The Derivative - Marrying Fundamental Factors into Commodity Quant with Patrik Safvenblad of VOLT CM

Episode Date: November 16, 2023

Thanksgiving is just a week away in the US, marking the end of an excellent year for The Derivative and we're inviting you to join us at the podcast table for a bountiful episode delving into the unique flavors of managed futures and fundamental expertise. As we carve into the Thanksgiving spirit, The Derivative is serving up a feast of insights with Patrik Safvenblad, the mastermind behind VOLT Capital Management. In this compelling conversation, Patrik unveils his approach as a fundamental specialist in managed futures programs, offering a unique perspective that distinguishes his strategies from the norm. We delve into the differences between pod shops and in-house multi-strats, explore the penchant for shorter-term models, and uncover the meticulous process of managing over 9000 signals. Join us as we navigate the complexities of commodity quant and gain invaluable insights into quantitative strategies just in time for the Holiday season — SEND IT! Chapters:   00:00-02:09 = Intro 02:10-05:30= Dark days in Stockholm 05:31-18:09=Professor to Quant: early skipping stones to “across the pond”, building a team & getting diversified 18:10-26:44= Shaping Fundamentals, short-term models & configurations in machine learning 26:45-41:49= Selecting signals, gathering data & adjusting portfolios 41:50-53:14= Efforts of running 9000+signal, portfolio level constraints & Podshops vs in house 53:15-01:01:50= Positive Skew, risk management & opportunities 01:01:51-01:06:33= Off the beaten track From the episode: Allocating to Alts with RPMs Alexander Mende on the Derivative Follow along with Patrik on LinkedIn and visit voltcm.com for more information! Don't forget to subscribe to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠The Derivative⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, follow us on Twitter at ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠@rcmAlts⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ and our host Jeff at ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠@AttainCap2⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, 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

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. Hello there. One week to go until Thanksgiving here in the US, which means we've made it to the end of another year on the podcast. Hooray. I hope you enjoyed the guests and the Thanksgiving here in the U.S., which means we've made it to the end of another year on the podcast. Hooray. I hope you enjoyed the guests and the talks and all the rest.

We'll be back in January with a new slate of guests plus some old favorites. Plus I may change things up a little bit, add some new segments. I'm not entirely sure. I'll noodle on that over the holidays. So enjoy this episode. Enjoy your holiday season. Okay, on to this one where we zoomed over to Sweden to talk with Patrick Salvenblad of Volt Capital Management.

Patrick describes Volt as a fundamental specialist and we dive into just what that means and how it makes them different than most managed futures program. We talk through the difference between pod shops and in-house multi-strats, why their models tend to be more shorter term, how long it takes to run 9,000 plus signals,

and more. Send it. This episode is brought to you by RCM's Managed Futures Group, which helps investors find unique managers like the one we're talking to today, Volt, which is managed futures, but not correlated to managed futures. Riddle me that.

Learn more on how RCM's team can help you find the right non-correlated piece for your portfolio at rcmalts.com. Now back to the show. All right. Thanks, everyone. We're here with Patrick Savenblad. Did I get the last name close? You got the last name perfectly close. Yes. I'm very happy with your pronunciation. How do you pronounce it in Swedish? I would say Patrik Savenblad. So it's kind of similar, but the vowel sounds are quite different in Swedish. Got it. And you are in Sweden currently. What part? Stockholm? Yes, I'm in Stockholm. I'm in the city center of Stockholm.

So on one of the main streets here in downtown. Great. And you're born and raised and spent your whole life in Sweden for the most part? Yes, I grew up a bit south of Stockholm and in the days when winters were still cold. Now it's much warmer than it used to be, but it's still dark during the winter. And then moved to Stockholm when I went to study. Got it. And give me, what do people not know that they should know about Sweden or about Stockholm? I think they should know that it's a nice place and they should also know that they should only visit in the summer.

Just too cold, too dark? It's too dark in particular. So the summers now are very, very pleasant and it's a bit warmer than London, for instance, and not as hot as some other places. So it's very nice to visit. The winters, you wake up and it's dark and you go to the office. It's still dark. You go home and it's still dark. You don't get these bright winter days that you have in Chicago say that, uh, uh, it's sort of all a little bit, you know, dark, dark. So, uh, no, no, I don't recommend it. It's nice for, for a day

perhaps, but, um, uh, you know, do living through, you know, 50 of these winters, I think I've had enough. Yeah. Where do you head anywhere for the winter? Get a little sunshine. Yes. We, we, uh, we tend to go to, to Italy during the go to Italy during the winter. So over Christmas, you get some light. So that's quite nice. And it always blows my mind. How many people, what's the population of Sweden in total? Sweden is around 9 million these days. So it's sort of the biggest of the Nordic countries. But of course, the Nordic countries are quite similar in size, all of them. Yeah, blows my mind of your Olympic teams are always very good, especially in the Winter Olympics. Right. And a population of nine million versus our 360 million, whatever the US is.

Yeah, but I think that we all feel that, you know, the Norwegians are out competing us in all the winter sports. So, yes, we have some good athletes, but the Norwegians are enormously impressive. They are as well. All right. Well, it's on my list. Next time I visit, I want to come see you. Yes, you should. Come during the summer, though.

Although I'm a skier, so I might want to come in the winter and get some skiing. Yes, but now we don't have snow any longer. So that was your 10 years too late for skiing around here. You'd have to go quite far north to get snow these days. Well, thanks for coming. I want to get into your program one up, but let's first talk. You've got a bit of an interesting past, professor, a few different hedge funds before starting your own gig here. So take us through a little bit of the personal background and how you got to where you are. You know, I obviously came to Stockholm to study and I enjoyed studying and I stayed, overstayed, became a PhD with the market microstructure as the focus.

So market microstructure, of course, that's trying to understand what happens when investors meet in the marketplace. How does sort of the composition of traders or investors affect how exchanges work. And, you know, I really enjoyed that. It's sort of a great learning experience. And I then became a finance professor for a number of years. And, you know, that was, you know, I enjoyed teaching, enjoyed learning things. But, you know, the producing research was a bit on the slow side. I think it's a very frustrating thing to do. And how old are you when you're a professor?

You must have been a young professor. Yeah, reasonably. I think I got my PhD at 30. So I was a professor like between age 30 and 33, or something like that. But so that was, so the question then is, being very interested in markets,

being very interested in market structure, what you do with these skills. And I became a hedge fund allocator. So I started out in Stockholm with a firm called RPM and I was allocating in the CTA space. So I was allocating to our trend followers and all these people that you also talk to. And I met many of all the people that have been on this podcast in that capacity years

ago now. And we've had, what's his name? I'm forgetting his name at RPM. We've had him on the pod. We'll put a link to it in the show notes. Alexander Mender, perhaps? Yes. Yeah. Yes. Yeah. Yeah. But I've met a lot of CTAs as well. Of course, Alexander is

a person I hired to RPM many years ago. That was definitely one of my best hires of people I've hired in my career. Very happy with him, he has a strong, a lot of experience and, you know, good insights. So very happy about that. Definitely.

So then professor, then RPM, and then on to where? Yeah, so then I became an allocator with the Norwegian state-owned bank. They were called DNB Norway in those days, now only called DNB. And that was running a hedge fund allocator team that allocated across all strategies. I was looking after macro and CTA myself, but I was allocating across all strategies. And you learn a lot from that. And I think the key, you know, learning experience in that job is that, you know, you can never, there's no, nothing that can compensate for like experience, you know, dedication to what you're doing, having a strong team, having integrity. And, you know, the other parts, you know, trying to understand the exact details of strategies turned out to be less important than figuring out, you know, who is good at

what they're doing, really dedicated to what they do. And that was sort of a good, you know, the biggest sort of takeaway from that period. But, you know, at the end of the day, I'm a quant. You know, the PhD damages you for life, I guess. And I needed to get closer to markets. And that's when I joined Richard Conyers and David Pendlebury at Harmonic Capital in London. And eventually

taking on the role as CAO there. And that's... What were they doing? What they were doing? Well, Harmonic closed down a couple of years ago, but Harmonic was a quant macro shop. And the quant macro shop is based on quite careful modeling and also using an approach that was borrowed from the fixed income space, namely being long,

short, neutral. So having equal risk long and short in everything that's Australian. And that was working quite well. And we had up to, I think, 1.9 billion in assets at peak. So that was a very quite useful experience. Unfortunately, the partners at some point, we started to have different interests and people wanted to perhaps spend more time with their family. Someone wanted to retire, etc. And at some point... The problems only someone with 1.9 billion under management can have, right? Well, you don't have those problems when you have 100 million under management.

No, you have other problems with 100 million under management. I think you're you're I think with one of the are a lot of lessons to be had from from that experience. And I think that, as you point out here, you're almost having more problems when you're doing well compared to when you are pressed against a wall kind of financially. Because you are using your excess cash to set up costs and set up structures that then become very difficult to undo in leaner times. Yeah. And looking back, I think the

harmonic would have been still around had we, as a group, not overinvested in hiring more people or taking on costs that made it tricky to handle in our times. And then coming back to your allocator days, you just piqued my interest there a little bit. Sounds like you're kind of saying that your quant mind didn't quite fit with the allocator. You could do all this work, you could identify which were the best performers perhaps, or the best risk adjusted, but then that might not have been the key. The key might've been just that they're dedicated and good people. Well, once you've identified them, then there isn't that much else you can do. You've identified that this is a strong team and they'll keep working 10 years and you can't really

improve on that by trying to be timing or something like that. That's, you know, that's fool's errand to try to do that. So in that sense, you know, running, even if, you know, running big multi-manager portfolios is relatively slow paced work. And it's, you know, when you are instead trading markets, every day you get a little bit of feedback on what you're doing and how well you were doing it. And that suits my personality better. And did you find it was hard? What was your experience on persistence of returns and some of the things in the allocator seat of some of the difficulties there? You found a great team.

They had performed well. Maybe they don't. And you have to have faith in that team to know that they'll come through the other side. Yes, I think that you have to separate two things. One is, do I have faith in the team? And that's going to be evaluated on things like, are they dedicated to the business? Perhaps they start having other interests. Harmonic was a good example of a setup where at some point people

started looking outside and having other interests. That's personal integrity and is the team stable? Those types of things. That's, you know, like personal integrity and, you know, is the team stable, those types of things, you know, that's where you decide whether you like the team. On the performance, you know, clearly there is a big picture, which is do I believe in this overall strategy? In my allocator days, I stopped believing in equity market neutral factor based, which was a big industry at the time. I was stumbled in 2008, I believe, or perhaps it was 2007, in this quant crash. And at some point, it seemed that that alpha had been, there was too much competition for

it. So that's a situation where you don't really care if a manager is good or bad. You just say that I don't believe in the strategy any longer. There could be other strategies like that where on the fixed income ARB side, for instance, there's certain subfields that we walked away from in the bank,

direct lending, for instance. And then when it comes to the strategy being implemented, I think the key there is that when the manager doesn't perform, it shouldn't hurt your portfolio. And you achieve that by allocating intelligently not too much to each manager. And also you're saying that, well, I don't care if you're good or bad, but if you're losing money, I'm going to be reducing your allocation and possibly to zero. But in a drawdown or a prolonged flat period, you can't be overly smart about or interpreting too much what's going on. But what

you can do is that you could protect your investor capital, much like we would do in trading where we have stop losses that stops us out of trades that we inherently believe in. But we also know we need to preserve investor capital. Stig Brodersen So somewhere along the way there, you became one of those people looking elsewhere at Harmonic and said, hey, let's start my own thing. Well, actually, I was probably one of the people that were more into keeping going. But regardless of that, I had a strong contact in Stockholm and the, we, you know, some, you know, I knew that there was a team being formed back home. And when, you know, Harmonic, well, when we dissolved Harmonic, I could then join that team and the friends from before. So I'm quite happy about that today with a group.

And I think that sort of a group of people with experience that all wanting to do things like the right way, if I put it that way. And I think that's been a very stimulating journey so far. And had you all known each other beforehand? So there are four partners, correct? There are four partners. We are five employees now in total in the firm. So five people total. And we kind of everyone knows someone, knew someone. So it's more like a chain rather than everyone knowing everyone. Of course, David here is our CEO. He has also worked at RPM, for instance, and told me who is our business development person. He used to be sitting in the same office as me at some point. And then everyone got to come back home, back to Stockholm. Or were those guys already there?

Yes, I was the only one moving back. But I think it's people with experience in the industry. And we wanted to do something that made a difference in the CTA space, identifying that there aren't that many fundamental CTAs out there and the ones that are there are quite much into the TA space. So we felt that this was a good business opportunity for us based on our previous experience. And we managed to get a local family office to support us with seed capital and working capital. So that was, you know, at that point

we just, you know, let's go for it. Let's go for it. Love it. So, yeah, let's dive into the, so what year was that? When did you guys start? 2017? Yeah, so the dive into the... So what year was that? When did you guys start? 2017? Yeah, so I think the process was sort of going on from 2016. The first trades were done in 2017. So let's get into the model, the strategy. So take us, you just mentioned a little teaser there, but take us what you guys identified as your potential edge of what you thought you could do. Yeah. So if I'm, I'll just make sure I know what I'm saying. No. So, I mean, the starting point here is that we want to be a fundamental specialist, right? So this is, you know, we know

that the world is full of trend followers, and that's not really a business to build unless you have, you know, you're already inside a big firm. We wanted to set up, you know, an independent, you know, classic CTA, if you like, right? Like the ones that were started 20 years ago, as opposed to being a team inside a bigger organization, good reasons for doing that. So we want to be a fundamental specialist. Of course, Jukka Harju, who's my co-PM here, he had experience of building fundamental models in his previous job, which was inside Links Asset Management. Now they are mostly known for trend following, but he was on a separate team that was building fundamental models. His focus was more on the commodity side. My harmonic, where I came from, was more on the financial side, so fixed income and FX equities. And that's the

sort of the skill set that we're bringing in. The thinking here was, let's look back and say, see, what are the things that haven't worked before? What hasn't worked where we worked before? And I take full responsibility for harmonics performance and what we did and the research process and all of these things. But there are clearly a number of things that I can identify

that I wanted to do different. I'm sure we'll get to some of those, right? And the same thing goes for Jukka. There are a whole bunch of things that he felt that we needed to do different. That was from a model standpoint, as well as we should trade these markets and trade. It was like a universe plus model standpoint. I think it's on the universe side, of course, we were constrained to the regular future space.

So that's not much to do there. On the signal side, we know that there is a lot of fundamental information out there that's not being used frequently in the CTA space. We wanted to use that and bring that to the market or to investors. But I think the

biggest, one of the biggest things that we both had seen in our earlier lives is that, you know, the world is changing. The world is always changing. You know, that's one of the constants in macro trading. You know, right now we have inflation. We didn't have inflation a couple of years ago. Now, that's a different state of the world. And the question is, how can we build a program in fundamental space that can handle the fact that the world is moving forward? When I was starting out in the CTA space,

you still had nat gas spikes in the summer. Now, that's just gone away, right? With this, you know, nat gas, the nat gas markets has just changed how it's operating because of all the shale gas and other sources and also, of course, better LNG availability in all kinds of places. So, you know, the world is changing and fundamental traders, you know, often have the problem that now the world has changed. My preconception about how the world works, especially if you're a discretionary, is not valid any longer. How do I move forward?

One thing that doesn't work is to do this in an investment committee setting, and that's something I experienced first hand. Investment committees you sit down, you have intelligent people, well informed well intentioned but the more you talk the less likely you are to change your mind. I guess we know this in all kinds of from politics and from other, you know, life situations as well. But, you know, as you are talking, you are convincing perhaps not the person across the table, but you're convincing yourself.

And you're convincing yourself to the point where, you know, your thought process become antagonistic. You know, my model is your model. your model, which one should we be cutting? If you feel it's your model, then you say, well, it has to draw down, but it will be a recover. If you think it's somebody else's model, you use it as the opportunity to stab someone in the back. Very, very unhealthy. As a CAIO, I obviously was part of this. And we were trying to find solutions. And the solution we were moving towards was to make more of a decisions formula based. Now, so that you had stopouts for models and you might have stop ins and you had certain conditions that needed to be met. And that fit very well with Jukka's thinking, because he had sort of taken

clean sheet design and said, well, how about if we just abolish the investment committee, we have our models, the models have observable characteristics, and we'll have a machine that handles all these allocation decisions. That's what we call machine learning when we talk about vault. And what we do is we do not sit in a committed setting and try to decide which model should be traded. Now we are discussing a model or sort of an overarching machine learning framework that will make that evaluation for us.

And in practice, of course, it's easy, or in theory, it's easy, but rather that how much returns do we expect from this model? How much does one model correlate with other models? And then you create some sort of of optimal allocation based on that. In my experience, the difficulty there is you crystallize the losses way more frequently than you get the gains, right? If you are always stopping models out, you kind of are crystallizing losses. That can be a difficult path. Yeah, you're definitely right on that. And I think that's where one of the

sort of the other innovations I would say that we have is we have a large number of models. And that means that when you're stopping out a model, you actually have something else that is similar that you can allocate that risk to. So it's a bit different from a situation like, so we had at Harmonic, we had the two models that traded emerging market FX. If one of those was stopped out, you'd sort of have half allocation to emerging market FX and you would miss out on the recovery. But in our setup, I'm not saying we're immune. This is definitely a concern. But by having a larger number of models, you have something else to allocate risk to. And you can do that.

And one of the nice things, by the way, is that in an investment committee setting, you would tend to allocate more risk to recent winners. And that's sort of the other side of loss, of losing money in a stop-out setting, right? Because you're stopping out something that then recovers is the same kind of loss as entering something that just made money and then loses. It's equally costly. It's just the flip side of the same problem.

But the framework is one that actually does not look at very recent performance as something that we'll be allocating more to. So those are things that you can control in a systematic setting, but that is very hard to handle in a person-to-person world. Right. I want to back up for a second. What do you mean exactly by fundamental? Because this is where it's always confusing me.

I'm a fundamental specialist but you're also a quant right so you're kind of bridging those two pieces so maybe a couple of examples of fundamental models or fundamental data that you're using to inform the models yeah so you know the the trading what we use it will look look quite similar to a checklist of a discretionary trader. It's going to be quite similar to that in logic. So if we are looking at something like crude oil, we would know what are the things that matter for crude oil. Well, first of all, of course, growth is good for crude oil. So you then go and say, well, how can we identify growth? Well, we can look at GDP growth, we can look at inflation, we can look at inflation expectations, things like that. Then we can look at things like supply and demand.

So we can look at how much energy is being used by, say, European refineries, by US refineries, by Chinese refineries. We can look at supply and demand in terms of storage. We can look at how much we can look at supply demand in terms of storage. We can look at demand in terms of the higher costs of the oil freight. If the price of over-rent inputs and outputs from refineries, so crack spread type indicators, that would be something that indicates more demand for oil. We can look at what investors are believing, are investors optimistic or post-pessimistic? So what can we learn from the equity market? What can we learn from the stock prices of energy companies? Well, I guess in energy, we have the heating days

and cooling days, perhaps that's more shorter term and it continues. Will Barron Right. So there's dozens. For crude oil, there might be how many? you know, shorter term and sort of it just, it continues, right. You know, For crude oil, there might be how many? Yeah. For crude oil, we have roughly 30. 30. And for something like cocoa, maybe five or something, right? Yeah. I think, well, I can actually check. I think it's more like 10 or so, but yeah. Okay. But yeah, but there's just not as many global. Let's split the difference. We have eight models in COVID.

There we go. All right. And so a normal factor model would say, hey, we've got these 30 factors, fundamental factors, but we've got these 30 factors. We're going to create a model that informs where prices are going, where we think it should be priced based off these factors. So you're doing something a little different than that, right? It's not just one factor model based on those, on that data.

Yeah, you could say that it's 30 factor models. You know, I think that's fair enough. Each model has its own prediction of where the world is going. And then we are using this sort of dynamic machine learning setup to allocate across these different predictions. But right, it's not one model saying here's the 30 and I'm going to wait. I guess it's somewhat the same thing. If I'm going to wait, factor 3, 75%, factor 8, 10%. In essence, it's going to be something along those lines.

For each factor you're looking at, there are two things that matter. One is how much or how profitable do we expect this factor to be? And that's going to be based on things like backtest performance, consistency of backtest performance, other things that we put into that side of the calculation. And the other thing would be to say is how much does this correlate to other things in our portfolio, both inside the market we're trading. I mentioned here for crude oil, we use inflation and we use inflation expectations. Well, obviously, those are very related factors. So as whole, we need to adjust to the fact that some factors are distinct but related. And that would be sort of a correlation type argument where you want to allocate more to signals that have fewer substitutes, if I put it that way.

And then it's fully dynamic with the machine learning. So what the weightings of these factors today might be totally different than they are six months from now or a year from now, or even a month from now? Well, the alpha is generated using the signals. The signals run with a holding period of 12 days. So we're quite a bit quicker than most systematic fundamental traders. And the machine learning runs more on a sort of, say, quarterly horizon, something like that, that the signal can go from fully allocated to zero in a month in some cases. But generally speaking, it goes a bit slower. And the idea here of machine learning is not that we want to override the signals. It's that we want to wait effectively to signals so that when the world

moves forward, one year from now, two years from now, five years from now, we make the portfolio sort of gradually move to the factors that are relevant at that point in time. Got it. And why the shorter timeframe? That's just how it worked out? Did you target that? Or as you were doing your modeling, it became shorter and shorter because you have more data points? First of all, we like that it's short, right?

You know, shorter term is helpful for risk management. If you want to have stop losses, and this sort of comes back to your question, isn't it costly to stop out models? One way of reducing that is that you have other models to allocate to. Now, the same thing goes on your trades. But if you have a stop loss and you were stopped out of your crude oil position or something like that, then it is useful that you have new fresh signals that you can allocate to.

So having shorter term signals means that you're less likely to stop yourself out and buying back exactly that same position. Like an FX carry trader say, if you're trading FX carry, you're guaranteed to buy back exactly the same position you were stopped out of, which means that if you are trading FX carry, you basically never use stop losses. Yeah, exactly. That's part of our due diligence process of like, okay, you use stops. When can you get back in the same position? Right away?

Okay, well, you really have risk then, which I've seen in real time. Like you kept buying it six months in a row and lose it. So it'll control that automatically by going into the different models. And so all these together, I think somewhere in your deck, there's like 9,000 plus drivers you call? Yes. If you take all the models times all the markets, you get a very large number of signals. And that's also one of the reasons we need machine learning to make the allocation for us. It wouldn't be efficient or we wouldn't be able to do this in a manual way with all the things that are moving. That said, a lot of these, of course, 9,000 signals sounds like a lot, but of course, they are variations on the

theme. So you'd have multiple timeframes of a similar signal, things like that, that makes the number look very large. Yeah. And then the end result is, it's how many markets? 70 or so? 70 markets. Yes.

And with a 12-day holding period, we trade roughly half of those every day. For people that know futures, we have 900 round turns per million on the portfolio as a whole. Got it. So the machine learning, it's giving you one portfolio every day and you're adjusting it, or is it considering each market separately or a little bit of both? So the machine learning is running a little bit of both, I think is the right, correct answer, meaning that you take both sort of the big portfolio and the sector into account. If you're trading energy, it will make a difference.

If you have a signal and it works in crude oil, if it works also in heating oil, then it will be treated in a slightly different way than if it doesn't, things like that. So we're trying to extract information about whether a signal will be profitable in the future, looking at all kinds of things, right? The consistency of the back-tested performance, but also, for instance, whether it works in other markets. And then, so each signal you have tested is profitable in its own right? No, actually, that's a very good point there because each signal is tested and we believe in it. We believe this makes sense. We believe that higher growth is good for energy prices.

We believe that's something that is true. We believe that we find ways of measuring this effect. And then if it makes sense to us, we will add it to the program. If it doesn't backtest well, the machine learning will basically say, well, I will not allocate to that. If it looks too bad, perhaps we'll take it out just to save on computing time. But roughly 10% of all the signals, perhaps even a little bit more than that, are zero allocated and in some sense are waiting for a different world or a different state of the world where regime where these can come in and be allocated to. And then how did you go about selecting all these signals? So do you have someone in-house who had that fundamental mind, right?

Like you need a different skill set to know the factors for crude versus the factors for cocoa versus the factors for Japanese yen. Yeah, no, I think that's a very good question. And it really just comes down to having been in the space for 20 years. So we look at each market repeatedly. Is there something we're missing out here in trading the Swiss franc or trading the cocoa or whatever it's going to be? And can we apply some concept that works in another market? Yes or no? And this is something that, in some sense,

is the sum of 20 years of work on my side and 20 years of work for Jukka. And you might not care, right? If it doesn't match up exactly with what the guy at Shell Trading is doing of his 50 factors or something, right? If they work for you, they work for you. Well, in some sense, I would even say, perhaps flip it around and say that, the person at Shell will tend to be quite, being discretionary, most of the time will be quite open about the factors used. So if you're listening to trader podcasts or things like that

with discretionary traders, they don't mind sharing things that work for them. And that's one of the ways you can learn something that adds value in our space without stealing their edge because they have a different approach. And that's also why I said here that it kind of looks like the checklist of a discretionary trader. Over the years, you identify things that matter. And that's definitely on the commodity side. That's following what other people do and talk about is very helpful. And then have you delved into alternative data? You're not getting satellite images of parking lots and things of that nature?

Not that well. We've tried a few times, but, you know, in the macro world, right, given that we are trading, you know, a handful of markets, perhaps, you know, 10 energy contracts or something like that, you know, a lot of the alternative data is so granular, that it's hard, you know, you have to kind of go through a very long exercise of sort of getting out refining a single signal. And we haven't sort of found anything that worked, but we tried a couple of times. We do use some indicators that are where someone else has done this work, right? They themselves have taken, I don't know, Twitter feeds or weather or something like that, created

an indicator that has predictive power. And I think in some sense that's preferable to us, because if you're building your own composite signal, you're going to curve it also in your signal, in your data construction, right? And if it comes from the outside, at least you're not overfitting your indicator construction. Right. But it seems right outside looking in, you must spend a fortune on gathering all this data, but is it all just relatively accessible via Bloomberg or something? Yeah, no, most, I mean, it's not cheap, but Bloomberg is our main sort of provider for run of the mill kind of data collection. It works well, it's a solid system. And then depending on what it is, we might

have either sort of FTPs and feeds and things like that to other providers. But by and large, Bloomberg works really well for the type of data that we need, stable platform and very happy with that. You mentioned computing time. How long does it take to run these models? And then compare that with what it would have been 10 years ago also. The running is relatively straightforward. I said we run with a 12-day holding period. We update the core signals three times a day. And the actual running might be an hour and 30 or something like that, 90 minutes, something like that. And I'm sure one could make that faster if we had to but what is taking a lot

of time is testing so we have more powerful machines that basically run you know not 24 7 but you know easily 12 hours a day testing stuff and I would and especially if you're testing portfolio construction type ideas you know that is you know very time consuming because you want to test for all the models that are live, all the models that ever were live, or some other models that you might not consider but are useful for testing, making sure that all the features of your portfolio construction make sense. Portfolio construction, of course, has all of this waterbed feature, right? You know, if you have a target risk, whenever you're cutting something, you're increasing something else to maintain your target risk.

And that's very time consuming. So generally, my day often starts as part of the research process by seeing, you know, what are the results that came out overnight from the machines? You know, you have read through the reports and you think about that and then you spend a day and perhaps towards the end of the day, you set up some new set of tests and then that can run overnight.

And then you come back the next day to see how that works. And the same thing goes for signals. Yeah. But is that frustrating? You want to be like, run it right again, right? I want to increase this parameter or change that and run it again. You got to wait all day. Yeah. Yeah, I know. It can be frustrating, but the thing is that you also can't just get a single reading and say whether it works or not. You need to look at a large number of dimensions.

And that's really what takes time. Because if we're now talking portfolio construction, clearly you have to think about slippage. You have to think about VAR and VAR constraints. You need to think about margin usage. And you need to make sure that things work in all these dimensions. So it's not that you can just sort of run it again and get the new result. That's perhaps more when you're testing the code that you can run it and it breaks and

then you can fix it and it runs again. And then how does that work? Do you still kill the investment committee? So if you have some new testing that looks promising, how does that work to turn it live? Well, we're three people on the investment side here. I'm the CIO, so I can decide, but generally we discuss and then if it makes sense, it goes live. I think it rarely, at this point where the program is mature, we've been running with the same ideas in different iterations for seven years. It's rarely something that is the source of conflict, if I put it that way. And I think the last thing that came in that was straightforward was that we introduced a separate value at risk limit for

grains. While we had one for each market and we had one for the sector, but we didn't have grains separate. And so that's an extra restriction. Now that type of thing, you know what it's going to do. It's going to cap your portfolio in certain situations, something else is going to get more risk. And if it's having these tests on sort of that run, you can quickly make the judgment call and say, it makes sense to do this, do it. And talk a little bit about those portfolio level and sector level constraints. I think we passed over that before. Yeah. So risk management, of course, it's key. If you want to stay trading for 10 more years, risk management is what you need. So I mentioned stop losses here.

Stop losses stop you from being in a position that loses money over and over, day after day after day. That's the purpose of a stop loss. You can have a one-day shot that continues, and we are in 1987 or March 2020 or something like that, you need to be able to get out. Now, the pre-trade risk management is really about balancing risks in the portfolio to make sure that you don't have too much risk on any single factor, if I put it that way. And we have risk limits, we use valued risk as our key tool here. We have risk limits on the market level, we have risk limits on the six sectors we trade, we have a risk limit on the portfolio. And roughly speaking, the risk limit is such that the max risk is not higher than 1.5

times the average. And if you think about trend following, trend following would tend to be much more opportunistic over time, right? So your max risk might be double or perhaps two and a half or three times the average risk as you have an empty book for a while and then you load up your positions. And some trend followers, of course, even gear up on profits and you can get even higher spikes. But we tend to and we target a stable risk taking. We want to make sure that our biggest one day loss is not too big, if I put it that way. We're looking at our own track record. I think we've had two or three

Sigma losses over the 1400 trading days we have. And that's compared to something like nine or so for a CTA index and perhaps even more for equities. And that's one of those things I'm looking after, looking for the portfolio on a really bad day. I think our biggest loss is something around 2%, shouldn't be losing more than 1.8 or so. Got it. And then on the, so that's market-based, you said, and then also sector-based. And then will it come adjust those on a portfolio level or it just gets there by adjusting the market and sector-based? No, and then we are capping risk also on the

overall fund level. But in some sense, there is a lower level. Everything needs to be roughly balanced. One level higher where it needs to be balanced and the top level where it also needs to be balanced over time. So question popped into my mind, how do you view this versus like a pod shop, right? Or when you were an allocator of like, I'm going to allocate to these 15 or call it 20 fundamental managers. I'm going to get that diversification. I can control it at the portfolio level by position sizing the risk. So compare what you're doing with that approach.

In some ways they're very similar, but in a lot of ways, very different. Yeah. And I think the biggest difference is the human side of things. right? You know, when I worked at DMV, we did have a multi-strat fund that I was involved in that had individual traders, or, you know, of course, someone trading energy because it's a Norwegian bank, and there's someone trading shipping because it's a Norwegian bank,

and there would be some other things going on. And, you know, the first thing you notice is that, you know, you can't turn off a trader and then turn him back on. You know, when you go to zero, you go to zero, and it's sort of you never come back. And I think that's a big, makes a big, sort of has a big impact on decisions. And the same thing goes, actually, when it comes to the, to sort of this research team approach where I mentioned that you have someone

who wants to push their individual model into the book. That if they don't get to trade what they want to trade, they'll be very upset and that's not helpful for the organization. And they might leave or something. They've spent all this time developing an option model and then they don't get to trade it and then they feel frustrated and they leave. So I think that's the biggest difference on the...

Right. None of your models are going to get frustrated and want to leave. No, exactly. For now, until AI really gets fancy. When the models have emotions, then we're in trouble. Yeah, could be. But so far, I think they are happy to live here in our private cloud solution. Just from a pure, do you think you get the same return profile? Assuming those fundamental traders could stay focused and everything's good, do you think you get somewhat similar return profile? Or I'll ask a different way. I think allocators, if I'm saying I'm going to allocate to Volt to get that profile, my worry would be like,

oh, there's too much similarity between the models, right? I'm not getting enough diversification because you guys created all of them. That's, I mean, that's an interesting take. I think that, you know, of course, we know that our models aren't that similar one to the other. Like, you know, weather model for trading wheat is not the same thing as, you know, trading crude oil based on what energy share prices are doing. So, you know, these are very, very different models. So you do have a lot of diversification.

However, thinking of diversification from an investor perspective, I think it makes perfectly good sense to say, here's Volst and they do one thing and I like the team and they do it well. And here's a trend follower that I trust and does things diligently and know what they're doing. And here's someone else who's a discretionary macro guy that does something useful for the portfolio and create diversification that way. On the future space, I think the examples of sort of just futures on multi-team

discretionary. I think that I don't have any recollection of someone doing that really well, actually. A lot of failures there. Myself included. What else do we miss here? No, I don't think we're missing anything. This is a snapshot of what Volt is about. You want to deliver returns to institutional investors. We want to do that in a way that is unique and it has risk control. We specifically have built the book so that we have good

opportunities in crisis. And that's why an extra reason to have strong risk management, because if you want to make money in highly volatile environments, you definitely need your stop losses. Talk through that for a second. So each signal basically has a positive skew on purpose? Or just by risking a little in the opportunity to make more? Yeah. So actually, what's really happening is that when you have shocks to the system, the world becomes a bit more predictable. So if you are walking into March 2000 or some similar event, on average, the world becomes a bit more predictable.

It's not clear where. In March 2000, we lost money in equities because we were long equities coming into the shop, but we made money being short energy and made money long-term income. For instance, the next shop that comes around might have another pattern. So it's important to be diversified. But what we have observed in our trading history is that we have had better than average chances of making money in bad months for equities or crisis in general. That the world is, when people are running for the exits, they are more predictable than you would call on a normal day, if I put it that way. St you're but it's also to me just a function of math right of like your stops are going to generally be about the same but the opportunity

is much larger right if if the range is greatly increased yes is that i yes i think you're it's um i mean in terms of skewness you know our best day is twice our worst day, and our best week is twice our worst week. So we know we have skew, and I think our best month is twice our worst month. So we know there is skewness in returns, and part of it comes from the stop loss type logic that if things go wrong, you reduce the position size. And the other part comes from when every once in a while, the opportunities cluster together. And that's true for us, and it's true for trend following, and it's true for discretionary macro. You are waiting for the setup where a lot of things move in the direction of your hypothesis.

Right. And that's a good point. The trend followers and macro, they're not necessarily designing the model to do that outside of the small stop profits run setup, right? But they're not saying we were betting against this or we're trying to get a crisis. It just sort of happens. Yeah. And I think that's something that is true also for trend following is that you think of something being, say, an equity crisis, but the opportunities, they come in random places in the book. Each crisis is different, and it FX program, and the FX program was a

perfectly good FX program. But every once in a while, you'd have this situation where the FX would just sort of miss out on some opportunities that appeared in the bigger book that were quite visible because that just was a shock that didn't make it into FX. And how do you think with the shorter timeframe, it seems counterintuitive that you could capture larger gains in a 2008 type extended move down crisis period. A crisis period tends to think of months long to years long down move in equities. When you're just 12 days, you have to be getting in and out successfully throughout that period. Yeah, I know. And then of course, in again and in again and in again. So it's really more thinking of that in the big space of instruments that we're trading, what's the best place to be in? And our winning trades tend

to be more like 20 days holding period and our losing trades are more like eight. So you do have an asymmetry that comes there as well. We are not forcing the models to have a 12-day holding period. This is more of a statistical thing. And that means that if you have something that is persistent in the world, the world is persistently going south or north, whatever we want to call it,

then our signals will naturally become more persistent. So, for instance, our best year was 2000, and we actually had a slower trading in that year than we had in average years because the world was persistent. The world was persistent in a crisis in the early part of the year and was persistent in recovery in the later part of the year. And then given all your experience allocating to trend followers

and other CTAs, would it be fair to say you wanted to, lack of a better term, have less negative carry, right? Have less bleed between those crisis periods in designing this model? Yeah, that's definitely right. You want to lose as little as possible, you know, as you're waiting for the next big move. Now, clearly, there are localized crises in these individual markets where we are trading. But, you know, the target here is to make money in every market every day. But the big opportunities, they'll line up every once in a while. But I think it's safe to say

that there is no automated bleed, if I put it that way, that trend following as an option replicating type strategy will have, you know, leave, you know, like the sort of long choppy whipsaw losses, those periods that they will be part of the strategy. And that's a feature. And we definitely try to, that. Yeah, I wanted you to say yeah, I was sitting in my office at these other firms screaming at the wall saying you're down again, down again. I think trend following is an interesting product because you obviously have to have investors that can handle these long flat

periods or long negative periods. And minimizing your bad years or bad months is key to keeping your investors on board. And in our way, we're setting up vaults. We're setting this up to trade for 20 years or so. And that means, or longer, that means that there will always be a new CIO at the firm that's investing, that comes in and wants to kind of clean house a little bit. If you're at the end of your drawdown, you're always at risk of being kicked out from the portfolio just before a big move. We've seen many of those things

happening over history, of course. Many. Your goal is don't stand out, right? Just survive in there. I think this is a sort of a general comment for anyone who's managing money right is that you you you don't want to stand out on the downside ever right you know you because you're you're really really at at risk even if you made money before right even if this is um you know, you're, you may, if you're, if you're down 20% a year, even if you made it up 40 of a year before, you know, people are going to feel this is too much for, to handle. And I think that's just, you know, keeping, you know, keeping,

keeping an eye on that or sort of having a strategy for how to handle the downside. I mean, that's very important. I think we'll leave it there. Thanks so much for your time and all the good info on Volt. We'll put the website and some other goodies in the show notes. Anything else you want to share? Where can they find you well they can um well they can google for me and send me an email that's easy as easy as that

or fly to stockholm yeah no i think that's actually one of the it's a good point actually but we we are in stockholm and uh it turns out we probably have more people calling and more people visiting than Harmonic used to have in London. Really? And being a little bit off a beaten track means that people are out there. They are looking for things that are different. Nowadays, of course, you do a Zoom call or something, but it's definitely not a disadvantage to be in Stockholm for that. I would guess you get few and far between visitors, but that's good to hear. Mostly Europeans, I would guess.

Well, I think perhaps it was more reflecting that being in London is not giving you any more, right? I mean, we are a niche player in a small industry, so it's not that we are being disturbed every week here. But we do get a good number of people calling us and a good number of Zoom calls and what have you. And I can actually reveal that I think our Nordic business is 25% or so of our total business. So we are in Sweden, but our client base is not Swedish. It's a broad European. We have clients from Australia. We have US clients. How do they handle that? Do they do it

in their local currency? The Nordic business? Yeah, well, with futures, of course, it's straightforward. We have funds and the funds will have Euro and dollar share classes. If you are running managed accounts, of course, you handle the funding in your preferred currency. So that's quite straightforward. We have eight managed accounts. So those clients, they know what they are doing and set things up in ways that work for them awesome love it well say hi to the team and uh

look forward to visiting you in person one day yes you're you're most welcome we we have never been plenty of nice well no there are plenty of things we can we you to and let you see and experience for sure if you come by. Otherwise, I trust I'll be seeing you here in the industry. Yes. All right. Thank you, Patrick. Great. Okay, that's it for the pod. And as mentioned at the top, that's it for the season. Thanks to Patrick and Volt for today.

Thanks to all our guests for the year. Thanks toff berger for producing and making us look good have a happy thanksgiving holiday season and new year and we'll see you back here mid-january peace 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 rcm Alts 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. 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.