of an impact for us i mean other than you know kids going to school and things like that um is there a favorite book or podcast or other resource of someone could learn more about ai and machine learning techniques and all this actually one of one book that i actually learned a lot from um was uh it was this uh well not but i i found it quite interesting um was i can't remember the name it's a book on jim jim simons on on renaissance technologies oh yeah um i can't remember the man that

the man that something i got it The man who solved the market. Yeah, yeah, yeah. I actually enjoyed reading that book. Yeah, the great part in there, right, is they were losing money and they just had something wrong in the code. Remember that little section? Yeah, yeah, yeah. We switched the sign on the e-mini price or something

and they fixed that and it was like... I i mean i think it's one of the things that you kind of learn when you when you build this business right that you realize that even jim simons for you know in the first 10 years there were many times that he almost lost everything as well um and you realize you know it's not just about you know he it wasn't like you know him he employed, you know, it's not just about, you know, it wasn't like, you know, he employed some, you know, Fields medalist or something like that. They kind of broke it, right? It was about getting, you know, a combination of people that managed to solve, you know, a series of problems

that eventually got it right. You know, it wasn't because he had the smartest guy in the world working on it. It was, you know, it's about the humility of the markets or the humbling experience of building strategies. And it's about the persistence of going through that, of weathering like huge drawdowns in the first few years,

but eventually kind of cracking. And, you know, like what they say, most things, people kind of give up. They kind of give up just before they kind of get the breakthrough. And I think for us, I mean, normally in normally in the experience you know when we went through various drawdowns you almost give up but you realize by persisting through you kind of you kind of finally start to get to these major breakthroughs right um you know i'm not to say that we're going

to be the next renaissance i mean i hope so but you know it's hard to come for anyone to kind of make that claim but um you know we certainly do believe that the right kind of dedication and persistence to the job is certainly the reason why we are getting the results that we get love it and then finally ask all our guests favorite star wars character yeah i mean i'm not really a star wars fan uh so it's a bit hard to say. But, you know, I think that there's a lot of kind of lovable characters in there. I mean, you know, it's kind of a strange one. I mean, I actually find myself kind of drawn to kind of, you know,

Darth Vader. I mean, I don't know. It's kind of a strange thing. But, you know, I think that, you know, he has a certain kind of charisma on screen and, you know i think that you know he has a certain kind of charisma on screen and and well um he's the ultimate right all the good movies have the guy who can't be beat right jaws um all those right it's like if he touches you basically you're dead um so that in those movies right that was the thing of like, Oh my God, this guy, this thing is unstoppable and unstoppable for us.

All right. This has been fun. Thanks. Hopefully it didn't take up too much of your day there in Melbourne. I'm starting to lose my voice midnight here in Chicago. So I'm going to get to bed and we'll, we'll talk to you soon. Great. Thanks for your time as well, bed and we'll talk to you soon. Great. Thanks for your time as well, Jeff. And we'll catch up soon. Thank you.

The Derivative is brought to you by CME Group. CME Group is the world's leading and most diverse futures and options exchange. For more information and educational resources about futures and options, visit cmegroup.com. You've been listening to The Derivative. Links from this episode will be in the episode description of this channel.

Follow us on Twitter at rcmalts and visit our website to read our blog or subscribe to our newsletter at rcmalts.com. If you liked our show, introduce a friend and show them how to subscribe. And be sure to leave comments. We'd love to hear from you. Thank you.