The Next Evolution of Trend Following ft. Nick Baltas

[00:00:01] Intro

Welcome to Top Traders Unplugged. In markets success doesn’t come from predicting what happens next, it comes from being prepared for what you can’t predict.

In each episode we go deep with some of the world’s most thoughtful minds in investing, economics, and beyond to understand how they think, how they prepare, and how they decide, and the experiences that shaped how they see the world. No noise, no short-cuts, just real conversations to help you think better and invest with confidence.

[00:00:35] Niels

Welcome and welcome back to this week’s edition of the Systematic Investor series with Nick Baltas and I, Niels Kaastrup-Larsen, where each week we take the pulse of the global markets through the lens of a rules-based investor. And let me also say a warm welcome if today is your first time you’re joining us and if someone who cares about you and your portfolio recommended that you tune into the podcast, let me also say a big thank you to you for sharing this episode with your friends and colleagues. It really does mean a lot to us.

Nick, it is wonderful to be back with you this week. It’s been a little while. How have you been?

[00:01:08] Nick

It is quite a while, actually. I’m doing very well and good to see you, Niels. I’m doing very well.

[00:01:13] Niels

Absolutely.

[00:01:14] Nick

It’s a very, very busy summer. We’re still busy. I’m looking forward to kind of the summer break in a week.

[00:01:21] Niels

Yes, I think we kind of took a head start. We’re wearing matching summer T-shirts. We are in our Friday.

[00:01:29] Nick

It’s a Friday. It’s a summery day in London. Yeah, it’s a summery summer actually in London.

[00:01:33] Niels

Yeah, absolutely. And I know that you literally just landed to get here for this recording. So, I really do appreciate the commitment to the show this week.

[00:01:48] Nick

Always, always.

[00:01:49] Niels

Good. We have a few papers, a few different topics. I’m sure we’re going to go off track and do other interesting conversations as well in the next hour or so. But as you know, I always love to hear kind of what’s been away from the topics, what’s been capturing your attention in the last few weeks that you have found kind of interesting.

[00:02:14] Nick

I mean, we always talk about kind of the markets or in a way, we talk about the markets, we talk about some of the asset class moves and maybe some client activity. Maybe I’ll just change the script, so, it’s been like 10 days, I’ve been kind of listening to another book. It’s an interesting one because the title, maybe you know it. And we did not prep that at all. So, it’s come as a surprise to you. It’s called Die with Zero.

[00:02:42] Niels

No, I don’t know.

[00:02:44] Nick

Yeah, a friend recommended that to me and said, listen to this, or read that, and kind of give me your view about it. And it all talks about how we always work long hours, and we strive to become better, and get paid more, and so on, and so forth. But the reality is that while money accrues interest, we as human beings accrue memories and memory dividends are ultimately what we care about when we grow older. And the whole point is, try to make the best you can to put out time and money to eventually invest in experiences. Right?

Which ultimately, you make your life, period. And therefore, kind of looking back, the more of those memories you develop at the right age, with the right people, or with the right, I guess, setting, the happier you’ll end up being as time goes by. So, it’s a bit of a reflection point because obviously we all grow up, we end up having kids, they used to be babies. Now they’re like no toddlers. At some point they will go to school and soon after that they leave the house. They leave the house. So, it’s an interesting one, frankly, it’s an interesting one. So, like, Die with Zero. There you are. That’s what is on my radar. How should I redial my day to day on that basis?

[00:03:57] Niels

Yeah, absolutely. And maybe it’ll be the audiobook that I’m going to listen to on my next long drive from Switzerland to Denmark.

[00:04:07] Nick

Do. Do and let me know what you think.

[00:04:08] Niels

Yes, I will.

I do agree with this thing about experiences. I was sharing with Yoav last week that actually I had one, kind of in that line, in the last couple of weeks because I did go with my family to a wonderful experience, which was a concert in Milan with Bruno Mars. And I like his music, but he’s an amazing entertainer. And it just was one of those experiences that you think back on and say, yeah, I’m going to remember this for a long time. So, I completely get it, yeah.

Speaking of experiences, speaking of memory, I’m going to change the subject a little bit because one of the questions I get a lot is how does AI impact CTAs and trend following? And I read an article, in the last few days, about one of the key things that’s changed this year is that the “best model” at the moment gets replaced like every 10, I think it was every 10 days or every 19 days. Unlike last year where the best model could last for three months or four months and probably much longer before that.

So, one thing is us as individuals, okay, we can probably just switch to whatever model we subscribe to and it’s not going to make a huge change. But if you’re sitting in a big corporation and you’re trying to manage, how do you get the most out of AI? How do you make sure you have the best tools? And especially I would imagine in finance, if it used to heavily to get some kind of advantage, I feel it must be very, very difficult now for firms to manage the whole process of how do you integrate AI if it’s changing all the time? Not from like, okay, you have one OpenAI model to another, that’s one thing. But these are different firms. These are completely different models that become best.

So, with your seat, sitting in one of the largest, most successful investment banks, I don’t know how much you use AI yourself, but generally speaking, and I know it can only be general, but how do you think about keeping up with AI, if you know what I mean? How do you even think about it nowadays with what’s happening in that field?

[00:06:32] Nick

I mean, I would say, just personal opinion, and you’re right in the sense that even taking out corporations for a second, even us human beings kind of keeping up with how those models progress is, in itself, a challenge. And maybe to your point, you might as well just use the version, I don’t know, T minus 1, you’ll probably get pretty much all you need without even having to upgrade to the latest model because the latest model will become your current model the month after. So, I agree with your point, just to start with.

Now, going into how corporations are like how, broadly speaking, research teams, or product teams, or organizations like ours are looking into the space. Of course there is a lot of deployment, there’s a lot of focus. There are technology teams that have been put together to work in those models and see how they could be incorporated in our day to day, obviously under supervision, and guidance, and human in the loop and processes.

These are important not just because they are important, but because also, by regulation, we’re required to make sure that everything works and operates in a sound and prudent manner. I guess the use of those models, the primary use, at least in my view and the things that we do that we see on the day to day, is more to enhance research. And there are ways that we utilize it, and I personally use it to enhance research. And that has to do anything from collecting information, synthesizing information, speeding up if you like, the thought process through a model.

I have my own debates, and these are more philosophical as to how the human brain will kind of continue evolving if ultimately everything becomes a prompt and the prompt spits out an answer. I think we’re at this important inflection point whereby we have been trained without it and we can assess how the output and we have the wisdom or maybe the experience or the expertise to call the bluff from the truth.

Obviously, the models become better. I remember, like maybe a year ago, I was asking one of the LLMs to do some research on the latest, of all things, trend following papers and it was just coming up with all sorts of hallucinations, random people, random names. But you could find that some of the names or some of the paper titles were actually validated but the association it was making between all of them had no ground. These days is way better, frankly. So, you see the evolution through time.

But I think the critical thinking is required, at least for now, to assess the output. There are cases whereby the output is of closed form in the sense of prove this to me and then you can actually read through, with sensible assessment of the output you can argue whether this is the right thing or wrong. But open questions like, give me an overview of performance of trend followers year-to-date. You know, I did that maybe three or four months ago. Had I not known what the performance has been, I would have like okay, that sounds reasonable. It was dead wrong. Right?

So, I guess very long winded answer to say that it’s a significant efficiency tool when it comes to coding. When it comes to researching, I mean we have to operate with human in the loop. And I think that goes beyond what Goldman Sachs or any other bank or organization would have to undergo.

I think it’s more like the social impact eventually this would have, obviously the token usage, the cost associated with it. These are things to come and it’s hard to predict how they will play out.

But yes, I think it’s one of the revelations or revolutions of our generation. I think the previous one maybe was Internet. Frankly I cannot recall such a massive change in the way that we operate in such a short period of time. If we think of ourselves maybe like a year ago or two years ago, it was actually quite different. It was different.

[00:10:38] Niels

I’m not a techie but I am interested in the space, and so at least what I’ve managed to do is kind of build a little agent that goes out every morning before I wake up. And I’ve given it certain things to collect for me, like performance of the main CTA indices, performance of the main tradition. Pretty much the indices I talk about on the podcast, by the way. And then also finally the five most interesting relevant stories about trend following and CTAs.

And I will say, I mean first of all it does not always get the numbers right. Maybe the websites are not completely updated, but most of the time it actually does now. And secondly, it actually has flagged many of the research papers we end up talking about. So, I’m fascinated by it and I need to come up with a name for it, but there we are.

[00:11:30] Nick

But I think maybe my last sentence on the topic, you said you’re not techie, but I think that’s the biggest beauty or maybe the amazing thing about it, right, it’s a significant enabler, but that’s where supervision requirement becomes more and more important. Right?

Because the output of it has to be somehow supervised and in the absence of some expertise, or experience, or both, it can be uncontrolled until maybe it becomes smarter than any of us. And supervision comes with an AI supervisor. But we shall see. We shall see.

[00:12:05] Niels

Now if we didn’t have AI, Nick, we would have to work a lot harder. So, I’m glad to tell you that the other point I found on my radar is a news story from the American Heart Association because they have concluded now that for adults (for most adults, I should say), actually, roughly three to five cups of eight ounces of coffee is actually safe and it doesn’t seem to have any negative cardiovascular consequences.

It might even be associated with lower rates of heart disease, stroke and failure. So, I guess that’s a benefit. Now it doesn’t mention anything about the risk of being around someone who just downed five big, large cups of coffee.

[00:12:53] Nick

I’m a buyer. I’m a coffee lover. I’m a buyer. I’ll take it.

[00:13:00] Niels

Anyways, let’s move on to some trend following updates. I mean we had a weak June. Well, I’m speaking, generally speaking. It might be much better for you. I know you’re having a good year. We had a weak June as an industry, we had a soft start to July. Now things seem to be turning around in the last week or so for CTAs and trend followers and it seems to be led by energies, currencies, fixed income. But what’s interesting, at least from where I sit, is it seems like it’s not all markets within a sector that’s participating. So, it’s like a trend pickers market if there’s such a thing (that’s probably not really a term).

So, we are definitely seeing an improvement. I think we’re seeing positive numbers for the main CTA indices now. And so, I’m curious to hear your thoughts on the trend environment at the moment. And maybe also (I know you can’t share anything confidential), kind of what you hear and what you talk with people about in terms of their interests. Do you see people getting excited about putting on maybe some diversifying risk given the fact that there are a lot of geopolitical risks out there and we have seen some weakness in certain parts of the equity space? So, just curious to hear what you are coming across and what you’re seeing.

[00:14:35] Nick

Yeah, for sure. So, I think on our side I would possibly reflect and confirm your points with regards to July. I think it’s been a good month, actually, so far. But this is primarily driven by rates activity in the last week, really. So, the selloff in the rates and obviously the short exposures has been helping quite substantially.

Some of the models that actually tried to put a threshold on how aggressive they would be with weak signals have been lagging a bit behind because, in all fairness, with a weak signal you have to also be lucky. And obviously, a short position that is not too large, in the absence of strong signaling, could be deemed as lucky if eventually you end up having a selloff in the market and you sort the market even by a small amount as the month kind of kicked off.

Commodities is certainly the other segment that has been helping both year-to-date but also month-to-date. Equities and FX are more like flattish or bit down month-to-date. But at least that is for us. That’s the seventh month of the year, right? That’s the fifth out of seven, assuming it closes positive with a good return.

[00:15:54] Niels

Sure.

[00:15:54] Nick

Setting aside, obviously, June that we had this reversion in primarily equities. And then what’s the other one? I’m just checking here – March.

So, it’s been a good year for the variety of kind of programs we have in that space, and it’s also been a very good year for diversifiers to trend following. So, for example, carry has been very, very strong both in the FX space and, as you would imagine, the commodity space with the backwardation we’re seeing on the back end of the geopolitical activity.

In terms of (at least in that space), in terms of investor interest, there is certainly increased interest across all fronts. There is increased interest perhaps driven by the recent performance. There could be, obviously, data points associated with some level of protection. And I know we’re going to discuss a few topics later. Again, back on the topic of how trend following can be seen as a defensive trade.

So, all in all it’s been a good year, and it’s been a good activity year for the trend space. And the entire pitch of this kind of muti strat, the enhancement, and so on, and so forth, I think we’ve discussed it quite a number of times. I think the reversion space, kind of skewness driven reversion space has had a very strong four or five months of the year. It’s been a bit more kind of mean reverting itself. But when we talk about diversifying entities and diversifying premia, it’s a month whereby trend following is doing very strongly, reversal not so much but the aggregation of the two is still a positive number. So, it’s been good validation of what I’ve personally been a big fan of for the last two, three years that we’re chatting.

Outside of the trend space… Or maybe should I make a pause here if you have any comments or questions.

[00:17:46] Niels

If you want to add one thing, if you want to add one thing as you go through this would be if there is anything that kind of surprised you about 2026 so far, anything that kind of stands out. But other than that, continue to talk about the other topics or the other things you wanted to bring up in this context.

[00:18:07] Nick

So, I mean, I would not say surprising, but I would say, this is something we have discussed, I don’t know, however many times, obviously here, not just with me and you know, with the rest of the crowd, you know the typical V shapes. You know, the one per year. So, we had one this year, right, in March.

[00:18:23] Niels

Yeah.

[00:18:25] Nick

But it wasn’t really the same as it was the years before because it was more of a fast recovery rather than like a fast drop and then a slightly slower reversion. But still in itself was a V shape. So, it wasn’t necessarily the environment for trend followers to operate. It wasn’t as impactful as it was for the years before.

But what I was, I wouldn’t say surprised, maybe happy with the validation of it is that some of the work that we have put into thinking about dynamic volatility scaling actually played out quite nicely.

[00:18:56] Niels

Okay.

[00:18:57] Nick

So, it was quite beneficial in this regard. Anything surprising? Nothing really surprising, as in there’s nothing that we had expectations for, maybe, at least from the way that we look into performance, it’s been a good year. And could we foresee a year whereby, I guess, the equity markets will be rallying as they have, obviously, with the temporary pauses over there?

[00:19:25] Niels

Yeah.

[00:19:25] Nick

Having those trends playing out, maybe yes, maybe no, But nothing has been a big surprise, really.

[00:19:35] Niels

All right, let’s go and run through the numbers. These numbers are as of Wednesday, even though we are recording Friday, but they have not updated as of last night. So, as of Wednesday, BTOP50 is up 1.34% in July, up 10% so far this year. SocGen CTA index up 0.89% in July, up 10.37% for the year. And the Trend index up 94 basis points, up 10.15% so far this year. Short-Term Traders Index, unfortunately, down about 0.6% and up 4.52% so far this year.

MSCI, this is as of last night, down 84 basis points in July, up 9% for the year. US Aggregate Bond index, down 96 basis points, and down 12 basis points now for the year. And the S&P 500, that’s definitely a wrong number that I would have typed. So, I’m not even going to mention it. I think it’s from last time I spoke to you. I’m thinking, no, no, we’re not down 3% year-to-date. Okay. All right. That happens when you have a busy schedule.

Anyways, let’s talk about the paper. So, we kind of divided the papers up to two in two groups. There are two papers that came across that we want to mention, we want to acknowledge, and then there are two papers that you kind of have a much closer relationship to, which, actually, Yoav kind of stole our thunder a little bit, in a good way, because he gave you very high praise for the conference you had put on recently and where these papers were mentioned. But since you are much closer to them, it’s great for you to come and walk us through them.

But let’s start out as we planned. And the first paper, it’s very short. It’s produced by our friends over at Graham Capital, and it’s a topic about Market Universe. The title of the paper is An Expanding Market Universe, and it’s written by Nastja Bethke and Thomas Feng. And yeah, can you say a little bit about what their main kind of… It’s about, as we’ve talked about many times, how many markets you need to trade, and how does the benefit show up, and when does the benefit kind of disappear? What’s your takeaway from their quick, small paper here?

[00:22:05] Nick

Yes, this is one of those examples whereby you know that, at least from stylized facts, this should be the outcome. But you have never actually seen the evidence for it. And like, yeah, of course it makes sense as soon as you kind of see the evidence.

It reminded me a bit of the work by Man that we discussed a few months ago. Right? And they were making the point (I guess for those that haven’t actually listened to that episode), they make a very interesting point which is a trend follower does two things: high level long-term positive returns and some downside protection – some crisis alpha.

Now, depending on which one of the two is your major objective, you can make the universe smaller or broader. So, if the focus is to have more crisis alpha or defensiveness, you should focus on a small core universe because in a downturn, the principal components dominate. Equity markets as a whole or duration are the ones driving returns. So, the more associated you are with those, the greater the return in a downturn.

But if the objective is to have a longer-term performance, more diversifying entities should come to your universe and therefore you should add more markets, perhaps at the expense of defensiveness. They would not necessarily hurt defensiveness, but they will not add to it. So that was what we had at the time.

[00:23:30] Niels

Yes.

[00:23:32] Nick

So, what this report now talks about by Graham is to say, hey, assuming you have those kind of core 60, 80 markets, let’s start progressively adding more in terms of kind of liquidity. So, you would progressively start adding less and less and less liquid markets by design, they become more exotic and therefore they are less liquid.

And what you end up finding is that, broadly speaking, as soon as you cross 70 to 80 markets, the less liquid markets do not really add much in terms of performance. But then they go one step further and say this result, it’s actually more of a kind of conditional result on what those markets are. So, they split those markets into kind of more macro oriented markets like equities, and currencies, and fixed income. And this is where they see no incremental value.

So, if you go to the less liquid markets, less liquid currencies, or maybe some emerging market equities, or maybe some, I don’t know, interest rate swaps, the underlying principle of return is still very macro driven. The principal components are there with a preexisting universe just by adding an incremental market of 1, or 2, or 3, or 4 ,or 10, or 20, it doesn’t really add dimensionality in your universe.

So, there is not too much of an enhancement. But there is some enhancement if those markets end up being more exotic commodities. And then they have non-Chinese commodities, showcasing that you can now expand the universe to more than 100 markets and still get some incremental value when it comes to long-term performance. So, it’s kind of related a bit to the Man work.

And, at the end, they also make the point that look, those markets come with lower liquidity. So, do not anticipate the performance increase to be as linear as you would expect it to be. There will be a bit of a hump coming from the lower liquidity. Right? So again, is it surprising? Not very much. Was it in line with expectation? Yes, it was. Is it nice to see it kind of on a piece of paper? I mean it gives some comfort that, yes, it’s exactly what we knew. So, that’s the whole thing.

[00:25:41] Niels

Yeah. And just out of curiosity, where do you sit on this? What do you think is a good range in your view?

[00:25:52] Nick

I mean we typically operate between the 50 and the 100.

[00:25:55] Niels

Okay.

[00:25:55] Nick

So, I mean, coincidentally, it seems to go kind of hand-in-hand with how we think about the world being around like the 20% asset class bucket.

[00:26:04] Niels

Right.

[00:26:05] Nick

It seems to be like another level that we kind of settle. What it does validate is that attempts to go broader either come at the expense of capacity or no impact whatsoever. Because you basically end up, if you have some kind of top level risk distribution, and you say my rates markets will contribute X percent of my total risk, the more rates you add into your program the less each one of the markets would contribute individually. Right? So that the sum remains the same.

So, if you’re adding pretty much the same principal component at lower liquidity flavors, you don’t really add performance, but you’re basically kind of impacting the liquidity of the overall program. So, it is, frankly, validating some of our results. And the bigger, I guess, value add is really coming from the commodity space.

[00:27:03] Niels

Yeah, yeah, okay, makes perfect sense, make perfect sense.

All right, let’s move on to another paper that came on my radar. It’s from the company called One River Capital. It’s called The Perfect Hedge. It’s by Patrick Kazley, he’s the author here, I believe. And it’s certainly playing on themes that we have seen before, just like the other paper, for that matter. I mean, sometimes, I guess, it’s also difficult to come up with brand new things to write about. And I should know because I’ve been talking about trend following now for 13 years every week. So, I know it’s difficult.

So anyways, it talks a little bit about kind of the first responder, second responder, and all of that. Was this again, for you, just a nice kind of confirmation that here’s another group that kind of buys into it or finds the same argument valid, or was there anything that you noticed that was a little bit different from, I think Meketa wrote about this, I think you wrote about this as well. Anything?

[00:28:17] Nick

Yes, so, as you say right now, Meketa has written about this topic, like the risk mitigation, for years and we have done very similar things in terms of devising frameworks to design defensive solutions on the systematic side. And to your point, they basically break down a defensive solution into kind of three major buckets, the first responders and that’s like more like kind of volatility option-based hedges. They come with low basis, high cost, reactive around flash corrections.

Then trend following kicks in as a second responder. And we know that quite well. It’s the crisis alpha but not the few day crisis but more like the multi-month crisis. So, no surprises there. And then, obviously, there’s a third bucket, not a responder anymore. And it’s more of a diversifying element that can moderate the cost of carry for the other two. And this is obviously the more alternative risk premia maybe or flat to negative correlation to equity markets. Obviously not convex, but certainly not concave either.

And you know, they present all those ideas, they try to describe the benefits as well as the downsides of each one of them depending on the economic regime somebody’s looking at. So, frankly not a significant level of surprises or misalignment. I was actually happy to see that some of those views are shared by the industry. And it also validates the fact that us addressing those needs for the last good amount of years suggest that there is interest, but there is also focus and institutions start becoming much more focused on protecting the downside. So, that was a bit of an indirect testament. Right? If somebody still, to this day, is putting out those reports, it is obvious that there is need and demand.

Some of the small snippets, that were actually very useful, and I very much agree with, and I’m a big proponent of, are when they talk about, for instance, what is the diversifying entity here that is going to help the other two not to bleed significantly through time? It cannot be a short gamma exposure.

And, at least from my experience in seeing the industry, there are times attempts to make defensive portfolios perform better in the long-term. But what you end up doing is that you’re almost recreating some downside risk from your diversifying entity to finance your hedge. And by all means there’s nothing wrong with that. But this is not, anymore, a defensive overlay. It’s a relative value trade.

So, there were a few snippets here and there and that was maybe the more relevant that I could very much associate myself to it and relate with it. Beyond that they have like a number of good analysis, for those that have the report would see some value very similar to some of the things that you and I have discussed in the past. They have some trend models and they show on a per year basis which model does better. They have different speeds for trend followers and they show which year which speed has done better. So, there are a few snippets as such.

I think there’s a third one that says should we have an integrated trend program or a portfolio of asset class trends, which one is performing better in which year? So, there is a bit of analysis, for the trend lovers, in the report. And I guess perhaps that’s part of the reason why we’re also chatting about it today. And that’s it really.

So yeah, there is nothing that I disagree with, frankly speaking. Yeah, pretty much aligned with my views and I think you’ve heard me speaking about that many times here.

[00:32:01] Niels

I have indeed. A couple of questions that kind of sparks from this, in general. One is this idea that maybe investors might think that if we made the trend following side more responsive, i.e. shorter-term, it will be a better hedge. But I think, I don’t know if this is in the paper in terms of a conclusion, but I think you and I probably would agree with that the challenge is that yeah, it might be better for a few days, but it ends up costing you more in the long run. Again, going back to this idea that short-term trend following doesn’t really work in the long run and it’s probably become worse in the last decade or so. But I think that that’s again, something where it might be a little bit counterintuitive for people to think that being a bit slower might be better in the long run. So that’s just one thing to think of.

The other thing I wanted to ask you, actually, is what has become also very popular, is portable alpha. I think a lot of investors probably would love to have something that can help them out when their equity portfolio is under stress. And I’m just using equities. It could be fixed income for that matter, but let’s just use equities as the underlying traditional asset here.

Now, what would you say if someone came to you and said, Nick, I really like this idea of portable alpha, but I come from a world where adding leverage to my portfolio by putting trend on top of my equity portfolio feels unsafe, risky. How do we best make it clear that adding “notional leverage” is less of an issue because the kind of extra leverage you’re adding is “the trend” which, historically (at least we can’t talk about the future), has been negatively correlated when your equities have gone into a longer crisis. How do we demystify the word leverage for people who are used to, no, you can only invest US$100 if you manage US$100.

[00:34:24] Nick

Yes. So, I guess two things on this one. So, I guess portable alpha has become, again, popular after like 15 or 20 years. But to your point, this whole concept of unfunded or stacked exposures has become indeed popular. And of all things, trend following is maybe one of the vehicles that is utilized in that context.

And then to answer your question directly, my take on that would go along the lines of enhancing equities with a trend follower. I think it should come from the perspective of perhaps reduced upside participation in the trend program. So, you don’t, by design, double up on your equities risk on the upside.

If it’s an equity/bond portfolio, I think the story is slightly different because ultimately your biggest concern is inflation spikes and equity/bond correlations. And in that context, at least in my view, a more unconstrained profile, at least statistically, is here to protect the main shock scenario for an equity/bond portfolio. The equity/bond portfolio is not facing its worst challenge when equities are falling. It is when both are falling.

And that is historically associated with inflation shocks, which is precisely those events that happened not overnight, but happen over the course of a quarter, if not more, which is precisely where you start observing those negative trends in both, which is exactly what your medium-term trend followers here to address. So, in a way being very vigilant about macro shocks, what an investor would do is to look into the macro regime and policy reaction and try to deploy hedges for the entirety of the equity/bond portfolio, which is precisely the times that whatever the trends are, a trend follower would capture them. So, that would be my approach of demystifying.

Now, what are the downturns that your portfolio is exposed to? In a pure equity is equities down. So, if it is really equities down, maybe moderate the positive equity risk you deploy in your trend follower. And that’s okay.

[00:36:43] Niels

Sure.

[00:36:43] Nick

But it’s not that the increased leverage is creating the concern here, it’s the increased leverage on the asset that you’re highly concentrated. And that’s how we’d probably approach that and look at it. That’s my take at least.

[00:36:57] Niels

Yeah, I think that that’s absolutely fair. Absolutely fair.

All right, well, let’s move on to one of the two papers that you have had some involvement with and that we mentioned briefly only last week. So, I’m excited to get into it in much more depth.

The first one is called Thematic Investing as Missing Factors, by Wai Lee. There might be another author, I’m not entirely sure, could be. Now, it looks at the construction of a theme specific basket. I think it could be an industry, could be a style. But anyways, I’m not going to steal any of your thunder. So, tell me a little bit about the paper and why we’re going to be talking about it today.

[00:37:44] Nick

Yeah, for sure, for sure. And, I guess, to clarify my involvement, it’s more about because it’s a topic I feel very engaged with, and I’m following it, and I’m spending time kind of doing a bit of research, speaking to a couple of people in the space. But I have not been a contributor to those papers by any measure.

And Yoav was very good, in terms of kind of going through some of the findings. He was indeed, here, in the conference we did. And obviously, kudos to Robert Kozowski for kind of allowing us to host the event here at GS.

And I must say, in passing, and we’re going to go to the thematics, maybe connected to your earlier statement, AI and the use of the machine learning and AI models has become, now, very, very common in pretty much all the topics we had. So, there were topics about portfolio construction or how can we basically utilize machine learning in that space, or how we can do equity market forecasting, or here’s like a nonlinear model to do so, or let’s look into sentiment and thematic associations, or here’s another kind of LLM model that we can utilize to do so.

So, in one way or the other, you start seeing all this financial economics research coming up with new technologies in terms of tools. So, I’m making that point because we did not focus on having that as a topic, but it was the overarching, if you like, thematic right through throughout the day. Right? So, it was a good day. So those that came, I hope they had a good time.

So, I’m sure if we go to the thematics, and the narratives, and however you want to call them, I prepared a bit of an intro to the topic and I’ll try to complement what you have said last week. So, at least we don’t basically say the same thing. So, that’s an invitation for people to go back to last week as well. Yeah, there you are, indirect advertisement.

[00:39:39] Niels

Thank you.

[00:39:42] Nick

So, let’s take a step back. We are doing quantitative investment. What do we try to achieve? Frankly speaking, we try to be profitable (whether that is in stocks or across all the asset classes) by capturing, broadly speaking, two types of scenarios. There is an underlying risk and we deploy that risk and we expect compensation for it. Think of a value investor, for instance. Right? You’re basically taking on downside risk, or some default risk, or however you want to call it, and then you expect compensation for it.

So, there’s a natural risk sharing mechanism. You don’t want to hold it. Okay, I’m happy to hold it as long as you pay me an extra premium. It’s very rational. So, there is, if you like, heterogeneity between investors and the way they perceive risk and the way they react to it. And in that kind of clearing activity, someone is holding the riskier asset, expecting compensation for it, and somebody else is paying for it. So that’s case number one.

Case number two is more like behavioral or structural inefficiencies. And I think trend following is the primary example here. But we can talk about low beta, for instance. So, low risk stocks are typically coming with a greater than expected Sharpe ratio, primarily because of leverage constraints. Pension funds, for instance, they cannot deploy any significant leverage, but they have a benchmark to beat and therefore they go for the higher beta or the higher volatility names and that creates an imbalance in the market. So, that’s more like a structural design that leads to systematic investing benefiting from getting into low beta names.

So those two pillars, and I’m not talking about the more kind of high frequency kind of start arb, I’m talking about systematic investing in the context of a medium-term frequency model. So, these are the two pillars.

Then, what you end up having, in systematic investing, is the necessity for a risk model. And the risk model is here to obviously explain what we’re exposed to and how do we do so? Well, we basically take asset returns and this can be stocks or multiple asset classes. We project them onto what we call characteristics; could be valuation, volatilities, momentum returns. And we managed to explain portfolio risk as a combination of factor risk or rewarded risk. It’s not always rewarded, by the way, but factor risk and then the idiosyncratic one.

I’m making this point because I think it’s critical as soon as we go to thematic investing. Risk factors are not always risk premia. Risk factors explain risk, some of which, when you’re exposed to them, come with some compensation, it is a risk premia. I think a good example is countries. Unless there’s a systematic reason why some countries should outperform others, there is no risk premia associated. But they are a significant component in a risk model to explain risk exposure (I’ll come back to that point when we go to thematics).

And last but not least, the risk models are just reducing dimensionality. You have a thousand assets, or whatever, 300 markets in a trend follower, you have a risk model and that breaks it down to, I don’t know, 10 or 20 factors. So, it also makes the portfolio, from a risk assessment standpoint, easier and more tractable.

Now why is all that relevant for thematic and narrative investing? Because in the recent years, and I would say maybe after we went past COVID, there has been a bit of transition in the investment thesis, more around thematics. I think the ‘stay at home’ theme that we had back in COVID is just one of the very many examples that obviously banks came up with.

There were thematic baskets and businesses built around that and hedge funds utilizing that product to take on some active bets or perhaps hedge specific exposures. But to me, themes are nothing more than a tagging operation. It’s just a categorical way of combining assets in a particular group that share some common feature.

In the same way as you have high positive trending commodities, or you’d have cheap stocks, or you have high profitability names, you also have names that belong to a geopolitical risk thematic or AI transition. And maybe some stocks have or are part of the AI revolution. Maybe some others have exposure and commonality in the thematic and they’re exposed to the thematic if the thematic kind of builds up. And hopefully you’ll see where I’m going with this one. I mean, I can give another example, and I prepared this example so at least we can have something to kind of compare against. Right?

[00:44:59] Niels

Sure.

[00:45:00] Nick

I can tell you that we built a portfolio of stocks that happen to start from the same name, from the same letter. All the stocks from A, and all the stocks from B, and all the stocks from C. Right? I mean, this is no different to just grouping stocks on a specific thematic. Right? It’s another tagging exercise.

And frankly, if people start talking about the theme A, theme B, theme C, theme D, I think the reaction would be that the correlation and the risk of those baskets will start increasing. Right? And therefore risk has to take that into account.

So, here we are today with a specific thematic, let’s call it AI. Just to pick one. Somebody could argue that this category has a fundamental connection. So, there are firms that are exposed in the AI hype or they participate to it, or the CapEx, or the development of the models, da, da, da, leads to higher correlation between them, which is obviously maybe perhaps a consequence of the fact they belong in the same sector.

These days we have the memory chips for example. Right? So, they are a sector, but also the underlying thematic is making that association even tighter. So, the cross correlation increases. And I think this is the consequence of now, investors thinking in that package manner.

So, the correlations, not only do they increase because they’re part of the same sector or maybe exposed to the same Microsofts or the same network of vendors and providers of services, but also people think, and investors think of those names as part of those groups and they allocate to the groups. And that activity, if the market is not elastic enough, it starts building some short-term price dislocation. And you know where I’m going, right? That’s what we call momentum at some point. Right?

Something becomes very hypey. And because it’s hypey, there is more allocation to it. And because there’s more allocation to it, some price path is starting being established, and then people with short-term momentum signals will capture that. And then one thing brings the other. And this kind of mini activity (I would not call it bubble by all means), mini kind of price pressure is delivering a trend.

Now, let’s start getting now into those papers. In a conventional risk model, what we try to do is explain the risk and return, risk in particular, but return too or return covariation (if you like), through either fundamental characteristics; momentum, valuation, profitability, so on, so forth, or some form of categorical associations; countries, sectors, industries. So, the truth is that some of those characteristics, if systematically being utilized in your alpha, some of the categorical, not necessarily but if you have good ability of timing country exposure, or maybe sector exposure, maybe you can grab some alpha.

So, where thematics now come into the picture is that now suddenly we have a new dimensionality reduction space which has both risk implications, because of what we’re discussing up until now, this association with a thematic, but also perhaps an outperformance association. If a specific theme is rising and you’re early enough, well done.

So, what those two papers talk about is taking the two concepts associated with thematics; one being risk implications, the other one being return implications, and go through analysis that introduces thematics and narratives into risk models. That’s the first paper. Or they utilize narratives and narrative momentum as the engine underlying potential outperformance, precisely because there is hype around thematics.

So, that’s a very long winded intro, only to argue that thematics, in my mind, is another way of reducing dimensionality that could have risk implications and could have dynamic allocation decisions, perhaps leading to outperformance, even if we cannot easily associate a risk premium to the thematics themselves. There is no particular reason why a stagflation macro narrative should outperform long-term. But there are times it can perform if the asset with it benefits from the thematic becoming more and more popular. Which is, again, a trend following mindset. At the end of the day, it’s the intertemporal demand for a particular asset that goes beyond risk assumptions. Right?

So, let me make a pause. But if you’d want me to, I can then go through some of the main findings or the analysis of the two papers. But I think it was important to frame everything that has to do with thematics in the broader context of how do we assess risk, how do we assess return, where return is coming from, and how we should explain this ecosystem of investment as a whole.

[00:50:43] Niels

Sure. Obviously, that’s super elegant way of explaining. So, every time I have to come up with a little question here. I feel that I’m probably starting at a disadvantage point to make it sound. But let me try.

From my background, from the trend following background, it kind of sounds like, okay, so we know we can group things together, we know they will be influenced by themes. And since this is not something that we do, so, I don’t have any experience with it, but what seems to be clear to me is that part of the success lies in the strength of the theme. Right? If some themes become stronger…

[00:51:33] Nick

Exactly.

[00:51:33] Niels

…then you would think that thematic investing group approach, whatever we call it, will perform better.

So, my question to you is, is there a way, or maybe it is something that people do, to actually instead of thinking about trend strength, to think about theme strength and to measure the strength of a theme and make it part of kind of the investment process?

[00:52:04] Nick

Yeah, so that’s, broadly, where I’m going with this one.

[00:52:07] Niels

Yeah.

[00:52:08] Nick

In the sense that think that you’re a trend follower and you’re allocating to commodities rates, currencies, and equity markets.

[00:52:15] Niels

Yeah.

[00:52:16] Nick

The way you express risk in that context is through a covariance matrix that looks historically into which markets correlated the most, and which are the more volatile, and so on, and so forth. What this perhaps is missing is that some of those assets, at a point in time, start creating clusters that go beyond what historical analysis can suggest.

Like, for example, there could be specific equity regions, and maybe currencies, and perhaps commodities, that are closely associated with a macro shock. Now, by deploying risk in those markets, we’re basically trying to build covariance matrices that somehow reflect what we expect to happen. But thematic association is short lived.

So, there is an argument to be had on creating better covariance matrices so that tomorrow the way we have deployed risk is going to reflect how risk is distributed and gets realized. So, what the first paper says or shows, the paper by Wai Lee, is that, if we don’t account for thematics in our covariance matrix, and they do it in stocks but I think we should think bigger than just the stocks here. If you do not account for those thematics, which are, again, sub clusters of activity that perhaps historically were not there, but today commonalities start arising. If we don’t account for those, then our prediction of portfolio risk is biased upwards. So, we’re more risky than what we should have been had we accounted for it. So, we’re not accurately distributing risk. That’s statement number one. And that’s for a trend follower. To your point.

The second one has to do with, we typically use prices and we argue that prices are enough for us to get to tomorrow’s prediction. And maybe there are those models, that you and I have discussed, and the street has now kind of become quite big on them, like the economic trends, we’d say, okay, there’s the underlying trend of a price path is looking for proxies. One proxy is the best path, another proxy is the economic regime. Another proxy is, what is this asset associated with it? Is it associated with thematics that grow? Does it have high exposure to them and does it have a positive exposure to them, or does it have a negative exposure to a fading narrative or a fading thematic?

Both of which should contribute to narrative strength for that asset. Why? Because, well, if something is rising and you’re positively associated with it, that’s an indication of you kind of following with it as long as it continues growing. So, the second paper called Narrative Momentum, suggests that, look, narratives do not come and go overnight. They come, they stay, and they fade.

And through that cycle of popularity picking up, at some point hitting the peak, and then starts dropping and losing, if you like, strength, you have short-term association of markets and stocks with that rising narrative. So, if you had an ability to measure that and capture that with some beta, you could, in theory, enhance your allocation model and say, you know what, not only does it trending upwards, or maybe not only inflation and growth are accommodative, but perhaps its own score to the prevailing narratives (because we can now rank them and cross sectionally find the more relevant ones) is very positive.

So, enhance your signal, add more to it. And I think there is something here at least worth exploring. And that’s what this paper is talking about. It says, look, the stocks that happen to have higher exposure on narratives that are becoming more popular outperform those that have less exposure to narratives that are fading.

And that’s the gist of that paper. They obviously look into too many other things. I think for us, the more relevant topic here is that this is not just price momentum. Right? So, it’s not that I’m packaging price momentum and I’m giving it to you like, in a wrapped paper, with a nice color on top of, and say, hey, that’s something new. I think the beauty here is to say there is a reason why price momentum exists.

So, if you go a few months back, before price momentum was there, there is an underlying engine that is creating excess demand so that tomorrow Niels and Nick will come and say, oh, here’s a positive trend. Let me just buy into it. So, what could be bad? It’s an underlying perhaps mechanism in the market that is creating this positive trend in the excess demand that comes perhaps from fundamentals or from narratives, Who knows. But there you are.

[00:57:10] Niels

So, another question that pops up when I hear you explain that is again, I come from this world where predominantly price is important, at least to what we do at DUNN. But I fully recognize that there are other ways that people exploit trends now, and so on, and so forth. But from a research point of view, the beauty about price is that it’s objective, it’s recorded, we can go back several decades and we can do our research based upon this. What I find more difficult to fully comprehend is how do you even backtest a thematical strategy? Because a lot of these themes just kind of pop out, and suddenly it’s a theme, and suddenly it’s gone, and so on, and so forth.

So, you don’t have to kind of spend a whole hour explaining, but is it even possible to test or how would you go about even thinking about testing a thematical strategy? Yeah, you can backtest, you can say, of course, if the theme is AI, okay, we know that today, so we can go back and see how it did. But seven years ago we didn’t talk about AI. It wasn’t a theme. So how do we even think about that from a research point of view?

[00:58:40] Nick

Yeah, that’s a beautiful question. And I think it has two sub questions. Number one, even if you could measure the thematics, what’s the underlying data?

[00:58:53] Niels

Right.

[00:58:54] Nick

And I think the second paper, this Narrative Momentum, and I know you, and you have discussed it last week, they have this point in time data pretty much, I think 13 plus million of articles.

[00:59:08] Niels

Yes, correct.

[00:59:09] Nick

Anywhere from like social media, I cannot recall… It’s certainly articles, corporate filings, trading blogs, news agencies, they basically have this information since I think 2013, 2014 (from memory) to 2012 or something.

[00:59:33] Niels

And they take the 13.3 million articles and they structure that into 347 narratives.

[00:59:43] Nick

Correct. And they have, obviously they have some robustness checks whereby they change the definition of those narratives. They have 57 driven by the so called JEL classifications in academic journals. And I can actually talk you through that.

Or they take some other narratives from some other academic papers just to showcase that the results are not dependent upon their definitions, but different attempts to have a collection of narratives has been popular. So, that’s the first point on the data and I think they have this uniqueness in the data which I think is very powerful.

The second point, which you very nicely kind of alluded to, is the one whereby some narratives only become relevant as soon as we happen to know them. And that’s a forward looking information that only in the presence of which we can go back and test. So, I think here what we end up having to deal with is not too dissimilar in nature to the factor zoo.

So, there’s this factor zoo. Now we have, I don’t know, however many in the single stock world in the hundreds of factors that frankly, between us, are highly correlated clusters. So, if I say it’s earnings to book, or book to price, these are both valuation characteristics, but certainly at some point they became known and post that time we can go back and test them. So, I think here we have like a narrative zoo. So, in some sense, again, and connected to my earlier points, it’s a dimensional reduction, but in itself it creates a challenge.

So, my straight answer to your question now goes as follows. Precisely because book to price and earnings to price correlate, and they are two different facets of valuation, we can make the argument that narratives are also correlated. So, when COVID is a theme before COVID and through COVID, healthcare could be another one. And you should expect a high level of correlation after COVID arrived. But we did not have to know COVID before COVID happened, had we had a healthcare narrative. And frankly, I don’t think that healthcare suddenly became a narrative. I think there is a way, obviously there are design choices here, and the bias, and so on, and so forth, but there are ways of characterizing the narrative space with more generic or, as one of the papers called them, evergreen topics that have been part of our ecosystem and will always be. Like, for example, I don’t know, trade wars, or stagflation, or capital expenditure, or maybe ESG (fine, ESG is like 15 years old), no, we cannot call it new today.

So, there is certainly a point in time that most of which became a reality. But I think there is an argument to be had that 30, 40, 50, 60 topics that have an everlasting nature can probably capture the intertemporal new thematics that come just because they happen to be correlated. We cannot have 500 of them. Possibly like 30 or 40 are enough. And that would be partly my defense.

Nobody knew the Straits of Hormuz, but people did know geopolitical tensions. So, I would expect the two to move together. And I think the counter to it is if you try to be very prescribed, you’ll find yourself finding COVID pandemic healthcare concerns and like five different ways of expressing pretty much the same topic, which is ultimately a health crisis. So, that would be my answer.

Yes, there are design choices and limitations, but if we think that you don’t need 500 of them, but maybe you need a dimensionality reduced core segment, in the same way as we have in factors, for example, there is probably a way out. So, yeah, that’s how I see it from a source perspective. You see what I mean?

[01:04:14] Niels

Yeah, I do. And I’m, again, always just trying to turn it back to kind of the world I know more about. And sometimes I wonder, when I hear and read all these reports, again, I can’t help wondering sometimes if this is just kind of Wall Street’s way of finding a new way of getting investors excited about something. And I’m not suggesting that there is no value to it, but I’m also thinking, how much different is it from just the way we have been doing it for 50, 60 years where we just say, yeah, energy markets are probably going to react more or less the same and therefore we call them a sector, so let’s just keep it simple, stupid, so to speak.

I know it’s a bit more nuanced today and I know there are lots of things, but in general, you’re not going to be completely wrong by putting healthcare companies together and just say, yeah, this is healthcare. This doesn’t have to be COVID or some kind of flu, bird flu, influenza, whatever. It’s healthcare.

[01:05:24] Nick

You’re absolutely right. My two points, my last two points on this one go as follows. There is an underlying price process. You need the best predictor. All we’re saying is that F of X, or G of X, or H of X, are transformations of what we observe. The price path in itself is not necessarily the best predictor, but it’s an easily observable, easily calculated, and empirically validated.

But in a prolonged, let’s say, growth regime, it is not unreasonable to argue that the macro regime says buy equities, the price path says buy equities, and maybe the thematics are pro-growth and therefore equities have high correlation and positive correlation to those thematics. So, I definitely agree with you that these are all facets of looking for the truth, where the truth is the real signal and the denoising of it. So, I don’t disagree at all.

I guess the other point, as you said, people look into different ways of reshuffling the story of investment. I think what happens, and obviously with the world of AI these days, there is limited attention, and there is too much dimensionality, and the ability of capturing news or capturing thematics reduces the impact of how we think and take investment decisions, specifically when they become discretionary. Obviously, the systematic world is systematic, but that’s how I would look into the need of reducing dimensionality. It’s just easier for the human brain to process and I think this is really where this is coming from.

[01:07:05] Niels

Yeah, well, I think the challenge for the trend following CTA world is to come up with some really sexy themes that we can talk about and get people excited about this strategy. This was beautiful, Nick. I really, really appreciate it. I know we left out two other papers that only came across my attention in the last couple of days, so I will keep them. And they are more trend following focused. Even some to do with turtle trading, I see is one of them. And there’s another paper about the science and practice of trend following systems.

So anyways, I’ll keep them maybe for a couple of weeks because next week we have a very special mix of guests. We have Rich coming back next week. That’ll be exciting. It’ll be even more exciting because he’s going to be joined by Dave Dredge and they are both very much into adaptive systems, and complex systems, and all of that stuff. So, we’ll dig into some of that fun stuff. So, if there is a question on that topic, by the way, by all means send them to info@toptradersunplugged.com.

But before you do that, you should definitely go to your favorite podcast platform and show your appreciation to Nick for all of these insights, preparations, and time spent on putting this together this week. We so much appreciate your support and a nice comment or a nice review will certainly help more people discover the podcast.

So, from Nick and me, thanks ever so much for listening. We look forward to being back with you next week. And in the meantime, as usual, take care of yourself and take care of each other.

[01:08:52] Ending

Thanks for listening to Top Traders Unplugged.

If you feel you learned something of value from today’s episode, the best way to stay updated is to go on over to your favorite podcast platform and follow the show so that you’ll be sure to get all the new episodes as they’re released. We have some amazing guests lined up for you and to ensure our show continues to grow, please leave us an honest rating and review.

It only takes a minute and it’s the best way to show us you love the podcast. We’ll see you next time on Top Traders Unplugged.

This podcast expresses the views of its hosts and the guests appearing on the podcast as of the date of its recording, and such views are subject to change without notice. Top Traders Unplugged do not have any duty or obligation to update the information contained herein.

Furthermore, Top Traders Unplugged make no representation to its accuracy and it shall not be assumed that past investment performance is an indication of future results. Moreover, wherever there is a potential for profit, there is also the possibility of loss.

This content is made available for educational purposes only and should not be used for any other purpose.

The information contained in this podcast does not constitute and should not be construed as investment advice or an offer to sell or a solicitation to buy any securities or related financial instruments in any jurisdiction.

Certain information contained herein concerning economic trends and performance is based on or derived from information provided by independent third-party sources. Top Traders Unplugged may believe that the sources from which such information are obtained are reliable.

However, Top Traders Unplugged cannot guarantee the accuracy of such information and has not independently verified the accuracy or completeness of such information or the assumptions on which such information is based. This podcast, including the information contained herein, may not be reproduced, copied, republished or posted in whole or in part in any form, without the prior written consent of Top Traders Unplugged.