Choosing Your Trend Following Battles ft. Nick Baltas

[00:00:00] Intro

You’re about to join Niels Kaastrup-Larsen on a raw and honest journey into the world of systematic investing and learn about the most dependable and consistent yet often overlooked investment strategy. Welcome to the Systematic Investor series.

[00:00:23] Niels

Welcome and welcome back to this week’s edition of the Systematic Investor series with Nick Baltas and I, Niels Kaastrup-Larsen where each week we take the pulse of the global market through the lens of a rules-based investor. Nick, it is wonderful to be back with you this week. It’s been a little while. How have you been?

[00:00:38] Nick

It’s been a little while. I think the last time was when we had the group session just before the new year. It’s been a very busy start of the year, I should say. Part of that is a lot of client activity, there was a bit of trouble, this EQD conference in Barcelona. It’s been well. So far, so good. How about you?

[00:01:03] Niels

Yes, completely the same. This is my busy travel time, so I can attest to the busyness. But it’s also been an interesting time in the market. I’m sure we’ll come to that. Now, we have a pretty full plate, a very solid lineup with a few different topics ranging from the state of the CTA industry to some hard-hitting research papers.

We also have a recent article from Risk.net. And if we can get to it, we also wanted to discuss a little bit about alternative markets, including Chinese markets. Of course, we also have a question that we’re going to be launching out within a few minutes. So quite a lot for us to be tackling today.

But before we get to that, I’m always curious in between our conversations. Besides the topics we’re going to be talking about today, what else has been top of your mind or been on your radar since we last spoke?

[00:01:59] Nick

So, since we last spoke, I guess it was the start of the year since we have spoken. I think what we have seen year to date, at least in my world, we have seen some of the non-trend components that we discussed a couple of months ago with Alan, performing quite strongly. I think some cross-sectional reversion dynamics and carry dynamics helped out quite substantially, and we see significant interest from our client base in these types of profiles.

I think there is still a bit of uncertainty in the market albeit, obviously the equity market hitting one all-time high following another all-time high and some sort of cuts being already “priced in” until the end of the year. I think there’s still some sort of uncertainty as to where we’re heading. I think the Risk.net article you mentioned has some good pointers. I’m going to leave that for when we’re going to discuss that.

But I should say it has been relatively busy in this regard. But yes, the good feedback about the non-trending components helping a trend following program is the one that has kept me quite busy with clients I must say.

[00:03:04] Niels

Sounds exciting. Sounds good. Now, before we move on, I think in terms of top of mind, I couldn’t help noticing something that happened yesterday. We’re recording on Friday, but yesterday, the Nikkei 225 Index broke its long-term record from the old Japanese bubble, back from December 29th, 1989.

Essentially, a 35-year drawdown has come to an end. That is, of course, good news for Japanese investors. But wow, certainly, our drawdown length in the trend following space pales when you see a 35-year drawdown coming to an end.

[00:03:51] Nick

Actually, let me tell you. It’s a bit of an anecdote. About four or five years ago, we started preparing a report for our clients that discusses how we think about defensive solutions. How can we put together a portfolio of systematic strategies to achieve a specific defensive profile, whether that will help you in a flash correction, a longer-term correction, and so on and so forth.

And we made this kind of supposition that investors think of drawdowns in two big clusters or groups. You have the Sharpe corrections that are much more event-driven, and you have those that are a bit more structural, or I guess longer-term, more sustained drawdowns. But even if that is the perception we have about drawdowns, the way they actually develop, it’s a bit idiosyncratic drawdown by drawdown.

So then we set ourselves a task. Can we use a data-driven process to separate drawdowns into those two clusters? No matter how much we tried, unless we found a way to tackle the corner solutions, we would fail because every single drawdown has its own features. One of those that we actually put in the appendix of the report was, well, look into Japan.

You cannot make a case that it has been in a drawdown for, like, 30 years, because how do you design a portfolio that performs over those 30 years? You have to look into a microscale at some point. So yesterday, my colleague that we worked together on this paper with, back in 2020, sent me the picture from Bloomberg saying, listen, whatever we had in this appendix is no longer the case. So, there you are.

[00:05:28] Niels

Yeah, that’s fun. It’s nice to dig out some of this old stuff. Now, before we get into our usual topics and questions, let me just say that from a trend following perspective, as of Thursday evening, so far, so good for the month of February. So far, so good for the year 2024. Strong trends across a wide range of markets and sectors. Perhaps with a few standout sectors such as equity, soft commodities, as well as the grains, and possibly a couple of standout markets like the nat gas cousins.

When it comes to fixed income, I think, from an industry point of view, it all depends on the lookback period. If you had a relatively medium to short one, you may have flipped your positions in Q4. If you have a longer one, you may still have a little bit of short exposure. That’s going to make a difference so far this year.

My own trend barometer finished yesterday at 39, which doesn’t really reflect the performance of long-term strategies. It’s not quite as strong as you would expect. However, it is more in line with the shorter-term strategies. Which of course is not surprising because actually, in that algorithm that calculates the trend barometer, we’re only using about a month’s worth of lookback for some of the indicators, so it kind of makes sense.

Anyways, as of last night, or maybe Wednesday night, I’m not entirely sure, the BTOP 50 Index up 3.46% for February, up 4.5% for the year. SocGen CTA up 4% in February, up 5% for the year. SocGen Trend up strong, 5% in February, and up 6.25% in 2024. And the Short-Term Traders Index is much more muted, up 1% this month and up only 16 basis points so far this year. MSCI World Index up about 4% as of last night, up 5% for the year. World Government Bonds down 1.2% so far this month, and the S&P 500 Total Return index is up 5% so far in February, up almost 7% so far this year.

All right. With that out of the way, let’s jump into a question that came in from Oliver. It’s a bit of a technical question, so I’m really glad that I have you on the show. Oliver writes, “I hope you’re well. Always greatly enjoying listening to you.” Thanks very much, Oliver. “This might be a more technical question, perhaps for Rob or Nick. Does it make sense for investors to use factor regression to evaluate managed futures? If so, which factor model do you recommend? For example, it seems that the Fung-Hsieh,” or whatever the name is, “seven-factor model is a risk factor model commonly used to evaluate hedge funds’ performance.”

So that’s the question. I think I butchered the name of that factor model, but I’m sure you know what it is.

[00:08:22] Nick

Yeah, that’s the Fung-Hsieh model, the seven-factor model. Two academics that came up at the beginning of 2000s with a factor model to explain hedge fund returns, but they were specifically focusing on trend followers. So, the factors came along with setting aside some of the more classical ones, namely, equity, bond, credit, and a size premium in the equity markets. These are the four more traditional ones in some respects.

Then, they put another 3-4 factors that would capture trend following performance, but this was done using the lookback straddles. This whole literature that talks about trend following signals following a delta of a lookback straddle really started from those academic papers. Because they came up with those factors in commodities, rates, and so on and so forth with a lookback straddle profile that would eventually explain what CTAs were doing.

So, they somehow connected the CTA performance and the trend following performance with a lookback straddle. So that’s the model. The question is, would we use such a model to decompose managed future returns? I guess the question can be asked in a slightly different way. Are we looking to attribute performance into specific profiles or are we looking to identify where risk in our portfolio lies?

I think those two questions are not necessarily the same unless we believe that any risk that we identify is rewarded. Or any rewarded exposure has an underlying risk embedded. Because that particular model partly suggested that yes, there are some risks like equity risk or duration risk or credit risk or possibly small-cap bias.

But then the other factors that were simply lookback straddles, in some respect, were simpler variations of a trend following strategy across every asset class, and therefore, it’s no surprise that they would capture the performance of those managed futures funds. The point is, should we look into the alpha and suggest that they don’t bring value? Well, we can debate as to whether they bring value or not because that alpha is not necessarily testament to their performance.

It is simply a way to identify which are the attributions of their performance across trend following programs across markets and asset classes. A different way of looking into it, I think AQR did something like that possibly 10 years ago. You can take every asset class, say commodities, equities, rates and FX, you can build a short-term, medium-term or long-term momentum, and now suddenly you have 12 factors.

But are these really the factors that if you end up regressing on a CTA will tell you that there’s no alpha or possibly you do bit of a performance attribution of hey, this particular CTA is capturing more medium-term equities than short-term rates for instance? So, I would look into that more as an explanation of, in some respects, using statistical measures, explanation of performance, return attribution, rather than this fund being exposed on XYZ, and therefore, there is no alpha left.

I think that’s always a danger with risk models, that somehow you end up adding more factors seemingly to explain the returns. But it’s not necessarily the case that picking up those return profiles, the fund is not performing. I think the truth is somewhere in the middle here. I think that’s the whole question, right?

[00:11:53] Niels

Let me do a little bit of a follow up here in my non-academic world. I don’t mean to be disrespectful to Oliver here asking this question. But sometimes when I listen to these things, and I think about trend following and I think about the simplicity that we want to instill in the way we design models as much as we can and so on and so forth. I fear that sometimes there’s a risk that all these fancy models to try and analyze what we do almost misses the point.

It becomes too academic, too theoretical to really be valuable. Meaning I’m not sure who would use this type of factor model to make any decision about whether they should allocate to a manager or not. Maybe I’m missing the point. That’s definitely possible. Maybe it’s more relevant for some strategies. But when it comes to trend following, I sometimes feel that we’re trying to do things in order to come to a decision about it that is way too complicated compared to what the underlying strategy is trying to do.

How does that sound to a professor like yourself?

[00:13:14] Nick

You’re raising the bar now. Look, I think in all fairness, having some form of a risk model in mind to at least be cognizant of what are the risks underlying, if you do manager selection for instance, I think it’s prudent. Now, beyond that, a model like this, which is just a linear regression model with a few factors might have some value when it comes to identifying different CTAs with seemingly similar profiles getting their return from different streams and therefore making sense to put them together.

That could be a way of me digging through a higher level of correlation that simply comes from different sources of performance, and therefore breaking that down from a top level into the statistically speaking constituents. There could be some value in it. But I would say from an academic standpoint, it all started 20 years back primarily to try and understand the profile rather than necessarily, how should I say it, reduce the “alpha.” That’s why I feel it’s more of a performance attribution model.

[00:14:17] Niels

Okay, that’s fair.

[00:14:18] Nick

That’s my view.

[00:14:19] Niels

I appreciate it very much and I appreciate the question from Oliver, of course.

Alright, let’s dive into something that’s completely different, something we never really talk about. But our good friend of the show, Linus Nilsson over at Nilsson Hedge, which by the way provides wonderful data for those who are interested in the CTA hedge fund world. Of course, Rich and I use his database when we write our monthly trend following report which people can find on the Top Traders Unplugged website.

But in any event, there was an article online, in his blog post. “The State of the Industry” was the name of it, or “The State of the CTA Industry”, where he’s essentially pulling some data from the regulatory bodies, the NFA in particular, that essentially tracks the number of participants in the industry. Whether it’s firms, different types of firms, whether it’s associates, i.e. people in the industry.

So, instead of me talking about this, Nick, I wanted to ask what you thought about it. We can share some of the trends, pardon the pun, from the data that he showed.

[00:15:32] Nick

Broadly speaking, what is reported in this short note is how the number of participants has changed since May 2021. So, we’re talking about a three-year change. It reports annual increases or declines. Broadly across the spectrum, there’s a decline of participants in this regard. If you focus on the CTA space, which I guess this show is more focused on, there’s a 3% annual decline in the number of CTAs. So, this is not assets that are managed, it is, rather, the managers themselves.

To be frank, I wasn’t hugely surprised by that. Not that I had any specific prior, but having seen at least how the industry has been evolving, there’s a good amount of consolidation or if you like this heavily skewed distribution of, I believe and I think you have more information on this one than what I have, but I think it’s much more skewed towards a few managers that manage possibly 80% of the total assets. So, it might reflect some attempts that didn’t play out.

The reality is, unless you’ve been in 2022, it has been quite a tough period. At the same time, last year wasn’t a great year either. Is that correlated with that drop? Not clear to me. But this is certainly an interesting statistic to look at. By the way, across the database, including introducing brokers that I see here or associates, to your point, everything is in a decline, give or take.

I think out of all reported managers, or if you like participants in this regard under the NFA, there’s a 3% annual decline. That’s what this short report says. What’s your take on it?

[00:17:16] Niels

Well, I guess like you, I’m not necessarily surprised that there is a decline per-se. I guess what I’m surprised about is that there was one category that had a small increase, and that was the brokers, the FCMs. There was actually a small increase. You would think that that’s also an industry that is getting tougher and tougher. To put some numbers on it for the people listening, back in summer 2021, it looks like there were around 1,360-70 CTAs registered. Now we are down to around 1,250. That is definitely a significant drop.

In terms of commodity pool operators, who are people who are associated usually with the fund vehicles themselves, we see a drop from around 1,260, so more or less the same, down to around 1,200 around this time. Then in terms of the total number of people or associates in the industry registered with the NFA, it stands around 40,000 now.

Anyway, the next topic we’re going to be talking about is an article that came in a paper, or actually was online, Risk.net, written by Faye Kilburn. I saw the title a few days ago, and the title was “Trend Following Struggles to Return to Vogue”. They went on to say, “Macro outlook for trend appears to be favorable, but 2023’s performance flop gives would-be investors pause for thought.”

So, at first glance, I have to say I didn’t really like that article. And then once I got a hold of it and I started reading it. It turns out that quite a few of our friends in the industry, people from AlphaSimplex, Aspect, and even to my surprise, the company I worked for, DUNN Capital Management, is also in this paper. They also cite some research from Man Group.

So anyways, why don’t you share your thoughts on the topic and maybe explain what the topic is, because I actually think the title was a little bit misleading. That’s why I kind of went in one direction initially. But the topic itself is a little bit different once you dive into it.

[00:19:33] Nick

I think the title is a bit more catchy in some respects. But the article itself is more about what happened that year. How do we observe 2023 and 2024, and how trend following has a place in institutional portfolios on the back end of this outlook and will 2023 performance completely reverse 2022 or, if you like, be not as amazing as 2022, have any effects on investor demand?

But I don’t necessarily think that it’s about a struggle. I think it’s more about a description of the market in the present moment. The article starts with an interesting observation, that being, that for quite some time now, using some form of risk indicators called the ClearBridge Recession Risk Dashboard. These are 12 indicators, and half of them signal, in the present moment, recession, whereas some others like jobless claims and credit spreads are more flagging some caution.

So, in some respect, it suggests that a dashboard of risk indicators doesn’t paint an optimistic picture about the short term. At the same time, you contrast that with S&P hitting new all-time highs throughout the year. Yesterday, we had the NVIDIA results obviously, and we have all been monitoring our screens as to what is the consequence of that certainly on the upside.

On the back end of this I will say a debatable macroeconomic environment, I believe that towards the end of last year, the demand for call options from VIX, certainly quite an expensive contractual hedge, was at all-time highs. So, they pose the question that in this market environment that you’d expect some sort of instability, disruption, is it the place for trend followers?

Then, I guess the article starts quoting some of the people you mentioned, all of whom in one way or the other effectively suggest that in these types of environments, trend following works pretty well. Now, at the same time, the most recent experience about 2023 was not the most positive in this regard performance-wise.

Last year, I think, we all felt this whipsaw year. I think Andrew has been quite vocal about that. And we had some significant turns in the bond yields not only in the first half, but also towards the end of the year. So, it was quite a volatile year for trend following performance itself, even if implied levels of vol have been dragging lower and lower and lower throughout.

So, I think the report is more about where we are heading now. Is the trend following the vehicle that we should be deploying in that environment? It’s no surprise that obviously our friends, as you call them, speak in favor of it. But there are a few stats, with regards to the performance of the SocGen trend index, that last year didn’t have a good year following one of the best years in the last 30 years, that’s 2022, and how the assets under management for the CTA industry and the managed futures industry, quoting Barclay Hedge here, haven’t substantially moved in the last 10 years.

There was a bit of market-to-market gains naturally in 2022 and some reduction in 2023, but the overall trend is relatively flattish through time. They quote a few people from the industry suggesting that yes, they were expecting a bit more activity in terms of requests for trend following. I guess the overall paper concludes by saying that in a market where we do not know where inflation is going to go yet, the 3.1 that we had recently was more of a surprise than an expectation. But then you get the jobless claims having, I guess, a different effect.

I think it’s a mix of macro data points that at least suggests some uncertainty about the future. Having said that, with four rate cuts being priced in, one could argue that hey, trend following has helped substantially where rates have been falling historically, so it might be the year. Obviously, there’s no conclusion. Nobody has a crystal ball. But that’s the conclusion that I got out of the report.

2022 was great, 2023, not very much. So, the current state of the market is a bit of a debate as to where we’re heading, but equity markets hitting all-time highs month after month. Rate cuts have been priced in. Is that the environment? Maybe. But I don’t subscribe to the title at the end, at least myself.

[00:24:18] Niels

Okay. Let’s start with the reverse. I agree with that as mentioned. I frame it a little bit differently. Whenever people say that 2023 wasn’t a great year, she’s using the words flop and all of that, I frame it differently. I think if you have a SocGen Trend Index that was up 27% in 2022 and it loses just 4% in 2023, to me, that’s a pretty good year actually holding on to performance and so on and so forth. So, my framing is different from the very start.

The second thing I would say is that I do understand why they want to write these articles and I do understand why they can get people to comment on it. But the starting point for me is a little bit awkward because you’re trying to essentially use some kind of forecast for what the economic environment is going to be like to try and forecast trend following returns. If there’s one thing we all say when we’re asked individually, that is, you can’t forecast our returns. So why even bother? Why even start?

Regardless of what’s happening, I guess the article came out a week ago suggesting that this year wasn’t a super year. Obviously, the numbers speak their own language. The year so far is actually quite strong. But it’s also coming from markets that operate much more independently than we’ve seen for a while. Some of them of course, we know the small soft commodities that are on a tear. And so really nothing to do with whether we are in recession or not, the price of cocoa is going to fly if there is a supply problem, and it keeps on flying, right?

I find it interesting that we keep talking about trying to predict the future when the whole premise of trend following is you should buy the strategy because it doesn’t try to predict anything and it has nothing to do with what you think the economic environment is going to be because you want to have it in your portfolio. Because it operates differently and it operates independently and it only looks at price, and it doesn’t try to look anywhere beyond today’s price and then whatever has happened in the past.

But of course, the people in the article are super well respected. The other theme I took from it is that, I guess, they’re trying to explain that when you have these transition periods and when there is uncertainty, whether we are in an inflationary environment or recessionary environment, that’s going to be challenging for any strategy really, but including trend following because the price trends change, and therefore, the models have to adjust accordingly.

So yeah, perfectly fine article. It’s just that the title, in my opinion, could have been different. And I think you’re right, you touched already on the article that was in the ETF trends paper that Andrew was part of where they talk about last year being a year of whipsawing. I think we’ve certainly talked about this a few times. So, let’s leave it like that.

Now we’re going to move into something super hardcore research stuff. It’s a paper called “Optimal Trend Following Portfolios” by Sebastien Valeyre, I think he’s called thereabouts. My French is not necessarily great. The paper derives an optimal portfolio that is based on trend following signals.

Now, I will be completely open and frank here and say to you Nick, as soon as I got to page three, I knew that this was for you and not for me to get into because the formulas are pretty hard core. So, I couldn’t be more excited about hearing your explanation of this paper. I’m sure there’s going to be some interesting things to take away from it.

[00:28:24] Nick

This paper appeared a few months back. It’s published in the Journal of Investment Strategies. As you say, it talks about trend following portfolios, but it takes a much more theoretical stance, at least in the first part of it. I tried to go through and summarize some of my points while maintaining a bit of a high level in my explanations. I don’t think it’s basically for the benefit of the show and the listeners to go through the heavy math.

The paper goes as follows. It effectively says that trend following should not exist in the first place in some respects. It’s basically going against market efficiency. But the fact that it exists and it’s one of the more prevalent empirical features suggests that we should model the range of assets using a specific model. I’ll come to that in a second. Assuming that this model is the one that explains returns in a world that exhibits trends. How should I optimally design my portfolio?

That’s how it starts. How do I model the universe of assets if I accept the fact that they exhibit trends? And if they do so, how do I solve for a weighting scheme that maximizes my Sharpe ratio? That’s, more or less, how the paper is put together, I guess, from a motivational perspective. So, they start by saying, “listen, we have a number of assets, it can be equities, it can be commodities, a range of markets, we don’t even have to name them, and their return process would have some drift.”

Equities typically go up because that’s the reward for taking on downside risk, for instance. They would have some noise, as any return stream would have. So, it’s an unpredictable component. But then importantly, there is what they call a stochastic trend. In other words, occasionally, markets exhibit trends which are inducing serial correlation in the return pattern.

So, by putting together the drift and the stochastic trend, they argue that not only are we rewarded for holding explicit exposures in markets that are risky, but at the same time, there are some trends that come and go that, occasionally, I can capture if I monitor the historical return. That’s, more or less, how they put it forward, or how he puts it forward. So, there are two important stylized facts that are mapped in that framing of the model.

Number one is the serial correlation, so, the momentum and the trend following activity. But importantly, there are cross-correlations between trends. In other words, when some of the equities are trending, maybe some commodities are trending because of some sort of interlinks in their fundamental modeling. So, they tried to incorporate those cross-correlations between markets and the cross-correlations of trends alongside the single asset serial correlation.

So, these are, again, some of the stylized facts that have been captured by this modeling. Given that model, here’s the question. How do I allocate between the markets assuming that I follow a trend following signal? What they end up with, what he ends up with is a given number of assumptions. You end up with four different portfolios whereby, optimally, you hold a linear combination of those four portfolios.

So, what are those four portfolios? They start by bringing some assumptions to solve the model because it’s, to your point, quite complicated. They say, number one, let’s assume (now all the features that were put into the model are now stressed), let’s assume that there is no autocorrelation. Well, if you have a market that doesn’t have any autocorrelation, the optimal portfolio to hold is the one that is effectively a mean variance portfolio.

So, you follow the expected return. You go long or short on the basis of that expected return. Which under some assumptions is also the most diversified portfolio, or under some assumptions is the risk parity portfolio. Again, a few little details here. But the point is, if I have no information about trend following, all I’m going to do is allocate my risk in a balanced manner. So that’s the first portfolio.

Assumption number two is, what if I don’t have drifts? What if I’m not rewarded for holding equities? What if I’m not rewarded for holding bonds but I have trend following activity? So, from this modeling exercise, those two components that were introduced, I’m now left with this stochastic trend. If that is the case, then what you have to do is build a mean variance portfolio that is following trends.

Conversely, the first one that is following longer-term mean variance with drifts, it does trend following mean variance with trends. That’s portfolio number two.

Portfolio number three, there are more details here and more assumptions. I’m going to keep them on the side. You end up getting what they call the agnostic risk parity portfolio. This is a portfolio that initially was presented in academic literature by the CFM group in Paris.

The agnostic risk parity portfolio says any universe of assets has principal components, or what you call technically the eigenvectors. So how can I deploy my weights so that I’m equally exposed in all principal components? It’s almost like equal risk contribution, but instead of looking into the assets, you look into the principal components. So that’s another solution of their model.

The last solution of their model is to do the closest to our businesses. You do trend following using a risk parity mechanism. Some sort of inverse scaling or covariance scaling of trend following positions.

To summarize, there are four portfolios that are sub-components or sub-solutions of their generic model. Those being mean variance using drifts (that’s more like what they call risk parity long only), mean variance using trends, diagnostic risk parity that you allocate across the principal components, and the trend following risk parity.

Of all these, then they introduce some of our classical cross-asset futures universes. They deploy those four portfolios into this universe. They do a backtest for the last 20, 30 years, and they come up with some realized Sharpe ratios. The better ones, out of those four, are the agnostic risk parity that solves for the principal component. In all fairness, it’s not too surprising in some respects.

If the asset classes are capturing different principal components of risk, if you’re well diversified – if you diversify across those principal components, as well as the trend following risk parity portfolio. Optimally, a combination of the two seems to suggest that you have your optimal trend following portfolio. At the end, they add a bit of long-only risk parity.

But I think the gist of it is that whether you go through this route of modeling and solving mathematically a complicated model, or whether, empirically, we have found that some form of inverse risk or inverse correlation allocation between markets in a trend following space gives you better results over the longer term.

To summarize, it’s an interesting paper because somehow it brings value to some of our, at times, heuristics. But boy, it’s full of math.

[00:36:08] Niels

Yes, I can attest to that part. That’s for sure. Okay, well, I appreciate that, and thanks for sending over the paper, Sebastian.

Let’s move on to another paper that I was only sent by you today, I think. So, I have not read it myself. But I know the author because I’m speaking to him right now, at least one of the authors. There were two authors of the paper.

The paper is called “Pick Your Battles & The Challenges of Factor Timing” written in 2019, I believe it was. So, tell us a little bit about that and maybe how it relates to what we’ve just been talking about.

[00:36:49] Nick

Why did I pick this work we did five years ago? That was with one of my colleagues at the time. She’s now at Oxford University finishing her PhD. What’s the background? The background is that recently, and I would say in the last month or two, I’ve been receiving (it might be just a coincidence), but I’m receiving more questions around timing. And that is not timing just for trend following purposes or for the market itself, I think it’s more generic.

I build a portfolio of systematic strategies, can I have dynamic tilts? Is that the right time for trend following? I think that is exactly what we discussed earlier on with Risk.net and the fact that we cannot foresee the future. When you look into alternative markets, and we might discuss that a bit later on if we have time, alternative commodities have had an incredibly strong year so far, but also, the last six months have been extremely strong. That is not just cocoa and coffee.

There has a broad spectrum of companies that have done very, very strongly from a trend following perspective. We get questions along the lines of, “Look at that, the performance has been astronomic. Is it going to continue?” And really, the answer is I don’t know. Somehow, I guess, also looking for good topics to discuss, I thought of bringing that over.

Even if it’s five years old, I think the shelf life suggests that it’s still relevant. Because at the time that we wrote this report, we approached the problem of timing in the following sense. So, timing I think principally is about getting the direction right, and as a secondary factor is about sizing it appropriately. So, the way we approached it is in the following sense. We said, you know what? Let’s just make it easy but let’s just try to follow timing from a more theoretical standpoint.

Let’s assume that we play a game and you’re only having two choices. You’re either going to go long or short, no sizing. The underlying strategy, it can be an asset like S&P, it can be a strategy like trend following. It can be any asset we can go long or short, just to simplify the exposition of the analysis. So, you can only go long or short, and then let’s assume we have perfect foresight. So, we have a distribution of returns, let’s call it normal for the sake of argument.

So, if we can get any negative return as positive because we’re going to short exactly before the negative return arises, then we have a perfect foresight and we can solve mathematically for what is the expected return if we get everything right. That’s point number one.

Point number two is that getting everything right suggests that increasing the Sharpe ratio we’re getting is going to be lower and lower and lower the better the strategy is in the first place. So, if I tell you to time a coin toss, the best thing you can have is always win. So that is going to get your Sharpe ratio from zero to something very positive. But if I give you a strategy that has a Sharpe ratio of 3, there’s not too many negatives to turn positive. So, the Sharpe is not going to increase that much. So, you have some nice mathematical analysis that suggests that the higher the Sharpe ratio of the strategy you try to time, the lower the incremental value brings, however good you are, and we’re talking about perfect foresight.

Why is that as the first example extremely important? Because I think there’s a misconception in the market that if I’m 50.1% right, I’m bringing value. And the answer is no. If you’re timing a coin toss, probably. But if you’re timing strategy with a Sharpe ratio of 1, you’re actually destroying value, more likely than not.

So, we wanted to bring it up as a nice statistical analysis that you should benchmark yourself and your skill with the skill of the strategy itself to perform without you touching it. Because if it gives you a positive return, more often than not, you’d rather just pick your fights. That’s why the paper is called “Pick Your Fights, Pick Your Battles”.

That was the first thing that we tried to do. To basically say in a perfect foresight world, what is the incremental Sharpe ratio you can bring? And that is extremely related to how well it is, as in the strategy on a buy and hold basis in the first place. Even more so, the frequency of you doing or calling the direction is very important.

Would you do it per annum, per month or per day? Because if it’s a high Sharpe ratio for a strategy you rebalance annually, you get something wrong, it’s not going to look good. But that high Sharpe ratio on a daily basis is very small because you have to scale it down. Getting every single day right is going to massively explode your Sharpe Ratio. I can put some numbers here because I can look it up here.

Assuming you have a Sharpe ratio of 1 annualized, if you get every single year correct for a Sharpe ratio of 1, and assuming normal distribution, you go to 1.45. So incremental, 1 goes to 1.45. That’s it. The same strategy on a monthly basis has a Sharpe of 0.3 because you have to divide by square root of 12. So, this 0.3, if you get every single month correct, you will be annualizing 4.6 Sharpe. It skyrockets.

On a daily basis, the Sharpe is 0.06. Tiny, right? Because you need to divide by square root of 252. But if you get every single day right, and the numbers now start not making sense. You make a Sharpe of 20.8. So, the point that we wanted to make is that number one, pick your fights, but also pick the frequency as a function of your trading costs. I think it’s very important to put that into context but also frame the problem of timing not just in isolation but in the context of the strategy you’re trying to time right.

It goes further as follows. We never have perfect foresight. At the margin, we might claim to have a bit of a foresight based on economic regimes or whatever, forecasting models and so on. So, we twisted the question then and we said, you know what, we don’t have perfect foresight. The question is, what is the skill you need to have to outperform?

Now, to solve that, I’ll go into a bit of detail but not too much of it. We just discussed the best-case scenario. What’s the worst-case scenario? That you get everything wrong. So, a level of skill, let’s call it 60%, will give you an average return which is the average of those two extremes, 60% of the good, 40% of the bad.

So, then you can mathematically solve what should be that probability, that skill that I have to make that combination of extremes better on average than me holding the strategy. So, you can reverse engineer the math and find the skill that you require to outperform. It’s no surprise that the skill you require increases with how good the Sharpe ratio is and also with the frequency.

In other words, the lower the frequency, the lower the skill you require. Sorry, the higher the frequency, the lower the skill. So, if you do it daily, it’s okay if you lose a bit in a few days if your skill is good enough. The problem is costs. That’s the problem. Then we expanded the analysis, and we went to distributions with tails. So, if you sell volatility, for instance, you typically get good carry. But occasionally, you get crushed.

Now, what’s the timing skill here? Well, the reality is if the strategy itself has like a 70% hit ratio. You really, really have to be extremely good to pick those down points. Statistically speaking, we ran a lot of analysis and simulations, it’s basically unattainable. It’s almost as if you’re saying I’m going to predict the five equity crashes. It goes beyond realistic levels of skill to be able to time those down moves.

I think overall, that was the first part of the analysis, looking into a single strategy or a single market and trying to assess what is a skill we require to improve the performance and creating a handicap for us or highlighting the handicap being the buy and hold Sharpe ratio.

If somebody says I’m going to time trend following, well, the first statistical point that I would want to bring is realizing over the longer term, obviously, things are becoming more conditional if you want to really think about it based on the economic environment. Anyway, a Sharpe of 1, for instance, will require a good amount of skill to marginally improve the Sharpe ratio.

So I think statistically speaking, the better the strategy, the more you let it roll on its own because a false positive is going to cost you, and that’s what we want to bring forward because we want it to go against this 50% chance. Later on, we did a bit more of analysis.

By the way, all that analysis relates to trend following as well, because in trend following, you need to get the direction right. Then sizing is an important consideration, but you need to get the direction right. Markets that, more often than not, go up in principle can penalize your conviction to go short, for instance, and I think it’s an important consideration.

That’s the first part. Should I pose any questions, or should I just give quickly what is the second part?

[00:46:27] Niels

Give us the second part quickly. I have a question for you afterwards.

[00:46:32] Nick

I’ll tell you. So, the second part goes into a portfolio. If you hold a portfolio of five strategies or five CTA managers. Everyone has their own portfolio to manage. For the sake of argument, let’s assume that the correlations between all of them are equal, the Sharpe ratios are equal. How can you time this portfolio?

Well, in principle, you would want to say, well, out of the five strategies I hold, I’m going to overweight that particular one that I feel is going to bring me more value and naturally reduce all the rest. So, I’m going to bring some conviction that is going to be reflected using an active weight.

What you find, and there’s a lot of little analysis somebody can do with regards to the Sharpe ratio of the strategies, the forecasting skill, the correlation, what you find is that (and I think that was the most interesting result out of all) a universe that is extremely diversifying in itself gets little value from timing due to the diversification minimizing the impact, diluting the impact of timing.

At the same time, too little diversification and too high correlation also has no value when it comes to timing. Because if things are very correlated, trying to bring any active weight, however good your active skill might be, it’s not going to move the needle. It’s only going to add costs. So, in some respect, you need lower levels of Sharpe ratios at the individual level of strategies. Going back to my previous point, as well as relatively moderate correlations. Not too high, not too little.

If I take two examples that we also have in that report. If you look into equity factors, equity factors are relatively low Sharpe ratio strategies, but they also happen to have relatively moderate correlations. So, timing that in some respect might create active deals, might be easier than having a set of volatility carry portfolios, or volatility carry strategies that tend to suffer occasionally but have a very high Sharpe ratio and a relatively high level of correlation among them.

So, this is the focus of that work. I wanted to bring it back and say, listen, timing is hard. It’s not just hard empirically. Statistically speaking, it’s hard. There are reasons why some strategies with high Sharpe ratios should be left as they are and avoid that temptation. Unless somehow the market environment or your confidence goes beyond those thresholds. So, pick your battles. That’s the title.

[00:49:12] Niels

Okay. I guess my question is, we know that trend following is not necessarily known for the higher Sharpe ratio. And frankly, that’s probably not why you should buy it in the first place. But just from your own overall experience and given everything you know, everything you talked about here, do you think there’s any merit in trying to time an investment into trend following?

I don’t mean as a trend follower. I mean an investor looking at trend following managers. I’ll just anecdotally just say that our founder, Bill Dunn, when asked this question, would always say, the best time to invest in trend following is at the bottom of drawdown, the second best time is today.

[00:49:58] Nick

I would say today, if that was your question. So, I’m halfway there. Is there a way to time it? Very, very hard. I think part of the reason why it’s hard is because you do get directional exposures. So, can we time equities? I’m not sure. Can we time treasuries? Maybe in some respect if you assume that the central banks are a bit more serially correlated. Can we time currencies? Not really. Sometimes they go hand in hand with global growth and the strength of the dollar. Can we time commodities? No. It’s hard.

So, in some respect, the fact that we don’t build cross-sectional portfolios but rather we build directional portfolios, timing trend, if you were to break it down, suggests that you somehow can time the underlying beta markets which themselves, if anything, definitely equities and rates, embed the most fundamental risk premium. The time value of money and the growth risk respectively.

I don’t think timing is easy. I think that’s the fundamental reasoning behind it, in my view. Now, it’s not like a value for instance which is purely cross-sectional. Somehow you can make the argument, look, the spread is very wide, maybe the opportunity set is better, maybe now it’s a better time to get into it. Or sometimes you look into implied volatilities, they are quite extreme, that’s a good entry point. Maybe it will be a bit costly, but historically, you do get some sort of statistical confidence from the short term.

I cannot see how, on a directional basis, somebody can argue that I can time trend successfully. I think we’ve discussed in the past the possibility of reducing some of the risk when you observe too many trends across the board. And I think this is when the risks in the portfolio become more concentrated. But that’s just a moderation in the top-level vol, but timing it is very hard.

I’ll give one snippet of analysis we’ve done. If there are no trends, what do you do? So, assume you count your models, you measure your trendiness across the markets, and you end up building a portfolio, but at the core, it has no trend. So in principle, you say, I’m going to stay out. Statistically speaking, you’re losing the start of a trend. You’re missing it.

You’re underperforming if you were to not deploy exposures when some trends are not explicit but they start forming. So not even in that case environment, in that case statistically speaking, it makes sense to stay out. I think it’s hard. I genuinely think it’s hard.

[00:52:49] Niels

I don’t think anyone, at the end of December last year, would have thought that oh, yeah Q4 2024 is going to be fantastic. There was no sign of that. We’ve seen it so many times. We did not know two years ago that things would explode given the return of inflation, a war breaking out, and so on and so forth.

But I do agree with you, Nick, that I am also discussing a lot of things about timing. Because then people say, “oh, well, we just had a good run in trend following so maybe we should wait,” or “I don’t think there’s going to be a lot of trends in the next six months so I’m going to wait with investing.” To me, as we talked about earlier, it’s completely the wrong way of looking at this particular strategy.

But I’m not going to beat that drum anymore because we only have a few minutes left. There’s one other topic, in particular, that I get a lot of exposure to in my travels. It’s this question about whether trading hundreds of markets, including alternative markets, versus small, let’s call it classical developed markets, whether one is better than the other.

I sense from the conversations that I have, that I think a lot of investors have a bias towards the narrative that suggests that having alternative markets in the portfolio is better. I will say from my conversations with some of my friends in the industry, I think they make a distinction about whether they are alternative but still fairly liquid or whether they are alternative and hard to trade. Maybe they do find some interesting properties in markets that are actually operationally hard, and of course, therefore, there will be fewer people trading them. Maybe that makes a difference.

I unfortunately sound like a broken record often when I say to people that as much as that can be true in a given year, maybe even over a two-year period, in the very long run, if I look at the well-established managers that’ve been around for a long time. Many of them are part of the mini CTA series that Alan and I did last year. The long-term returns don’t strike me as they support that view that those managers, who have alternative markets in their portfolio, have done better in the last decade.

I just don’t see that. But I do recognize that they are different, of course, and will perform differently at different times.

It’s not that I want to make it a conversation about Chinese markets, per se. But I did see some headlines, and I know it may not be relevant for CTAs, but I did see some headlines this week about China cracking down on certain hedge funds because they had been too active on the sell side of Chinese equities.

Which brings me back to this point about, what are the risks of going into some of these countries to get that alternative exposure? Can we completely ignore that? But that’s a slightly different conversation maybe? What are your thoughts?

[00:56:18] Nick

From a pure statistical perspective, the marginal gain you gain from additional markets drops once you have X in place. Once you have that, then the X plus one and X plus two marginally reduces the contribution it can bring, let alone the costs. So, I think expanding a universe with more markets, “alternative” or “exotic” or however you want to call them, is a consideration between marginal gain versus cost.

To your point, whether it’s illiquid, or harder to get hold of, or you need an onshore broker, whatever the complexities of access can certainly increase the hurdle. Maybe a component of a program that only isolates positions in those markets could have some value added. I would look at it more as a satellite component that can be added on top of a core market allocation.

Frankly, I’ve been convinced, in some respect, that some of the alternative commodity markets have value to bring in a trend following portfolio partly because now we see some of the more benchmark-oriented markets not necessarily bringing too much. Except for occasional spikes like in the first quarter of 2022.

I think the trend following in classical commodity markets has not necessarily been the best in years. I think both 2010, 2011, 2012 have been less so over the outperformance within the cross-asset part. But markets that sit outside of them, of the benchmark, in some respect because we’re talking about commodities which in their family are quite idiosyncratic even among them. There’s more diversification between them and I think there is more producer-consumer dynamics rather than speculator dynamics that can allow some of the trends to last for longer.

I think if we look into cocoa that’s been in the news, or Robusta Coffee, or nat gas to your point and winter being one of the warmest winters we’ve had. I think it’s been the warmest winter on the record so far in the northern hemisphere. I think there are some sort of fundamental reasons why we observed those trends, and those strengths are quite sustained.

When we talk about what drives trends and you take not one step back, but 100 steps back, that’s how we all started. One of the reasons that we bring this forward is that there are market structures that can make specific demands for a particular market or supply, I guess, that create serial correlation of the returns. These can be efficiently captured. I think in those markets, you do observe those dynamics.

I can take cocoa, the fact that we know when we observe, and we document at the present moment how constrained supply is. The growing conditions are poor, shipping disruptions. We know that the supply constraints will be there. Some of the trends we’re seeing, that actually continue until now, as we speak, could be driven by that. I think there’s value in those, but there has to be prudent portfolio construction and sizing. That’s my view. I’m a believer in those markets.

[00:59:42] Niels

Yes. I get the sense that perhaps you and I think of alternative markets slightly differently. Maybe you include some markers that I actually don’t think are alternative anymore like cocoa. I think that’s a perfectly liquid classic. Well, I say liquid but it’s a classical market that’s been in trend portfolios for decades. But of course, there are some contracts, and some of them are even hard to pronounce, that are clearly alternative and so on and so forth.

But I think it’s a very balanced view. I actually think that if you were an investor that had to pick, say, four or five different trend followers, because that’s probably still prudent if you have a large amount of capital to choose, then of course you would- at least in my opinion, you would choose managers that are structurally different. They all have to have certain base criteria that they live up to in terms of quality and all of that stuff.

But clearly, there’s a structural difference between picking a manager trading 300 or 400 markets and picking a manager to trade 75 markets on top of different types of trend models, different types of risk management, etc. So, I think there’s room for everyone. I think that’s the good news here. There are risks and rewards with all of these choices.

But commodities in general, if we’re referring to them as the most alternative markets we can find, are wonderful. Cocoa is up 136% in the last 12 months, and natural gas is down 51%. What’s not to like for a trend follower in those markets? So yeah, it’s a conversation that’s going to continue without a doubt. But at the end of the day, it’s the performance numbers that will essentially give us the evidence whether one is better than the other, even though it might be driven by narratives right now. So yeah, interesting.

[01:01:40] Nick

I’ll leave you with one final statistic which I find astonishing. I think today is the 40th business day of the year, or you can call it 39th or something, and if you take out some of the holidays here and there, 38th, it doesn’t really matter. But give or take, we’ve had like eight weeks since the beginning of the year.

One of our trend following strategies, the one that is focusing on some of those commodity markets we discussed, it’s been having a stellar year, no surprises. I think it has had out of those 40 days, 8 with a negative return. Obviously, this is not a reasonable statistic in itself. But this is partly a testament of the strong performance of those markets as of today.

But even more so, this is when some of the questions arise as to should I get in or should I wait? And my point is, listen, statistically speaking, let’s look at the data. Past return, at the strategy level, does not predict future return, statistically speaking. Then let’s look into the markets. Are there structural reasons to believe that those markets that are driving performance are going to crash?

So, it’s one thing to go sideways, it’s another thing to crash. I think that’s how we should be looking into those questions. Quantitatively but also qualitatively in some respect and finding the reasoning behind why those trends exist and whether there is reason for them to continue existing.

But lastly, I agree with you, I think our definitions are not necessarily 100% compatible purely because we would have to draw a line at the point that liquidity becomes too little but not too little to trade for the managers. But for us, building indices on the back end of those markets, we have to make sure we guarantee the close and all that. Certainly, there’s a difference in the definition here, but I think the principles are the same.

[01:01:47] Niels

Yes. In fairness and to be balanced about this, I don’t know this for a fact, but it could well be that if you looked at the last 40 days of trading and you look at your equity sector, that that might also actually have been up most of those 40 days. So, this is the wonderful thing about trend following that we have these very strong trends that capture a sector.

In 2022, it was like bonds. For days and weeks and months, you almost made money every day. I wouldn’t say every day, but it was so consistent. Then you have periods where it’s equities. And then as you said, we have periods right now where it seems to be in the commodity space. That is just the wonderful nature of us being allowed to play in a sandbox that is much bigger than what traditional portfolios can offer.

Hence, this is why we deliver return streams that are different. And on top of that, actually are incredibly complimentary to the traditional portfolio. Although you should be careful saying it’s a win-win-win, it is a pretty compelling opportunity.

Nick, it was also very compelling to speak to you today. As always, I really, really appreciate you and your time, as I’m sure all our listeners do. One way of showing Nick how much you appreciate that is to go and find your podcast platform of choice, leave a rating and review, tell the world how great Nick is on these episodes, because it really greatly helps us, and it shows appreciation for the hard work that all the co-hosts do every week.

Next week, another wonderful person is on the show, Cem. If you have some questions for him, probably more in the volatility or global macro space, send them over info@toptradersunplugged.com, and I’ll make sure I’ll tackle those with Cem. From Nick and me, thanks ever so much for listening, and we look forward to being back with you next week. Until next time, take care of yourself and take care of each other.

[01:05:47] Ending

Thanks for listening to the Systematic Investor podcast series. If you enjoyed this series, go on over to iTunes and leave an honest rating and review and be sure to listen to all the other episodes from Top Traders Unplugged. If you have questions about systematic investing, send us an email with the word “question” in the subject line to info@toptradersunplugged.com, and we’ll try to get it on the show.

Remember, all the discussion that we 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.

Thanks for spending some of your valuable time with us and we’ll see you on the next episode of the Systematic Investor.