The Derivative - Modeling Markets and Accessing AI with Robert Rotella and Jag Prakasam

Episode Date: March 18, 2021

When you hear news about artificial intelligence (AI), it might be easy to assume it has nothing to do with you. But if you use face recognition on your phone? Check. Google auto-fills your search bar? Check. And increasingly – the asset manager you may have managing your money is using some combination of machine learning and AI. Today’s guests were early adopters of that tech and applying it to trading algorithms. Jagdeesh Prakasam, Chief Executive Officer and Robert Rotella, Founder & Chairman of Rotella Capital Management have been in the game for a combined 50+ years. Today we’ll be getting into all the mechanics behind AI and Machine Learning as well as Robert’s beginning at Commodities Corp, mass customizing machine learning, how a lot can go right and a lot can go wrong in AI evolution, weak learners, early machine learning models, human bias in machine models, machine learning natural limits, systematic trend following, the balance of math and art in AI, Rotella’s models, and trying to stay on the leading edge of technology. Chapters: 00:00-02:39 = Intro 02:39-32:28 = Two Chemical Engineers Walk into the Trading World 32:29-48:16 = Rotella as a Whole 48:17-01:08:20 = Digging into the Models & Flight to Safety 01:08:21-01:18:10 = Q-Deck 01:18:11-01:35:15 = Machine Learning & A.I. 01:35:16-01:39:48 = Favorites Check out the COVID 19 charts mentioned in the podcast and follow along with Rotella Capital Management here. And last but not least, don't forget to subscribe to The Derivative, and follow us on Twitter, or LinkedIn, and 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. So in other words, the models that I was using back in the 80s and 90s, they don't work anymore. And so then the question becomes, well, why is that the case? And that's because the fundamentals, I believe, have changed dramatically. I mean, you've got an on-market that's at all-time low yields in the entire history of the country.

And you can't expect the same results because of that. That's where I think, that's where I was talking before about how you need to also have a qualitative analysis of the whole situation as opposed to just thinking, well, this is going to, trend following has worked in the past, it's going to work in the future, and that's it. It can work in the future. I believe it will, because I think the trends will continue, but it doesn't mean you're going to get the same results as you did 10, 20 years ago. Happy St. Patrick's Day, everybody. Although it'll be a happy day after St. Patrick's Day, everybody. Although it'll be a happy day after St. Patty's Day once you're listening to this. I'm wearing my green for everyone regardless and have a little bit of the luck of the Irish with us today with not one but two great guests.

Number one is one of the OGs of systematic trading model research and development, Robert Rotella, who is one of the early traders at Commodities Corp and founder of the cleverly named Rotella Capital Management. And number two, sorry, sorry, Robert. And number two is actually the new number one at Rotella with CEO Jagdish Prakasam. Did I pronounce that right? Oh, that's perfect. Who we just called Jag when we talked to him. But Jag joined us as well, who came up through the ranks of Rotella from researcher in 03 to its CEO today. So we're here to talk about what it was like back in the 80s and 90s, as well as what it might be like in the upcoming 30s and 40s with the advent of machine learning and AI and all that good stuff. So welcome, guys.

Thank you for having us. Yeah, thanks for being here. Happy St. Patrick's Day. They do anything out in the Pacific Northwest or not as much as Chicago, I'm sure. No, I didn't notice anything in my area. Not really. It's not like Chicago downtown. I know how the river gets dyed green or what have you. Nothing much happening here. Yeah. And you both had spent time in Chicago, right? Yes. Yeah, actually, it was a great experience in Chicago. And we were at that time named the Sears Tower. I think the 91st or 92nd floor. It was an interesting experience, as well as the Prudential Building. The us,

true Chicagoans still call it the Sears Tower no matter what it's called. It's technically I guess Willis Tower but then they were talking about even renaming it something else. Yeah it was quite a view from from I was almost like being in a plane sometimes and some of the storms there you were in the storm sometimes and it was quite an interesting experience. And I've heard from people it'll actually sway yes it can sway as much as my understanding uh easily uh 12 inches uh and uh there were some people that actually had a lot of trouble with that on a very windy day they they just did not like being there i actually am comfortable with the sway, but some people, they prefer the rigidity, I suppose.

That's like classic Nassim Taleb stuff, right? Like you want it to be not perfectly strong and to withstand anything, but anti-fragile, not strong. Were you out of there before 9-11 or were you in there after 9-11? Thank God. Yeah. More luck than anything, but we got out of there before 9-11? Were you in there after 9-11? Thank God, yeah. More luck than anything, but we got out of there, I think, in 99. Not exactly sure, but – and then that was not the best place to be after that happened. Yeah. Scary.

Yeah. And so, Robert, tell us a little bit, you started out with Commodities Corp, which is well known for, uh, a bunch of the hedge fund managers that have shot out of there. Um, how'd you get involved there and what was it like back then? Uh, I would say that a very important aspect of, uh, Commodities Corporation was Elaine Crocker. She was the president at the time. And she went on to become president of More Capital. But at the time that she was at Commodities Corporation, she did a lot to, I think, improve the success of the company by hiring some very, very good traders. And she also ran the company very well, too.

I think she was an important part in helping me to succeed. But if you want, I can start just a little bit back before that. Sure, yeah. Because there was a lot of history there and I started trading I guess around 1980 just getting out of college and and virtually all my trades were losers and I was going off a fundamental trading reading the paper and so on and finding out that what you read in the mainstream media wasn't necessarily going to get you rich very quickly. And so I started realizing that if I

did the opposite of what I actually thought I should do, I actually might be more successful but that's not an easy thing to do psychologically and so um i went on to the floor of the new york futures exchange it's around 1984 and um i was yeah new york futures exchange i think that even predates my now what were they trading there yeah that was it was called the knife future and it was a future new york stock exchange and basically it pretty much correlated with the s&p 500 god so not the new york border trade sugar cotton and stuff but that's the knife all right i think i have traded knife way back when yeah exactly and uh i found out that I was a really, really bad pit trader. And so I realized that if I wanted to succeed, I'd have to do something else.

And that was the beginning of the computer age, 1985. And where I think I was very fortunate is that I was able to, I realized there was a lot of, you have to understand, people were trading like off of Quotron screens and basically just by the seat of their pants, looking at the data on the screen and then making a decision on how to trade. And what I started to do is, because I had an engineering degree, I had programming skills and I started looking at the data and analyzing it and doing pattern recognition at that time and so I was probably one of the first people to incorporate pattern recognition

and but doing that on a systematic basis using programming and doing statistical studies to try to develop trading systems. And then your engineering was from RPI? That's correct, yes. Yeah, I have to throw that out there because I went to Union College, so right across the river there. Not far away. Somewhat rivals, but in hockey. Hockey, yeah.

Anyway, sorry. Yeah, in chemical engineering. But I will say, and that mathematics helped me a lot, the programming there, but also the statistics course that I took in my graduate degree at getting an MBA that helped as well. Understanding that things are probabilistic, especially in the markets. So I started programs in trading mechanically as opposed to on a discretionary basis. And lo and behold, I found out that if you did things without too much overfitting of the data,

it actually might work and um so in 85 i went on my own uh as a four member uh trading and then uh so went to and then started trading the s p 500 futures bond futures and uh went on from there i was making like 100 a year 1985. That's not braggy. I'm just saying that's the way the markets were moving at that time, 85, 86, 87, 88. It was really an incredible time in terms of trading, a lot of movement in the markets and a lot of clean directional movement as well. And then I guess it was in 89, I broke even, but I had a $50,000 account and I was making 100% a year in New York City after taxes, that doesn't get you anywhere. So $50,000. So basically I had to go get a job again and I became a clerk on the floor. And that's when I sent my resume to Commodities Corporation. And fortunately, they looked at it.

They came and interviewed me at my house or my apartment, should I say. I was going to say, you weren't doing too bad if you had a house in New York. No, not exactly. So anyways, they interviewed me, and they decided to give me $50,000 because they felt it was sufficient because that's what I was trading in my personal account. Unfortunately, I had lost all my money because I was really happy that they supported me because my father was a school teacher and my mother a housewife. They didn't have a lot of money themselves. And so they let me about $50,000 and that kept me going. And that along with the allocation from Karate's Court, I was very fortunate to be able to trade and do well for them. Yeah. So a few things. One, I'm just interested.

When you first started saying, I'm going to do this mechanically and statistically, there weren't any programs, right? Like you had to actually program yourself or there's some early things where you could run statistical analysis? You know, there might have been, but they weren't sufficient for my needs. Or they might have been a million dollars a year or something like that. Whatever, yeah. And so, yeah, and as you probably know, when they have these programs that they try to be generic and work for everybody,

but then you have specific needs and they didn't meet the needs that I had. So you're correct. I was doing programming in Pascal and in C. And at first I started with Excel spread or Lotus 1, 2, 3 spreadsheet, I suppose. But then you found out pretty quickly that that didn't meet your needs. And so you needed something like a much more powerful programming language. So yes, I did all the programming myself. Wow. And then secondly, I'm just thinking back, right? Early 80s, John Henry heyday, the turtle traders were making hay.

Did you know of those guys and know of this systematic kind of trend following world? Or were you kind of approaching it independently on your own? I knew virtually nothing about it. And just to give a little background, as I mentioned, my father was a school teacher. I came from a background of nobody ever worked for themselves. They all worked at a factory or they worked as a school teacher. That was it. That was the only option you have, or maybe like got a clerk at a store or something to that effect. And so I had nobody that could give me some kind of guidance in terms of help in either starting a company or in trading. And all trading was to me was something like a brokerage

firm. And that I had no idea that there was another aspect to it until I read Reminiscences of a Stock Operator by Edwin Lefevre. And that was probably one of the most important books that I read. And where I found out that there were actually people that made a living trading the markets. Stock operators. Exactly, yes. And that was basically a book probably many people don't know about Jesse Livermore, famous stock trader in the 20s and 30s. It's been mentioned on this pod a few times as people's favorite books. Yes.

I read it long, long ago, but yeah, it's great yeah that and extraordinary delusions of a madness of crowds uh by i think mckay that's a very good book too as well it needs to be uh they should come out with the updated 2021 version and include crypto and nfts now you realize things haven't changed yeah um and then i want to come back and ask you some Commodities Corp questions but let's bring in Jag and Jag tell us a little bit about your background. So my background actually is at least education wise it's similar to Robert as well I have an engineering degree in chemical engineering and I developed an interest in programming primarily to build some neural nets to apply to heat exchangers and that was my original thesis I you know I wrote a paper on it and

things like that but in finance I mean in chemical engineering one thing you realize is things move really slowly I mean if you can implement one algorithm in a decade, you know, you're doing pretty well for yourself in some sense. And that was the state anyway then. So I implemented that. And then I realized I was looking to move to a place where I can analyze more data. And finance was a natural fit at the time. And I applied for a program in Chicago. I did my master's in financial engineering.

And I started as an intern, actually, at Rotella Capital and analyst. And since then, I've been trading a variety of different portfolios and futures as well as equities. And 18 years later, under Robert's mentorship, I'm still here. I think alongside him and benefited from his mentorship. And on that note, I should tell you, I don't know if Robert remembers this story, but I got to tell you one of the early, early days, I, you know, I was trying to build a system. And until then, I was supporting the research rather than building my own systems. And somewhere around 05, I started building my own systems. And I, I created the system, the very first day I started building this, I suddenly saw a sharp of too sharp.

And I was so excited. You couldn't believe it. I took a walk out. It was, you know, it was a beautiful day in Washington. I took a walk out and grabbed a cup of coffee and came back in just to think, I don't know what happened. The first day I hit it, you know. And then, of course, Robert walked by and he came to my office and he got behind and said, Oh, let me check if you have your slippage numbers right.

He clicked that button into that. And that model fell apart. It was so bad. I said, Oh, my God. I thought you were going to say he was going to come to you and be like, It's not bad, but we need a sharp of eight in order for this to be up to seven in real time. I'd be happy with one. But I'm saying like the difference between the test and the reality is like,

I need it way higher in order to end up at 0.7 or something. Yeah, it was 18 months after that, before I could come up with a tradable, you know, model. But that was, yeah, there are some good learning curves in this business, especially when you go from a different industry into finance. And why, any commonalities you think? Why both chemical engineers? So you're writing the formulas or what's special about chemical engineering? Anything or it's just your brains work similarly?

I don't know. I think it could be just the training of an engineer itself, probably. You know, we look at a system and we look to optimize. Is there a more efficient way to, you know, capture these relationships? That could be it. Like, for example, in the heat exchanger, why did I apply neural nets on the heat exchanger? Because what is a heat exchanger real quick? So it's, it's, it's basically cools down liquids in a plant, you know, tries to maintain a liquid in a certain temperature band in any part of the, you know, petroleum petrochemical plant in this state.

And so what it was doing is they would run these plants for 24-7. So you always had a night shift as well. So the plant engineers would fall asleep or something would happen. They would typically miss controlling the temperature. And there was a lot of energy wastage. So it was a good sort of use case to apply and control it. Whenever the temperature went up, you kind of brought it down kind of thing. And it gave you a good framework to get other inputs,

environmental inputs as well. So in that sense, I don't know, maybe engineering is well tuned to trading a little bit because you see a lot of engineers who get into trading. Yeah. And chemical engineering, maybe because you're not just fiddling, like electrical engineering seems like you're fiddling and figuring it out. Chemical, you have to more create a theory and formulas in order to make it work, right? If anything, maybe mechanical engineering, because I actually equate the markets a lot to movement, like fluid dynamics, things to that effect. But I suppose you could look at chemical processes as something akin to markets. But the bottom line is, is you're looking at movement, you're looking at momentum, and you're trying to detect where the momentum is and whether it's going up or down.

So there's some engineering ideas in that. We used to have a partner out of, I'm going to forget the name, but now outside in Tennessee where they built the nuclear weapons. And his job, he was an engineer in stochastic engineering to test the groundwater for nuclear contamination, but you can't literally test all of it. So we had to come up with formulas and models for, you know, figuring out what the true level of contamination was. And he, he became quite good at helping us develop a bunch of models on the side as a side business. And so Robert, when you,

when he first came as an intern, were were you like this guy's going to be leading the company one day or what was the thought then do you want me out for this part of the yeah uh you know it's almost like uh we we tried to hire as many researchers as possible and uh I actually had a lot of respect for Jagdish. I really liked his humility, which I think is an exceptionally important attribute in creating trading models, as well as getting along with people.

And so I can't say that I particularly singled him out, but I definitely feel that he has some very great character skills, which you realize there's more aspects to trading and running a business than just being a good trader. That's one of the mistakes I made, actually, in my business is not understanding that soon enough. Oh, yeah. Well, this world's littered with great traders who are terrible businessmen. Yes. Right.

And then, Jag, same question in reverse. When you were sitting there getting hired as an intern, you're like, I'm going to run this place one day? You know, I, to be frank, at the time, I was, you know, this was a new space. I was a kid in a candy store. I was absorbing everything. I wasn't thinking ahead, you know, this was a new space. I was a kid in a candy store. I was absorbing everything. I didn't, I wasn't thinking ahead, you know, so much.

But my goal then was, I was just fascinated by the markets. To be frank, I just was, I felt happy. I was given an opportunity to get a chance, you know, to even get into the field, you know, explore models, modeling in the financial markets. So I was actually grateful for that, that in itself, sort of, it has an entrepreneurial nature. You know, the markets give you an entrepreneurial nature way before the technology world ever gave it to you. And even now I feel trading is the closest you can ever get to the bottom line, like in terms of P&L, you know, you, you kind of have a direct connection,

you know, exactly whether you're making money or you're not. And, you know, sometimes you have these tech companies and others who have thousands and thousands of employees. I doubt, I know they see stock appreciation, but I doubt each one of them. If you ask them, how do you directly contribute to the bottom line? It's very great There's nothing like finance. It's a brutal space, but it's very black and white Data scientist types they love the as you said before they love the quick turnaround, right?

I don't have to wait months or years for my experiment to play out. I can I can see it work out in near real-time right Yeah my experiment to play out i can i can see it work out in near real time right yeah sorry go ahead no i was going to say we started it as day traders so we had at least my trading was day trading so it was the type of thing where you needed to make money rather quickly otherwise you couldn't survive in a situation like that so yes your results are pretty quick. And so the Polaris program at Commodities Court, that was day trading? Initially, that's how it started out. That's because that's the way my trading was. Again, the influence from the floor and so on.

And the psychology, it suited me quite well. It's very important to be able to suit your trading to your psychology. And so we were doing the day trading. And then once we were allocated more and more capital, we realized that that would be a problem in terms of just liquidity and being able to allocate all that money. So that's, excuse me,

that's when we started going from day trading to longer term trading. And then how much money, that's when we started going from day trading to longer term trading. And then how much money were they running when you started? Like how much was 50 grand was what percent of their overall budget? My guess is that it was less than one tenth of one percent. I honestly have no idea. And were they doing this with hundreds of traders, thousands of traders? I can't say how many they were doing it with.

I simply don't know. But I think it's not so much the number of traders. I think they had significant allocations to some larger traders, maybe someone like Louis Bacon or Paul Tudor Jones, somebody like that. But I wouldn't really know the allocations myself. And did you cross paths with those those guys at all no um i uh um i'm sorry i i met lewis one time uh via elaine and it's very nice for her to do that but otherwise um i really haven't met that many other traders in the business ever ever yeah and that's partly

my fault as opposed to I so I was very very excited about the business and I'm probably going too quickly but we started raising money we were very successful uh very fortunate i should say and um but then life got it got in the way shall we say i had got married had three children and i started realizing that there's there's more to life than just trying to make money every day and so um that's when i started shall we say broadening my interests this was in the decade of 2000 yeah and um after 9-11 and so just started changing and not getting so involved with the industry conferences and things that i think becoming more shall we say, introvert internal. Good for you, right?

That's the, it should be the goal, right? Not to just knock, bang your head into the wall day after day. What's really important, I think people have to know what they want in life. Like I remember hearing about a pretty famous money manager. I don't know, he's probably in his 80s. And still he spends virtually every day of his life trying to find the next best stock. And don't know. He's probably in his 80s. And still he spends virtually every day of his life trying to find the next best stock. And that's fine if he's happy. But you start realizing that there could be more to life. And it's not to say that, look, you got to survive unquestionably.

But what I didn't, I've always had other interests like art. I believe trading, we talked about programming and being able to analyze the market from a mathematical level. But I also believe there's a subjective level, an artistic level, shall we say. And so I tried to apply both art and science to analyzing the markets. And so in the same way, I believed that I needed to also expand my mind, not just from a quantitative level, but from a qualitative level. And that meant being involved in other things like traveling and so on. And I do love photography. It's always been an important part of my life as well. Yeah. Before we got on, you said that background is your own photo, right?

That's correct, yeah. And to that end, what I can tell you is that I had mentioned to you that I had gone broke, I think it was 1990. And again, it wasn't because I lost money, it was just because I broke even for the year. So anyways...

That's a podcast in and of itself, how to go broke, breaking even. just because I broke even for the year. So anyways. That's a podcast in and of itself, how to go broke, breaking even. Yeah, right, right. So, and that's when I got the impetus to write my book because I needed the money. I needed, I was, I think I was given like a thousand dollar advance or something. And that was a lot of money for, I mean, I needed that money pretty badly. And so I said,

okay, I'll write a book for that. I didn't realize that it would take me two and a half years. I probably made about $5 an hour for that. But in any case, I still remember after I was given a lot more money from commodity corp, this is probably 1994.

I was living in Singapore at the time. Commodity Corp wanted me in Singapore to train other traders. They wanted to duplicate the same kind of idea in Asia. And so I went to Singapore and lived there for three years. But anyways, one of the allocators who originally visited me in New York, I talked to him and I said, do you recall the first meeting we had when you allocated that $50,000 to me? And he said, I don't really remember much about it. I said, do you have any idea how desperate I was at that time? Because if I didn't get that money, I probably would be driving a taxi like Bruce Covener, but without the, you know, the allocation. I said, I was in pretty desperate straits. And I was, what I wanted to find out was, did you sense that at all from me? And he said, no, the only thing I can really

remember about the entire interview is you had all these pictures you had taken on your wall. And I really liked those pictures. And I felt that if anybody could take pictures like that, they deserve a chance. So the point is, is not to speak so highly of my work, but photography, but the point is is that i think it's important to be focused but also to have other aspects to one's life and so what i can tell you is that i think the photography and the art has really helped me a lot in other aspects of my life including the trade because a lot of this is pattern recognition that's a great right like here you thought you were impressing him with your trading skills and your experience and your knowledge and he threw the money at you because of the photo. Exactly.

Good lesson there. Always wear a good shirt or have good photos behind you. Yeah, right. And so at the peak, how much had Commodities Corp allocated to Dean? I don't remember, but certainly, and it wasn't all commodities court, but back in the early 2000s, we had over 1 billion. I think it was like 1.2, maybe 1.3, but I don't know. At our PKUM, we were 1.6 billion. Oh, really? Okay. Well, thank you for that. Yeah. And as a trader, does that ever, like most people know it's, I'm doing 50 contracts now I'm doing 500, right. It doesn't really matter,

but at some point it's got to get into your head of like, man, this, this, I scrubbed this trade. This is a lot of money. Yeah. And you realize the responsibility you have for the people. And you know, for some people like a Bernie Madoff, he doesn't care. But for me, that responsibility was enormous. And not to say that Bernie Madoff was a traitor, but anyways, the point I'm trying to make is that you have a responsibility to your customers.

And that weighed on me a lot. And what I realized was that it went against my philosophy of what was most important to me. And one of the things that was most important to me was independence. And so what I thought is by raising all this money and making all this money, I become more and more independent. That's not the way it works. We actually become more dependent for better or worse. Someone had a great – right on Twitter, everyone was cheering Melvin Capital blowing up with the GameStop and whatnot. And someone piped up and was like, hey, just so you know,

90% of their assets most likely is pension funds and endowments and like sovereign wealth funds. Like it's not just, it's affecting some real life investments. Exactly. That's exactly right. And so that really weighed heavily on me. affecting some real life investments. Exactly. That's exactly right. And so that really weighed heavily on me. And not only that, but I was probably one of those, I don't want to say few, but a lot of money managers don't really trade their own capital. I grew up trading my own capital, taking responsibility for that.

And I really loved, you know, for me, I could lose millions of dollars of my own capital. Not happily, but I could do that. But if I lost a million dollars of somebody else's capital, that really bothered me far, far more. Where it was the opposite with a lot of other money managers. And so that's when I started focusing more and more on just being a proprietary trader as opposed to trying to raise outside money well we'll come back now to rotella as a whole and jag you can give us kind of the uh the elevator pitch of of what Rotella looks like now? Yeah. As, as we just discussed, you know,

Rotella has been a quant systematic you know, outfit since for the over 30 years now. And we have evolved quite a bit in those 30 years in terms of technology, in terms of the types of models, and we can get into it in a few minutes here. And right now we have six different strategies across both trend following, short-term momentum, as well as volatility space. And we have a fintech product, QDEC,

which we can discuss briefly as well later on in the talk. And between that, we are primarily based out of Seattle. But we do have employees spread over Chicago, UK. I just wanted, so 30 years, and actually we have Roy niederhofer on next week which he's been at it for around 30 years as well but like that's probably have back-to-back weeks on the podcast the two two of the longest track records out there right like yep it's insane to me that you've been at it for 30 years so what does the program look like now? Like the flagship program or there is no flagship versus 30 years ago? Yeah. Polaris is still our flagship program and it's doing phenomenally

well. The 2020, we outperformed the benchmark by over 20%. Wow. And, but so what, how much of that is still the same as it was in 85? Like 5% or 80%? This is an actively managed. Is the DNA the same? Right. So this is an actively managed portfolio. So what happens is the objectives that we are trying to capture in this portfolio remain the same.

So the objective hasn't changed from what Robert wanted to do in the late 90s through mid 2000s through now. But the models that have gone into it is an ensemble of a variety of models. 70% of the risk is in machine learning models and 30% is in technical models. And we can get into the evolution of the program as well, but the objective itself has been consistent. And as we know, just from that answer that it's at least 70% different because there wasn't machine learning back then, right? Besides what was in Robert's brain. You know, what it is is see, we actually, as Robert also mentioned, he wrote the book on

technical trading. So a lot of those, we built on a lot of those indicators internally and built a lot of models in the technical analysis space. So what we did is see after post 2008, what happened is there was a market structure change. There was a fundamental change in the market structure, which we picked up on earlier on in 2011, 2012, we picked up on earlier on that there is a fundamental market change. Until then, our models were working in a vacuum on each market, one market at a time. The reason we were comfortable with that framework is that we assume consistency in market structure across the wide variety of markets that we were trading.

But then what happened is we brought in, we realized that the market structure has changed and we needed to explicitly capture those inputs as well. So we created machine learning frameworks that could actually enable information exchange between these markets on our technical indicators. So actually, this is a very beautiful evolution of our research from technical to machine learning, because it actually builds on all of the work that we had done in that space and captures, exchanges information between the markets and between the technical levels across these markets before it makes a trade on one market. So essentially, instead of making a trade on using just one market data, it actually

gets information from other markets as well and then makes a trade on one. But the information that it uses is actually the technical indicators that we had built on before. The evolution of it is beautiful and it's very much an efficient framework to capture that information. And I can explain what that framework is, but I'll leave it to you. Yeah, we'll come back to that. So I want to ask robert

back in the early days what were could you give some examples of what some of those early models were were they like simple volatility breakout uh contrarian all the above yeah i i think the easiest example could be a simple pattern recognition let's say the market is up three days in a row what's the probability it'll be up the fourth day in a row? And so there's two things you're always looking for. One is the percentage times it's up versus down. But hand in hand with that is the magnitude of the move. So for example, if it's up 60% of the time on the fourth day and down 40% of the time time it still may be a good trade to go short if the magnitude of the mood is 10 yeah exactly right and so ultimately that's what you're doing you're looking at percentage um in other words are you looking for non-randomness non-randomness essentially

um do we have something that does better than 50%? And if it does, the magnitude of those moves are important. So it could be, like I said, three days in a row. It could be patterns like that. But yeah, volatility breakout systems as well. And then just to tie it back to you, Jag, so now in the future, in the past, maybe it was if S&P is up three days in a row, we have good confidence the fourth day's move is X magnitude and we're going to try and capture that. Now is it if bonds and crude oil and Euro dollars are do something three days in a row, I think this is going to reflect in S&P? Sure. Those are the types of complex relationships that we want to capture because trend itself has become a little more complex to capture. That is our

feeling and that was our hypothesis and it's panned out that way. It's played out that way. So, you know, there are two different frameworks that we built. One is a neural network framework and one is a random forest framework. And that, I'm going to interrupt you real quick on that. And after 08, you're saying, was this direct result of the Fed printing and all this stuff of like the markets getting manipulated or was you saw correlations moved to one? What was the 08 catalyst to kind of switch? Oh yeah, no, this was after the Fed influence, you know, it took us about two years to recognize that this was sort of almost a semi-permanent market structure change. You know, I mean, 2010, 2011. You nailed that.

It's definitely become permanent. Yeah, it has become permanent. And actually, Robert, you know, internally, I remember this. He came up with this thing saying he actually almost called the market in 2009. I remember this when he said, you know, equities is going to be the place to be in. Managed futures is probably going to have a difficult time for the next couple of years if this continues. Thanks for the call, Robert. That's when I was going full throttle into managed futures.

He actually, you know, surprisingly enough, he actually wrote a letter to all our investors. It's out there. It was not something that we did a 2020 analysis and fitted this viewpoint. He actually wrote a letter. Everyone has it who was an investor at the time. So, so anyhow, so that was that, but we did, you know, we had a curiosity on machine learning before it became, now today it's a fad a little bit when you use machine learning and ML and AI. But, you know, going back to 2005, we tested out neural network based strategies to alter the leverage over our technical models, right? We actually have done a lot of research there, but we never found anything convincing. So convincing from a supervised learning standpoint, where supervised learning isn't, there is a neural net is a good example of supervised learning, but an unsupervised learning would be a clustering technique like K-means clustering. So those kinds of techniques we used, we used a lot in the mid

2000s, but supervised learning like K-means clustering and things like that. We used a lot of those kinds of technologies in our analysis and portfolio construction. But in terms of using the supervised learning like neural nets, we didn't find a whole lot of success primarily because it felt like the amount of data was not enough. So for those kinds of powerful algorithms, even 30 years of data is not big and And it can memorize regimes very easily. So it had the huge risk of overfitting. And we were extremely uncomfortable at the time to roll it out on a live portfolio just for that reason. But quite recently, especially when you've had like the 30 years was all bonds going up and rates going down and things like that.

Right. So it's very easy for it to memorize. It's an easy problem for it. Right. And but in the past decade, we in the past five years, we found some incredible success. And in terms of using neural nets, as well as in terms of using random forests to trade, incorporate, and reduce the risk of overfitting. And it's played out and out of sample. So far, so good. And Robert, did they have to twist your arm to move into this, or were you twisting their arm?

I've been leading the charge. I've always been an advocate of using technology. And so unless you disagree with that, Jagdish, I feel like I've been. No, no, you are very much at that. I'm ahead of the curve every time. He's actually pushed us surprisingly, funnily enough, you would think that we would have to convince him. On the contrary,

there is an element of frustration every time he sees us like, you guys are not ahead enough. He's always saying, you haven't yet given me the prediction model, you know, or the forecasting model. We have actually gotten to a point where we are able to predict persistence, but not real pure prediction. But Robert's still pushing us to get to the, you know, predictive modeling, especially in a low yield environment. That's going to be a big part of, you know, what's going to be required here. And then start out. Sorry, go ahead. No, to that end, Jeff, I think what's very, very important is that people understand that the landscape has changed dramatically. So in other words, the models that I was using back in the 80s and 90s, they don't work anymore. And so then the question becomes, well, why is that the case?

And that's because the fundamentals, I believe, have changed dramatically. I mean, you've got a bond market that's at all time low yields in the entire history of the country. And you can't expect the same results because of that. That's where I think, that's where I was talking before about how you need to also have a qualitative analysis of the whole situation, as opposed to just thinking, well, this is going to, trend following has worked in the past. It's going to work in the future. And that's it. It can work in the future. I believe it will, because I think the trends will continue, but it doesn't mean you're going to get the same results as you did 10, 20 years ago. Right. They could be, which we've seen the last 10 years, they've been much choppier trends, right?

They've been volatile trends. So you could look at a chart and, okay, crude oil went from here to here. It looked like a downtrend. But compared with the 80s, like you mentioned, it was much harder to trade. Right. Exactly. And then if you fit it to these past 10 years, the next 10 aren't going to look anything like that. Right. And I think also fit it to these past 10 years, the next 10 aren't going to look anything like that. Right. And I think also it's your asset classes.

If you think about it, you have a lot more options 10, 20 years ago in terms of being able to invest, say, in interest rates and commodities and so on. But I think that's changing. And I think a lot of people are becoming very, very frightened about their allocation to interest rates. And I think that's only going to continue in the future. You're saying because of globalization and because there's not really a difference between the currencies, it's all play on the Fed? Well, I guess... All the above, everyone's easing. You know, there was a time

when you could get five even ten percent on a T-bill yeah and now when you're getting virtually zero percent

on a T-bill when you're getting a few percent on notes and bonds that's not gonna solve your

financial issues especially pension front and so on that have all these liabilities. And then I want to come back. So when you first started digging into this, was it more a way to, hey, I've got all these ideas I want to research.

We don't have the manpower, womanpower, whatever. Like, let's get machine learning going so we can do all these ideas faster? Which is kind of the supervised learning and just like brute force methods. Hey, let's get this research done. That comes back to my failure as a CEO in terms of running the business. So I just thought, you know, I got all these ideas and just hire people. And if these people have the same interest that I have, that they're going to do some really great work.

And the problem is if you're not managing that properly, all it becomes is a chaotic environment and nothing gets accomplished. You didn't check their photo portfolio. That's right. She said, all right, you can work on this project first. Let me see your photos. Exactly right. So that's where Jagdish has been enormously helpful and productive and not just on the research end, but also in managing the company and so that's made a

big difference in terms of i believe hopefully being more successful going forward okay so the bottom line is is not only do you need to be able to understand the trading process but you also need to manage that process as a business it's a very different skill set that's required yeah let's dig into the models a little bit jack so we've got the polaris is still there in the flagship yeah you guys have launched some other strategies as well? Yeah, we have launched a couple of more. We have the short-term momentum as well, which also has a substantial amount of machine learning in it, which has a track record from 06. And we do have a couple of volatility strategies. And one of them you're probably more familiar with but i i can get into it but that also has primarily used as machine learning to

uh hedge uh the equity beta um so the short-term momentum who's that designed for it's like as a hedge or just as a pure absolute return see uh with the managed future strategies, they have an absolute return mandate. But they are uncorrelated to the more medium to long-term traditional trend-following returns that you expect. They are meant to be uncorrelated to that. So, for example, we use internally the SOC Gen CTA index as a benchmark and we look at that as, from a correlation standpoint, we don't mind if Polaris is correlated to it,

but we would prefer that the short term index is uncorrelated to it, but also, just not uncorrelated, but also, better on the upside. So that's important. So that has a dual mandate from that standpoint. And then, so back to Polaris. So it's roots and DNA would still be trend following. Is that fair to say? Yes. It's the roots and DNA is still trend following. It still has a spectrum of holding periods from 20 to 60 days.

And the framework itself, as I mentioned before, previously you can think of these models as just single decision trees. And what now we have done is we have used random forest to bring it together. So a simple example would be, let's say we use 30 years of data and there are 200 trading days needs. So there are 6,000 days. And what we have done is we have

subsetted that data into, this is an example, not exact parameters, but we have subsetted that data into chunks of 200 randomly. And then we take these decision trees and fit it to these random chunks of data. The beauty of that is you have a set of weak learners. So it's very difficult for any single decision tree to memorize the whole data set, but they have this random pieces of data that they have memorized, right?

So essentially, we are increasing the noise deliberately. We are not giving them an easier task. So what you have is a set of weak learners. So we don't trust any one of these decision trees, but together in an ensemble, we take an average of them and figure out is there a bullish sentiment or a bearish sentiment on the market and then we use a position sizer and replace the trade. So essentially what this does is it reduces the risk of overfitting. It kind of it's very forgiving from that standpoint because of this multiple weak learner concept rather than trying to achieve the most optimized model you can ever get and so each of those are making their own decision in terms of look back period and so things like that universe of it

so we look at the whole universe of it so So let's take, for example, we had S&P, the 10 years and crude oil in the portfolio. It would do it for all across all of them. And then they would look at the influence of each one of those markets on the other. Okay. So it's not necessarily deciding parameters, but essentially throw it all in there and tell me whether I'm supposed to be short bonds or not. Right. So any kind of market structure change, the way it is designed is it should be able to capture it. And it would not be in a situation where you are watching that in real time after two to three years and then you realize, okay, I got to go and change my model or replace it. So the framework that we have created is meant to adapt on the market structure change. And this is more for medium to long-term basis, right?

So the objective function is to capture trends in the medium to long-term space. So 20 to 60 days, but in the short-term portfolio, in the short-term momentum portfolio, our objective function reduces quite dramatically. So it would be between five to 20 days is what we are looking to do well in or to see if we should be bullish or bearish. And then are there any constraints on top of it of like, we don't want to have a long bias or we, uh, long bias or things like that? Right. So there is no explicit bias.

So we are not tilting the portfolio one way or the other. We have an equal risk budget across all the four sectors. We don't want to add additional parameters. Uh, if, you know, I know some managers still credit a little more to equities when that was their getting out of the managed future slump and things like that. But the problem with that, we know the problems with that in terms of sudden market crashes and things like that.

They don't land up providing you the diversification you're hoping for in the portfolio. But in our case, we have tried to reduce actually the number of parameters by doing this by keeping it giving it an equal risk budget so there's no I mean there's no explicit bias anywhere but if it if the trades from time to time it could get long biased but that's natural course of market yeah just in my brain right if I had the perfect machine learning tool and I turned it on in 09 it would quickly learn like okay lengthen your

term and have a long bias on equities and then in theory as things started to switch the other way it would shorten my term and take away that long bias from equities. Right yeah those are some of the things that I, I'm actually pushing to work in that area, a predictive type of work. And I think that, um, there's a lot of things that can be gained by that. Um, but a lot of things that can go wrong as well. Unquestionably. Yes. Right. Right. Right. Cause sometimes you get like inverse cycles like you think you're getting a top but you're actually

getting a bottom and so on so um but that this that said if you can find a turning point that still can be very very helpful in terms of the training that you're trying to do and then it also always in my perfect world AI machine learning tool, it'll ping you and say, you should be going long Toronto real estate or something, right? So there's like also a limit, a natural limit to what the inputs are. So in a perfect world, the robot's running it and looking at the whole universe of possible investments and shooting something out. But there's practical limits to that, obviously. Well, I think people see this massive universe of investments, but I try to whittle it down to five or six at most. And to me, all of these other

offshoots are really derivatives of... So for a very simple example, when you have long-short equity trading, it looks great from a diversification standpoint versus being long in the stock market. But a lot of things may be diversified in a bull market, but they become quite the opposite in a bear market. So the point I'm trying to make is that even though you have long, short equity trading, to me, that's just a variation on going long stocks. And you'll get

the non-correlation in a bull market, but you don't really need that non-correlation. In fact, I don't always feel that correlation or non-correlation are that critical. It's more getting the market right than anything else, quite honestly. And that long, short, the last couple months have shown us that's a very convergent trade as well, right? You're also betting on things remaining relatively stable. So even if the market doesn't go down, if there's a huge shift from momentum to value, you're getting taken out on a stretch.

Well, that's one of the reasons why we're hoping that possibly momentum trading, which is what I see something like Polaris or managed futures becoming more important in the investment landscape, simply because unless you're expecting to make, you know, one-tenth of 1% of 1% basis points on your interest rate portfolio,

there's got to be another way to do it. And should we enter a bear market equities, there's not a lot of alternatives out there right now. So to me, momentum trading could be very promising. I've said that here and elsewhere before, like your bonds were there. They're going to be a flight to safety. They still, even with negative rates, I think there'll be a flight to safety. So that's all still good with them. But what do you do in the meantime? So like, I'd rather have

something that has a chance of making money in a normal environment. And then that is most likely going to be there in the flight to safety kind of crisis period versus bonds at these levels, you're almost guaranteeing you're going to lose money until the crisis period. Yeah. And let's not forget, you know, the bonds have been a flight to safety for 40 years. It doesn't mean that that'll always be the case. And you know, the state that you're living in, I don't know if I want to buy Illinois bonds as a flight to safety.

So yeah, I was was i was telling everyone biden's gonna get elected they're gonna like forgive some of the illinois debt um that's a topic for another time but we've seen some of that with this stimulus right they gave money to the states right so we're holding out for like a all states debt reset because illinois will make out way better than like arizona or somebody or yeah california or something that but the bottom line is is that um that safety may not be there anymore and so to me that's another argument why momentum trading might be important going forward and then let's touch real quick on the volatility programs.

Yes. Sorry, go ahead. One of the things with, you know, the volatility programs is a natural fit with momentum trading in some sense, because a lot of people have invested in classic trend following as a crisis alpha strategy. But there is so many different sort of interpretations to crisis alpha. And Jeff, I had a chance to see your interview with Katie Kaminsky as well, where she was defining crisis alpha. And she was actually trying to clarify the point because many people think crisis alpha is exactly tied to the equity markets. Right. Tomorrow, if we're down 1%,

where's my crisis alpha? Yeah. Managed futures has to kind of diversify. And that's not really the case. Right. And, but so what we, what we did as part of our research, what we arrived at is also a pure direct hedge to the equity markets. Let's get a direct hedge to equity markets. So we have a long-only S&P and VIX futures strategy. And essentially what it does is it plays off of the convexity between S&P and WIX. And what we are trying to do there is there manages the amount of positions we need to hold in S&P futures as well as WIX and V stocks and Euro stocks. And so it's at its base it's like a long

long so it's long the VIX and long the S&P as a head. Exactly so you'll never be long shot WIX at any point a lot of the wall traders in the past five years, as the markets were going up in a straight line, landed up trying to shot Wix and got caught on the other side of that trade. But here is a strategy that we would never be going shot Wix. It's always long and long. It plays off of the convexity. So in calm markets, we would need very little exposure to the WIX contracts, but have a lot more exposure to S&P. And at that point, if there is a market crash, for example, in Feb 2018,

WIX went up 300%, the market went down 5%. And you don't, there is considerable amount of leave-away for you to hedge that risk and actually provide a positive P&L. And the second challenge in that that we optimize for is to retain those gains. One is pay as little as possible for that insurance, but also retain those gains when that happens. And as we all know, that helps greatly in portfolios, especially in the compounding equation of portfolios. Yeah. And how do you view like April, May of 2020 when VIX is at 60? Right. So it's hard to implement that strategy, but that's where the AI comes in and says,

this is what the hedge ratio should be. Correct. So that is part of what it does is we try to lose as, pay as little premium for this insurance as possible. So one of the disadvantages is that if you're stuck in a high VIX environment and the markets are going up, you know, you're going to be not as hedged as

if the VIX was low and the markets were going up and then you had a crash, right? So there is going to be that feature, but luckily for us, those phases don't last for extended periods of time, as well as we try to reduce the premium that we pay for the Wix contracts as much as possible at that point. And why not, or do you do this? Is this part of Polaris and the other programs? No. So Polaris has its own. Why not? Because it seems like it fits the MO, right? Right. So we do offer, what we have done is, we can get into it. I spoke about mass customization. So we are offering custom portfolios, which are multi-strats with Polaris, as well as

these wall products together. But we want to maintain Polaris as is and the wall as is, because the reason is, see, we have a long track record in all of these strategies yeah and as an investor i appreciate that i'd rather have you be like different flavors that i can pick at the ice cream shop the problem is if nobody's buying those flavors and you only have four right from a business standpoint that's the issue so some managers say i'm going to put it all together in my best combo flavor another one others say no we're sticking you get to pick your flavor right well a lot of these arise simply because i'm looking for something to help the portfolio that we're trading on a proprietary basis so um i see

a need for it or we do and um so that's how the strategy gets implemented. So, and if it works, then it probably has application in the outside world as well. But it's very, very important not to stray too far from the original strategy because customers don't want, okay, the bottom line is, is you can do anything you want when you're right. But when you're wrong, that's when people start getting upset with what you're doing and so um if you have a product that let's just say for

example is short-term in nature i did it because i was successful at it i went from short-term to long-term but as a general rule you want to avoid things like that because if if it doesn't go well then people will start complaining well why don't you just stay with the short-term aspect to it so on definitely and then is there any um then i'm going to ask about q-deck in a minute but just on the portfolio level and you mentioned like okay we needed this so it's a new program is there anything that looks at hey we need to add bitcoin futures or ethereum futures or things that. And you have the overall, the machine learning is looking at different opportunities and saying, add those to the portfolio.

Yes, unquestionably. I'm happy for you to answer this as well, Jagdish. But to me, that comes back to what are the major asset classes that we can invest in? And to me, interest rates are dead as far as I'm concerned, in terms of a choice versus, say, 30, 40 years ago. Equities I'm becoming somewhat concerned about. You've got commodities, which Jim Rogers talked about a lot, essentially precious metals and so on.

There's a lot of debate right now whether commodities really are a true giving positive returns. right now whether commodities really are a true Increase Giving positive returns. Okay as an asset class like versus equities. Okay, and that's that's a Topic for a whole other discussion. Yeah, that's would say not really right? Yeah, that's over the last 30 years would say Unquestionably right and there's you know And there's good reason for that from a practical standpoint. If commodities keep going up, then people can't use them anymore. So something like corn or whatever. But let's just say that maybe things like precious metals are a good long-term positive carry or positive return investment.

Then you've got currencies, which basically, that's another interest rate play essentially, as far as I'm concerned. So then that leaves real estate and cryptocurrencies. And I'm hoping that cryptocurrencies are going to be a very, very important part of the landscape going forward. And how do you view that? Because they don't really fit in any asset class either. So is it just pure, it's the speculative asset class? So it's just a pure momentum and pattern recognition

trade? Medium of exchange currency. I view it as a digital currency, just as valid. So you've got the fiat currencies. But now what we have are digital currencies, and they're providing people question, well, what's the inherent value of a cryptocurrency? I think there can be tremendous inherent values if used properly. We shall see. To me, they're not really being used as exchange of value, but maybe a store of value so more in like the precious metals bucket to me yes right right but that's partly because of of the um the demand right now is uh the supply actually is not as great

in other words there's a lot of people having more and more demand for that. Now, you can argue how much of that demand is speculative and how much of that is using it actually as a means of exchange. That's hard to say, but clearly there's increasing demand to use it as a currency. All right, Jack, QDEC. So the last piece of the business, you got all the models, you got the strategies, you shoved Robert to the side. Respectfully. You went after my job, Jeff. I don't know about this.

So tell us about QDEC and what you guys are doing in FinTech. Yeah, you know, this is, you know, we looked at as we are, as Roberts pushed us to look for predict to work, do predict to work and capture trends as early as possible. We continuously keep an eye out even on the business front to, you know, catch trends early. And there are a couple of things that's going on that we picked up about four to five years ago where we saw that there is a big, this is way before COVID. COVID, if anything,

just accelerated this model of adoption of FinTech. But right before that itself, we could see that there is a trend of mass customization in the financial industry. And this actually originally started out, one of the most successful ones is Starbucks. Starbucks was incredibly successful in rolling out mass customization. Otherwise, who would go to a coffee store and pay $5 for a cup of coffee, right? I mean, they figured out a model where they could sell it in mass and customize it and feel every client, you know. Yeah, neither my wife or I drink coffee and we probably spend like $3,000, $4,000, $5,000 a year there or something.

Exactly. Yeah. Right. So that, and you see that trend across the space today, right? In retail, in sneakers, for example, you can go on a New Balance website and pick, you know, you can customize your shoe. I mean, who knew you could customize your sneaker, right?

I mean, I'm not the most artistic guy, so I don't, but I have seen people do that. And I said, oh, wow, you could do that. Right. And this is a big trend. And what we noticed is there are a couple of things. One is we figure that if you see the wire houses itself, the wire houses, their market share has gone down from 42% down to about 34% in the past 10 years, and it's expected to go down below 30%. And the RIAs and the independent, the hybrid RIAs, their market share is expected to go up. They've already gone up from 16% to 24%. They should be at 30% in less than two years time. So they gaining market share the wire houses are losing market share and a parallel trend is happening here where banks the number of branches at this was you know looking at some FDIC numbers the number of branches bank branches in the US went up from 18,000 to 82,000 and this data is a little dated couple a couple of years old, but

in a 10 year timeframe from 2004 through 2015. And when you think about it, all of that, when that kind of growth happens, a lot of FinTech investments happen. Okay. And in the venture space, what happened is it went from 5 billion to 22 billion invested in fintech investments by 2015. And from there on, it has only increased because the 2021 data, I was looking at it the other week, and it came out that every 91% of your banking clients, retail clients have made at least one transaction on mobile phone. And they expect 78% of the millennials to be digital banking customers by next year. If these RIAs and hybrid RIAs want to capture that market, they pretty much have to adopt, you know, a digital technology. Yeah, they need an app for lack of a better term, right? Exactly. They would need to adopt this. But what we are trying to do is given our success in these, you know, our core strength is research and in machine learning models and models that can generate alpha.

So why not take it, offer it in a mass customization model and both to asset managers as well as RIAs. So in the RIA world, what we are working on is trying to build robo advisors for them so that they can capture these millennials and give them the tools that is required to grow their asset. And on the other side, on the asset management side, we are truly giving them tools to build their own custom solutions. And they could use our solutions or they could use a hybrid approach where they take solutions from different managers and bring it together. But they can do that with QDAC. And that's one of the trends that we have captured on.

And if there's an investor listening to this podcast and they have a CTA allocation and if the CTA doesn't have a FinTech product, they should definitely be concerned, I would say, because there are two things from just from a research standpoint. Actually, we are able to attract far better talent. The truth be told, we actually compete with tech companies for talent, not with other financial industries, because majority of the talent who are good are going towards tech companies. And they would rather go to a fTech company than, than a hedge fund. Yeah. A hedge fund, a traditional hedge fund. Right. So, so from, even from that standpoint, we are equipped very well to generate,

I think far better quality of research. At least the odds are in our favor by, by this venture that we have. I love it. And any worries on investors of like, hey, stay in your lane. Are you a tech company or are you a hedge fund? Like, I don't trust that you can do both at the same time, walk and chew gum.

Actually, they have actually embraced it. I would say that they actually see the need for this. There are, all of our investors are QDAC clients as well. And they have embraced it in a way that they see the value prop in it. They see the value of building custom solutions. And I don't think, see, this is a SaaS product, really. QDAC is a software as a service product. And what makes it powerful is when you combine it with our data science scientists. That's what makes it really powerful, right? And they see value in it.

It's a method of delivery, right? It's no different than saying I'm using the internet to deliver a solution. And if anything, they get far more transparency into any strategy that they built. And, and if you have an investor who invests in one of our products, they can have access to QDAC and see intraday PNL. They can see the models that are live on it. You know,

so that is great. They get far more transparency from a due diligence standpoint. They know exactly what, you know, so that is great. They get far more transparency from a due diligence standpoint. They know exactly what, you know, what trades are being made on any given day, which is, which is the sort of transparency you would get if you had a managed account, for example, in the holidays. Yeah. That's, that's exactly what you get here for them. Any worries if you give them too much, it's like a loaded gun, give them too much, they'll shoot themselves in the foot. You know, they don't have,

well, we are not- Right, I've liked in terms of, you're down today, I'm gonna panic, right? Private equity, I've always joked, has the best gig in the world because they're like locked up for four years and they mark to market or they mark to their own mark.

So the investors are just forced to have this lengthy holding period. And I think that's part of a good portion of why they perform so well. Yeah. No, this is more of a B2B business. So we are not going direct to consumer. So for someone like an allocator, you know, like for example, the Mutiny Fund can have access to it and take a look at it and look at our IPL.

That's perfectly fine right I don't think any of the team at Mutiny would panic like that so in the same way b2b we work with RIAs that have the similar sentiment they would not worry about it because they actually appreciate the tools and they can actually build a robo-advisor to attract more clients, which is great. They actually see it in a way we are doing the heavy lift. So we want to be the operating system of the next generation asset managers.

All right. Just as you were speaking there, your image makes it feel to me like there's a satellite about to crash on your head. It looks like a satellite with its solar panels expanded there. I'm actually sitting in the International Space Station. Yeah, I love it. Without a mask. That's the only place I can go without a mask so I decided sit there the uh you think they have to court you think they have to get tested and do the temperature check before they step onto the space station probably that would be odd

they might I don't know you know so I want to finish up and talk with with back to machine learning and AI and just kind of a 30,000 foot view on all of that. So I don't know where to start, but one of my questions, the smarter and smarter the machines get, will, will it converge at some point, right? So if you guys are building all this stuff, if say you have 10 companies just like you as processing power becomes more and will it converge at some point? Right? So if you guys are building all this stuff,

if say you have 10 companies just like you, as processing power becomes more and more and you kind of leave them unsupervised more and more, does the end result converge in the future because they're all coming to the same result? You know, this is, this is sort of obviously a very hypothetical and realistic question, and will it happen or not, right? And this is an opinion, obviously. I mean, who knows what would exactly happen.

Yeah, we won't hold it to you in 40 years. Right, right. But, you know, the thing is, see, in the technology space, typically their problems tend to have a high signal to noise ratio and academic machine learning models work great. But financial data, as we know, there's a lot of noise in there. The signal to noise ratio is not very high. And it is going to, there are going to be winners and losers. And it's going to be extremely important to have what I'm saying as AI interpretability. So sometimes in AI models, you don't completely understand what's happening inside them,

but then at least you should be able to understand, okay, given a certain scenario, how does it perform? So you should have enough of those scenarios to understand and feel comfortable that, okay, I know that this works for this case. There is a strong sort of objective function here that it was optimized to and it meets that metric. If you can do that, I think that is more important than exactly what AI technology you're applying, or even if you're applying AI. See, AI is not a must. See, internally, that's why, you know, majority of our programs, we do not call it an ML-based

program or anything like that. You've seen our tar sheets and things like that. It doesn't say that. The reason is simple. Because if a simple model can achieve the objective we are looking to achieve, we will take it every single time. We are not applying machine learning for the sake of applying machine learning. But if it gets complex and there is a place for it, we use it. So we are very careful about when we make that decision to apply or not. And that's what you'll see

because there are a lot of people out there, a lot of managers out there, their starting point, especially who come from a tech background, pure tech background, their starting point is machine learning. And then they kind of try to peel the onion and go back to what end, right? And they quickly realized that it's actually the signal to noise ratio in financial markets is far lower than what they would have experienced in any other field that they had applied it to before. And so, Robert, you mentioned before, right? It's the math plus the art of it all.

So if there's this AI that no one really knows how it works, how do you square that with like needing to kind of know the fundamentals behind it or the, not the fundamentals, but like if I don't have a reason for it to work, how do I trust that it works? Yeah, well, I think ultimately we have to remember what we're studying. And to me, we're studying human human nature and unless you're talking about in prices yeah exactly right and that's people think uh like the stock market is is simply it is what it is because that's what the economics is no it's not the case what it is is people's perception of the world, which is very, very different than the pure numbers behind it.

So the bottom line is, to add to this conversation, is that ultimately I believe what we're doing is studying human nature, which in some ways doesn't change over time, but it does. It's unique to each specific situation. But in terms of all the programs converging on one solution, anything's possible, but I'm not concerned about that. Because number one, your assumptions all have to be the same. And we all, a lot of us have different assumptions about the world. I'm certain that a lot of the assumptions I have about the world are vastly different than other people that you interview.

And so that said, I think you're going to come up with different responses. And ultimately, human nature is quite often a mystery and very difficult to detect with a lot of clarity. And so that's a good point too, of like, okay, but human nature isn't a clockwork machine that we can map, right? So if the machine learning's job is to map that nature, but it's unmappable, how do you square those two things? Or you try and get as close as possible to being mappable? Right. I mean, quite often what you'll hear is like, you know, somebody gets into office, elected office, and you expect them to carry out certain policies and the exact opposite policies occur

sometimes. Okay. And then sometimes they do carry out the policies that they originally intended. So I guess the point is, is that it's not always easy to know what the long-term consequences are of what you're doing and even trying to predict that. So it's much more unknown than we realize. And I kind of think of it also, and AI hasn't quite solved poker yet, although I read a few things that it's getting close. But right at a poker table, you don't know what the guy has. You don't know what he's going to do.

It's human nature. But I can assign probabilities to it, right? And then I can do my own bet sizing based on the probabilities of what hand he has or put him on a set of hands. So do you guys kind of view it of that? Of like, we're just looking at prices. We're going to set a range of possibilities, what could happen, and then set our own bets, basically, according to those range of possibilities. I think that's a good example.

But let's just say that we did the same thing that we did in the markets. Let's just say you have a very good poker player. And we analyze how this individual has played in the past five years, and we see where the person is faking and so on, then you have a good idea of how that person might work in the future. But if that individual knows that that's what you're doing, then that person might try to do the exact opposite. So the point is you've always got these other scenarios that can occur that will screw up the whole analysis. And speak a little bit, Jag. So do you guys kind of internally think in terms of probabilities like that?

They do think of things like that. But one thing you can always make, observe that the trading model represents the person's personality. You know, it's funny how that happens. I know that they code it up and I know how they're, how the risk management technique or what have you, what kind of drawdowns they're willing to take and things like that. You will see that reflected in the models, you know, and, and even in, even in a machine learning model, it's not that much of an autopilot when we apply it to the financial markets. It's a, you know, it's a lot more, there's a lot more control than that. You know,

there's a lot more of you in it. It's a lot more of the creator in it or the creator in it. Even on the unsupervised models, cause you have to start it off, start it down the path. Right. See, the unsupervised models are not used for as much around direct decision making and trading and things like that. It goes in on the input level, but not really at those critical points. But again, choice of making that using a model is again made by the person themselves. That's one because they have some sort of an objective in their mind they want to achieve i mean that is whether they explicitly state it as as this or not uh it's fine you always have an objective that you want to achieve um that is unstated maybe but yeah yeah no and i always say that like you could turn the you know sophisticated a on ai on and it just says buy today and

sell in 100 years right like that's my maximum long term. If you don't tell it, you don't want this much drawdown, this much volatility. There needs to be some inputs or you're going to end up way away from where you want to be. That comes back to what I was saying is that you have to have a certain set of assumptions that you can't just create models without any kind of framework or assumptions in it. And those assumptions quite often reflect the builder of the model. And so a very simple example would be my psychology was far more amenable to short-term trading.

There are other people that you've talked to, like say turtle traders, that their way of trading was long-term. It's not a question of what's right or wrong. What can you stick with? Exactly, right, right. And for me, the long-term trading wasn't as appealing, but it doesn't mean it's any better or worse.

It simply didn't fit my psychology. So that's the difference. We have a client who's an ER doctor and he's like, yeah, I don't make as much money as the other guys, but when I'm on, I'm on and I'm in there. I don't have any, I don't go home and I'm on call and things. So it's like just a lifestyle choice. I think some of those trading things are similar, right? I don't want positions over the weekend. I want to enjoy my weekend and not have to worry about it. Yeah. And, you know, I talked about the reminiscence of a stock operator. You know, supposedly, I don't know how true it is, but supposedly Jesse Livermore preferred going short markets versus going long markets because

he felt that you got a lot more quick action when you went short versus when you went long. And, but, you know, mentally most people would prefer buying versus selling. So there's all these different aspects to what you want in a trade. It's not just your, your timeframe, but whether you want to go long. And then obviously there's another very, very important point is that, and that's whether you want to be part of the crowd or not. And that's an exceptionally difficult uh part of trading which people don't realize um and and partly especially

in the culture that we're in right now where if you're not part of the majority um it's not necessarily good in the markets sometimes it's a big benefit to be part of the minority yeah and then how do you guys view if you're on the edge of this machine learning and ai like for overall society right like i think back to the uh presidential debates and andrew yang and i was like he was talking about everyone else is on stage arguing about this coal mine and they're going to try and save 200 jobs and he's out there saying hey trucking is the biggest employer in like two thirds of the states.

Those are all going to be self-driving in 10 to 20 years. Like how are we replacing all those jobs? So that got a little political, didn't mean to go there, but just in terms of like, what is the near term and far term future look like for AI and what it could do to our whole society?

Any thoughts? I'll let you answer that first, Jagdish, because I have some thoughts too. Sure. You know, see, with any technological invention, right? I mean, this is what probably they said when Henry Ford brought the car out when there were buggies, which were popular. And, you know, they probably thought it's going to take away jobs or, or what have you with any kind of, if anything, we've been able to build on it. We've been able to do more with it. Right. I mean, in a sense, I feel it's good. Any kind of invention is good because we are, we are able to build bigger things off of it. There may be

a lot more things in our future that we don't know yet that we can build but we don't need to restrict technology from happening. And that's what I was I wasn't necessarily saying it's a bad thing. I think it could be a good thing. You could eliminate poverty. You could say, hey, we can build anything we want for anybody at super cheap cost um right it kind of fixes the world but then it opens up a whole another slew of issues of like if people don't have jobs what do they what do they do it's kind of like understanding the the essence of the atom nuclear power and uh it can either do great things for us or it could destroy us. That's what I see right now. Yeah.

And knowing human nature, well, let's be optimistic. Yeah. Well, and that's Elon Musk and all those guys came out and said, like, this is the biggest future threat, right? Exactly. That's exactly right. Yes. And so it sounds, Robert,

like you're on the side of Terminator days coming at some point. Jag, you're more John Connor, we can save the world. I'm more optimistic. You know, I have a bias towards optimism. But I'm fascinated, truly fascinated by this technology. I'm naturally curious about this. I'm just fascinated by what, how far we have come. You know, a simple example is even this, you know, I, granted we've been, we have been in a pandemic, but some of the positives about the pandemic has also been,

if you see that technology adoption has gone phenomenally fast, like, you know, I can tell you in the medical profession, a lot of visits, you don't need to actually go in person. They're doing it virtually. And that could have been done, you know, 10 years ago. We had Skype 15 years ago, but it took an event like this for a change in structure to change, you know, a shift to happen, or you can deliver education, you know. I think MIT's entire coursework is available online, right? Right. Personally, I'm sending my daughter to school. I'm rooting for the schools to open, but I must say, I mean, I'm just saying to places where it's underserved or unreachable or, you know, where we can get, you know, nobody wants to go there, live there and teach those kids.

You know what, this is great if they were able to get technology down there. But things like that I'm optimistic about. But yeah, to be determined, you know, a classic example about the AI one was, I think it was the Cortana, was it I think it was Cortana the Microsoft bought that assistant yeah so it's initially I think in the first as soon as they launched it people were having fun with it and someone put an algorithm to teach it a bunch of different things and it became super inappropriate because it is an AI algorithm that was learning this, right? So sure, you know, there are those risks and then they had to kind of put

some guardrails around it and saying, okay, could we remain within this sort of thing, you know? I guess to clarify, I'm optimistic about the future but doesn't mean there won't be drawdowns in the future and furthermore societal drawdowns exactly right like a Great Depression things to that effect unquestionably it's no different than saying look I I'm very very keen on cryptocurrencies doesn't mean you just blindly take out a mortgage and buy Bitcoin right now or whatever. Although I think some people have. Any other big AI thoughts?

What's the coolest thing that's going to happen next in our lifetime? Self-driving cars are already there. We're going to Mars. Robots doing surgery. I think, I don't know if it's going to be AI. I think it's going to be learning more about ourself, actually. And the things that we can actually develop internally, like from a meditative standpoint. That's what I'm hoping for, but we'll see.

All right, we're going to end up with some of your favorites. We'll do quick. So you both spent time in Chicago. Favorite Chicago pizza spot? I like the deep dish Pizzeria Uno Oh yeah Uno

Uno yeah I don't know I don't know if they're still there but India House I like their pizza I like the Indian food there it was good India House pizza

I don't think I've ever had their pizza no no they don't but they had great food there. And you got, I forgot to mention, you guys have been doing all those COVID charting and whatnot since the beginning of the pandemic. Right.

We can link to it. What are some of your, it's a weird thing to ask, but do you have any favorite stats or charts that have come out of that? Come out of that. That is more a question for our data side. I'll tell you what I, what I like is that the fact that Florida and Sweden, which don't have the,

all the mandatory restrictions and so on are doing just as well, if not better, than other countries or states alongside of them. So that would give you an idea of what I think about the whole thing. You have a favorite vaccine? Have you guys gotten vaccinated yet? No. Not going to touch this stuff. Really?

Not yet tested. There's a lot of guinea pigs out there um favorite seattle spot for dinner any none or bellevue no um is the office is it in bellevue or inue? It's actually in Bellevue. Bellevue. Yeah. You know, I like... You know, Robert, you took me... You know, it's been a year since I've gone to restaurants. Right. You're like, restaurants?

What are those? What are you talking about? You know, you hit me with that one. But, Robert, the Italian place, I've gone a couple of times after you took me to that. Oh, yeah. You know, I haven't been to the restaurant myself for a while either. Yeah.

We'll skip that one. And finally, we ask all our guests' favorite Star Wars character. I'm going to throw you a curveball. Joseph Campbell. Say it again? Joseph Campbell. Joseph Campbell. That you a curveball. Joseph Campbell. Say it again? Joseph Campbell. Joseph Campbell. That is a curveball.

He's not a character. He's an actor? No, no. He's kind of the basics of how Star Wars began. The whole idea of the myth. Ah. I'm going to go research that. I like it.

He helped George Lucas with the whole thing. All right. You met him or read about him? How'd you learn about him? No, he's written about myths and how important they are and the whole idea behind them. That's what Star Wars is based on. Yeah, yeah.

Love it. Jag? No, I don't. I'm not into it. I should read up on this I mean I know what Star Wars is but I don't have a favorite character how about R2D2 he's kind of a

or Darth Vader's emotions on his face alright guys it's been fun thanks so much for your time we'll put in the show notes where to find out more about All right, guys, it's been fun. Thanks so much for your time. Okay. We'll put in the show notes where to find out more about all your programs and QDEC and all the other good stuff. 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.