Return Stacking: Form, Function & Friction ft. Corey Hoffstein and Adam Butler

[00:00:07] Corey

They often say, well, it’s totally divorced from trend following. Our bottom-up process, which is the majority of the weight (we weighed at 30% top-down, 70% bottom-up), is a trend following strategy. Our replication approach is truly applying trend following signals. I think a lot of people get that wrong when they talk about replication.

We just find that the top-down does seem to work; does seem to work just as well as the bottom-up and provides process diversification. And so, we find it’s better if we blend them together.

[00:00:37] Intro

Imagine spending an hour with the world’s greatest traders. Imagine learning from their experiences, their successes and their failures. Imagine no more. Welcome to Top Traders Unplugged, the place where you can learn from the best hedge fund managers in the world so you can take your manager due diligence or investment career to the next level.

Before we begin today’s conversation, remember to keep two things in mind. All the discussion we’ll have about investment performance is about the past and past performance does not guarantee or even infer anything about future performance. Also understand that there’s a significant risk of financial loss with all investment strategies and you need to request and understand the specific risks from the investment manager about their products before you make investment decisions.

Here’s your host, veteran hedge fund manager Niels Kaastrup-Larsen.

[00:01:32] Niels

Welcome to another episode in the Open Interest Series on Top Traders Unplugged, hosted by Moritz Seibert. In life as well as in trading maintaining a spirit of curiosity and open mindedness is key and this is precisely what the Open Interest series is all about.

Join Moritz as he engages in candid conversations with seasoned professionals from around the globe to uncover their insights, successes and failures, offering you a unique perspective on the investment landscape. So, with no further ado, please enjoy the conversation.

[00:02:08] Moritz S.

Hey everyone and welcome to episode number 16 of the Open Interest Series on Top Traders Unplugged. Today, as a special, my Takahē Capital business partner Moritz Heiden will join me as co-host, and we’ll be speaking with Corey Hoffstein and Adam Butler, from Return Stacked Portfolio Solutions, about trend following and portable alpha ETFs.

Corey and Adam started the Return Stacked ETF business together with Michael Philbrick and Rodrigo Gordillo, from ReSolve Asset Management, and launched their first ETFs in 2023.

Since then, they have been expanding their suite of Return Stacked investment products significantly, now spanning a variety of portable alpha solutions from trend following, multi asset carry, and merger arbitrage, to Bitcoin and gold. We’ll focus on Corey’s and Adam’s approach to replication and the complexities that go along with it. And we’ll also discuss some of the challenges associated with putting an active trend following trading system into an ETF wrapper. So, Corey and Adam, welcome to Open Interest. Great to have both of you here.

[00:03:10] Corey

Thank you for having us.

[00:03:12] Adam

Thanks. That was a great intro.

[00:03:15] Moritz S.

Well, thank you. Well, we learned from the best, from you. We were just, before we started, Moritz H. and I, we were reminiscing, like, when we first got in touch with each other, I think. I ventured to guess that it was probably 10 years ago. Moritz said, no, it’s probably not 10 years. It must be less, maybe six, seven. But I remember we were reading Corey’s stuff from Newfound research, and we’ve been chatting about the bitcoin basis trade, and rebalancing timing luck, and liquidity cascades. So, it’s been a couple of years and it’s great to finally have you on.

[00:03:46] Corey

Time flies fast. I would guess it’s probably 10 years, now, we’ve been chatting. So, yeah, it’s been a long time.

[00:03:51] Moritz S.

There you go. All right. And it’s really great for me to have Moritz, the two Moritzs, the two quants, as a co-host. You know, Moritz has helped me to come up with the smart questions, which he’s very good at. So, why don’t you kick it off?

[00:04:06] Moritz H.

Sure, thanks. Hi, guys. Yeah, great to have you on the show and also great being here. Yeah, now, we’ve already heard a little bit about the things that you guys have been doing.

I was specifically interested in the background on the Return Stacked idea and the strategies, maybe also giving the listeners a little bit of an introduction into what return stacking actually is, and where it came from, and how you came up with the idea, and also why you launched that product in the end, or that suite of products, actually.

[00:04:39] Corey

Yeah, so, I will do my best to breeze through this in a timely manner. So, as you mentioned, return stacking is synonymous with this idea of portable alpha that’s been around since the 1980s. Pimco was really the progenitor there.

It’s been an institutional concept. And the idea is, basically, how can you keep your beta in your portfolio (the core stocks and bonds that institutions are building their strategic asset allocation around), but search for alpha in more fruitful places? If you’re allocating to stocks, do you have to try to find alpha and picking stocks? Or can you just keep the beta and use financial engineering; use capital efficiency through futures markets, and swaps, and other capital efficient instruments, to free up capital in your portfolio and allocate to more fruitful sources of alpha?

That concept has been a mainstay for institutions for the last 40 years but has largely been out of the reach for most individuals and wealth managers who aren’t going to manage a book of derivatives themselves. And so, return stacking is our attempt at encapsulating that idea into tickerized solutions so that people can buy a product that gives them that implementation without having to manage any of the behind-the-scenes process. We are going to manage that from an institutional perspective and give them a variety of flavors.

As you mentioned in the intro, so, Adam and I are both technically from different firms. Return Stacked Portfolio Solutions is a joint venture between Adam’s firm, ReSolve Asset Management Globa,l and my firm, Newfound Research. And we both really came to the idea of adopting portable alpha, as a solution, independently. You know, the same way we’re talking about that I go back a long way with you guys, I go back a long way with Adam and his partners as well – probably almost 15 years, at this point, sharing collaborative research.

And we both came up in the 2010s offering alternative investment strategies and just struggled against the nonstop outperformance, particularly here in the US, of a core US 60/40. And sort of by the second half of the decade, realized that perhaps there had to be a better way. Rather than telling people to sell some of your core stocks and bonds to buy alternatives, and have that constantly be a disappointing approach, we started exploring other ways of implementing alts and portable alpha came up as this solution.

And so, in fact, I actually was invited to a Barron’s roundtable about ETFs in 2017. I said, I thought capital efficient ETFs would be the future, in 2017, and gave this exact example of hey, if you have a capital efficient ETF that gives you simultaneous stock and bond exposure, you can use it to introduce alts without giving up core stocks and bonds. There was a large conversation going on in FinTwit at the time. I don’t know why it took us five or six years to actually get an ETF out the door. Sometimes things are easier said than done. But I think both of our firms came to this idea as a solution of solving both the practical and behavioral issues of trying to get investors to adopt and stick with alts.

Now, Adam can add some extra color here. His firm has been doing this stuff for a long time. A lot of their mandates have been futures based. They’ve been blending active strategies with risk parity solutions for most of their history. So, they’ve been implementing this idea of portable alpha and return stacking in a different way.

I think what we’ve done, with our ETF solutions, is try to distill it and simplify it into core building blocks that wealth managers can use to build whatever sort of portfolios they want for their clients.

[00:08:23] Moritz S.

And I think you’ve done a great job educating the space and kind of like putting it on the front because, as you said, portable alpha return stacking, which is the term that you’re using and you’ve coined it, core satellite. You know, this stuff has been around for decades. Sometimes people didn’t even know that they have it in their portfolios.

For instance, through QIS strategies, some of them have this portable alpha concept and a risk parity type of bond and equity allocation depending on which level of vol you run it on, is something like a portable alpha product at the end of the day. So, you’ve done a great job putting it out there, like you said, creating a tickerized solution and building suite around that.

And what we now see, quite interestingly, the US market for active ETFs and your types of products, trend following ETFs in particular, is growing very fast. There are more and more firms entering the space.

It’s kind of very different here in Europe where ETFs function in a little bit of a different way. They don’t have the tax advantage that you guys have in the US, for example. We’re lagging behind.

But as a kind of like 30,000 foot views, how do you see it developing now that BlackRock, I think, has launched a managed futures ETF? Is it going to put all the mutual funds out of business at some point? Who knows? I mean, this may be as a precursor before we get into the more technical stuff.

[00:09:50] Adam

So, we run both mutual fund and ETF based return stacked strategies. The mutual fund strategies are able to trade a slightly wider and more diverse group of futures markets. Because just operationally, if you want to venture outside of trading just futures markets that trade on US exchanges, then if you’re trading an ETF and people want to make a market during the day, then some of the markets, that you’re going to want to make a market in to create, or destroy, or redeem units, are not very liquid, as you know, when the people want to create or destroy those units.

And so, it’s just not as operationally convenient and you end up with the potential for larger kind of intraday slippage on trades, for market making, et cetera, if you trade much outside the US markets and down the liquidity continuum. So, there are definitely some operational features that you can deploy in a mutual fund structure, because of the end-of-day always-traded NAV component of it that is more difficult to replicate in an ETF wrapper. But again, the ETF wrapper has some tax advantages, it has the intraday liquidity, it has the ease of platform access, et cetera. So, it’s just a bit of a trade-off and what people prefer.

[00:11:29] Corey

Just to add a bit of color to that. There’s what I would call strategy versus structure. And when you use the ETF structure versus the mutual fund structure, there are constraints; there are benefits, but also constraints and Adam highlighted those. When they’re trading on the secondary, you know, when it’s exchange traded, there are market makers who are trying to hedge their exposure and keep tight bid/ask spreads. And when you’re underlying, or illiquid, or not trading, or whatever it is, it makes it very hard for them to do that.

So, what you tend to see is that, yes, we are seeing more active products. We’re seeing more derivative based products (and we can get into the regulatory environment that’s allowing that), but most of the futures based products that are coming out from Alpha Simplex, and Man AHL, and Invesco, and BlackRock, tend to be a subset of more liquid US and European traded markets than you see in their mutual fund or hedge fund products.

[00:12:24] Moritz S.

So, that is a great segue into one of the questions or some of the questions that I wanted to ask. So maybe, Moritz, can you hold your thought? Because I know that you wanted to ask questions about the replication methodology, but now that we’re already speaking about the topic of bid/offer spreads, and I think, Adam, you mentioned slippage, that is exactly one of the things that comes to mind. When you put these markets, let’s imagine a trend following trading system, when you put this active strategy into an ETF wrapper, there is kind of like this limitation. Well, maybe theoretically there shouldn’t be a limitation on the number of markets and the liquidity of the markets, but in practicality there is.

Because if you’re putting too illiquid markets into it, or if you’re putting too many markets that are trading in different time zones (some of which may even be closed when the ETF is traded in the US), you’ll just have a very wide bid/offer spread. And you cannot trade the ETF at NAV. You have to cross bid offer when you buy and sell.

I mean can you put some color on how wide these bid/offer spreads are? What do you see there? I mean, what’s the usual range there?

[00:13:39] Corey

So I’ll talk to that a bit. So, first I will say you can do NAV based trading in ETFs. For wealth advisors who have access to an institutional desk through their broker, they can normally NAV based quote that’ll settle at the end of the day.

You can do this in Canadian markets, you can do it in US markets, and normally you’ll get a quote, and the market maker might say, I’ll fill you at NAV, plus or minus a few pennies when you ask for a two way market. Similarly, you can ask for a risk quote, an intraday. They’ll tend to give you mid-price plus or minus a few pennies.

So, if you’re trading substantial size, what we have found is, actually, that going through an institutional desk, you can get incredibly tight fills. What’s harder is when you’re actually not trading substantial size and you’re not going to get a risk quote for a thousand shares. And that’s where you can see bid/ask spreads of 20, 30 basis points on average, I don’t think are unusual. And during chaotic markets or economic news events, you can see it blow out further than that.

And so, I think that will be a fairly persistent feature of these types of funds. I don’t think you’re ever going to see real penny bid/ask spread as long as you’re trading markets that are less liquid, and you have daily trading, and the market makers don’t necessarily have full transparency into what the basket is.

The market makers, whenever there’s less transparency, they’re going to widen out and thin out. And so, I think you tend to see that as a persistent feature of these types of funds.

That said, I always like to say, well you should only be entering the fund and never redeeming. So, it’s a one-way trade-and-hold-forever and that cost sort of smooths out over time.

[00:15:20] Moritz S.

Yeah, sure, you know, you should have a longer hold period, of course, and not day trading any of these products, but staying with a number of markets and liquidity. I mean, ETFs, in general, are designed to scale into the billions. I mean, what constraints does that impose on the ETF? I mean, imagine you’re starting a strategy that includes orange juice futures, and palm oil, and lumber, and then you get to like 100 million or 200 million. You could probably remove those markets and make the underlying portfolio more liquid.

Is there a way, say, you scaling it into 5 billion, 10 billion, is there a way where, at some point, you could force redeem investors or say, we’re closing the ETF, we’re no longer taking any more money? Because if there’s not, and the ETF can theoretically go into the gazillions, then at some point you’ll have two or three markets left to trade with.

[00:16:26] Adam

I mean, we started with a universe that provided what we felt to be the right balance between ability to scale and sufficiently spanning the diversity of the potential bets within the futures space. There are also things that you can operationalize to expand out the number of markets that you can trade with sufficient liquidity to get quite a bit more scale. So, for example, a lot of the markets have sufficient liquidity to trade within, for example, an ETF basket that we expect to scale into the billions. But there are CFTC limits that limit the number of contracts that can be held in the front month.

But if you’re able to trade back month contracts, then often you’ve got 10 times (or sometimes considerably larger) more bandwidth or capacity that you can trade in those markets that allow you to operationalize substantially greater breadth. But I mean, there’s no getting around this trade-off.

At some point you’re going to hit capacity levels where you’re not going to be able to trade a meaningful proportion of the available markets at size that are going to make a difference to the P&L of the strategy over time. That’s true for hedge funds, it’s true for regular mutual funds, and it’s also true for ETFs.

Just there’s a trade-off between accessibility, liquidity, and the diversity of bets that are available. And, you know, all things equal, as we all know, the greater the diversity of bets in expectation, the higher the expected performance.

What we do find though is, and I’ll admit to being a little bit surprised about this when we first sort of began to do the research on this, you can get pretty darn close in replication space with a much more finite universe of markets. As long as you’re spanning the equity index, bond index, commodity, and currencies, as long as you’re sort of generally spanning those, even within a fairly liquid basket. You know, we use kind of 26, 27 markets and ours. You can get really close, from a replication standpoint, to the performance of the full universe. And it’s probably no surprise because everybody is impacted by the same set of constraints.

In aggregate all of these futures players can really only trade these markets in proportion to their liquidity. So, if we’re trading 26, 27 markets, they represent something on the order of 70% or 80% of total global futures liquidity. Then you’re just going to get pretty close to the average of what all of the different players are trading at any given time.

[00:19:40] Corey

So, to answer one of your questions really quickly, that Adam didn’t touch on, is you can’t close the ETF. So, we have to be able to run it at, in theory, infinite scale. And one of the choices you have, as an asset manager, is to run the strategy differently at low size versus large size. I think that is a change of track record.

We wanted to make sure that someone looking at the fund at $1 billion knew the fund had the same strategy at $100 million, rather than trading markets that we were then going to take away. The reality is if this got to $20 billion, I think it would sort of self-correct. By definition, the performance would degrade substantially and people would start to take their money out.

So, there isn’t the ability to close. We built it for substantial scale, but at the extremes, $100 billion in a strategy like this, it would just self-correct with performance.

[00:20:30] Moritz S.

Yeah, the market forces take care of it.

[00:20:32] Adam

Right.

[00:20:32] Moritz S.

And what you’ve mentioned, it’s not technically style drift, but you are correct. If you’re starting with kind of like 100 markets, and this includes lumber, and orange juice, and then at some point you remove these markets, it’s a different thing all of a sudden.

[00:20:43] Adam

Yes.

[00:20:44] Moritz S.

You’re buying the same product. Exactly right.

Great. Moritz, do you want to talk about replication in more detail?

[00:20:56] Moritz H.

Sure. Back to the technical stuff. Yeah, I mean it goes into a similar direction, and we already touched upon the subject of the trend component, which is part of many of the products that you have on the shelves.

So, speaking about trend, I saw that you are doing both top-down and bottom-up replication, and I was specifically interested in, is it kind of like coming from your proprietary trend system that you had before or is that not usable in that context? Why did you decide, basically, to do both of the approaches? How do they work and which index do you, for example, use? So, which one is the one you’re actually replicating?

[00:21:36] Adam

So, I’ll start with the bottom-up because I think it addresses your question about whether it mimics the sort of legacy trend strategies that we have been running for many years. I mean, what we did was try to span as much of the potential space of how one might follow trends across the basket of futures markets that we intended to trade.

As we all know, actually, there are a variety of ways to define trends but it is still finite. So, there’s the kind of (call it) classic quant time series momentum type approach. There are some moving average cross type strategies.

The moving average crosses can be on or off. So, something is above or below a moving average, or above or below the cross of the moving average. Or it can be continuous where the signal, the size of the trend is a function of how far above a moving average of prices, for example.

And then you’ve got these sort of breakout style strategies. Which I know you guys are very intimately familiar with. And within the breakout style strategies, now you get into all kinds of degrees of freedom. You’ve got the… Are you scaling vol all the way along? Are you only scaling vol on the entry? Do you have a trailing stop? If so, how far?

So, once you sort of enter into the event driven style of trend, which you guys are more familiar with, the degrees of freedom explode. So, what’s interesting is that within the sort of time series and moving average style trend, well, it’s kind of like you can model a moving average cross by weighting, for example, longer term moving averages positively and shorter term moving averages negatively.

And so, just the weight that you give to the different trend lengths ends up allowing you to model a variety of different continuous style trend strategies. So, what we did was we just took a time series trend approach.

We applied it to each individual market, and for again, for each individual market a variety of trend look backs from kind of 5 days out to, call it, 300 days. And then we regressed all of these individual market, individual trend length strategies on the Societies General Trend Index.

Over the very long term we’ve got daily returns going back to 2000, and we saw what proportion of the strategies, at different trend lengths for different markets, were required in order to provide the best long-term fit to the trend index. And what we found, actually, is that when you… And keep in mind, the trend index is an average of 10 funds. All these funds are doing things slightly differently. Some of them are going to be more sort of breakout style, many of them are more continuous quant style.

But as we model the fit across the average of all of them, the bottom-up strategy actually does just a really good job with an average daily correlation in the neighborhood of kind of 0.8.

So, why don’t I stop there on the bottom-up because you probably have questions already and then we can get into the top-down afterwards once we clarify.

[00:25:34] Corey

Can I do a 30 second summary for the less technically inclined? The 30 second summary here is, when you talk about trend, we’re all generically typically talking about the same thing. But the implementation details really create a lot of dispersion in the space.

Anyone who’s looked at the trend space knows there’s a ton of dispersion and that comes from what markets are you trading? What trend speeds are you trading? How much risk weight are you giving each market? What sort of constraints are you putting on? These are all choices that you can either make discretionarily, you can try to do some sort of fit, but that’s what leads to the dispersion in the space.

What we did in designing our bottom-up model, which is just a trend following model, is all those parameterization choices came from the objective of how do we create a long-term system that looks as close to the SocGen Trend Index as possible?

So, how much risk weight we put in oil, or gold, or the euro, or dollar, or the S&P. What trend speeds were trading oil, and gold, and the S&P? All of that comes from trying to fit the long-term performance. And what you get out, I think, is actually, it’s not a secret. I think it sort of rhymes with what you could guess.

You get intermediate term trend speeds, and you get 25% of your risk weight pretty much in equity indices, in bonds, in commodities, in currencies. And within commodities that we trade, it’s split pretty evenly between metals and energies. And my guess is (you guys have lived in the space long enough), you would guess the average generic trend follower is putting equal risk weight in the major sectors and categories. And that’s what we found came out of our fit. So, when we talk bottom-up, we’re running a trend program. It’s just all the parameters were chosen to try to fit what sort of the average trend follower looks like.

[00:27:26] Moritz S.

Would you, at some point, recalibrate that fit? Like, you know, this is when you think about the opposite, which is the top-down replication, you have that responsiveness to changes as underlying managers change. Hopefully you’ll be in a position to pick it up. With the bottom-up approach, you don’t necessarily have that response function. It’s not, it’s not built in.

So, it could be that some of the trend managers that are included in the SocGen Trend Index start trading differently, or start trading different markets, or I don’t know, become shorter-term again. Who knows? We don’t have a crystal ball. Would you rerun the model, recalibrate it annually, or is it kind of like set in stone, you’ve disclosed it, that’s it?

[00:28:13] Corey

I was going to say it’s a non-trivial problem because I 100% agree. What if every trend manager goes from long-term to fast trends over the next year?

When you fit a model like this that has so many parameters, you need a big depth of data. And the other thing you need is for all the markets you’re trading to have gone through meaningful periods of contributing positive or negative P&L to the system.

And what you find is like, if you were to just train on 2000 to 2010 and compare that to just training on 2010 to 2020, you will actually get some pretty different weights, risk weights in particular for different markets. And it’s because the system is having trouble figuring out whether the trend followers simply didn’t trade a market or whether the market just never really contributed to P&L over that period.

And you can have something that contributed a ton of P&L in one decade and no P&L over the next decade, meaningfully contributing to the variance of the system. And depending on the period in which you fit, it will look like trend followers don’t even have that in their system.

And so, what we found is to get a good long-term average sense of what they’re actually trading, you really do need 15, 20 years. And so, we can keep rerunning it every year to try to update, but it’s not going to meaningfully change the weights.

[00:29:41] Adam

We also, when we were investigating this, did analyze the extent to which the weights on the different markets and trend speeds changed depending on the time period that was used for training. And we also ran what you guys will be familiar with, a variety of different cross validation type tests.

And yeah, I mean to Corey’s point, you did have some periods where one or two markets didn’t represent as high a weight, but the sector weight was still very consistent. A good example is the 2000 to 2010 period. ES actually had a very low… Like, just ES systems, S&P 500 had very low or lower weight in the portfolio.

And I don’t know anyone who lived through that period knows that the S&P just wasn’t really trending in that period. And so transistives didn’t hold a lot of S&P because it largely went sideways and was trendless. And then from 2010 to 2020, the S&P was massively trending, mostly trending up. And so, the trend systems had a much larger weight in them.

But, if you train on 3/4ths or 80% of the data, no matter how you slice the data, the trend weights, both in terms of sector and individual markets and the weights on trend speed all look astonishingly similar. There actually isn’t a meaningful difference.

And if you try to evaluate whether there’s been any sort of systematic bias in, for example, trend length over time, because I think there was quite a bit of hypothesis that the trend length for systems between 2000 and 2020, for example, increased systematically over that time period, we actually didn’t notice any meaningful evidence of that. So, we perceive that as kind of largely mythical. So, you know, in general, we found that it just generalizes very well.

[00:32:03] Moritz H.

Did you run into some trouble with using a limited set of markets? You mentioned 23 to 28 in the beginning that you’re using, but you’re replicating an index which, I don’t know, might have managers who are substantially in other markets. Because we know from some managers were saved, in some of the years, just because they traded in sugar, for example, which is something I guess you don’t have in the portfolio. So how does that show up? Does it make a difference overall or kind of like is it in aggregate, just not that much of an impact?

[00:32:35] Adam

I mean, the thing about the trend index is it captures the performance of the largest managers by AUM. And therefore, as we’ve already discussed, as you grow in AUM, the less liquid markets that you trade end up having a smaller impact on total portfolio P&L.

So, you know, by virtue of these funds being so large, these managers, I guess, have actively chosen to not constrain AUM, to build strategies that scale. And so, we’re replicating the strategies of managers who’ve already built strategies to scale.

And so, I think from that perspective, we would expect there to just be less contribution, in aggregate, as you replicate the managers in that strategy, on average. There are just less idiosyncratic bets that are materially impactful. And, you know, obviously, from month to month, quarter to quarter, we do see some of that dispersion.

You know, coffee was a big one in 2024 and 2023. Some managers obviously did just incredibly well. But, on average, in the index, the managers just didn’t have, in aggregate, a very large bet to it.

So, we did see some dispersion because that bet has now rolled back and a lot of managers who did benefit from it are now giving back a substantial amount of those gains. So, now we’re seeing positive dispersion where we saw negative dispersion before. So, I think what we’ve observed is just on average, it balances out over time.

[00:34:21] Corey

One way to think about this is, again, these are large managers. Let’s say they trade a long fat tail of markets, 70, 80, 100 markets, 80% of their risk weight is probably in the same 20 markets we could all rattle off the top of our head, and 20% of their risk weight is in this fat tail.

Unless their signals are all perfectly correlated to each other, though, when you start to average those weights together, it’s going to attenuate towards zero. And so, what you end up finding is… And then you have to consider, does that fat tail have zero correlation to the markets we’re already trading, or does it have some overlap?

And so, as you sort of say, okay, there’s probably some correlation with markets we’re already trading. It’s not perfectly independent. And those weights are going to attenuate towards zero because not all the managers are perfectly, you know, trading the same signals. You end up with a situation where that 80% really does explain a huge proportion of the variance of, again, their average performance.

We’re trying to deliver the beta of the category. And we’re trying to outperform that over the long run through fee differentials. For managers that are trying to truly add alpha, I absolutely understand the desire and want to trade the fat tail of markets. And that’s exactly what we do in our mutual fund offering.

But in the ETF, that needs to be scalable, and we’re just simply trying to deliver trend beta, which we think is an incredibly compelling proposition on its own. What we have found is the beta can largely be captured with just a handful of the common names you would expect.

[00:35:50] Moritz S.

And to capture the beta efficiently, I think you guys run two models, right? You have a top-down, and you combine that with a bottom-up. Why did you, what was the thought process behind that? Do you think that just, you know, the combination of the two is going to generate greater stability in the replication process or is it kind of like, oh, you don’t know which one is the better one? So, kind of like, you know, place two bets, see what works.

Because you could have made a choice, and said like, okay, we’re doing bottom-up, that’s it. Because as you said, Adam, you have 0.8 correlation to the space, which is, you know, great. You could just go with that. So, maybe give us some background on why you combined the two approaches.

[00:36:34] Corey

Yeah, so, the top-down, just to get everyone on the same page, is a regression-based approach. Where we’re going to look at the most recent performance of something like the SocGen Trend Index and run a robust regression to say what mixture of futures markets, trading long and short, what basket would have given us very similar performance to that, to the index, or a mixture of managers?

And then we’re going to assume we can hold that one day forward – that these managers aren’t making substantial changes day to day, that the prior weights would have been close for the next day. And we’ll keep rerunning that evaluation on an ongoing basis.

The pros of this method are that it’s incredibly flexible. And you touched on this earlier, Moritz. We don’t actually care how the managers are coming up with their weights. They can innovate all they want and we should capture that innovation over time. We’re just trying to say what is the result of their process, on average? What do the average weights look like across all these managers?

The other really interesting thing about the top-down approach is that it can capture the return of markets that we’re not necessarily trading because of the way regression works. So, for example, we don’t trade the Swiss frank, but if the return of the Swiss frank can be pseudo approximated by a linear combination of say gold and the euro, and P&L is coming from the Swiss frank, our regression process might buy a combination of gold in the euro to pseudo approximate that Swiss frank. And so, it’s a very flexible approach.

The con of this method is it is, by definition, rearward facing. So, you have to look at 20, 30, 40 days. Now, most of the time trend followers aren’t changing on a dime, but every once in a while they do. The regional banking crisis of 2023 is a good example where managers went from very short bonds, to almost flat, to long bonds, in a two or three day period. You’re not going to pick that up immediately in that lookback process because you’re looking over those 20, 30, 40 days.

The other issue with this type of approach is that (and we touched on this in building the bottom-up system), if a certain market doesn’t contribute any P&L, or contribute to the variance of returns over the last 20 or 30 days, the system has no idea how managers are positioned. So, just as a hypothetical example, imagine coming out of 2022, every manager is short bonds and then bonds flatline for 20 days. Those managers would likely still all be very short bonds. But if you’re just looking at the last 20 days using a regression process, the regression is going to go, I have no idea if they’re long, flat, or short. Now, there are some things you can do to sort of manage around that. But that is one of the problems.

And so, what we said was the top-down approach works. We can build this, and we have great conviction that it works, but it’s also untethered mechanically from trend following. What if we complement it with another system? And that’s when we built the bottom-up system.

And we think the pros and cons of the bottom-up system are almost the exact opposite of the pros and cons of the top-down. And what we find, from a tracking error perspective, and a correlation perspective, is that the two systems are almost identical.

They’re statistically indistinguishable as to which one is better at achieving its objective of sort of tracking that average trend beta. But when they have tracking error it’s fairly uncorrelated – like, I think a correlation of only like .2, .3. And so, by definition, we are better to combine the processes together to bring the aggregate trend tracking error of the system down.

So, what we do is every day we run that top-down system, regression based system, get a target set of weights. We run the bottom-up system, which is a true mechanistic trend following process, get a set of target weights. We blend them together and that’s how we come up with the target weights for our total system.

And so, when a lot of people talk about replication, and I listen to this podcast quite frequently, they often say, well, it’s totally divorced from trend following. Our bottom-up process, which is the majority of the weight (we weight it at 30% top-down, 70% bottom-up), is a trend following strategy.

Our replication approach is truly applying trend following signals. I think a lot of people get that wrong when they talk about replication. We just find that the top-down does seem to work. It does seem to work just as well as the bottom-up and provides process diversification. And so, we find it’s better if we blend them together.

[00:41:08] Adam

If you had to choose just the bottom-up or the top-down, the bottom-up has a better track record of stronger average fit. But the top-down sometimes completely overwhelms the bottom-up in terms of capturing probably idiosyncratic systems or methods that are being applied by these individual funds that are just not captured in the continuous time series trend that we’re modeling in the bottom-up strategy.

So, in a very real way, the top-down is an expression of epistemological humility. We just know that there’s constant innovation happening in the space. We don’t know everything there is to know about how best to follow trends or trade futures markets in general. And so, we want to be able to capture some of that innovation, some of that dispersion in thinking among the individual managers. And the top-down gives us an opportunity to capture that.

[00:42:15] Moritz S.

When you run the top down on the SocGen Trend Index, I think, Corey, you’ve aptly used the term it’s kind of like ‘tethered in trend’. I mean at least you’d assume the underlying managers to follow a systematic diversified trend following trading strategy.

Now in the case of the SocGen CTA index, not the trend index, the other one, the big one, trend does play a significant role, but it’s not the only source of returns. And managers have changed during the past 20 years, trading VIX contracts. Maybe there’s some more mean reversion going on.

It’s kind of like there’s different source of returns, maybe a noise term or something like that, where you now enter into the regression calculation. Do you think that makes it easier? Because I don’t really have the answer to that. It could, actually, make it easier to work with a top-down approach and find the best fit.

[00:43:08] Corey

I’ll say it makes it harder to do the bottom-up, for sure. The reason we…

[00:43:11] Moritz S.

Exactly. Harder to do the bottom-up, no doubt about that.

[00:43:14] Corey

Yeah, the top-down, I would have to think a little bit more about that. It depends on some of the speed of the models they’re using. Like, if they’re using faster mean reversion models, it could be very difficult to pick that up because if you’re trading, you know, 20, 30, 40 contracts, even if you’re doing a robust regression method, you still need enough data to get that fit. And if they’re trading daily, you know, multi day mean reversion models, it’s just going to be very hard for you to capture that effect.

And so, maybe you just leave it aside and you are ultimately just capturing predominantly trend, which is still highly explanatory of the broader CTA index. But that may not be the case over time. And so, I think like our mutual friend, Andrew Beer, runs a strategy that replicates the CTA index. He only uses a top-down approach.

I don’t think that’s wrong. I think it’ll be very hard for him to build a bottom-up approach that captures trend, carry, mean reversion, intermarket signals, all this stuff that is happening in the broader CTA space that gets captured in that index.

[00:44:20] Adam

I think it’s an interesting question though, and area of exploration. I mean we run systematic global macro multi strats. So, we run skewness strategies, carry, relative value carry, seasonality of various types. So, we have all of these underlying signals that we can use as explanatory features in a bottom-up context.

We’ve run those features in a bottom-up context. They do help to improve the explanatory fit of our bottom-up model. They improve the explanatory fit against the CTA index, to a larger extent, than they improve it relative to the trend index, to your point, because the CTA index, the managers in there are using a wider variety of signals in addition to trend.

There’s some question about if we’re trying to express a trend following index, whether we should include non-trend features because those features do improve the fit. The good news is, against the trend index, the improvement in fit is marginal. So, it’s not really that hard a decision. If we were trying to replicate the CTA index, we’d have a harder decision, I think, about the breadth of the different feature families that we might want to include in the bottom-up replication process.

[00:45:50] Moritz S.

I think the way I view it, with all the sometimes critiquing that goes on about these ETFs, we’re not critiquing anything here. If anything, we accept it as a new and fair benchmark, to be quite frank. If you are a hedge fund that is charging performance fees in that space, and you’re supposed to produce alpha, you better beat these ETFs who replicate, as you say, or who give access, cost efficient access to the beta of our industry without charging performance fees. So, I guess that’s just, you know, that’s the bar.

If you want to be in the hedge fund business with a performance fee, charge 2 and 20, or whatever it is, you better jump over that bar. Maybe not every year. You know, there’s always a luck component to it, but over the longer run you just need to be better than the ETF, period.

[00:46:40] Corey

Or offer something different. Quantica had a great piece that I revisit very frequently, that talks about the benefits of adding a longer tail of markets. And they have this interesting trade-off of, the more diversification you have and the improvement to the Sharpe ratio makes the strategy look more absolute return. But actually, because you’re putting so much of your risk budget in this longer tail of markets, you don’t necessarily get the same crisis offset that CTAs have historically delivered.

I hate that phrase, but I’m going to use that because you’re not putting as much into the markets that may necessarily be correlated with the big economic factors. And so, I think one of the really interesting trends we’re seeing in this space is people saying, well, the beta is the beta. The beta is pretty cheap.

I mean, I’ve talked to institutions who try to get the beta in QIS from banks for 10, 15 bips. They really look at the beta as being like equity, like beta at this point, because of how trivial a very naive trend strategy is to implement.

So you go, okay, well what else can you do? You can look at spreads, right? You can look at, do you use breakout systems? Do you design your system to be faster? Do you design certain constraints so it has better diversification properties? Do you do what Florin Court does and do pure alts, where maybe it’s a very vanilla trend strategy but just getting access to that fat tail of markets is actually very operationally difficult?

I think there’s a lot of innovation that can happen in the fat tail of the space, that there’s still huge room for hedge funds to continue to charge their 1 and 10, or 2 and 20, whatever they want to charge. But I think those hedge funds that are doing very generic trend, I think you’re right, are just going to get squeezed out by lower cost, more efficient data implementations.

[00:48:26] Moritz S.

Yeah, it’s market forces at play. And I think that’s just fair and square, to be quite honest. You know, we like trading some of these, what we call fatter tails or potentially fatter tailed markets, alternative markets, whatever. I don’t necessarily like that term. I think they’re all markets, they are just different markets. Some of them are a little bit more operationally heavy lifting to access than the S&P mini futures contract.

But then it’s also spreads, synthetic markets, stuff like that creates a difference. And we also think that the punchiness of returns creates a difference.

You know, most of the funds in our space have lowered their volatility profiles in recent years, in the recent 10 to 20 years even, to a level where it’s, I don’t know, between 8% and 12% volatility. And volatility control just no longer looks like the CTA returns in the 1980s or early 1990s. But those were, I think, very valuable, especially during times of stress.

So, the question is, how much bang for the buck do you want to get when you need it? Obviously, last week we would have liked it. I mean we’re recording today on the 15th of April. You know, 10 days ago we had a little bit of a drawdown event, I guess, in the CTA space.

We’d like it to be different, but who knows? I mean, maybe this is only the early innings of something that’s happening in the market where maybe in 30, 40, 50 days we’ll be out of our drawdowns and making good money.

[00:49:54] Corey

Well, this brings us, actually, full circle back to portable alpha, in a way, because a very high vol trend strategy can just be thought of as a more capital efficient low vol trend strategy. If you’ve got a 50 vol trend strategy, I can allocate 1/5 the capital to that and effectively get a 10 vol strategy versus just buying a 10 vol strategy.

And so, at the extreme, an incredibly high vol trend strategy, ignoring sort of the knockout leverage risk there for a moment, you take it to sort of infinite vol and it becomes a portable alpha overlay.

The reality is though (and I think Cliff Asness has talked a lot about this) most clients can’t tolerate more than 10 vol. So, even if you say to them hey, this is a 50 vol implementation and you only have to put 1/5th of your capital in and that that gets you a 10 vol overlay, the line item risk is still too big in this industry and those products tend to struggle. And so, that’s why I think you’ve seen all the major players move to the 10 to 12 vol because that’s what the institutions and the institutional process can handle.

[00:51:07] Moritz S.

It has become that. You know, I think 20, 25 years, it’s wasn’t like that.

Moritz, I think you want to, in a second, answer a question or put a question to what Corey just said about the blowout risk, which we think is very interesting, a very interesting thought in that space.

But I agree with you, Corey. The pain of 20% volatility, or 25% volatility and the potential drawdown that that could come with is something that most investors can no longer tolerate. Whereas 25 years ago, the S&P 500, most equity markets, they kind of had 20% volatility all the time and everybody was fine with it. Something changed. Something changed in the institutional allocator space where 20% percent volatility is no longer fine.

It has to be dialed down, it has to become less. But obviously that is not without consequences and side effects. But let’s speak about the risk of ruin, Moritz, what do you think?

[00:52:04] Moritz H.

Yeah, Corey already mentioned it a little bit. How far can you go, actually, with stacking different strategies that might be correlated, especially if one of them is like, for example, higher volatility? So, how do you manage that?

Especially in an environment like last week where we have equities going down and then trend also doing what trend does, also being down. So, is there a limit to kind of like what you stack? I guess you would not combine things that are closely correlated.

So, usually we know that equities and trend are not too correlated. So, it’s a good diversifier. So, it lends itself to stacking. And there are some other categories where you would probably avoid stacking them.

[00:52:45] Corey

Adam, why don’t I talk about regulatory here and then you talk about like practical margin management and what we saw, maybe last week. So, from a regulatory perspective, there are actually some limits here when we think about implementing these ideas within an ETF or a mutual fund.

And I’ll try to make this part fast because I don’t want to bore the audience and put them to sleep. But there’s a rule called the Sec 18F4 rule. This is also known as the derivatives rule, that was put into place three years ago, now, I believe, that governs how ETFs, and mutual funds, and closed end funds can use derivatives.

And if you’re going to use derivatives, like options or futures, you have to follow this rule. And this rule is a VAR based risk rule. And it basically puts limits at, say, you can only have a 20 day 99 percentile VAR of either 20% or 2 times your relevant benchmark. And so, depending on what product we’re bringing to market, sometimes we use the former rule, sometimes we use the latter, but it does ultimately apply regulatory constraints on how far you can push this stuff.

I think from a true operational perspective you can push it much further than what the regulators allow. But there is a regulatory cap on how far this stuff can be pushed and how safe the SEC wants to keep these ideas.

So, Adam, maybe you can take it from there and talk about a little bit of our experience last week of maybe specifically stacking trend, a 13.5 vol trend system on S&P 500 and how we manage the margin there.

[00:54:23] Adam

So, the strategies are rebalanced every night. So, imagine you’re starting at a hundred dollars, you’ve got the strategy experience is a $3 loss. Now you’re at $97.

Well now you’ve got $97 worth of S&P and you’ve got, let’s call it a stocks plus trend type strategy. You’ve got $97 in the S&P, and you’ve got $97 in a 13.5 vol targeted trend strategy.

So, you’ve lowered your exposure to both in recognition of the fact that the total value of the portfolio has declined, but you’re also maintaining the balance between the $1 of S&P and $1 of trend. So, because we’re constantly rebalancing both the underlying exposures within the trend strategy and the relative exposures between the beta and the alpha that’s stacked on top, the risk of ruin is fairly negligible.

And also, I mean, we only need, at maximum, kind of 25% or 30% of the total portfolio value to collateralize the trend overlay at any given time. And we keep 40% to 50% in cash, and to the extent we want to, for example, have 100% in the S&P 500 as our beta, well, a portion of that is going to be allocated to an ETF to get that beta. And then a portion of that is going to be allocated to cash, which is going to collateralize the futures trading, but also we’re going to have what equates to another 50% exposure to the S&P 500 via S&P 500 futures.

So, there’s always more than enough collateral. There’s a balance between… You know, we want to have as much allocated to cash funded beta where it makes sense, for example, in equities where it’s more cash efficient as possible while making sure that we’ve got plenty of collateral available to collateralize the futures strategy in the event that we have a drawdown in one, or the other, or both, like we experienced last week.

And so, you know, if you go back and audit the performance, the margin consumption, the relative balance between the beta and the overlay of the ETF strategies, you know, there was just no dislocation at all. There were some days where we had, going into the day, we had slightly more S&P on the books or slightly less S&P on the books just by virtue of the fact that the value of the S&P component had moved overnight. But these are just the normal course of operational realities of managing a fund.

[00:57:37] Moritz S.

All of this is super interesting and we’re speaking for almost an hour already. Even though trend following is near and dear to, I guess, all of our hearts, I’d like to speak about… because you also have multi asset carry, you have different types of return streams that you can stack.

One that I’d like to speak about is, I’m just picking, that is the merger arb piece, which I think is probably one of your later additions. You don’t have that around for too long. If you could just explain, I mean, merger arbitrage, risk arbitrage, if you could give us kind of like, you know, how do you do this? Is it systematic? How do you pick the targets? Just give us the high level overview of how that works.

[00:58:21] Corey

Yeah, so, really high level merger arb, for those who aren’t as familiar with it, the idea here is that when there’s a public announcement of a public company being bought, there’s typically an announcement of a share price at which that public company is being bought out at. And the price of the stock will jump up towards the buyout price. But it typically doesn’t get all the way to the buyout price. And there are two components of why it doesn’t get all the way to the buyout price.

The first is simply the time value of money. The deal is not closing tomorrow. It’s probably closing sometime in the future, at some unknown time in the future, though generally somewhat scheduled out. And so, you’re going to discount that future known value back to the present value at the appropriate rate. And so, that’s part of it. There’s a time value money.

And then there’s sort of a credit like feature which is, well, what is the risk that the deal falls apart? And the less certain the market is that the deal is going to happen, the larger that spread is going to be.

And so, merger arbitrage, in its most naive form, is when a deal is announced, merger arbitrage strategies will go in and buy the deal post announcement, and basically try to capture the spread over time, taking on the risk that the deal falls apart. And so, it’s often most frequently looked at as a true risk premium.

People will offload the shares, after the announcement, to people who are willing to bear the risk of the deal falling apart. And so, you should earn some sort of risk premium. Very generically implemented strategies have historically earned somewhere around 200 bips over SOFA, over the last say 20 years, 250 bips over SOFA.

The returns of a merger arb strategy are, interestingly enough, pretty lowly correlated with other popular strategies, whether it’s the equity risk premium, bond risk premium, or credit risk – probably about a correlation of just 0.5 to the credit risk premium. And so, as another diversifying return stream, we think it’s a pretty interesting diversifier.

We found a team out of Israel called AlphaBeta. They had been running this type of strategy in a hedge fund for a couple of years, and we worked with them to take their approach, which is systematic and it’s a machine learning based approach, to actively screen US listed deals and turn it into an index. So, we then license that index, it’s called the Alpha BetaMerger Arbitrage Index. And that is how we implement our strategy.

It is a much more concentrated approach. They basically define the universe – US listed only deals, it has to be certain deal size, deal type. And then after that they’re using a machine learning model to look at, effectively, the quality of the deal – how much spread is left, what’s the acquirer look like, what industry is it in? All those sorts of features go into the machine learning model to say what the expected return is pricing out at, what’s the probability the deal closes versus the deal falling apart, what does that imply for the expected return? And the expected turn has to overcome a certain hurdle rate to be included in the index.

[01:01:36] Moritz S.

I think you implicitly answered one of the follow-up questions I would have had which is, you know, are you ‘hedging this’ with a short leg of the of the acquiring company? Which I presume you’re not doing, the way you’ve explained it.

[01:01:50] Corey

We actually are.

[01:01:51] Moritz S.

So, you are.

[01:01:52] Corey

The vast majority of deals are cash deals, in which case you don’t have to buy the hedge’s short leg. When the acquirer is a public company and there is a share component, we will hedge the appropriate amount by going short.

[01:02:09] Moritz S.

Okay. That is an interesting part inside an ETF because you have to locate the stock, borrow fees, the short leg might be called away. What if GameStop wants to use its shares to buy MicroStrategy? You’d be long MicroStrategy short GameStop. I mean, that could be a life ending experience.

[01:02:32] Corey

I suspect that would get screened out. What’s interesting there is our flexibility because of who we partner with to bring these products to market. We can do specific short location, but we can also get these names on swap. And so, in certain cases we look at whether it’s more effective to actually explicitly short or get the short on swap.

[01:02:54] Moritz S.

Excellent, great. Look, before we come to an end, Moritz spotted, on your website, that you have something going on with Golden Bitcoin. So, maybe the last two or three minutes we’ll touch on that. That’s always interesting and exciting.

[01:03:06] Corey

Yeah.

[01:03:07] Moritz H.

And I have a specific question to Corey. How did this product come alive and why? I have a theory, and my theory is kind of like from the TwitterVerse.

Corey missed out on the Hyperliquid token drop and wanted to get as close as possible to launching something. So, it’s kind of like mixing in gold and Bitcoin is the best thing that you can do in the return stacked world. Is that true or is it kind of like completely the opposite, you just want to mix digital gold and the real gold?

[01:03:37] Corey

Yeah, I think it’s painful when everyone brings up that I missed Hyperliquid, given how early I was to that. It’s a bit of background lore on me, for those who know what Hyperliquid is.

No, this for us, as we continue to look to expand our suite of products, we actually had this product on our roadmap very, very early. We probably should have launched it earlier than we got around to launching it.

But what we’re really looking at doing with the return stack ETF suite is bringing these building blocks to market. We have an equity plus trend. We have a bond plus trend. We have a bond plus merger arb. We have equity plus multi asset carry, bond plus multi asset carry. As we continue to look at what we think are interesting diversifiers, there are diversifying strategies and there are diversifying assets. And I think both gold and bitcoin, particularly with the thematic narrative of store of value, are interesting potential diversifiers that are hard for people to carve out room in their portfolio for, but may make interesting overlays to the portfolio.

And so, for us, we say, well, what if we bring a product to market that provides, you know, you give us a dollar, we’ll give you a dollar of US equity plus a dollar of a golden bitcoin portfolio, actively managed but not actively (in like a trend following sense) more in just balancing the risk between those two components so that people can introduce both gold and bitcoin as an overlay to their portfolio. And so, for us it just falls into that diversifying asset bucket. It’s not really a diversifying strategy the way we’re doing other approaches. But we, again, we think gold in particular, and I think bitcoin is proving some of these features, can be a very interesting long term diversifier for a stock/bond portfolio.

[01:05:23] Moritz S.

Excellent. Well then, thank you for joining the two Moritzs. For the first time on the podcast today, we really liked it. I hope our listeners will like it too. It’s been a pleasure speaking with both of you and, as usual, listeners, we’ll include the most important points of today’s discussion in our show notes and should you have any questions, we always welcome them.

We also welcome listener feedback. Please send us an email to info@toptradersunplugged.com where we’ll pick it up and respond. Thanks again for listening and we’ll see you next time on Top Traders Unplugged.

[01:05:54] 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 iTunes and subscribe to 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 in iTunes. 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.