TTU140: Bruno Gmür, Founder and CIO at Quantica Capital

[00:00:06] Bruno

Trend following as we all know is a very painful strategy because most of the time, you can be underwater and you can have periods of one, two, three years that you are underwater, which makes managing and also, from an investor view, makes it very difficult. Then you start questioning your approach. You’re maybe forced by investors to change things although in the long-term view, nothing has changed. This is, I think, the big challenge everyone in our industry has. But on the other hand, we also see a lot of resilience in our industry because most of the competition of the successful trend following CTAs stay around for 15, 20, 30, 40 years.

[00:00:59] 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 will 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 product before you make investment decisions.

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

[00:01:54] Niels

Hey everyone, and welcome to another edition of Top Traders Unplugged where today, Alan Dunne and I are joined by Bruno Gmür, Founder and CIO of Quantica Capital, as part of our mini-series focusing on the one investment strategy that beat everything else in 2022, namely, trend following and managed futures more broadly.

First off Bruno, it’s great to have you on the podcast. Thank you so much for joining us today. We have been looking forward to this a lot, and as you know, I’ve been looking forward to this for quite a few years. I hope you’re doing well.

[00:02:23] Bruno

Thank you very much, Niels. We are doing well. As you know, we are sitting in Zurich, just around 5100 meters from the epicenter of Swiss banking. As you can imagine, it was all but a boring week in recent times. Thank you.

[00:02:43] Niels

Indeed. I’m sure we’ll probably touch on that a little bit. But anyways, before we dive into all the topics that we’re going to discuss, I would very much like if you could just set the stage a little bit for the audience and give a few minutes of background to Quantica and perhaps sort of share a few highlights, and also what type of strategy, from a 30,000 feet point of view, that you run today and where the business stands as we have started 2023.

[00:03:14] Bruno

Sure, Niels. Thank you. Quantica Capital is a Swiss systematic investment management boutique based in Zurich with a rather small and focused team of 14 people; 10 in the investment team. We are managing, at the moment, $800 million in our flagship product, which is the Quantica Managed Futures Program that has now just completed its 18 years track record. Our QMF program, as we call it, is the only thing we do. We are not diversified in terms of different products, different approaches.

The QMF program can be considered, and is considered, as a style consistent medium-term trend following program that is fully focused in liquid markets. So we trade only futures markets, no OTC products, no options on the liquid space (maybe we can discuss this later), just having the most effective and efficient implementation with the least cost and implementation costs possible.

As I said, the program has successfully completed an 18-year track record. Our client base is highly diversified with regards to regional and type of investors. We have different vehicles although we only manage one program. It comes in UCITS format, it comes in the Cayman format. Obviously, we also can do managed accounts that allow for more specific restrictions in the implementation of our program. That’s basically a short introduction.

[00:05:02] Niels

Absolutely. That’s perfect, Bruno. I very much appreciate that. Now, as you may know, Bruno, Alan and I have a list of different topics that we like to discuss with our friends and we alternate a little bit in terms of who leads out with one specific topic. So as we always do, I’m going to let Alan kick it off today. So over to you, Alan.

[00:05:02] Alan

Thanks, Niels. Good morning, Bruno.

Maybe just to get started, you mentioned I suppose some of the features of the Quantica managed futures, medium-term trend following and pretty much a focus on liquid markets. Could you give us a sense on what was the journey to bringing you to that from an investment philosophy perspective? Was that based on your experience in markets, that you settled on trend as the core of the program, or based on academic research? How did your philosophy evolve to have a focus on trend following?

[00:06:06] Bruno

Thank you, Alan. I think that’s a good point to start.

To understand our philosophy and our investment approach and our differentiating factors, it’s probably helpful to look back at where we are coming from and what Quantica’s history and background was. First and importantly, no one at Quantica has ever worked or has ever had any touching points with other CTAs or trend followers. All that we did was develop completely independent of any similar peers or peer group experience. In fact, from my background, it’s actually coming from the complete end, or opposite spectrum, of trading.

My background was more in academics. I spent more than 10 years in the ivory towers of universities, first studying mathematics and doing my master’s with a focus on probability theory, mathematical finance, and statistics. Then I was attracted more by application of those models. So I did my PhD at the University of Zurich, also doing research and doing lectures for graduate courses for more than 10 years in the subjects of financial markets, game theory, econometrics, and all kinds of other quantitatively related stuff.

Basically from that, I was grown up and raised under the dogma, of course, of mathematical models and of efficient market hypothesis. In my first job that I had in the financial industry, I was hired in the chief investment office of a large Swiss private bank and my job was to develop quantitative models to improve the tactical and strategic asset allocation.

That was the first job, and of course, coming from the more theoretical model type of things, a tactical overlay was already questionable from an efficient market hypothesis view. But we did the tasks to optimize tactical allocations coming from a strategic long-term allocation. What you’re doing when you have tactical allocations, basically, is you are looking typically at quarterly investment horizons. The tactical overlay works usually or is implemented at banks or at larger investment firms on a quarterly basis.

When you implement a tactical overlay, be it an asset allocation overlay or a risk allocation overlay, what you’re effectively doing is that you are creating a spread. So think about the tactical allocation of a 60/40 portfolio. If you change that, you’re basically overweight equities, underweight bonds, or vice versa. So what we did, at that time, was just look at those spreads on a risk-adjusted basis between different asset classes and different instruments.

What I found there, independently, was that, indeed, those risk-adjusted spreads between different markets or market segments, that’s what I then called a trend following market inefficiency. So indeed, they had a kind of path dependency. So our thesis then was that actually, yes, markets that outperformed other markets with a typical quarterly investment horizon did continue to outperform, which was basically a contradiction to the market efficiency hypothesis. That was the starting point.

After detecting that, I kind of converted from being a believer of market efficiency theory in practice to trying to implement those kinds of (call it) inefficiencies into models. This is still what is at the basis of our investment process or investment philosophy is to detect trend following inefficiencies on a quarterly or medium-term basis, but on a more risk-based and on a more relative basis.

So we detect relative risk-adjusted trends and this makes us, most of the time, highly correlated to the CTA industry, but in other times, our returns can differ quite a bit because they are based on a relative approach or in modern terminology, you could also say, cross sectional momentum. So it’s not only time series momentum, it also has an element of cross sectional momentum. But this is still the philosophy that we have.

Actually, when we started the Quantica managed futures program, think back 20 years ago, CTAs and trend following strategies were considered more a black box. So nobody really understood what we were doing and it was never our objective to do just another trend follower CTA. In fact, at that time, I was not even aware of time series momentum or other things.

The only thing that I detected was our approach that was designed to build a tactical overlay on a strategic risk allocation was uncorrelated, basically, to everything except for something that was called managed futures or CTAs. So this, I think, is important to understand where we are coming from and what our differentiating points are still.

[00:12:33] Alan

You mentioned that differentiation, one possible differentiation being the focus on relative trends, which is not something necessarily shared by all CTAs. So maybe it would be useful to explain that in a little bit more detail. Also, some people might question, is it the same philosophy as to why you would get directional trends or as to why you would get relative trends?

[00:13:02] Bruno

Yes, that’s a good question. We can highlight that in two different ways. So the easiest way, if you just think about what (let’s call it) a traditional CTA or a time series momentum model does, it’s just looking at, usually, daily returns of the different markets and it’s just analyzing, let’s say, trend signals, moving averages, breakout models, or whatever kind of trend signals on each of those markets individually. So our approach on relative trend following is just expanding that universe of potential markets to consider.

So think about that maybe in terms of if you’re trading or analyzing 100 different futures markets, at the first stage in our signal generation process, we normalize those daily returns by risk. We do it with volatility. (We can discuss that maybe later.) So out of those 100 markets that trade on very different volatility regimes or specific volatility environments, we smooth out or extract that market-specific volatility and normalize those 100 markets to what we could call 100 risk factors that all trade on the same risk regime.

Now, when you have this, you can now enlarge those 100 markets by looking at each spread. So basically, out of those 100 markets, you can generate 10,000 spreads. This is basically our data universe where we want to detect the trends. Basically, you could say, all right, instead of just looking at trend filters on those 100 markets, we look at trend filters in terms of outperformance on these 10,000 markets, giving us a much more robust approach or much more stability.

The other way you could look at that, and what actually is the difference on that approach, is that already on the signal generation stage in the process, the state of the covariance matrix or the state of the covariance structure of the market comes already in the signal generation stage.

This is the biggest difference compared to a time series momentum approach when the covariance structure, maybe at the end, comes in the portfolio construction process but not in the signal generation process. So the outcome of that (to simplify maybe a little bit) is that whenever the correlation matrix or the correlation covariance structure in the markets is more stable, our trend signals move slower. If that covariance structure changes rapidly, our relative trend indicator can move faster.

This is how we usually relate our differences to traditional trend following performance. In times when the covariance structure is very stable (and that can also be in risk-off markets), when, usually, traditional trend following or time series momentum can have a difficult time because there are always whipsaws, our signals are moving faster, and hence (we believe) allows us to extract a more stable risk premia or premia out of the markets.

[00:17:11] Alan

Maybe just one more on that. From the perspective of why you get these trends, it’s often described about the speed of reaction from investors, people observing and behavioral biases like anchoring and then underreaction and overreaction, which intuitively is very obvious in respect to individual from the time series perspective, but maybe it’s less obvious on the relative trend, I’m guessing, because you’re basically detecting trends in these relative performances that not everybody is observing. So then, how do you get comfortable that those types of moves are as grounded and will be as persistent?

[00:17:54] Bruno

I think for us, economically, the hypotheses were key because coming from this tactical approach, when you think about the large investors of the big money, if they change their strategic allocation over tactical, what are they doing? They’re actually exactly moving from one market segment into another one, hence, creating flows from one market segment into another one. As it happens, and this could be my economic explanation of those quarterly trends, it would be you are just detecting and following money flows.

The explanation would be that obviously, large investors, very large, large investors, they cannot implement their tactical moves within one or two days. So they would use a couple of weeks probably, or even months to implement those, maybe risk reduction or their regional focus. Our approach by analyzing these relative market spreads allows us better to identify divergence between the different markets or markets that decorrelate from other markets for whatever reasons.

But typically, I would attribute that to money flows. Money flows are always relative. So they come somewhere and are going to somewhere. So we believe this is, in a way, the economic rationale behind it. It also makes sense that you will detect those types of trends then on a multi-week, multi-month period and not on the intraday or let’s say a couple of days period.

[00:19:45] Niels

I’m sitting listening to you, Bruno, and I’m trying to get my head around this relative trend, maybe coming from this pure side for so many years, and I’m trying to come up with some really clever questions about it. Some of the things that I noticed you said is that, for the most part, your performance is highly correlated to the trend following index, yet you are taking this additional step where you’re taking 100 markets, you’re turning them into 10,000 spreads.

So just for me to understand, in one way, do you know why then the difference isn’t bigger? Because you’re doing, certainly, a lot of extra work here compared to let’s call it the more simple way of doing trend following, and if it isn’t bigger in terms of difference, why not just do it the easy way?

[00:20:44] Bruno

That’s obviously a good question. So I first made a comment on correlation between different trend following models. We all know that correlation does not directly explain performance. Correlation in trend following models, in our view, is determined by the direction of a signal. So when everyone is short bonds, as last year, you’re short bonds, you’re short equities, you’re long commodities, obviously, you’re highly correlated. But then the difference in performance really depends on the sizing of your positions. Even if the sizing is fundamentally different, correlation is high.

Our approach of identifying trends on a relative basis, of course, if a market like let’s say the S&P 500 starts to outperform all other equity markets (not only US markets but also European markets), then typically it also has an absolute trend. So directionally, our signals and positions are probably pretty much (or we know they are) the same because obviously we run an adaptation of our models that just ignores that relative aspect and, apart from that, are doing exactly the same thing. So, [it’s about quantifying] what is the additional benefit and what is the additional value by looking at that.

Over the long term, we just conclude that, in terms of Sharpe ratio, the improvement in taking into account this covariance structure and this expansion of the universe, you can improve your Sharpe ratio by a magnitude of let’s say 0.2, around. This doesn’t sound a lot, but over the long term, obviously, it is meaningful.

So, our hypothesis is that the effect of adding this covariance structure in the signal generation process can increase your efficiency in terms of Sharpe ratio expectation of the program by a magnitude of around 0.2. That does not mean, as we all know, 0.2 is a dangerous statement, because it does not mean that you will always outperform. It can take a very long time until this outperformance is kind of manifested.

[00:23:32] Niels

Those are great points, Bruno. I appreciate that. Now, you mentioned the Sharpe ratio and I have actually in this series maybe asked I think everyone about it a little bit because of the paper that Cliff Asness wrote along with some of his colleagues from AQR about whether we, as an industry, are becoming too focused on Sharpe, and so on and so forth. How do you see Sharpe as something to strive for?

Let me give you a little bit more to sort of lead you a little bit in terms of where I’m going. Are we doing this, do you think, mostly because this is how potential and current investors think about it when they think about risk-adjusted returns, knowing full well that when you have lots of upside volatility that it penalizes your Sharpe? How does that play a role, because clearly, it’s something you’ve thought about in terms of what you’re trying to achieve?

[00:24:34] Bruno

Yes, of course. The discussion about volatility and Sharpe, maybe one should differentiate that. In terms of is volatility the right measure of risk, that alone is a good question. I would say it depends on what markets you’re looking at and what strategies you’re looking at. In terms of medium-term trend following and positions in futures, yes, I would argue yes, of course, volatility is the right measure.

Then the question, or the problem, is not is volatility the right or wrong measure. The problem is, we don’t know what the real volatility is so we have to approach to estimate, to evaluate what the right volatility measure is to build our positions. That’s also true for Sharpe. So if you ask me, is the Sharpe ratio, on a trend following program, the right measure in the long term to measure its success? I would definitely say, yes, of course. What else should it be? Because our investors want performance and long-term performance.

So our mandate of our diversified program is clearly stated. Yes, it should maximize the Sharpe ratio while being, over the long term, uncorrelated to any or to most traditional and non-traditional risk factors. So this is our mandate. If now a client of course wants to adjust that, if he wants to have let’s say a trend model that should not have extreme beta exposure against equities, then in a second step, yes, of course, you can do that. But this comes at the second stage and it will, by definition or by our understanding, it will decrease the long-term Sharpe because any restrictions that we allow to enter will almost by definition or by design decrease the long-term Sharpe.

In that sense, yes, who am I to disagree with Cliff Asness? But in that sense, I would disagree and say no, our main mandate is to generate absolute returns and to maximize Sharpe ratio while diversifying (and we can maybe discuss that later, what diversification really means), while diversifying their portfolio holdings. Yes, we can do that in a second step, but it’s not our main mandate.

[00:27:29] Niels

Okay. I think that’s fair. You say, who am I to debate with Cliff Asness? You’re the one with the PhD, so who am I to discuss with you? But I do have one question that I’d like to pose, and that is, well, what about the fact that you could have trend following strategies that maybe are more pure with a much lower Sharpe but a much higher long-term return, and actually, perhaps with a more extreme trend following profile, which means that actually as a portfolio tool, as a player in the portfolio of traditional investments, it might even have a bigger benefit to the investor.

I know that we don’t design our programs for what investors are going to use it for, we design it the best way we can. But I guess, maybe with my limited knowledge of the math that you could say that if you just want pure trend, that maybe having the best Sharpe isn’t what’s going to give you the best returns over time, but of course, you’re going to have to accept more volatility and more drawdowns for sure.

[00:28:40] Bruno

Yes. I would kind of disagree. Because as a CTA or managed futures program, in our case, we exclusively invest in futures and we use a daily rebalancing mechanism. So we adjust risk on a daily basis according to our target risk of each position and of the portfolio. So by that, as you know, as a managed futures investor, typically in our case, we just use a margin of around 10%, maybe 15%. So there’s a lot of room to scale up your exposures.

We run 2x or 4x strategies of our program, and actually, by our hypothesis, we can achieve exactly the same Sharpe in terms of a new measure, return and volatility in the right way. So you can level that up. If you can achieve a 10% return with a 12% volatility, you should be able to achieve a 20% return with a 24% volatility. So that’s doable. We know that.

It can be leveled up by a daily rebalancing mechanism that really takes care of rebalancing in a short period. You cannot do that with, let’s say, a monthly or quarterly rebalance strategy because of the effect of compounding. But if you rebalance on a reasonably small scale, you can multiply your return with the level of volatility.

[00:30:35] Alan

Maybe just to pick up on some aspects of that and maybe just to give you the opportunity to talk a bit more about it, you mentioned how with the relative approach, that the covariance matrix is embedded within the signal generation. So if you could maybe just explain that in a little bit more detail.

One thing I’m thinking about is if there is this greater sensitivity to the covariance matrix and correlation. In times of stress, and from the perspective of crisis alpha (we’ve all heard of correlations going to 1), does that become problematic in that, okay, you’re not saying you have a crisis alpha mandate, but if there was a perception that that might be part of it, is that a challenge in delivering that type of profile in crisis periods?

[00:31:28] Bruno

Yes. We are still model-driven. So we always use models, theoretical models, to justify what we are doing. From a pure model perspective, we have the hypothesis that if we knew the market parameters, if we knew the covariance structure exactly, and if we knew the expected returns of the underlying markets, we know what we would come up with as a strategy assuming no transaction costs, and we want to come up with the best estimated or closest solution to that.

The big problem, of course, is how do you estimate your covariance matrix? This is, of course, the difficult task, and even more so, how do you estimate your expected returns? Because we all know that in financial markets, your signal-to-noise ratio, which is basically expected return to volatility, is very small. So there’s a lot of noise and this noise will always be there in the market and there will always be shocks in the covariance matrix that per definition are shocks and are unpredictable.

So this is the real challenge of our investment approach is how to model, how to estimate those covariance matrices that go into our signal generation process unlike traditional approaches (that have the same problem by the way when doing the portfolio construction of course). We know that you can always be wrong when shock comes into the markets and disturbs your covariance matrix.

Typically, that comes along at the same time with a trend reversal because your positions have all kind of adapted to a covariance structure and then a shock comes and then obviously, usually, you are on the wrong side of the trade by definition of being a trend follower, because typically, these shocks come after an end of a sustainable trend.

This is what is the difficulty and the pain in trend following because it’s something you have to accept because you cannot do anything about it. But we know that taking into account that covariance structure can improve the trend following returns over the long term. To your question on how does it come into the equation, just think about again on risk-adjusted spreads between different instruments.

When you normalize these risk-adjusted spreads, you’re measuring something like call it an information ratio or generalized Sharpe ratio which is then dependent on the volatility of your spread, which is by definition, a function of the correlation. So when this correlation changes, when you have a decoupling of one market against all the others, this comes typically together with a change in the correlation structure.

Our approach is, what we believe, quite useful to detect this divergence or this change in the covariance structure very early. But again, you’re right, estimating the covariance matrix is basically the biggest challenge.

[00:35:14] Alan

Okay. Well, that’s maybe a good time to pick up on what’s been going on in markets recently. Obviously, we’ve had a shock, as you say, in the form of banking stress and strains. We’re recording in March of 2023 so we’ve had a big reversal in interest rate expectations. So maybe give us a sense on how the system deals with that type of shock. Obviously, volatility in interest rate markets picked up very substantially.

From a correlation perspective, we saw some shifts there as well. But what was your perspective on whether it was more or less difficult running the relative trend in this environment?

[00:36:02] Bruno

In this environment, it was actually not, because after quite a long time with sustainable trends, our positions were what we believe quite similar to a traditional CTA because the trends were kind of very consistent across all asset classes. So of course, the big trend, the big, confirmed trend, was being short bonds, being short interest rate markets, having some concentrated FX positions maybe.

Then what happened last week was a perfect example of what we’ve discussed before, that suddenly, you have a shock in the covariance matrix, everything changes. So, correlation between equities and bonds just reverts from one day to the other in a massive way and market prices are reverting across all asset classes. So this is a perfect example, and in that magnitude, also quite a significant example that, of course, hit all medium-term trend followers because everyone had concentrated positions, being short rates, long some commodity markets, and then this shock leads to a reversal across all markets.

In a period like that when the covariance matrix has been quite stable for quite some time, our approach did not lead to substantially different positions than the traditional approach. So I would say our relative approach did not outperform and did not underperform in that specific situation.

But what is important in explaining the behavior of trend following performance in those scenarios is that, and we actually quantified that effect and we came up that more than 50% of the risk reduction that followed (because as a trend follower, you have to reduce your risk positions because volatility exploded), so more than 50% of the portfolio changes after that shock was driven actually by volatility and not by trend signals.

This is also something we also highlighted in one of our publications earlier, that risk management and portfolio construction have a significant impact on the position sizing. So, it’s not only about signals, it’s a lot about risk management and portfolio construction. In that event as well, the biggest change in positioning (and this is what I expect will be the same throughout the industry) was actually triggered by the jumps in volatility and even less so by the methodology that you apply to measure your trends in the portfolio.

[00:39:04] Alan

In terms of the methodologies for measuring volatility and covariance, there are differences across the industry and obviously different lookbacks. Do you see that as a source of alpha in your approach? How do you think about what’s the right way of viewing that?

[00:39:23] Bruno

Yes. Actually, we don’t see that as a source of alpha. Effectively, we believe what should your volatility or risk measure, what should tit result in? Actually, being or having a medium-term or quarterly investment horizon, you want to forecast or manage medium-term volatility. So, over a couple of weeks, what will be the volatility or the realized covariance over the next couple of weeks? So it’s not really important what it will be over the next two days or over the next three years. The typical investment horizon is around a couple of weeks to maybe a couple of months.

What we found, and actually, this is, I think, very interesting, is the methodology that we use in general (we use some adaptations) has not been changed over the last 20 years, which is, I think, quite surprising. So, our lookbacks and our typical lookbacks to estimate a couple of weeks volatility or covariance matrix are still, we find, the best way of looking at it is using an exponential weighted moving average across. Surprisingly, the decay factor that was initially, in 1998, proposed by JP Morgan, I think it was using a lookback decay factor of 4.94 has still generated the most robust predictions of volatility and covariance throughout the last 20 years.

I find this remarkable because people always think, yeah, everything has gone faster and market structure changed a lot. Yes, market structure changed a lot on a high frequency. Intraday markets are not comparable to what they were 20 years ago, but those intraday moves have no predictive power or no effect on medium-term performance. So the market structure, in our view, over a medium-term horizon has not changed over the last 20 years. It’s still the same.

Back to your question, yes, you can use any reasonable risk methodology that is able to kind of forecast or predict medium-term volatility. But we all know, and again, our volatility forecast is not the real volatility in the markets. It’s just a forecast. If you are a little bit wrong, it does not have an impact. By definition, you cannot predict shocks to this volatility to the real volatility. So in that sense, it should be just a robust methodology. If you have a robust methodology, you have to obviously adapt and adjust for certain periods.

For instance, in the last couple of years, of course, we had some extreme low volatility in some of the markets, and then of course, in order to protect your positions from big shifts, you maybe work with more floors or with just other measures or you use nonparametric extreme events and just limit your potential exposure by past observations or whatever methodology you can use. But still, in fact, a very robust exponential weighted moving average approach to volatility has been superior.

[00:43:16] Niels

Can I follow up on that? Because it was actually one of the things I was thinking about, and then you kind of already spoke a little bit about it. So clearly, what’s happened in the last, and I don’t even know whether it’s the last 10 years or 15 years, but you’re right. Because of the changes in the volatility regime, managers said, well, hang on, we need to put in a floor to volatility. You can do that in a different way. You can just put a floor but you could also just say we have a long-term volatility estimation, we have a shorter-term, fine, okay.

What about correlation, Bruno? I actually don’t know the answer to this. Does it make sense to have two different let’s just say lookbacks for correlation? Because clearly, we know that, in the last couple of years, correlations between equities and bonds have been very different from the longer term. Let’s put it very simply. If you get an extreme change of correlation (and let’s just for argument’s sake say that the current correlation, positive correlation between stocks and bonds at least up until recently, was unusual. I’m not even sure it is, but let’s just say it was), then of course, you could have some even more shocks if you’ve based your covariance and your correlations based on that correlation and then it turns out to be completely “wrong” at some point. So I’m really curious about this because I don’t know the ins and whether you do it or whether it even makes sense to look at correlation also in more than just one timeframe.

[00:44:54] Bruno

Yes, it depends on where your covariance estimator goes in or plays a role in your investment process. In our case, it plays a role in the signal generation process. In the portfolio construction process, it goes into just managing the overall risk exposure of the portfolio, which are two completely different things. Yes, it might be that you use different timeframes or different approaches to these kinds of volatilities and variances.

But at the end, again, my approach to that would be you always want to have some parameters or an approach to estimating that is quite robust. So your overall long-term portfolio should not be impacted a lot whether you are off just a little bit or even a bit more, your approach to measuring that risk. It should be stable and it should focus and target a very robust approach, and then it’s a reasonable method.

We all know that, especially when covariances and correlations go in, typically they enter the investment process of decisions with the inverse. So you have to invert that and then just mathematically it gets, I would say, interesting to model that. But you have to be very careful because when you just apply that very naively, you can get disastrous results.

[00:46:39] Niels

Actually, that would be another topic that I’d love to bring up with you because I’m sure you can help our audience think about this a little bit. One of the things that I feel that we as an industry (and I know there are obviously other quant-based strategies that would also apply here), but I think one of the things, one of the key skills we really bring to the table is that we take ideas, we formulate rules, we do thorough research, but we do research in a way where we have a high expectation that when we apply or deploy it live, it looks similar, meaning that we avoid a lot of the pitfalls about backtesting and all of that because we know there has never been a bad practice that’s saw the day of light, so to speak, or light of day, I should say.

So can you talk a little bit about, obviously without giving any secret sauce away, but just generally, what are the things that one needs to be careful about? You already alluded to some of it, but what is good research and how do you then turn good research into a good live implementation?

[00:47:51] Bruno

Yeah. That’s definitely the one-million-dollar question. Our approach has always been that what we do in our models should be very robust. That means they should have a very low number of free parameters. In one way, how we achieve that already in the first stage, that I quickly mentioned, by risk adjusting or risk normalizing all the markets that we trade, we already reduce the dimension in the sense that we treat all markets after that exactly the same.

So we do not need different models for commodities or for equities because everything that we trade is just a risk factor that is already risk normalized and we apply the same methodology to all markets. The only thing that then is different can be the correlation structure between the markets. This actually depends on the specific state of the markets. But to answer your question, our approach to research has never been just let’s add whatever stylized fact in the market that has worked in a backtest.

Everything we do should make sense from a theoretical perspective or should be explained by some market behavior. For instance, as we said before, we believe we can rationalize the case for medium-term relative trend following by the fact that it will follow money flows from different investors on the same scale. Everything we do should have a rationale to represent a theoretical model.

Because in trend following, the big challenge (and I’m sure you’ve discussed it before) is that diversification is key. We need many different instruments. We need many different sources of returns because when you apply a trend following model and whatever systematic model on one particular instrument, on a medium term, you can never expect a Sharpe ratio higher than maybe 0.2 or 0.3. So it’s usually even lower than the Sharpe ratio of a buy and hold.

The only way you can inflate that Sharpe ratio is by adding many, many different sources of trend following returns and by the fact that those trend following P&Ls have a much lower correlation than just long-only positions. So by just the law of diversification, by adding more and more markets, you can lift up your overall Sharpe ratio, and then you know how hard it is. Everyone can create the backtest on an instrument giving you a Sharpe ratio of 0.2. That’s nothing. A Sharpe ratio 0.2 would already qualify to enter your model. In our case, we would never do that until there is a clear rationalization or model; what does that additional signal bring in a portfolio context?

If it brings something, we would research it and then we would enter it. But in our case, again, our focus of research is not necessarily always in adding new signals, new approaches. It’s even more important that we research implementation techniques, because we know that if we have a better implementation by smoothing out unnecessary trades, we can lower the fixed costs.

So we always operate under the assumption that we want to implement our systematic models with a threshold of overall implementation costs of less than 0.5% per year. This needs a lot of research. You can apply your research and you see an immediate effect of it because it’s almost deterministic over time. When it comes to adding new signals, it’s much more complicated because you have the tendency to overfit.

So our research is actually almost more focused on implementation, on risk management, on the portfolio construction, and of course, also on signal generation. But we do not favor signal generation over the other aspects.

[00:53:11] Niels

That makes sense. We’ve obviously heard from other managers that they do spend a lot of time focusing on execution and all of those things. I fully understand. I sit back with the impression that a lot of the things that you do today are things that you discovered 20 years ago, so to speak. However, I am curious if there is one or two things you would say, over the 20-year period that has passed, where you say, yeah, we did make a couple of important discoveries that kind of changed the way we do things. Would you say that there have been a few of those as well?

[00:53:53] Bruno

First, you’re right. Our basic philosophy is still the same. This has been unchanged. But nevertheless, we have learned so much, over the last 20 years, to understand better where it comes from, how we should improve, how big can any incremental improvement of any specific strategy, how big that can be. In that sense, yes, we learned a lot and we are learning new things every quarter, every week, every month, and this makes the whole thing so exciting.

So to answer your question, yes, we believe we understand, much better, this kind of the market structure of why does our approach work? Why do we believe that our relative approach is a bit better suited over the long term to generate the higher Sharpe ratio? Why is diversification so important? What can we do with implementation? These are all things that basically, 20 or 15 years ago, we did not really care about, and we care, now, much more. So I think our understanding of what we are doing is exponentially higher than it was 10, 15 years ago.

[00:55:28] Alan

Maybe just to build on that but also to go back to the start where you talked about that experience of working in a private bank, because obviously, people tend to allocate to trend following as part of a multi-asset portfolio. So from that perspective, some of the conversations we’ve had with other managers around when you include trend in a traditional portfolio, a client may want to make some adjustments like, say for example, capping the equity beta, etc. Have you looked at that?

I suppose my question is, if you were now back in that original role you were in and you were applying the Quantica managed futures in the context of a multi-asset portfolio, would it be just as it is or would it be adjusted in any way when it’s applied within the context of a larger portfolio?

[00:56:24] Bruno

Yes, it depends. If you count from a typical or usually I would say a rather stable portfolio that those clients have, then I would say, yes, you should add it in the diversified way. If you use, for instance, a very active strategy of steering or managing your equity exposure and you want that to be pure to your approach, then yes, it might make sense to exclude that beta exposure from our exposure. So it depends what the client wants to do with it.

If the clear objective of including trend is (and some call it) risk mitigation or things like that, then yes, it might make sense or in the overall portfolio context to do some restriction on our trend approach. But overall, I would believe that even in those cases, the long-term effect on the overall portfolio would be very similar. Again, with every restriction, you are just reducing your expected Sharpe of your trend allocation, which, again, is not something that I believe pays off over the long term.

We can maybe discuss this kind of notion of investors wanting crisis alpha or they want tail risk protection. Yes, they want that, but have you seen any manager surviving with a tail risk protection product? Are those investors really willing to pay a premium? In a way, you can get crisis alpha, but you have to pay a premium. This is why we do not anymore use the term crisis alpha to promote our strategy.

Instead, I would say crisis alpha is something describing a tail risk protection or more a short-term effect. We all know that you pay the price over the long term. Again, this is over the long term not what investors should target because they should, as well overall, optimize their long-term performance, of course, subject to some risk restrictions. But again, in the past 20 years, I’ve not seen one happy tail risk investor or tail risk protection investor and not one happy tail risk hedging manager, but I might be wrong.

[00:59:14] Alan

Partially, some of these considerations are in relation to timescale, and obviously, the characteristics of managed futures and trend following look different at different time timescales in terms of convexity and skewness, etc. Obviously, you started off with that kind of quarterly medium-term horizon. So does that timescale perspective inform part of that portfolio construction process?

[00:59:42] Bruno

Yes. So this is something which I believe is very important and not widely understood or people usually talk about returns or trend following returns or crisis alpha without even specifying the period that defines, let’s say, for instance, a crisis or that define some characteristics of your returns. Unlike traditional investments, trend following investments or trend following P&Ls have very different characteristics over different timeframes.

The usual connotation with trend following is yeah, trend followers are offering skewed, right skewed returns, are offering convexity, and things like that, and this is not really true for all frequencies. On a daily basis, trend following returns are heavily left-skewed. On a weekly, monthly, that skewness slowly moves to the right and you see the full benefit of trend following returns on a quarterly basis and not on a daily or weekly basis. This is exactly what I think is very important.

You see the effect of trend following diversification and benefits. The portfolio is best seen again on this tactical horizon in one quarter. This is why we usually attribute that, in my view, the biggest benefit of trend following on a quarterly timescale is what we would call smart diversification. So that means, yes, over the long term, it’s great to be uncorrelated. But in fact, a trend follower is not always uncorrelated. He has directional positions.

The great effect of trend following almost by design is that you have a negative correlation or a negative beta in a bear market environment, and you have a positive beta and positive correlation on the right tail of the distribution on the bull market. This is how we try to describe what the biggest benefit is, that the performance on a quarterly basis of trend following is surprisingly regime independent, and we call it regime independent or smart diversification.

In my view, the term crisis alpha has basically two problems. The first is the word crisis, because it’s not really defined in terms of magnitude and in terms of frequency. For some investors, a 5% drop in a week in the S&P is a crisis. For some, a 20% drop in three months is not a crisis. But by definition, a crisis should be a rare event or tail event so it should not happen too often. My second concern with crisis alpha is alpha because alpha is kind of a constant or usually describes the constant term in the regression. So it’s kind of a constant premia that you get in a crisis. I think this is probably true for a tail risk hedge strategy, but not for trend following CTAs.

So if I would term it, I would rather replace the alpha by negative beta (you should call it). If at all, call it ‘negative crisis beta’, but of course, that’s too complicated. Then I would replace the crisis by a more wider regime of quarterly returns that could be termed like a bear market, and then we can talk about negative beta in a bear market that comes together with the positive beta in a bull market. But maybe this goes too far in that discussion.

[01:01:20] Alan

I don’t think negative crisis beta is going to take off.

[01:04:17] Bruno

So do I.

[01:04:19] Alan

Niels, what do you think?

[01:04:23] Niels

Well, I think it’ll be cheap to get the Google AdWords for that particular phrase. Let’s put it that way. By the way, what I will suggest to people listening to us today is actually to go back and listen to the episode a few weeks ago we did with Katy Kaminski because I brought up exactly those kinds of concerns. Not exactly the same, but the concern or the challenge that I have had in the last 10 years with the word or with the term crisis alpha.

Actually, Katy did a great job explaining what she meant with it originally. So I think it’s probably lost in translation and if you listen to what she says now in terms of the origin of that phrase, it actually still makes perfect sense. But I’m not going to rehearse that now because I do want to go somewhere else, and that is, Bruno, people have been listening to us for one hour and three minutes thereabout at this time.

Clearly, with everything you’ve explained, people must sit back with the feeling that trend following, it’s so easy. Why are they getting paid so much for doing so little? But, as we all know, and hopefully has been made clear today, it is not easy and there are a lot of things that go into it. So I can’t stop myself from asking you a little bit about those who have come out in recent years and had great success last year in saying, well, there’s no real need to pay much for trend following or CTA returns, we can replicate that. We can just do a linear regression on returns and we do that.

I say that in a loving way because obviously, one of the people who does that is also co-hosting on the podcast so we have this back-and-forth debate about this. But I do want to try and address the different aspects of this topic because I think it is important, especially for investors who are maybe looking to enter our world and who may consider replication versus going to a few managers. How do you see this and what are some of the dangers you may see in replication versus let’s call it the real thing to quote a Coca-Cola advertisement?

[01:06:46] Bruno

Yes, obviously this is an important question, Niels. As I pointed out, 20 years ago, trend following was considered a complete black box. No one had any idea of what a trend follower is going to do. It’s all a machine, it’s kind of untransparent. That switched tremendously probably 10, 8 years ago with the publishing of papers that just explained how you can replicate trend following correlation, which we all know is extremely easy. You have a first year student and he will program you a trend following program that is applied on five instruments and he will come up with something that correlates 0.8 or 80% to CTAs.

But replicating correlation is not replicating performance. Again, just the great advantage of those replicators is obviously transparency, so people can understand positioning better because they can look it up. When it comes to cost, yes, maybe, but still, usually they are not as cheap as one would think. But obviously, they have advantages of transparency and cost. However, and again, relating to our models, they have clear disadvantages.

The clear disadvantage is typically the risk management or rebalancing process, because I don’t know those products all in detail. But some I heard they rebalance every month maybe every week, which is already clearly from a model perspective, not optimal. The second one that I mentioned is diversification. Some of those replicators, they replicate the CTA returns or industry with 10, 15, maybe 20 of the most liquid instruments.

Yes, with those instruments, you can get a correlation of 0.8, but you give up this theoretical or the real result of diversification benefits by adding just more markets. My estimate for that alone would be that will result in a lower Sharpe ratio of around 0.2 over the long term. But again, of course, if you have a Sharpe ratio of 0.4 compared to a Sharpe ratio of 0.6, you can still outperform over one, two, three, four years.

So I think that the great thing on that product, and especially last year, was in a year when the trends have been really, really strong in all asset classes, such a product can immensely outperform and that this is great, but I have some concerns in the value proposition in times when those products underperform after costs because they shouldn’t underperform over a couple of let’s say one, two, three years because otherwise, they would have a difficult value proposition.

This will definitely happen as well because you cannot, with any approach you have, you cannot outperform all the time. I think in terms of underperformance still, I would say a diversified or valid research product has much better arguments than just a replicator.

[01:10:37] Niels

Yeah. Absolutely. Alan, maybe the last round of questions from your side before we wrap up with Bruno.

[01:10:44] Alan

Yeah, maybe just a final one. You touched on how you’ve enhanced or you see the potential for enhancing risk-adjusted returns with the introduction of the boats types of trend following. But in general, when clients ask how to think about returns for the space, how do you answer that? Beyond looking at what different types of trend following have delivered historically, is there anything else? How confident can we be that that kind of historic return profile should be there going forward?

[01:11:21] Bruno

Our outlook or let’s say our long-term outlook for trend following unlike others has not changed over the last 18, 20 years. Our long-term outlook with the approach like ours should lead to a long-term Sharpe ratio of approximately 1 on a gross basis before fees but after implementation. Obviously, there can be times with a much smaller performance and you will have drawdowns, and all of this you can assess by simulation or by other techniques. But, in the long-term view, nothing has changed in that approach. But you can only achieve that or, let’s say, stay at this level by adding some incremental improvements. Because otherwise, our assumption is that your long-term Sharpe, if you just don’t change anything, will gradually decrease and you have to make up for this gradual decrease with your research process.

So what we see in that direction is obviously, again, expanding the universe, adding more diversification adding, I don’t want to call it alternative markets, but just adding other sources of trend following returns to the existing ones, improving on the risk management techniques that could also maybe lead to small incremental increases. But overall, that hypothesis has not changed in our view on the long term.

But still, trend following as we all know is a very painful strategy because most of the time, you can be underwater and you can have periods of one to three years that you are underwater which makes managing and also, from an investor view, makes it very difficult, and then you start questioning your approach. You are maybe forced by investors to change things although on the long-term view, nothing has changed.

This is I think the big challenge everyone in our industry has. But on the other hand, we also see a lot of resilience in our industry because most of the competition of the successful trend following CTAs stay around for 15, 20, 30, 40 years.

[01:14:11] Niels

Yeah, I think we can all relate to those points, Bruno, so well said.

I just want to finish off with the two questions I always finish off with so maybe you already are prepared for what I’m going to ask you, Bruno. But we do want to hear from our friends, what’s the one thing they hear about trend following that they disagree with the most? So I’m curious whether you have one pet peeve that really gets to you when you hear it.

[01:14:42] Bruno

One statement about trend following I’ve heard, and I’m not the only one I’m sure you’ve heard the same over the last 20 years, probably in the frequency of two or three years, people come up and say, look, trend following does not work anymore. It’s just done. It’s dead. The Sharpe ratio expectations are at zero, they are much, much smaller, markets got more efficient and trend following is dead.

This is the statement that I would disagree the most with because a year like last year again is a very good counter-example to that statement. But we all know all these events are probably rare and the best advice would be yes, just look at it as a long-term investment, stay in it, and live through these more difficult periods. But this is the thing I would disagree with the most, the simple statement that’s trend following is dead.

[01:15:46] Niels

Absolutely. Final question maybe would be, and I know we’re still early in the year 2023 but a lot of things have already happened, so to speak. So certainly not boring so far. But when you look a little bit into the future, and I don’t mean from a return point of view, but just generally speaking, what are you most excited about overall? Are there any things that kind of give you a source of concern?

One thing we haven’t even talked about today but you did a recent paper on, another great paper from your firm, and that is the footprint of CTAs liquidity. I’ve talked probably for quite a while, some people might say for too long about my concerns about liquidity. Although many people might think about liquidity as just can we execute our orders, but actually, as we started out by referencing the recent events in the banking system, is another kind of liquidity event to some extent.

So anyways, I didn’t mean to force you in a certain direction with your answer, but anything that stands out when you look into the future of things that you’re either excited about or concerned about?

[01:17:00] Bruno

Yes. So what I’m most excited about in the situations that we are in now is that we have a fully systematic strategy. I wouldn’t want to be in the shoes of discretionary managers that have to assess the situations and take bets on their own assessment of the situation because in a situation like that where uncertainties are again, I would say, bigger than normal, it’s just great to have something proven, a systematic approach that we all know we can go into drawdown but we understand these kinds of mechanisms, we are confident about the approach. But this is the first point that I would like to mention.

The second one that goes into the same direction is I’m very excited about that our investment universe is within the futures space because when you talk about liquidity, what we need as a CTA, is we need access to, I would say, risk adjusted liquidity. So if the spreads are going up but volatility is going up, then it’s not a problem for us because we always trade in units of risk and we have never really seen a degradation of risk-adjusted liquidity in futures markets even in the most distressed periods.

On a risk-adjusted basis, liquidity was always provided in the futures markets, and this is what is hopefully here to stay. But in situations like we are now, this is definitely, I believe, a huge advantage of a CTA approach compared to a different type of macro exposure.

[01:19:01] Niels

Yeah. I like those answers. Actually, I want to repeat it. I want to rephrase the last point you made because I think it’s so important, and that is for CTAs, the most liquid markets are the most volatile markets. People don’t realize that, I think, for the most part, but that is a very important point that you made there, Bruno. Thank you so much. It’s a great way to wrap up our conversation.

Bruno, thank you so much for being on the podcast, sharing your thoughts, your insights, and your journey and we hope we could do this sometime in the future again. To all of you listening in today, I hope you were able to take something from today’s conversation onto your own investment journey. If you did, please share these episodes with your friends and colleagues.

From Alan and me, thanks so much for listening. We look forward to being back with you on the next episode of Top Traders Unplugged as we continue our deep dive into the CTA industry. In the meantime, go check out the show notes for this episode and all the other resources you can find on our website, and not least, of course, take care of yourself and take care of each other.

[01:20:06]] 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 on 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.