226 Systematic Investor Series ft. Rob Carver – January 15th, 2023

[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 Rob Carver and I, Niels Kaastrup-Larsen, where each week, we take the pulse of the global market through the lens of a rules-based investor. If you’re new to the show, I hope that today’s episode will trigger your curiosity enough to check out the back catalog and listen to the past episodes that you may have missed, like my conversation with Alan last week where we introduced the new amazing mini-series where I think we’re doing something that has never been done before as we have managed to line up most of the largest managers in our industry to discuss the most important topics about their strategies, the business, as well as the industry as a whole.

We can’t wait to start sharing these conversations with you, which I believe we will start doing on Monday. Also, if you missed our Wednesday episode, I really would encourage you to listen to the midweek conversation we had with Susan Shirk, who is probably the foremost expert on the US-China relationship, and who takes us behind the red curtain to explain how China essentially derailed from its peaceful rise. So I really encourage you to go back and listen to these episodes if you missed them this week, but not until you finish listening to Rob and I today of course.

Rob, it is wonderful to be back with you in the new year. So much happens every time we are apart, so to speak, in our conversations. How are you doing? How are things on your side?

[00:01:51] Rob

It’s wet and raining outside as I speak so it feels like proper wintry weather now. As we were discussing before we started the recording, I’ve got a bit of a cold which will explain why, I don’t know, I think my voice sounds a little bit sexy like this, a little bit gravelly. So hopefully, the listeners will enjoy the change in tone.

[00:02:11] Niels

I’m sure they will. They might even enjoy the content as well. You never know.

[00:02:14] Rob

Oh, that would be going a bit far now, surely.

[00:02:17] Niels

All right. Well, we do actually have some good topics in our unbiased opinion, of course. But before we do that, here’s a little market wrap because there obviously have been a few things going on this week.

Before I do that, maybe as it is the first SI episode in the new year, let me just acknowledge and say a big thanks for all of you who left a rating and review in the last few weeks on iTunes, Spotify, Amazon. First of all, they’re really nice, and secondly, we do pay attention to them and we do appreciate them very much.

Now, this week, it was all about the December CPI report. It was released on Thursday, and it was a pleasant surprise for long-only investors. The headline CPI fell 0.1% month-over-month, and the year-over-year measure fell to 6.5% from 7.1% in the previous month. Now, core CPI (the measure that excludes food and energy) rose 0.3% over the previous month, a slight uptick from 0.2% in the previous report. Core CPI has fallen to 5.7% year-over-year from its peak of 6.6% reported last year.

The market breathed a sigh of relief as witnessed by the massive rally in the long bond on the day of the release, closing nearly two points above the previous day’s close. Much of the rally seemed to have been driven by short covering and driving the yield to maturity down to below 3.6%, I think. At that yield level, it’s hard to justify, perhaps, buying on fundamental reasons only.

Also on Thursday, Philadelphia Fed President Harker said the days of the 75 basis points hikes are surely behind us and that 25 basis points hikes will be appropriate going forward. Those comments and encouraging CPI release has driven consensus to expect a 25 basis points hike at the February 1st FOMC meeting. He also said that a soft landing was still possible even if it was bumpy. What he probably means is that the economy may avoid recession but that the unemployment rate will rise. Reading the outlook for 2023 that has been published so far, that view is a bit out of consensus. Pundits are expecting a recession later this year, and Fed fund futures are predicting a rate cut later in the year as well.

For their part, no Fed speakers are hinting at rate cuts anytime soon. Although this is not market-moving and not really mentioned in the press, the Federal Reserve Open Market portfolio of holdings fell below $8 trillion for the first time since its peak at nearly $9 trillion last April. As a reminder, the Fed is letting the portfolio run off to a tune of $120 billion a month right now. Next week, we’re going to get a measure for how the holiday selling season faired when retail sales are released on Wednesday.

All right, Rob, what has stood out to you in the last, I don’t know, six weeks since we last spoke?

[00:05:29] Rob

I was just looking at the year-to-date numbers actually, going back as far as six weeks. I guess it’s been a fairly smiley positive time for the markets. Both bonds and equities are up, vol is down. If I look at my market monitor report that takes all the moves across all the futures markets (I monitor and vol adjust it to put it on the same scale), the biggest fall is V stocks and VIX. So the vol markets, obviously. On the plus side, I see a lot of equities; Euro stocks, DAX, and some European stocks have done particularly well. The rally in fixed income actually is not showing up in that report so it’s not that significant.

Unfortunately, for my portfolio, it was rather significant. I have a short bond position on, and I’m still short bonds actually so I haven’t closed that position yet. That lost me a little bit of money. I’m actually down for the year-to-date 38 basis points, which is not very much but it’s only been a couple of weeks so pretty much noise. Without those bond positions, I would have been up so that was a bit painful because I’m short vol so that made me money. I’m also long various equity markets so that also would have done pretty well, I guess.

Looking forward slightly, I’m still trading with roughly low risk, running at about a quarter of my long-run expected risk target at the moment. Biggest position is still short bonds, short US 20 years. My biggest long position is… My long position is actually a bit more dispersed. There’s nothing that stands out. But I’m long equities, as I’ve said, I’m long metals as well as it happens. I’m short vol, which is like being long equities if you like. So I still have a very much long equity short bond position on, which with the way the markets have moved over the last couple of weeks, this beginning of the year is fairly balanced. Those two things have pretty much hedged each other off and they’ve had small losses. So we’ll see what happens if this bond rally continues.

Obviously, I’ll expect to close out that bond position and have a fairly procyclical position on. Who knows? If inflation has peaked and is falling and if all recession risk is priced in, then yeah, maybe that will turn out to be a good year. We’ll see how it goes.

[00:08:02] Niels

Absolutely. Yeah. It’s interesting. You mentioned volatility a couple of times. If my memory serves me right, I thought I noticed that I think vol or VIX is kind of hovering around just above the low of all of last year. It’s really been pushed down. If you think about it, we’re heading into a new year. A lot of the crises from last year are completely unresolved yet volatility in the market or in the equity markets, at least, is not suggesting any concerns really. So yeah, it could be interesting.

[00:08:33] Rob

Just to say, I think there’s some weird stuff going on in the vol markets, actually. So it has been a bit of a puzzle that in a year in which equities sold off pretty radically, vol hasn’t really been elevated that much at all and now, as you say, is very low. It certainly drew my attention. If you look at the put-call ratio, which I think Bloomberg publishes, it’s at a historic high.

The other weird thing that’s going on is there’s been a big increase in people trading very short vol, one-week options, which in technical terms, means there’ll be a lot of gamma hedging going on. I’m not a vol guy. You have vol guys on the podcast who’ll be able to express their opinions better on those, but the vol market, certainly for the last 18 months or so, to me, has felt really strange and almost disconnected from the equity markets. So yeah, that’s an unsolved puzzle for me.

[00:09:30] Niels

Yeah, it has been a conundrum and actually, the way that Cem and I think Hari has spoken about it is really that last year, investors just seem to have been much better hedged than before and therefore, there’s been much more supply and therefore, we’ve seen these unusual periods with a very little spike in volatility when markets were trading off. Now, in terms of trend following, I will say the first couple of weeks have been a little bit soft, I think softer than your portfolio. I think your portfolio actually is doing pretty well.

I think CTAs in January are giving back a tiny bit of the profits from last year, which of course was a record year, let’s not forget that. On first sight, I would say this week, probably not a lot happened even though we had a little bit of excitement from the CPI report. But in a diversified portfolio, I don’t think that it played a major role. But when you then look and drill down a little bit, I did notice that things like aluminum had a bit of a turnaround and suddenly had like five or six pretty strong days in a row against the longer-term downtrend at least, if you’re a longer-term manager. I think it made a new recent high every single day of last week. So that’s going to cause a little bit of pain for longer-term trend followers.

But offsetting that, I will also say that another market that had a pretty good week was the Mexican peso. It’s been in a big uptrend for a while now and I think that certainly helped out most trend following portfolios last week. Otherwise, when I look through a diversified portfolio, I would say fixed income, as you suggested, that’s where most people would have lost their money last week as it went higher on the back of the CPI data, but maybe also in metals. Although I did notice that you said you were long, but I think a lot of people are still short. Equity is probably mixed depending on which part of the world you’re looking at. I think energies probably made a little bit of money, so did softs perhaps. Grains and meats were pretty flat.

My own trend barometer, though, has picked up. Now it’s 55, which is getting close to a nice level. Of course, people might think that if trend follows are down for the month and the trend barometer is up in a strong level, how does that work? Well, the trend barometer is a much shorter-term timeframe in terms of how it’s calculated. So I think that if it stays up for a while, then maybe trend followers will also start to do better, or longer-term trend followers will do better.

Anyways, in terms of data, it’s super easy because month-to-date and year-to-date is the same, so I just have to mention one of them. BTOP 50, down point 0.6%. SocGen CTA Index, down 1 (changed). Trend Index, down 1.32% and the Short-Term Traders Index is up 0.25% and that is of course as of Thursday evening. Other than that, MSCI World, obviously having a strong start to the month, up more than 5%. Government Bonds up more than 2% and the S&P 500 Index up more than 4% so far.

Well, let’s jump into some of the questions we had coming in, some of them a few weeks ago, but because of the holiday schedule, we’re going to bring them up now. So, the first one for you was from Andrew. It’s about his moving average crossover journey. Andrew writes, “I was using the 16 and the 64-day exponential moving average crossover but found that I was getting in and out of trades at nearly the exact same price after long journeys in both directions. So I switched to a 5-day and a 20-day EMA crossover. Using 5 and 20 EMA, I was using stops at 1.5 ATR at entrance to stay in the trade, but I found that I was getting stopped out, but that the trend would continue. So I stopped using stops to wait for the crossover to occur and then I found that I was taking 7% to 12% losses, which is not cutting losses short. Also, sometimes, I was entering the trades late based on the trend following study that said it’s better to be in the trade than to wait for the next entrance. Questions: Should I use a stop or wait for the crossover to occur before I exit? Second question, should I only enter trade right at the entrance?”

[00:14:08] Rob

I have to say, this is interesting because I think this is a very common story for a lot of people. If I just summarize what Andrew has done, I’ve got this system and I started trading it but it didn’t do what I expected or it lost some money when I started trading it or I wasn’t happy with it so I switched to System B. The same thing happened again. It didn’t do what I expected or it lost some money so I switched to System C. Now what Andrew seems to be saying to us is what’s System D, what’s the next thing I should be doing?

But to be honest, Andrew, what you should be doing is just sticking to one thing. If I look at your first system which is a 16/64 crossover, that’s a pretty good system if you’re going to trade a single crossover. It’s not too fast but it’s fast enough that it captures trends fairly quickly. In my own empirical research, I actually find that’s the best out of a particular set of crossovers you can use. Just to add this, the difference is not statistically significant. The other crossovers either side do pretty well as well, and to be honest, you’re probably better off trading a blend of crossovers. We’ll come to that in a minute.

So there’s nothing wrong with System A. You should have just stuck to it and carried on with it and be prepared for the fact that over a relatively short period of time (and he doesn’t say the time period between each of the decisions). If he was trading the first system for 20 years, then okay, perhaps fair enough. But my suspicion is that’s not the case.

[00:15:44] Niels

Also, he doesn’t mention how many markets he trades, which I think also is a very important factor.

[00:15:50] Rob

Yeah, definitely. Because obviously, if you’re trading a small number of markets as a retail trader typically is, it might be that over a relatively short period of time, you might just get unlucky and the crossover you’re using just doesn’t work. Moving on to the specifics of his questions, which crossover is best? You’re better off using a blend of crossovers. If you were to use a single one, personally, I’d clump the 16/64. But really, the key message here is to define something and stick to it. As to the question of whether you should use stops or wait for the crossover to change when you get out, this is, to be honest, a status preferences to be honest.

Again, in my research, I found a small improvement if you use a crossover on the stop. But there’s nothing wrong with using the stop and it makes the system simpler and easier to understand as well and easy to run manually. So, go ahead and use a stop. Just be careful that the stop is calibrated such that it’s getting you out of the trades at about the right level. He says 1.5 ATR, I would say that’s way too tight probably for the crossover he was using. For the original 16/64, I would use probably 7 ATR as my stop, because that will give you something that will… It’s not about optimizing the ideal exit, it’s about making sure that the stop is aligned in time with the holding period of the underlying trading signal if you like.

If he goes on my blog or buys my third book Leveraged Trading, there’s a big discussion in there about how you can match these things up if that’s the route he wants to go down. But yeah, Andrew, the key thing here is you’ve got to find something that you’re comfortable with and you understand and just stick to trading it. I feel like you’re almost, I hate to be rude, but you’re quite a bad advert for systematic trading, because what you’re doing is not systematic. You’re switching from one system to another system to another system, based probably on insufficient evidence and understanding of what you’re doing.

So yeah, find something you’re comfortable with, make sure you understand the characteristics it’s likely to have and then just stick with it.

[00:17:58] Niels

Yeah, just one more thing to add there and that is, I think you should also think about what your backtest looks like because there must have been some evidence of what you’re seeing in real-time in your backtest as well. So it shouldn’t come as a surprise. Again, I have a suspicion that it could also be to do with the number of markets you trade. Anyway, let me just make it absolutely clear that of course, whatever Rob is saying with specific things, this is not investment advice. We take no responsibility whether these things work or not but this is just ideas to go and do your own due diligence and checks and all that stuff.

We move on to a question from Joseph. Joseph writes, “A huge fan of all the work. Thanks for your generous contribution to the trading community. Question: Is running your own trend following portfolio a reasonable idea in 2023 given the vast number of professional funds available and the importance of an ensemble approach? Said differently, do you think the professional fees paid to external managers are easily offset by the ability to take greater risk in an ensemble and the benefit of being a passive investor?” Then he says the impressive track record of DUNN and others set the bar extremely high for new managers and DIYs. We’ve talked about this before, actually.

[00:19:22] Rob

We have, yeah. This is one of those how-long-is-a-piece-of-string questions, to be honest. If you take someone like me who’s obviously quite comfortable and experienced in running their own trading system and has a reasonable amount of capital to do it with, then it probably wouldn’t make a lot of sense for me to go and take my money and go and invest in a bunch of funds. I wouldn’t get my checks to return from doing that. I have so much fun trading my own money as well so there we go. But at the other extreme, if you take someone who is not as experienced and not as comfortable and has a much smaller amount of capital, so perhaps, a few thousand dollars or euros or something like that, you’re going to struggle to trade a diversified set of instruments that’s going to give you much of the performance of a diversified portfolio.

Let’s just talk very quickly through some theoretical numbers, and then I’ll pass it back to you, Niels, for your opinion. So, I would say, very roughly, it seems okay to assume that a very large, diversified fund can achieve a Sharpe ratio of perhaps 1 to 1.2 before fees. An individual investor trading a small number of markets is probably looking at maybe a quarter of that. If you assume you’ve got a 20% performance fee and a couple of percents management fee where they’re maxed out, that individual investor would be better off in putting their money in a fund. Even better, if they can do it, maybe via UCITS funds, into an ensemble. If you find a dozen CTAs with UCITS put into a dozen UCITS funds, that’d be great.

To do direct investment into a large number of funds normally requires quite large amounts of capital. So that’s a different story. As you get more of your own capital and you are comfortable with what you’re doing, there’ll be some kind of break-even point where, mathematically, at least, you’re unlikely to be getting much benefit from investing in an ensemble of funds rather than your own fund. So somewhere between a few thousand dollars and not much experience, and several million dollars and plenty of experience, somewhere along that point is a break-even in which it makes more sense, if you want to, obviously, because not everyone wants to manage their own investments, mathematically at least, rather than putting it into a bunch of external funds on which you’ve got to pay fees. So that’s my take. It’s a purely mathematical take if you like.

[00:22:02] Niels

Exactly. I was just going to say, that’s the math part. Funnily enough, we’re going to talk about math later on today, but that’s the math answer which I don’t disagree with at all. I think there’s a bigger question that people always have to ask themselves and that’s life-work balance. Because if you want to do what you’ve done, Rob, people need to take it very seriously and they need to devote a lot of time. Not everyone wants to do that. They want to play with their kids, they want to go out on a Sunday instead of sitting in front of their computer putting code and having to defend that to their spouse maybe because they’re not spending time with them. All of that stuff, you have to take into consideration. So this is why I’ve often been an advocate of actually using our great colleagues in this industry for making investments. So just keep that in mind as well.

Now, here’s a nerdy question for you, Rob, a very short one from Bart. Bart says, “Can you discuss the importance of kurtosis in managed futures? I understand the benefit of skew, but the kurtosis is still a mystery to me.” I’m sure it is for quite a few people.

[00:23:16] Rob

I love the nerdy questions, Bart, so thank you for that. Let’s just quickly define what kurtosis is, because not everyone will be familiar. It’s just a fancy way of saying a fat tail distribution, basically. So if you compare a distribution of returns to a Gaussian normal distribution, then you’ll see much more mass in the tails of the distribution than you would expect. So if you look at something like just being long equities, S&P 500, with a normal distribution, there should be a relatively small number of days in which the index loses more than say 3% or 4%

. If you actually look at the history of returns to the S&P 500, there has actually been many times and more days when the index has lost 4%, 5%, 6%, 7%, all the way up to the 20% loss that we saw in 1987. So, it’s another way of saying fat tails. The relationship with trend following, it’s back to our old favorite subject of outliers. These fat tails, the idea is that as a trend follower, you want to be on the right side of them. The idea is that basically, by staying in a position when it’s in your favor, you’ll hit all these nice fat tails and then if you get a fat tail against you, you’ll cut your position and be out of it. So the idea is that you can transform a distribution that’s unpleasant like the S&P 500.

If you can be short on the days when there are negative fat tails and vice versa, then the distribution of your returns or your trend following strategy will look like a nice positively skewed distribution with lots of big fat right-hand tails; so lots of big fat positive days and a smaller number of days in which you lose a lot of money. The way I like to think of trend following is you’re taking an underlying market’s distribution, you can think of it as a piece of clay, and you’re squeezing it and massaging it and changing it into something that you’re happier with, and making it more positively skewed.

If you have fatter tails in the underlying distribution, then you’ll end up with fatter tails in your trend following distribution. The idea is that hopefully, there’ll be more on the good side than on the bad side and that will [result in] lots of nice, wonderful, positive skew. So, that’s the general idea. If you imagine that you’ve got a portfolio of normally distributed returns, so lots of markets that behave exactly as the finance textbooks said you’d expect them to, and you look at the diversification benefits from holding that portfolio, they’re basically limited by the amount of diversification you can get measured through linear correlations, basically.

If you go from one market to a hundred markets, the improvement is about 2 to 2.5 times, in terms of improvements in risk-adjusted return, which is nice. It’s better than nothing. But if you actually look at what you get doing that, the improvement is more like 4 or 5 times. It’s much better than you’d expect given the underlying distributions being normal. It’s the fat tails in those underlying distributions that are doing that.

So, the short answer is we like fat tails because we like outliers. That gives us a higher diversification effect because trend following a single market (as we discussed briefly with the first caller), is not that profitable. So you need to have lots of markets and this outlier, this fat tail effect, allows you to get additional diversification from them.

[00:26:52] Niels

Cool. Good. The last question is from Russ. Russ writes, “For decades, trend followers prided themselves on one set of parameters across all markets in their systematic models. Whether it’s X days for breakouts, or X days for moving averages, or a mixture of both, etc., Jerry and CTAs are adamant against no optimization for each market traded or optimization much at all, if at all, to get their parameter selection curve fit. I know Jerry prefers longer and slower parameters. Question one: how do you all pick your parameters that you do trade? Throw darts at the wall and take the one that the darts hit? Do you choose a 200-day breakout based on your experience in markets, etc.? Why and how?

With that stage set, a little more background. There are probably shops out there that game or hunt CTA stops and parameters. They know what a lot of CTAs use and game them to death and take easy money at times (stop them out essentially). This, among other possible reasons, has led to something I’ve noticed in the discussions in your podcast. I’ve heard statements or hints that managers and shops are now using short, medium, and long-term system parameters to avoid fast crashes or enemy firms gaming their models. Fast medium or slow parameters has felt like fancy talk and dancing around what really might be curve fitting.

Why not just confess the sin and optimize the parameters and get something that will probably be the average of the fast, medium and slow parameters anyway? Question two: Is fast, medium and slow parameters in models a recent development or can you confirm if this has been done during the ’70s and the ’80s?”

[00:28:54] Rob

Goodness me. What a question! This actually relates a little bit to a paper that we’ll discuss in a bit. So yeah, there’s a bit of a theme to this episode, definitely. This actually comes back as well to the very first question about which crossovers should I trade. So 16/64 is kind of fast-medium, I would say. Andrew said he switched to something even faster and then he wasn’t sure what to do. So, yeah, it all relates together. Well, how do I pick all the parameters that I trade? Is it just throwing darts to the wall?

[00:29:28] Niels

Don’t you actually have a dart in your shed? I see it right there.

[00:29:31] Rob

I do. Obviously, this being a podcast, people listening can’t see this but if I move my head to one side, there is a dart board behind me. But that is not how I pick my parameters you’d be pleased to know, or maybe not pleased to know, maybe that would be nice. In all seriousness, you can do tests where you completely randomize your parameters. It’s not a bad test, actually, of the fitting of your trading strategy because you basically want to see the difference between what you’ve actually fitted and what a set of random parameters would look like, because with some markets, there might not be that much difference to be honest.

If you take something that’s done very well with trend following in the backtests like, say, bond futures until quite recently, then randomizing your parameters, you wouldn’t necessarily see a lot of difference. Anyway, I’m getting a bit off-topic there.

So, I do a mixture of things. I’m a big fan of keeping things as simple as possible. Actually, in my production system, what I’m actually trading right now, if I was to take the example of fast, medium, slow, in fact, that’s three things if you like. I’ve actually got six things at least in my core trend following system. There’s a lot of other stuff in there as well. I’ve got different ways of doing trend following. But to keep things simple, my core moving average crossover strategy, I’ve got maybe six speeds.

What I do is very simple in my production system. I basically just say, “For this particular market, which of those six things are cheap enough for this particular market to trade?” So, for something like S&P 500 or NASDAQ, I could trade all six of those. For something like…

[00:31:15] Niels:

Cocoa.

[00:31:25] Rob:

I don’t trade cocoa, actually. Let’s stick with the theme and say dry milk futures. They’re very expensive to trade. I can maybe only trade one of those crossovers. Then what I do is I’ve got either one crossover, or two, or three, or four, or five, or six, and then I just take an equal weight across them.

That’s actually not the empirically or theoretically best way of doing it, but it’s easily the simplest and is obviously very robust to accusations of overfitting, which is obviously what is being thrown at us now. So yeah, that’s how I fit that particular thing. I have other things I fit like risk weights for markets, and so on and so forth. What I’m trying to do is basically, most of the time, my philosophy is, what’s the simplest way I can do this, the most robust way I can do this? Is there sufficient statistical evidence to move away from that and do something that’s a bit more complicated? If not, I stick with the simplest approach, basically.

So yeah, if you think about a slightly different problem which is allocating risk weights across markets, with that, if I was to just give each market the same share of the risk capital, that would mean, for example, I would be massively overweight equities because I have a lot more equity futures than I do have other kinds of futures. I would be underweight vol futures because I think there’s only two vol futures I actually trade. So that would have a much smaller weight than they actually get. So that’s an example where there’s very strong empirical evidence and just common sense that the maximally diversified portfolio is going to have less equities and more vol than the simple equal weight across markets would.

Now, let’s get a little bit more into the idea about whether you can actually game CTAs. The logic here is, I guess, that all CTAs are pretty much doing the same thing. That’s the first point. What follows from that is that if you can work out what it is they’re doing, then you could front-run them effectively. I think front-running them is a more accurate way of describing what you would do. What you would do if you knew that CTAs were likely to be buying tomorrow based on what’s happened in price over the last few days, you’d buy today and you’d hold that position and then close it out. Russ talks about stopping out CTAs. I don’t think you can necessarily do that, stop out the entire CTA industry. If you could do what people were doing, you could definitely try and front-run them. Absolutely.

Personally, I think that if you try and do that, what you end up doing is trading like a CTA, but faster, pretty much. You end up just trying to trade faster trends. We know that fast trend following works okay. We’ve been talking about fast versus slow quite a lot in this series. To be honest, using a mixture of speeds is the most robust thing to do and we already said that to the first guy. Picking a single speed is going to potentially lead to you being front-run especially if everyone was doing it. It’s definitely more statistically robust. It’s definitely more profitable out of sample. It’s definitely what you should do. I’ve already described, that’s what I do, definitely.

As he says in the question, why not just optimize the parameters and get something that will probably be the average? Yeah, if you do a really hardcore statistical optimization and you do it properly, you end up with something that’s pretty similar to what I described already, a simple mixture of fast, medium and slow with some account for trading speeds.

I’ll just quickly address the second question. Actually, Niels, probably in terms of historical background, you probably have a better idea about this than I do. I can only speak to the experience at AHL where, obviously, I was working between about 20 years ago and about 10 years ago. But also, I knew something of the history of how their systems developed over time and I do know that their first systems were just one speed, if you like, but they added other speeds quite quickly. That would have been done probably in ’89, 1991, that kind of time period. So they weren’t trading that very simple one entry, one stop system for very long if you like. They added the additional speeds quite early on. I can’t speak to the wider CTA industry in terms of developing.

I think it’s probably quite unusual that anyone is trading a single speed. Even the simple strategy, one entry, one stop, you’re probably doing more than one of those. You’ve got fast entries, medium entries, slower entries. I say this a lot, but at the end of the day, you end up with things that behave relatively similarly and that’s why returns in the CTA industry are correlated, because everyone’s doing everything a little bit differently. At the end of the day, if you’re trying to find trends and you find a trend, you stay in a trend till it ends, your return profile is going to look pretty similar to someone who’s doing something similar to that but maybe with a fancier approach.

[00:36:56] Niels

Yeah. A very comprehensive answer. I don’t know that I have that much to add to it. What I will say, just a couple of observations to Russ. One is, of course, that when you hear people say, there’s absolutely no optimization to what I do, it doesn’t necessarily mean that there’s absolutely no optimization to what they do. We all base our choices somehow on research, otherwise, there’s no point in doing any research, and so on and so forth. So I would just say that as a little caution there. I think most people will choose parameters based on some kind of observation, whether it’s hardcore, data-driven, or whether it’s a visual thing. Who knows?

I think fewer people would use just a visual interpretation of the markets. They would probably have some kind of data to back it up. But I do think that there may be two ways of deciding then what the actual model should look like. You can either go by a common sense approach where you would sit down. Maybe it doesn’t happen that much anymore, but it probably did a few years ago, where people would by committee maybe decide whether we should have 25% short-term models and 50% medium-term and 25% long-term, and they would move those weights a little bit every year depending on what they found in the research. That’s one way where they use some discretionary element.

Another way would be to just simply do it by the data, let the data decide how the parameters would be picked. That can certainly be done. So that’s another way of doing it. But my overall sense and answer back to Russ is that the whole thing about gaming the CTA industry, this is something that I’ve voiced certainly over the years and I find it incredible that some of these big investment banks are so obsessed with trying to tell the whole world when CTAs are likely to trade and often when you read the headlines, it’s actually not at all what’s happening in the portfolios that we come across.

So, I would be careful, generally speaking, and also, I would say, there’s a lot more going on than just the signal that actually decides whether you have to trade or not on a specific day (the whole risk management side of things), and also just the way managers implement their systems. I think fewer and fewer managers use stops once they become of a certain size, and therefore, it’s not like you pick their stops. It’s not how we manage these portfolios anymore. It’s not to say that you can’t, it’s just to say I don’t think that’s where most of the money today is invested. I think it’s a fool’s errand to try and do that.

But if you’re some kind of market maker or short-term speculator not trying to do a trend following system, but just trying to say let me see if I can buy today and as Rob said, maybe tomorrow there’ll be some more bias because there are some signals that are generated, and then I can get out later in the afternoon. Okay, sure. If that’s what you want to spend your time doing, go ahead. There’s always going to be some times where it works and other times where it doesn’t work.

Your last question about the shorter-term nature back in the ’70s and ’80s, I think there’s some truth to that. I can’t remember what we were doing back in that early phase of DUNN’s career. We know from Richard Dennis’ experiment, that certainly the timeframes they were using initially were shorter term than what they are typically today. So sure, the industry has evolved and timeframes have changed. As I’ve mentioned before, we do run a study every year to see which lookback periods based on a classic trend following system would have done best for every single calendar year. There are years where shorter-term timeframes would have done best, for example, in 2008, even though it was a good year for everyone.

Actually, 50 days in our experiment here was the absolute best, but you would have made money even at 200 days lookback. So I think also lookback periods very much depend on the type of trend following approach you use. I think breakout will have certain properties in terms of lookback, and then other types of trend following might have completely different properties in terms of parameter picks.

Do you want to add anything, Rob?

[00:41:22] Rob

Yeah, two things actually. Firstly, briefly on stops. The reason why stops don’t make sense beyond a certain size, if you think about what that implies, it means you’re going to close your entire position on a single day. If you’re running like $5 billion, that’s probably going to be a bad idea and going to really hurt you in terms of slippage. So yeah, it’d be insane for a multi-billion dollar fund to be running at a single speed and a single stock. That would make no sense at all.

I just feel I should defend optimization slightly. Let’s take an example of two managers. One has done a proper backtest and optimization with only backward-looking data. The other one has said, “Based on my market experience, I know that a 200-day moving average is best and that’s what I’ve used. When I backtested it, I found that it’s better than anything else. So I haven’t outfitted at all.” Actually, that manager has effectively done in-sample fitting implicitly, because they know from their market experience what the best moving average is likely to be.

All the backtests done have confirmed that they were correct in that. That just means that their mark experience is accurate, their memory is accurate. For them to say that they haven’t done any fitting is actually incorrect, because they’ve pulled the best parameter out of the air. That’s no different from running and in-sample optimization and finding that 200-day moving average that way. Both of those approaches are inferior to someone who’s done a proper optimization that uses any backward-looking data. That won’t result in an unrealistically high Sharpe ratio. So properly done, optimization is probably better than not doing optimization at all.

[00:43:05] Niels

Yeah. I think that’s very valid and I think actually what distinguishes firms from each other is, to a large extent, the people who probably know how to do a good optimization without all the mistakes you can make, and people who don’t, I think that’s what shows up over time in the data. So I think that’s very valid, for sure.

Now, we’re going to shift gears to something completely different for a little while and then we go back to something that’s very common for people listening to this, and that is the US debt ceiling because this week, Treasury Secretary, Janet Yellen notified Congress that the US will reach its statutory debt limit next Thursday, and asked House Speaker Kevin McCarthy to either suspend or increase the debt limit. Janet Yellen wrote that the Treasury Department will begin taking certain extraordinary measures to prevent the United States from defaulting on its obligations. She also told McCarthy, it is unlikely that cash and extraordinary measures will be exhausted before early June. Now, you brought up this topic, so I don’t actually know where you want to go with it. But these are just some of the things that I had noticed during the week.

[00:44:26] Rob

Yeah, I’ve mentioned the US debt ceiling before. I think I first became aware of it in 2011 when I was running the fixed-income portfolio. We had this issue with the debt ceiling and also the US being downgraded by the rating agencies. The reason I want to mention it is I just think it’s completely bonkers that the world’s most economically advanced country is in a situation where they can’t actually properly manage the government finances.

On the one hand, those of us who don’t live in America could be like, well, this is quite amusing, but actually, it has very serious ramifications because US government debt effectively underpins the entire global financial system. So this is potentially quite scary. It does feel like certain politicians in the US have had this ability to repeatedly point a gun effectively at the head of the global economy and say, “Give us what we want or we pull the trigger.” So that’s my general despair at the situation.

The political situation in the US with taking all of this time to elect a speaker in the House of Representatives, it’s almost like there are three parties now in American politics. There’s sensible republicans, democrats and these people who still believe that the previous president should be president, who in fact, technically is president because he didn’t lose the election, right? It’s a very, very bizarre and weird situation. It’s a fascinating juxtaposition between the fact that America is a very advanced economic country, home of all the best tech companies on the one hand, and on the other hand, they seem to have the political level of a much poorer and less developed country.

The other thing that amuses me about the US debt ceiling is every time it comes up, various people come out of the woodwork with all kinds of creative ways to solve it. So, in the past, we’ve had mint-the-coin, the idea being that they could mint this huge coin that would have a notional value of $10 trillion dollars or whatever you needed. This time around, we’ve had a new one, which is the idea of issuing debt at an extremely non-par level. The reason being that it’s only the notional value of the debt that counts towards the ceiling and not the actual market value.

So the idea is that you can basically issue a bond that had a notional value of $100 but a market value of $200. You could sell it for $200, and $100 of that would count towards the debt ceiling. You could use that to recycle existing debt and then you’ve basically effectively borrowed another $100. So these things are quite interesting. Creative finances is something that I find interesting. But yeah, it’s just a bit depressing that we’re even having to have this conversation, to be honest.

[00:47:21] Niels

It’s somewhat topical and you wouldn’t know why I’m saying that. That is because, in a few weeks, I’m going to publish an episode that I recorded yesterday where Kevin Coldiron, who’s the host of the Ideas Lab Series, actually interviewed Perry Mehrling who has written these books about empires and money and basically a description of Charles Kindleberger who was one of the foremost economists in the United States. It’s all about really, the Fed and the role of a key currency country, etc., etc. So yeah, there were definitely some interesting points and a very interesting history lesson being shared yesterday. But as I said, it’ll come out in a couple of weeks.

Now, let’s get back to something that we are much more familiar with, which is trend following. This is inspired by another interesting paper by our friends over a Man. It’s a conversation which ties into what we talked about earlier today, namely, different speeds of trend following. They call it the Need for Speed trend following. They start out by quoting, I think it’s Maverick in Top Gun in 1986, where he says, “I feel the need, the need for speed.” So they’ve done a good job in tying it into something that we all can relate to. But they also explain what we mean by speed and they actually do a pretty good job in highlighting different things within it. So without me taking any of these spoilers away from you, Rob, tell us a little bit about why this is an interesting topic and paper.

[00:49:05] Rob

Actually, it reminds me that my last blog post was called Scream If You Want To Go Faster, which is also a quote, but this time, from a pop song. So yeah, I won’t publicize my blog post, because it’s not relevant to this topic although it is to do with trading speed, but it’s about something a bit different. Obviously, you can go and look at it if you want to.

Yeah, so this was quite an interesting paper. It relates to the discussion we’ve had actually throughout the program so far. Which crossover should I use? Should we use a mixture? One interesting thing is if you actually look at the performance of fast trend following empirically, it’s not as good. It’s not as good. It hasn’t been historically as good as medium and slower trend following. That’s particularly true in equity indices, and it’s also particularly true after about 1990. So before 1990, it did pretty well, and then subsequently, performance pretty much flatlined or even went down slightly depending on exactly how you measure it. So based on that, you might think, well, what’s the point of trading quickly?

Interestingly, if I think back to 2009, Winton famously very publicly slowed down their system on the basis that they wanted to reduce costs, and they didn’t really see any evidence that faster trend falling was actually of any value. But this paper takes a more nuanced approach. Actually, the interesting thing about faster trend following is it has two very good properties. One is it’s a much purer kind of protection against the market falling. Owning a trend following strategy is a little bit like owning an adoption straddle in that it will do better if the market moves away from where it is now, away from the strike, and the faster you trade, the more pure that relationship becomes.

Faster trend following has much better positive skew than slow trend following. It has a lower correlation with the overall market as well, and that kind of makes sense if you think about it. So you can imagine, if you were trading really slowly, you would pretty much have been long stocks for the last 40 years, if you’re trading slow enough, because there would never have been a drawdown that was long enough or big enough for you to actually stop your position out. Even if you speed up a little bit, there would probably only be about three or four occasions in that 40-year period when you would be short or flat, and the rest of time, you’d be long. So your beta to the stock market would be relatively high and it would look like you were just owning stocks, basically. The same could arguably be made of bonds or any other market as well.

So faster trend following, what this means in principle is that faster trend following is actually a much better diversifier to a long-only portfolio; so a classic 60/40 portfolio than the medium and slow trend following is. That actually overcomes, effectively, the disadvantage it has from the fact that its net performance isn’t as good. So what these guys do in the paper, which is reasonably thorough and quite nice, I have to say, they look at the characteristics of trend following different speeds, which, for me at least, is not new information, but it’s still nicely laid out.

What they then do is look at the different behavior of these different speeds during sell-offs in equity markets. In this paper, I guess because it’s aimed at investors who’ve got 60/40 portfolios, they’ll always bring things back to the equity market or the 60/40 portfolio and say how do things compare to that. They’re saying it’s definitely the case that in the very worst, in sell-offs, you’re better off having more of these faster signals in them. Now, of course, the downside of trading faster is it costs. It costs you more money. So quite rightly, they highlight the point that execution is very important because costs are going to form a much bigger proportion of your net returns for faster signals than for slower signals. So, you can’t just willy-nilly trade these things, you have to have good execution.

Then the last section of the paper, which is probably the most novel and the most interesting for me, is what they basically do. If you’re following along at home, this is Figure 4, which is quite near the bottom. They look at the drawdown profile of a 60/40 portfolio which obviously has been pretty terrible recently, but also didn’t look great in, for example, 2008-09 because the sell off in equities more than overcame the positive returns that were made in bonds. There was a similar massive drawdown in the early 2000s for similar reasons.

So what they do is they take that drawdown profile and then they say, what happens if we add that to a blend of momentum that includes more or less fast and slow stuff? They find that drawdowns get much smaller if they use an equal blend. So going back to what I was saying about fitting, I have an equal blend of different speeds of momentum. That obviously has fast momentum in it and that does much better than another blend that completely excludes fast momentum and therefore is much slower.

So yeah, the bottom line is that, on a standalone basis, fast momentum (take it or leave it), it’s not probably going to add a lot of value to a portfolio. It may even detract from it. But done properly, it can actually improve the diversification properties of trend following versus 60/40. Given that most people are investing in trend following alongside a more standard long-only portfolio, even if it’s not exactly 60/40, that is obviously a good thing to do.

[00:55:14] Niels

Yeah. Obviously, a topic we’ve discussed from time to time. I think that, like you mentioned earlier today about volatility, because of course, the argument has been made that shorter-term strategies will do better during “an equity crisis”. We’ve certainly seen examples of that, where the first few weeks, clearly short-term strategy should, on balance, do better, not so much if the crises continue. But then we had this year, I say this year, but then we had 2022, and actually, it didn’t work.

This is also why I think what you’re saying is that it’s always probably a good idea to have a little bit of everything, right? You shouldn’t put all your eggs in one basket and say, “Definitely, this is what I need and I rely 100% on shorter term timeframes going to save my portfolio.” They don’t always do that. There’s a lot of other things, but it was another good paper from them.

Do you want to talk about optimal trend following with transaction costs, given the fact we’re already at an hour point or do you want to talk about math?

[00:56:24] Rob

Let’s do that really quickly because it relates to the previous discussion we’ve been having. I said there were different ways of deciding what parameters you should have in your system earlier when we were talking about fitting. I said there’s pick the best number based on your market experience, there’s proper fitting if you like, there’s in-sample fitting. There’s actually another way, which we haven’t discussed, which is that it’s actually possible to theoretically derive the best parameters of what your trading strategy should look like. The way you do that, in principle, is you make some assumptions about market behavior. You will in a lot of very, very fancy maths.

You then say, given these assumptions, what’s the optimal strategy to maximize my performance here? The nice thing about this approach is you don’t need any actual data to do it. So there’s no real fitting going on. A slight caveat here is that if the assumptions you’ve made are themselves based on observed market behavior you’ve actually empirically calibrated your model, then you could argue this as a fitting by the backdoor. But if you can minimize the amount of that that’s going on, then that’s obviously good. You can theoretically end up in a situation where you basically use no market data at all to say what’s going on. Now, it’s a nice approach to use. I’ve used it myself.

I thought I’d mentioned this particular paper, which is called Optimal Trend Following With Transaction Costs. It’s available on SSRN. We’ll put a link on the page. What these guys do is basically they take this approach. They say given some assumptions about market behavior, without any data (no fitting involved), what’s the best trading strategy to capture trends given that we have transaction costs? Because without transaction costs, obviously, for example, faster trading will be better than slower trading.

The nice thing is that they recover something that looks very much like the moving average crossover that I, in particular, am extremely fond of. So, it’s just a nice, neat little model, not really of any practical value, to be honest, but it’s nice to have your models confirmed by another source of evidence, if you like. So as I said, different sources of evidence. Some of those, you can accuse people of overfitting, but this particular approach uses no data at all. So there’s no fitting involved. It’s a really clean ticking the box of yes, this is a good way to capture trend.

[00:59:03] Niels

Fantastic. Good stuff. Well done. All right. The last point is something which is a little bit different, because we’re into the British politics of some sort, but then we’re combining that with math. So tell us a little bit about what we’re going to talk about.

[00:59:21] Rob

Well, our prime minister this week is Rishi Sunak. It’s quite hard to keep up with who is prime minister in the UK at the moment. It’s a bit like living in Italy, but both the weather and the food is worse. He gave a speech recently and one of the things he said was that everyone at school in the UK should study maths until they’re 18. Most people at the moment have the option at the age of 16 of switching. Most people in the UK, they basically do A-levels, which means you’re doing three or four subjects, so obviously narrowing things down much further. Not many students subsequently do A-level maths, which would be studying maths up to 18.

There are obviously various arguments to be made as to why this is a good thing. But it was brought into focus for me yesterday, actually, because yesterday, I gave a talk to a local secondary school about careers in the financial markets and trading generally. One question I always get asked is, how much maths do you need for this? How much maths do you actually need to do what you do or to work in the financial markets? It’s actually quite a difficult question to answer because there’s a huge range of jobs in this industry. I’ve worked with people who stopped doing maths at 16 and went on to do something entirely different.

To take an extreme example, I’ve worked with people who did Latin or Greek at university, so very much at the other end of the scientific spectrum, but they’re working in jobs where the main skill they need is the ability to deal with clients and get on with people. That’s something they obviously do very well. On the other extreme, you’ve got people who are, I have to say, much cleverer than me, with PhDs in maths or physics, who are working on either particularly advanced trading strategies or working as quants and doing things like derivative pricing and risk management, and so on and so forth.

Personally, I think I sit, roughly, in the middle of that spectrum. I did A-level maths and then I did economics, which is obviously not a real science, but it is a subject that has a fair degree of mathematics in it. So to bring it back to trend following, to just follow a basic trading strategy that somebody else has already developed without needing to worry about backtesting, optimization, all of that kind of stuff, yeah, high school math is, I would say, more than enough.

Actually, one of the things I’ve tried to do in my writing is to do things that don’t require advanced maths as much as possible. I don’t think I’ve ever used, for example, the integral sign and anything I’ve ever written. I think, to a degree, that’s not necessary. It’s definitely helpful if you are backtesting to have at least an intuitive understanding of statistical significance and potentially, ideally, to have more advanced statistical knowledge if you’re really going to get into the weeds of this stuff.

There are use cases where you would need very advanced maths, but generally, the amount of maths you need to do this job is probably quite overrated. That’s my opinion anyway. I’d be curious to know what other people think.

[01:02:40] Niels

Yeah, absolutely. It’s a good point. I actually think you’re absolutely right. In the industry as a whole, there’s definitely room for lots of different talents because we can’t all be super quants and just look at the screen all day. So it’s a roomy industry, and it’s a fun one. It’s never dull and always changing. Actually, because it ties into whether you’re systematic or not, it all ties into what’s going on in the world because that affects how markets are moving. I think there’s something for most people to enjoy.

[01:03:19] Rob

I would say, actually, that it’s useful to have people from a different mixture of backgrounds. If I was employing, if I was signing a hedge fund, I’d probably, for example, employ a historian because I’d want someone who has an understanding, and I’d probably employ a politician, sorry, not a politician, but someone with a background in politics and someone with a background in geography because all of these things feed into how markets move definitely. I’m trying to remember. Is your background physics, Niels? I can’t remember.

[01:03:49] Niels

I wouldn’t say it’s physics, because I didn’t study anything after high school. So I have absolutely no credentials to this. It proves that you don’t need a lot of math to do this. Yeah, absolutely. I am learning by doing. There we are. All right, good stuff. I think we’re going to wrap up for today. This was fun as usual. Thank you so much, Rob, for all of that.

Now, if you haven’t left a rating and review until now, this is your chance. If you would take five minutes to do that, it really does help. If you don’t know how to do it, you can go to toptradersunplugged.com/review and then there are some instructions that I hope you will take us up on. As usual, you can send questions to us. That you can do by email info@toptradersunplugged.com. Next week, it’ll be with Rich. I’m sure he would love to tackle some of your questions. So by all means, send them through to us.

With that said, that’s all from Rob and me today. Thanks so much for Listening. We look forward to being back with you next week. Don’t forget, by the way. Check out the new, in my own unbiased words, completely amazing series of conversations with the absolute best minds in our industry starting on Monday that I’m doing with Alan. I think the first one up in the series is someone that you know very well as it is the CIO at Man AHL.

From Rob and me, as I said, thanks for listening. Take care of yourself and take care of each other.

[01:05:33] 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.