222 Systematic Investor Series ft. Rob Carver – December 11th, 2022

[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 weeks edition of the Systematic Investors 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 catalogue and listen to the past episodes that you may have missed, like my conversation with Alan last week where we discussed if there are strategies that can be used to diversify your trend following allocation, such as shorter-term strategies, or if the cost of smoothing out returns this way is just simply too high in the long run.

But also, if you missed our Wednesday episode, I would really encourage you to listen to this midweek conversation we had with Shawn Hackett, which I think was one of the most interesting conversations of the year, not least, because it’s kind of outside my own area of expertise, and perhaps, it could be one of the most important ones for people who believe in having commodities as part of their diversified portfolio.

We know that weather can have a massive impact on commodity markets. But in a period where inflation is already elevated, the coming weather volatility that Shawn sees coming may well end up also having an effect on bond yields, on stock prices, and currencies in the coming year. So as mentioned, I would encourage you to go back and listen to these episodes when you’re finished listening to Rob and I today.

Rob, great to be back with you this week. So much continues to happen between our conversations. How are you doing?

[00:01:58] Rob

Yeah, it’s quite good the listeners can’t see the video feed because you’re sitting there in your T-shirt looking nice and toasty, and wearing a baseball cap, and I’m here wearing about 17 layers and a very warm fleecy hat because it’s minus 4 here. My wooden recording studio, unlike yours, is not very well insulated. It’s sunny, which is nice.

Obviously, people from Scandinavia and Switzerland will say that minus 4 is summer. For us feeble Brits, this sudden plunge in temperatures has been a bit of a shock, I think it’s fair to say. That along with all the strikes we’re having. We’re trying to keep upbeat going into Christmas, but the weather and the strikes aren’t making it easy.

[00:02:45] Niels

I think you’re doing a good job. I also think people appreciate now, the length that you go to in order to have these conversations. So I’m sure we’ll find some topics to warm up your inner self, so to speak, that we can really get into. Speaking of this thing about the coldness, yes, I am Scandinavian, and yes, I do live in Switzerland.

I will say there is a difference between minus 4 in Denmark and minus 4 in Switzerland. It’s not the same. It’s significantly colder in Denmark because the weather is just more wet, and usually also a little bit more windy. So yeah, it’s not as easy to compare these things anyways.

[00:03:25] Rob

It’s the same here. Minus 4 in the UK is definitely a lot colder than minus 4 if I’m skiing in Austria, or something in the mountains, which is much crisper. Anyway, this is not a weather podcast so we should move on.

[00:03:37] Niels

It isn’t. No, no. Absolutely. Anyways, December is a tricky time for the capital markets as well. We have banks, brokers, and investors all trying to close out the year with their respective portfolios 100% invested. Carrying over cash over the turn, so to speak, that’s the year-end that’s also referred to as “the term,” is not acceptable in many capital markets, and as a result, markets can become volatile to the point of seemingly irrational.

This year, we’re seeing this play out in the treasury bill market. For example, if you look at the soon-to-mature December 15th treasury bill, we see that it yielded 3.15% at the close of November, but it ended the day yesterday yielding only 2.39%. Now logically, that makes no sense. The overnight Fed funds rate window is 3.75 to 4%, and the Fed repo program offers 3.8% for institutional investors who qualify for that. Yet the near-term bill curve continues to be in disarray as we approach the year end.

How can this be the case? Well, it turns out that between December 6th and December 13th, the Treasury is paying down $76 billion in bills, and that’s to say that they sold $76 billion fewer bills than the amount maturing. In effect, the Treasury tipped the supply-demand of treasury bills out of balance, which has resulted in this wild gyration in the bill market.

Now also talking about gyrations, Friday morning, producer prices came out and will likely be a disappointment to the Fed as the month-over-month inflation measures came in above expectations, despite the year-over-year measure tricking lower. Traders reacted to this by selling into the close of the week with a 30-year bond trading back above 3.55% as we closed off Friday.

The volatility is sure to continue into next week as we have the consumer price index released on Tuesday, FOMC announcement, its rate decision on Wednesday, and the November retail sales report is released on Thursday. Now, the Fed is widely assumed to raise the overnight rate by 50 basis points. The CPI and retail sales are sure to influence their decision-making in the new year, and will therefore be closely watched.

By the way, before I close out, in other news, I thought it was interesting that investors, they cashed out $26.66 billion of domestic US equity mutual funds and exchange-traded funds over the last seven days through Wednesday according to Lipper Funds, and that is the biggest one-week outflow since April of 2021.

Another piece of news I thought was quite interesting, and that is that Blackstone will limit investor redemptions from its $69 billion Blackstone Real Estate Income Trust in December after being hit with heavy withdrawals in October and November. That doesn’t sound so great, in my opinion.

Anyway, let’s talk about what has been standing out to you, Rob, since we last had you on the Systematic Investor Series. I know we really just recorded something a few days ago, but that is a secret. So far, it’s in the vault, but it’s not been released.

[00:06:49] Rob

I think it’s been about a month since I’ve been on the show probably. So just looking at one month performance for me personally has been sort of down a little bit. I think I’m down something like 4 or 5%, something like that. So not a great month to be honest. Losses coming in fixed income, US 20 years, I can see that. Also in crude oil, there’s been a bit of a reversal of the bullish kind of energy trade that we saw, particularly at the start of the year, but also over the year generally.

The other thing I can do is look at the movements in all the markets I track, not just the ones I trade, and adjust that by risk to get a kind of idea of how significant the moves are. If I look at that, then for example, the gasoline contract total last month is down about 18%. ETO is down about 17%. Those sound like big moves, but they’re moves of about 1.5 standard deviations. The markets moved against, but not massively. ETO is down about 17%.

Markets that have done well over the last month, most of which sadly I don’t have a position in, but anyway, let’s list them out. So a Euro-CAD is up a couple of standard deviations. That’s pretty, pretty high. There’s been some rebound in some equity markets. Obviously in China, there’s been very interesting political situations with the riots going on there, so that’s a very volatile market, but at least over the last month, the eight shares are actually up. So that’s kind of interesting.

Taking a longer view, it’s pretty much the end of the year now, right? So I think it’s kind of okay to sort of think about year-end P&L. There’s only a few weeks left. So far for the year, I’m up just a tiny touch under 20%. So the pattern of that performance is basically over the first few months of the year, I obviously made a lot of money from trend following as pretty much everybody did. Then there was a few fallout months, then made some more money in the summer. August and September were very good, hit a new high watermark in September.

Since then, things have kind of tailed off. So I’m down roughly 12%. In roughly a 12% drawdown at the moment. From a peak return of about just over 30% in September, I’m now looking at about 20%, which is better than pretty much every long-only asset class. So on a relative basis, obviously great. It’s not as good as 2019 was for me, but it’s better than- it’s about the same as 2015 so that just gives you an idea of where that return sits in the pack. I’m doing the same exercise, so looking at the whole year now, and looking at risk-adjusted returns across the whole series of markets.

If I was to look, for example, just at returns, not at risk-adjusted returns, it would be perhaps no surprise that the worst markets were Ethereum and Bitcoin, which were down over 60%. Because those markets are so volatile, again, that doesn’t really show up in the risk-adjusted numbers.

Looking at the risk-adjusted numbers, US 2 Year was down over four vols, four standard deviations, your Euribor as well shot, so all fixed income markets and all fixed income markets that have a relatively short duration as well. So the front end of the curve, obviously when rates policy is in a tightening environment, the front of the curve tends to get hurt more and the yield curve tends to flatten left. There you go, that’s a fixed income theory for you, or perhaps more accurately, fixed income empirical results.

Then on the long side, interestingly, the kind of energy markets that were dominating things earlier in the year are no longer appearing in those top markets of the year, at least in vol adjusted terms. So interestingly, the best market, the ones I track is Swiss-Yen, which I don’t trade and I guess would only be useful if you, Niels, want to go to Tokyo anytime soon. You might be interested to know, that’s moved up 3.5 standard deviations, and bizarrely for me at least, but at least over the year-to-date period, the Russian ruble has done very well as well, which obviously, is extremely counterintuitive.

You can argue about whether the- I mean, this is the future I’m talking about. So it’s a market that’s kind of trading but not necessarily attached to the underlying in any meaningful way because the underlying spot ruble is effectively embargoed and very hard to trade, but for what it’s worth, the future has done very well. So it’s been an interesting year, but not the very clear story of the first half of the year where it was all about the energy markets.

If there was any clear story this year for me, it’s the fact that fixed income has done particularly badly just looking at this list of markets. But otherwise, we had some very clear trends at the start of the year, and since then things have definitely been a bit more muddled.

[00:11:55] Niels

When you say fixed income has done “particularly badly,” do you mean that the markets have done badly, or that your performance has?

[00:12:00] Rob

I mean the markets have done badly. If I now turn to my own performance in the year to date, let’s quickly do that just for clarity’s sake. Looking at the P&L from worst to best interestingly, gas, natural gas, has been one of my worst markets, which is kind of interesting. I guess it’s been very volatile and not in a good way, very choppy. Then there’s a bunch of equity markets that didn’t do very well, and some FX markets that didn’t do very well. So quite a mixed bag on the short side.

Let’s have a look at the biggest profits. My biggest profit actually is still in energy markets. Gas is kind of an anomaly. So I’ve made a lot of money on things like heating oil, gasoline, Brent, and that would all have been the first few months of the year, and then kind of holding steady since then. I’ve done well in currencies, Brazilian real, Euro-Yen. They’re all up there. Corn is up there and US 20 is up there. So actually, one of my best markets was fixed income markets. Being short that fixed income market was obviously a good trade. I can see Korean 10 Year there as well.

So actually, interestingly, thematically in terms of where I’ve done well and where I haven’t done so well, long energy. The long energy trade is the only real clear standout signal there. Even then, as I said, there’s an energy market, and we have gas specifically where I didn’t do so well. That’s kind of interesting. So where are we in terms of the current situation, in terms of my positioning going into the end of the year? Risk is very low, which is kind of what I’d expect in a drawdown when trends are not so clear as they have been perhaps.

My biggest long positions, interestingly, are in soy, both soy meal and soy oil. I’m actually now currently short on US gas. We’ll see how it plays out. I’m also short VIX, which is effectively a risk on trade. So that’s kind of interesting. None of those positions are particularly big. My kind of risk is running at maybe a third or even a quarter of my long-run risk target. It’s very much a question of keeping, metaphorically, my cards close to my chest, and not putting too many chips on the table at the moment, and just waiting to see what plays out in the new year.

[00:14:29] Niels

Were you happy with the way your dynamic market selection, I can’t remember what you call it-

[00:14:36] Rob

Dynamic optimization, but dynamic market selection is actually probably a better name for it because it actually describes exactly what it does.

[00:14:44] Niels

Were you happy with the way that all panned out? I can’t remember when you implemented it, but obviously, essentially for people who may not know what we’re talking about, you’re trying to have a bigger universe of markets that you effectively can trade and then optimize for selection of those markets that would essentially replicate the optimal portfolio best.

[00:15:04] Rob

This is actually quite an interesting discussion because it’s quite hard to objectively measure the benefit that this strategy has. You can say, “Well, what am I going to compare it against? Should I compare it against the 100-plus markets I could trade if I had sufficient capital?” Well, on that basis, it would have underperformed in expectation. It might have outperformed over the last year, which is roughly how long I’ve been running it. Over the entire backtest, it will underperform, because it just obviously can’t do as good of a job. The underperformance isn’t big, it’s about 20% of Sharpe, which still leaves me with a huge improvement over what I had before. So I’m happy with that. You should compare it against the alternative, which is a static selection of a number of markets.

Now, the issue with doing that is that in that static selection process, you might by luck pick markets that are doing particularly well on average. Let’s take an example. My best-performing market over the last 12 months is Brazilian real/US dollar. Now, let’s suppose at the start of the year, I’d only been able to choose one market to trade in my portfolio as my alternative to dynamic optimization. This is obviously an extreme example, but let’s just suppose that. After a year, if I’d looked at my performance over the last year, not quite one year, but year-to-date is about 20%.

If I’d just been trading Brazilian real, I might have made 40%. From that, should I conclude that my dynamic optimization is a complete waste of time? Of course not. I’ve just been really lucky with my choice of market. Now I am going in April, because April is when I normally do my very thorough P&L risk calculation. I normally do a blog post about it as well, because it’s the end of the UK tax year for anyone wondering why April, which seems very arbitrary. It’s April 6th as well, the start of the tax year, it’s very arbitrary. It’s not even April 1st. There’s no logic to it at all.

[00:17:11] Niels

April 1st would be a bad day to choose since it’s April Fools.

[00:17:14] Rob

It would really be a bad day. One thing I’m going to do then is to actually compare my system as it was before I started trading dynamic optimization, which would be a fixed set of about 30 markets, and see how well that’s done with the caveat that those 30 markets may by luck have done really, really well. So that’s the issue. It’s very hard to say, except by looking at long periods of history, and by adjusting for this effect of luck, which is obviously an exercise I’ve done, to say whether it’s behaved as I expect.

In terms of general behavior- because I think one thing that I like to say to people is doing a backtest isn’t just so you can identify which is the best trading strategy to trade. This is a common misconception. Actually, a backtest is often the worst way of doing that, because it can lead you down the path of overfitting, which those of us who do this for a living know all about from personal experience. Another good thing to do a backtest for is to give you an idea of the likely behavior of a trading system.

By behavior, I mean things like holding period, number of times you trade per day, trading costs you incur, typical leverage values, typical numbers of positions you take, blah, blah, blah, blah, blah. You can measure all kinds of metrics. Then when you actually start trading, you can compare what you’re actually achieving against those. If you’ve got the tools to do that, you can do this statistically.

So you can say, is this outside of what I’d expect for the distribution of this over a year? You can do that with P&L as well. So you can say, “Okay, I’ve lost 20% this year, is that unexpected, or should I shut my system down? Have I done something massively wrong? Has the market condition changed dramatically, or is that just roughly what I’d expect looking at the backtest?”

You can do it with P&L as well, but you go beyond that. If I look at the behavior of the system in terms of the numbers of trades it’s done, holding period, costs, all of these things, then I find that’s well within my expectations. So has it behaved as I expected? Yes, but if you’re asking that in the very narrow sense of is it better than what I did before? The answer is, well, frankly, you can’t really tell that just by doing a straight year-by-year comparison of P&L.

[00:19:37] Niels

I completely agree. It’s interesting because when you explain how your year panned out, and you’re doing it in one specific way, when I look across the industry, when I look at how the year panned out for us, or is panning out for us, and so on and so forth, we all do things slightly different.

Broadly speaking, trend following this year has had kind of the same pattern. Most managers probably made their last new high around September, end of September, and have had a little bit of give back, maybe October, but thereabouts. So it’s been an interesting year, but just maybe to finish off on the trend following update as we normally do.

I think this week was pretty uneventful from a P&L point of view. As Rob already mentioned, energy bills were put under some pressure yet again, long-term trend followers may still be long in the energy sector, but of course, some other systems will be shorter this time. Who knows? I think currencies, I would say are probably mixed, although there were some big moves in things like Mexican peso. I don’t follow the real, but Mexican peso certainly had a big move. Then of course, with the higher PPI coming in, we did see a little bit of help to those who were still short in the fixed-income space.

I think soft and grains with the moves they had probably did okay, maybe with the exception of soybean oil. I think that went a little bit the other way this week. Yeah, metals, equities, again, depending on the timeframe how it did. What’s interesting to me also is that the trend barometer that I publish every day, that’s been moving up now. I think the low was like 23, which was really confirming the poor, poor month of November, but it’s been moving up and it finished at 48.

So now we’re back into neutral territory. Are we going to see an improvement into the last couple of weeks of the year? Could be. That would, of course, be good news. Year to date and month to date for the indices we track, BTOP50 down about 1% for the month, up 13.1% for the year; Soc Gen CTA Index is down 1.3% for the month, up 18.6% for the year; SocGen Trend down 1.75%, up 25.5% for the year; and the Short-Term Traders Index flat for the month, up 11.5% for the year. That is as of Thursday, and Friday was definitely a good day. So the numbers are probably a little bit better.

Let’s move on to some questions that came in. Always when Rob is on, there’s so many questions, which is great. We appreciate it. The first one that came in is from Elie. Elie writes, “Hi Niels, hope all is well. As always, I’m enjoying the weekly podcast and looking forward to Rob’s show this coming week. A few questions for Rob. I’m in the process of testing Rob’s continuous trend systems, and it seems that daily vol targeting is an inherent feature of a continuous system, since the system doesn’t look at the market in trades,” meaning it doesn’t look on a trade-by-trade basis per se, “but instead calculates the appropriate position each day. Is there any way to run a continuous system without daily vol targeting? For example, only calculating vol when the signal change sign?” I’m going to stop with that because there are like three or four questions in total.

[00:23:21] Rob

There’s basically two main components to my system that make it different from a kind of classic trend following system, traditional, whatever words people who use these systems like to use. Basically, I’m continuously adjusting my opinion on two things. Firstly, what I call my forecast, which is effectively my expected risk-adjusted return and in a trend following system, that’s purely the strength of the trend. So when I initially go into a trend, I’ll have quite a small position because the trend will have not been going for very long, or the market will be coming out of a choppy period and the trend is only just really making itself evident.

When a trend has been going for a long time then the position will be bigger, because if you actually look at the relationship between trend strength and future return, generally speaking with some caveats, and you’ll need to read my new book to find out what they are, but generally speaking, the stronger the trend is, the higher the return you expect the future and therefore the bigger your position. So I’m constantly changing my opinion on the strength of the trend effectively.

The other thing I’m constantly changing my opinion on is the size my position should be relative to current volatility. If you think about a very simple static system, what you’d basically do is on day one, open a trade. At that point, you’ve got some opinion about the strength of the trend, and effectively, you will probably treat all trends as equal. There are other ways of doing it, but generally speaking, let’s say you’ve got something simple like a moving average crossover.

The moment that short crossover moves over the slower one, you say right now I’m going to be long one unit of risk, and then whatever happens to the trade after that, I’m staying long one unit of risk until you get closed out either by the crossover reversing or stop loss being hit. Also, at that point, you calculate the size of the position you need depending on the current level of volatility.

So if volatility is low, you’d have a bigger position and vice versa. Because you’re holding that position constant throughout the life of the trade, you’re effectively ignoring any changes of volatility that happen. That’s the way that a lot of people do it. The way I do it is to continuously evaluate those two things, strength of trend, level of volatility, and adjust my position size accordingly. I say continuously, but it’s effectively on a daily basis.

Now, what Elie is saying is, what happens if you take out one of those components? The word is you take out the fact that you’re adjusting for the strength of volatility throughout the life of the trade but effectively keeping the other component in there. So you’ll still be trading every day as the strength of trend increases, but you won’t be adjusting the size of your volatility.

Now, if I look at the improvement in return I get from a static system to the system where I’m looking at both these variables every day, very roughly, about a third of the improvement I get comes from the fact that I’m changing my trend forecast every day, and about two-thirds comes from the volatility adjustment. So volatility adjustment is actually more beneficial to me than adjusting the forecast for trend strength.

Doing what Elie suggests, of course you can do it. Of course you can do it. Absolutely, you could do it. You could put a trade on and be adjusting that continuously depending on the strength of trend, but ignoring changes in volatility, but you’ll lose roughly two-thirds of the benefit from doing that. It’s also a bit of a perverse thing to do because one huge advantage of the simple system where you just go along five contracts, hold the position, hit a stop loss, close, that’s a much simpler system to run than mine. You can run it much more easily manually on spreadsheets, and so on and so forth.

Now, doing what Elie suggests, which is this hybrid kind of halfway house system, actually involves pretty much as much work because you still have to re-estimate the strength of the trend. It’s not really a lot of extra work than to estimate the change of volatility as well. So it’s possible, but it will not be as good, and it seems like a bit of a strange thing to do.

[00:27:42] Niels

Yeah. I will say that this is of course also because you’ve chosen moving average crossover. I think a lot of the quote-unquote “early” trend followers- obviously, the earliest trend following that I’m aware of is actually volatility breakout. Price breakout came about 10 years later in the ’80s, of course, with the turtles. But vol breakout is how I think both Keith Campbell, Dunn of course, and some of the other guys from the ’70s. I wonder a little bit.

Certainly, I can see that if you were using breakout systems, I can certainly see. If you took the static route, which is perfectly fine, then certainly you cut down on the maintenance work. There’s very little to do other than do your trades for a new signal, so to speak. But I take your point that it might be different if you’re using moving average crossover.

[00:28:41] Rob

The issue is more the fact that it’s quite hard to construct a breakout system that produces a continuous forecast. I’ve actually done it. I call it breakout. Because breakout is pretty much a binary on-off thing, right? Whereas a movement average-

[00:28:57] Niels

You would need a lot of parameter combinations in that system to do it. This is partly what we do on our side, so I’m pretty familiar with it. So you can do it, but you’re right, it’s a lot of work.

[00:29:07] Rob

Niels, are you revealing the secret sauce?

[00:29:10] Niels

Not really, but there we are.

[00:29:12] Rob

We’ll put that on Twitter anyway. “Niels Finally Spills The Secret Sauce.” Listen to this episode.

[00:29:19] Niels

Good stuff. Well, that’s just the first question that Elie had. Then he goes on to say, “Rob calculates signal scaling based on average across many futures contracts. Can this be done with in-sample data due to the vast amount of data being used?

[00:29:36] Rob

What Elie is talking about here is that basically in my system, because I’m combining different forecasts together, and because also the way my position sizing works, all my trading rules are constructed in such a way that they have a consistent scaling. In my world, zero means flat, plus 10 means an average size long position, plus 20 means a position that’s twice as big as that, and I don’t actually go above plus 20, minus 10 means an average size short position, minus 20 is twice as big as that. Then obviously, in between that, you’d have plus 3.6 or minus 19.2 or something.

Now, of course, the issue is how do you- I use many different trading rules and almost none of them automatically produce a number that’s in that scaling that has that meaning. So what I do is I look at the historic data and I basically measure the natural scaling of something, some trading rule, just think of it as a black box that produces a number. So all I do is I look at the distribution of those numbers and I say, “Okay, what number, what scaling factor should I use to translate that?,” It’s just a linear factor, just a multiplier, “from the numbers that are coming out of the black box into this nice minus 20 to plus 20 scaling?”

That can be done in various different ways. In my backtest, I do it on a rolling basis looking only at historical data, so there’s no forward-looking thing going on there. Actually, because this number is relatively stable, because it’s not really adding any alpha to the system, there’s not really a lot of harm in doing it in sample, as Elie says, to be honest. You’re kind of cheating in a way, because you’re using forward-looking data, because it’s in the sample. In practice, it’s fine because you can use almost- the slower your system is, the more data you need to get a reliable figure. You don’t need much data generally to get a number. You can almost use the same number all the way through your backtest.

Actually, for a lot of trading rules I use, you can actually derive these numbers either using artificial data, or even analytically. In other words, you can literally, if you know how to do things familiar with a kind of option pricing theory, literature side of things, you can do things like say, if I assume that prices follow some process, I can actually calculate with a pen and paper what the scaling factor should be and you find nice results.

For example, with moving averages, if you double the length of the moving average, the scaling factor changes by the square root of 2, which you kind of expect if you’re familiar with this financial mathematical theory. So it’s one of those things I say to people, don’t. I do it the proper way, backward looking, always out of sample, but in practice, it’s one of these things that makes almost no difference. So don’t lose sleep over it.

[00:32:45] Niels

All right, cool. Now we move to the dark side, because the next question is, “Rob has mentioned a mean reversion system with a high Sharpe. Is there anywhere to find out more about this or do we have to wait for Rob’s new book?

[00:33:00] Rob

Elie obviously already knows the answer to this question. I will say that in the book, it does have a high Sharpe, it’s got Sharpe over 2. I do say in the book, I never trust anything with a Sharpe over 2. This is no exception. Just because I built it, doesn’t mean you should trust it. It probably still has a positive Sharpe, but I doubt it’s as high as that.

There are all kinds of caveats with trading mean reversion systems. So Elie, don’t think you’re going to buy my book and there’s going to be this guaranteed Sharpe ratio 2, because you should know by now that that’s not the kind of thing that I do.

[00:33:41] Niels

Absolutely. Well, Elie continues down the dark side, because he then goes on to say, “Can Rob join the podcast more frequently? I think that for individual investors trying to design the system, he’s an especially important co-host. Perhaps you can start a series with him, a Series Two.”

Now, Elie, you clearly don’t know or maybe you’ve forgotten how long of holidays Rob takes every year. So having him on more frequently on the podcast is really going to be tough, but I do appreciate the comment, and of course, Rob is an important contributor to all these.

[00:34:18] Rob

I appreciate the comment as well, but I should point out that this is not like one member of my family, this Elie person- one question about my new book and another question saying I’m the best co-host on the podcast, people may be suspicious.

[00:34:35] Niels

He didn’t use the word best. Let’s be clear.

[00:34:39] Rob

That’s fine. Thank you, Elie. Much appreciated.

[00:34:42] Niels

Moving on to another question from Matthijs, who’s written in before, so nice to see you again here. “Thanks again for a wonderful podcast your team is making. I really enjoy working my way through the entire back catalogue.” Great to hear. “A one-inch issue that frequently comes up in the context of trend following is that while it has a tendency to perform quite well during slow equity and bond sell-offs, it doesn’t offer as much protection. It’s less diversification and quick sell-offs. To protect against that sort of shock to your portfolio, people often seem to look towards options strategies for tail protection and other alternatives. I was wondering if you or Rob could share how you think about the role of such tail protection in your portfolio. If one has a portfolio of say, one-third buy and hold stocks, one-third buy and hold bonds, one-third multiple asset class trend following all scaled to the same vol, how important is it to add tail protection to this mix? Presumably, it becomes more important if you scale things to a higher vol. Is the drag from the tail protection more than paid for by the higher vol you can run the rest of your portfolio at? Finally, is there any easy way to access such tail protection either through a DIY solution or some external manager? Thanks for considering the question.”

Thanks, Matthijs for this. It’s kind of funny because we have that coming up in the year-end conversation, I would say.

[00:36:17] Rob

Yeah, we do. We talk about a couple of papers. One from a tail protect fund which has a certain slant, and one from AQR which is, in my opinion, a better paper and kind of agrees with your point that basically the faster the sell-off, the more likely is a tail protection fund. For those who aren’t familiar, a very simple tail protection strategy would just be continuously buying out of the money put options on equities or some proxy for your portfolio.

So it’s a strategy that basically most of the time has a negative return but has a high positive return in a market downturn. That’s obviously a very pure way of buying insurance on your portfolio because it’s pretty much guaranteed to pay out. Whereas, of course, trend following is something that empirically has done okay in sell-offs. We can point to sharp sell-offs, like COVID 2020 most recently, where it hasn’t done as well and where a tail protection fund would potentially do better.

Now, is it important to add tail protection to a mix? Should you run things at higher vol? Questions like these. This is one of those things where it’s not really- I can’t really give you a theoretical answer. It’s going to depend really highly on the character of the tail protection fund involved. So if you just do something really simple like selling out of the money puts, that strategy does so badly most of the time.

[00:37:48] Niels

Did you mean buying?

[00:37:49] Rob

Sorry, buying out the money puts, of course. You know, it’s been a long time since I’ve traded options, Niels.

[00:37:55] Niels

I was just wanting to make sure we’re not giving people-

[00:38:02] Rob

Buying out of the money puts, definitely that. So if you just do that naive strategy, then you’re probably not going to get any benefit from adding that to your portfolio in any size, because its performance is just so poor that- the old times, when it will obviously gain you extra money. Perhaps the only exception would be if you were currently running your portfolio at very high leverage and where that particular strategy would then mean you actually were able to preserve capital in a downturn that would otherwise wipe you out.

So if you are running your portfolio at such a high leverage, let’s say, I don’t know, a 20% downturn in the S&P would completely wipe you out, then yes, adding tail protection to that, even a poor tail protection strategy, would make you money. You shouldn’t be running your portfolio with that kind of leverage anyway because that’s insane. I’d say for a good tail protection fund- and Niels has obviously interviewed tail protection managers, and I think he could probably answer the second part of the question better, which is, how do you get the tail protection? Yeah, let me just finish and then you can come on to that.

So in the abstract sense, if I can buy a tail protection fund that’s not costing me too much money in premiums, because it’s not just naively buying out of the put options, it’s doing something smarter, and selectively buying or using some kind of indicator to indicate when the vol is relatively cheap or relatively high, or perhaps even dynamically switching between buying actual options and buying synthetic options like through fast trend following strategies.

There are all kinds of ways to juice up your tail protection fund and make it clever. Then yes, there will be a value for adding tail protection into your overall fund. I would just caution though, about then saying, “Oh, this is great. I can now pull the leverage lever, whack out my leverage because I’ll be protected from a market downturn.”

Unlike the very simple tail protection strategy where you’re continuously buying put options and you’re definitely going to be protected from a market downturn, assuming that the counterparty risk isn’t a problem, then any situation where the tail protection is not perfect, which is true of trend following, a good tail protection fund that isn’t costing you as much money in premiums, you’re going to be giving something up.

What you’re giving up is that certainty of a payoff in the event of a tail event, and that means you should not then necessarily say, “Oh, great, I can go and get loads more leverage now.” Absolutely not. I think that would be a very silly thing to do. But yeah, over to you Niels.

[00:40:47] Niels

Well, first of all, I wanted to say that obviously Matthijs is kind of right in saying that usually, trend followers, we do better when there’s a slower sell-off. Clearly, if something changes from a peak in equities, for example, where like in February of 2018, yeah, trend follows, we’ll have a problem because we’re going to get caught usually in every single sector on the wrong side. However, I would say actually that during COVID, I think we did quite well.

Certainly, at our fund, we were up in March pretty strong, and had the crisis continued, we would have been absolutely correctly positioned for that crisis. It just didn’t materialize. I think also with tail protection funds, and I’m not an expert, I’ll proclaim that for sure, is that a lot of the money that was reported as being made during COVID, a lot of it had also to do when they were basically monetizing these options. Did they do it at the right time? Or a week later, they would probably not have made that much money, and so on and so forth. So I think there’s lots of things to consider.

Of course, then you have a year like this year where everybody was hoping that their tail protection funds would be working, and it’s just been the complete reverse. In fact, we’ve seen some of the most well-known long vol managers basically close their shop. So what I will say instead, Matthijs, is two things. You should go back to the 12th of January of this year and listen to the conversation that Harry had with Dave Dredge. That’s a great conversation. Then you’re going to look forward to a conversation I recorded yesterday, where we had Dave back but this time with Cem.

So again, two vol experts talking Sharpe here and me trying to keep up. That was a fascinating conversation, not necessarily specifically on your question, but on vol and on what’s been going on in that space this year and so on and so forth. But great question and we appreciate that.

All right. A comment from Richard, and a couple of questions, more questions. Another family member, it seems like. “Can’t wait for your new book, Rob. I have two questions about carry. Question one: In the original carry paper, the authors described how the carry strategy works in all asset classes but that the mechanics is varied. In equities, global bonds and credit, carry predicts future price appreciation. In currencies, future prices are independent of carry. In commodities and options, the market takes back part of the carry. Have you considered conditioning your carry forecast based on asset class?” First question.

[00:43:38] Rob

Okay. I’m just actually quickly pulling up my own asset class figures just to kind of give people a flavor for- so when they say, “gives back carry,” what they mean is the thing about carry is it’s quite a nice trading rule because you can actually work out how much money you expect to make very directly. I think that spot effect is a simple example of a carry strategy. If one currency is yielding 3% more than another, then your expectation is you’ll make 3% a year or whatever leverage you’ve got. You actually calculate the number. Whereas for something like trend following, you need to do a regression on realized versus expected trend strengths, like I was talking about earlier, to give you that figure. It’s much more precise.

The issue is, of course, that most trading strategies don’t necessarily do equally well on all asset classes. So for example, equity indices don’t seem to be as good with trend following, typically class trend following that’s on their asset classes, which is kind of interesting. The same is true of carry as well. So actually, I have this thing in my book I call “the percentage of carry realized,” which is- so effectively, if you think about the simple example where you expect to earn 3% a year, if carry realized is 50%, that means you’re only earning 1.5% a year. If carry realized is 200%, that means you’re earning 6% a year.

Carry is actually not just telling you how much you earn from carry, but it’s actually saying, well, this also means the spot price is going to appreciate as well you’ll make some extra money. So on average, carry realized to me is 115% across all asset classes. In other words, carry delivers the 3% I expect, plus a bit extra on average. The authors of this very seminal paper on carry found that was not true necessarily of all asset classes. Interestingly, carry seems to work best in financial, so equities, vol, bonds.

In FX, you get about the same amount of carry as you want to get. It’s about 109% there. It does pretty well in metals, but it doesn’t do as well in energy and AGS where you only get 77% and 46% respectively. So the market takes back some of the carry, but carry is still a profitable strategy if you like. So if you think about AGS, it’s about 50%. So if the carry is telling you you’re going to earn about 3% a year, in practice you make 1.5% which is still positive, it’s still good, which is nice, but it’s not 100% of what you expect, and that’s also the finding in the academic paper.

Now the question is, what should one do about this? Should one then say, “Well, actually, I’m going to trade less carry in say AGS markets like corn, and wheat, or soy, and more carry in financial markets like S&P, or VIX, or whatever?” Now, I’m always very reluctant, as we’ve discussed on the podcast before, to make these kinds of things where I start singling out individual instruments or groups of instruments like asset classes, and trade these differently from how I trade everything else, because statistically, the evidence really is not there.

So although it may seem like this is a very striking finding, that if I compare say, I don’t know, AGS, which is running at about 50% realized carry, versus equities that’s running at over 150%, it looks like it’s three times better. If you actually look at the statistical significance of the numbers, it’s pretty weak. So I’m just very weary of overfitting and so on and so forth. The benefit I could get to my system from doing that, from putting more carry in equities and less carrying in AGS, would be in the order of a few basis points of Sharpe. So my Sharpe, or my whole strategy might go from like 1 to 1.05. To me, that’s not enough of an improvement to risk potentially overfitting. So I personally would not do this.

[00:47:33] Niels

All right. Final question from Richard. He writes, “To calculate the raw carry forecast for seasonal instruments, I used the spread of nearest contract to nearest plus 12 months. However, I’m aware that this gives a weaker forecast compared to the usual method of nearest versus next nearest. Should I scale up my forecast to account for this? If so, by how much?”

[00:48:00] Rob

Carry is one of these things where in some cases, it’s quite hard to estimate when it should be. So obviously, there’s seasonality. There’s also weird things going on in it. For example, in the German 10-year bond market and a couple of other European bond markets, there’s actually a strong seasonal component there because of the issuance cycle. Even in equities, you would expect seasonality in equities, but actually, basically in Europe, most companies pay semi-annual dividends and one dividend is much bigger than the other. So you tend to see spikes in dividends which affects carry as well.

So carry is subject to this kind of seasonality issue. Hopefully the last time I plug my book, but there’s a whole chapter in my book which goes on about this and different ways to get around this problem. For this specific question, I would be very reluctant to do this. My attitude would be perhaps to say- some people would say, “If you can’t estimate carry accurately, then just don’t try to carry on that instrument.” I think that’s fine if you’re consistent with that. I’m less confident about saying, “I’m going to cut my carry forecast in half because I’m not so sure about it.”

To me again, this seems a little bit like overfitting potentially, without really looking at hard empirical evidence just based on a kind of slightly whimsical view about that seasonality. Maybe I’m being harsh. My general view is to do everything the same for every instrument unless there’s very strong, compelling evidence both statistically, and also a good reason why you should do anything different. It makes life simpler because you’ve got less chance of overfitting.

[00:49:37] Niels

We’re going to move on to- I think maybe we’ll just bring up this one, even though there were two things. One we forgot from last- no, we didn’t forget it. But from last time, there was a paper. What have we learned things since Jegadeesh and Titman’s paper came out 30 years ago, that article. I think you brought up another one today that I’m not personally familiar with, another AQR paper. They seem to be producing all these papers that we can talk about forever, but that’s about portfolio optimization as far as I remember, mean variance optimization. So you pick where you want to go.

[00:50:23] Rob

Yeah. Let’s briefly talk about the AQR paper on optimization. I like to think that there are a lot of people who listen to the podcast and want me on every week because I have the view of the retail investor. But I’m hoping that also there’s a small community of nerds who are listening who like me on because I do occasionally like to talk about the really nerdy stuff. So this is for you nerds, okay?

One of my main academic interests is in portfolio optimization and how to do it robustly, and so on and so forth. I do quite a lot of research and writing on it. So I’m always very interested whenever anyone comes up with a new take on the subject. This is particularly good because it’s a paper by AQR, who of course are generally very good. Specifically by- I’m probably going mispronounce this, so you’d do well to correct me, but by-

[00:51:15] Niels

Lasse Heje Pedersen.

[00:51:16] Rob

Lasse Heje Pedersen. Okay, I would have got it correct. Who is a Dane, an extremely good person who has a classic book called, Expected Returns, a more recent book actually, which I keep meaning to buy. So if anyone’s listening, Christmas is coming. So it’s about investing in drawdowns or something like that, no doubt it will be very good, because he’s a very good writer and a very, very brilliant guy. He’s written this with a couple of co-authors, who I think are also at AQR, but I could be wrong.

[00:51:47] Niels

I think that’s true.

[00:51:48] Rob

That’s true. These are all AQR people. It’s very interesting for two reasons. Firstly, it’s kind of using a way to do portfolio optimization that deals with the problems of portfolio optimization, which are well known. That it produces extreme weights, that small differences in the input can produce radically different weights, but it does so in a relatively simple way. I say relative because of course, the average person will look at the equations in this paper and be like, “Rob, really? That’s simple?” But it is relatively simple.

Therefore, it’s very intuitive to understand what’s going on. For those of you who are familiar, it’s kind of a take on the Bayesian Black-Litterman paper, which must be 25 years old now, possibly older. So it’s come and gone, and going back to that, and doing some tweaks, which is nice. I like people using robust methods that you can easily understand, rather than some fancy machine learning algorithm which is much less intuitive and probably less robust. Relevant to this podcast, is what they then do is say, “We’re going to alter the expected returns according to momentum.” Okay, so basically, the portfolio will overweigh things that have had positive momentum and underweigh things that have negative momentum.

For me personally, there are two ways I use momentum in my investing and trading life, if you like. There’s a very direct futures trading, classic CTA-type strategy that I run, that I talk about on this podcast, but most of you will know that I also have a long-only investment portfolio. Mainly of ETFs, but also with some individual shares that’s long only. I can still use momentum there by giving a higher weight to assets that have had positive momentum and a lower weight to assets that have had negative momentum.

The simplest way of thinking about that is purely the bond-equity mix. Let’s say my actual long-term target is 80/20 not 60/40. If equities have done particularly well relative to bonds, that could go to 90/10 and vice versa. I do that in quite a fairly cack-handed way. A very, very simple way. The nice thing about that paper is it has a similar kind of framework, but a much better theoretical grounding. So for the nerds out there, it’s definitely worth reading.

I have to confess, I only skimmed through it briefly because it was sent to me on Twitter by somebody yesterday. So I’ve not had a lot of time to look at it. If you’re into this stuff, it’s definitely worth looking at.

[00:54:23] Niels

Yeah, absolutely. Do you want to just then go straight to the trailer for the year-end group conversation? You had a few thoughts and things that you wanted to highlight.

[00:54:34] Rob

Let’s talk about the Jegadeesh and Titman paper. Let’s do that because that’s left over from last time. Remind me, was this something that someone had sent to us or something that came to your attention?

[00:54:44] Niels

No, I think this was one of your own topics, and I think you found the article in Zero Hedge, or maybe something like that.

[00:54:52] Rob

Niels, wash your mouth out. I do not look at Zero Hedge.

[00:54:55] Niels

Alpha Architect.

[00:54:56] Rob

Alpha Architect! Zero Hedge? I’m wearing a fleecy hat not a tin hat. Zero Hedge is not my natural home. Jegadeesh and Titman, for those who are not familiar, was the original paper where academics looked at momentum. Although, of course, like with a lot of these things, the industry had been using it for decades before that, as we know. In 1993, which is nearly 30 years ago, which is a terrifying thought, the academics finally caught up to the fact that trend following works.

To be pedantic, time series momentum works, which is not quite the same, because it’s a cross-sectional relative effect. It’s not about looking at whether things are going up and down independently. It’s about saying, “What’s in a basket of stocks? What’s gone up a lot?” Buy those. “What’s gone down a lot?” Sell those. Trading that factor portfolio. It’s very much in the spirit of the Fama and French papers. These two things are linked, but they’re not quite the same. This paper was written over 30 years ago. This new paper, which is- now again, you’re going to have to correct my pronunciation, Tobias Wiest, is it?

[00:56:18] Niels

Oh, I don’t have it in front of me.

[00:56:20] Rob

It’s W-I-E-S-T, so it’s German. Anyway, these papers come out in financial markets and portfolio management. It’s what you’d call a survey paper. It looks at a lot of papers that have been written in the last 30 years and kind of reviews the evidence, because obviously, one thing that academics really worry about is that they find some effect and then it disappears.

Why does it disappear? Well, there are two main reasons. One is that they data-mined the effect, it wasn’t really a robust finding to begin with. The other possibility is that everyone finds out about it and starts trading it, and then it goes away. There’s been papers done, and I think AQR might do one actually, where they look at a lot of these anomalies that people have found in the past. One of those two things has happened to them, and then they try and identify which is which.

Momentums, luckily for us of course, is not in that category. Ours actually kept working. There’s some interesting stuff in there that’s perhaps a bit more nerdy about the explanation for momentum. Is it behavioral? Is it risk-based? A lot of practitioners maybe don’t care about that. They’re just happy that it exists, and maybe that’s the right attitude.

Personally, from an academic perspective, I do find interesting the explanations, and I’m very much in the behavioral camp for what it’s worth. That’s probably the most interesting part of the paper for people listening. If you are interested in that stuff, go and read the paper. There’s a very interesting graph in the paper which shows the momentum for different periods, and asset classes, and geographic regions, and does so over time.

If we look at the original paper, which was like a lot of academic papers purely focused on US equities, because US academics seem to believe that there’s only one stock market in the whole world, and that’s the S&P 500. They found that you have a very good statistically significant effect, and you also have an average monthly return on this factor of 1%. Which doesn’t sound like much, but then that’s effectively another 12% a year over what you were earning, and that’s neutral of any market effect. So it’s 12% of pure alpha effectively, which is very good, obviously.

Now, what’s kind of happened since then is that a more recent paper was done by Ken French, overseas, The French in Fama-French, and that came out this year actually. Another paper I’ve read, by looking at the survey findings, they found that the momentum has actually reduced in the US equity market. It’s reduced by about half what it was effectively, so it’s no longer as strong there. A lot of papers have also looked at things like commodity futures, which obviously is very relevant to us.

I found very strong momentum there, and found strong momentum in European equities and Japanese equities, and so on and so forth. You could argue that this is a pretty standard story that in a very efficient market like the US equity market, an effect has been found and is starting to dissipate as more and more people trade it. That could be the case. In less liquid markets, the effect is still there. That could be one explanation. The problem is if you believe that story, then eventually all momentum will disappear, and it will no longer work, at least on this cross-sectional basis.

Maybe us absolute momentum trend following guys will still be okay. I’m not sure I buy that. I think for me, this sort of hammers in message, which we hammer in a lot on this podcast, and which I personally believe is one of the most important things to worry about, which is diversification. If you are just trading US equities with your momentum strategy, it wouldn’t have done as well recently as it did 30 years ago. To me, the answer to that is not to stop trading momentum, but to be trading lots and lots of different asset classes and geographic regions, and getting that diversification benefit.

[01:00:23] Niels

I think, actually, what you mentioned earlier and I agree with that behavioral biases, or behavior is really important in understanding why this probably works so well over so many different decades. It hasn’t been arbitraged away, so to speak. I’m talking about trend following in this case, of course. So interesting. Well, thanks for a quick update on that paper, which is called Momentum Factor Investing: 30 years of Out of Sample Data, as far as I recall.

Any quick few words about the topic that’s coming up in the- when I talk about the year-end group conversation, it has been recorded and it’s being published on the 24th of December and the 31st of December. That gives us a little bit of a break in our recording schedule, but it was a fun conversation. Is there anything you want to tease from that or have we already talked about it? Some of it was regarding tail hedging which we kind of ended up talking about earlier with a question we had.

[01:01:33] Rob

It was obviously fun recording the year-end episode, always is. One thing that came up, and actually we talked about it a bit earlier because we talked about tail hedging earlier. I said there were two papers that we were going to talk about which relate to the way you put tail hedging and trend following in a portfolio. One of those was a paper by a tail hedging manager, and they’re a bit biased, and that paper had a number of, shall we say, shortcomings. I obviously just disagree with the conclusion, but I’d like to think that wasn’t the only reason why I didn’t like the paper.

Now, you’ll have to listen to the year-end episode for a full critique of the paper. There is one thing in there I would like to just pick up on, because we didn’t really go into detail in the year-end episode about it and it has something I’ve talked about before, which is the relationship between leverage and CAGR, or annualized return or geometric return, which are all basically the same thing.

Now, basically, what will happen is if you apply leverage to a portfolio, or to a trading strategy, or whatever, your classic annualized mean return will increase linearly. If you double your leverage, you’ll see it, your mean return will also double, as will your standard deviation. That will mean that for example, your Sharpe ratio will remain constant. The definition of Sharpe ratio is basically what’s the return invariant to risk that allows you to compare things that are different.

If you assume you can use leverage, if you’ve got infinite leverage, you can basically just choose the strategy with the highest Sharpe ratio that’s kind of done the portfolio maths. Now CAGR works a bit differently, because what will happen is you apply leverage. If you say double the leverage on a portfolio with a given CAGR, with a given geometric return, the geometric return will not double. It will increase by less than that. If you apply leverage again, and again, and again, the CAGR will level off and eventually reach maximum CAGR, which is also the so-called Kelly criteria.

The important thing here is that there is a relationship between leverage and geometric ratio, whereas there isn’t between leverage and Sharpe ratio. That means if you’re using a geometric return as your measure of how well or badly a portfolio is doing or whether you should make a decision like in this paper as to whether to include say, a trend following index or tail hedging strategy into your portfolio, you need to be really wary of what will happen if you include something that has a low volatility into that mix. That low volatility will dampen down the arithmetic return, but it will also dampen down the geometric return. At least most of the time, unless your leverage overall is very high.

That was the error that was made in that paper, I think. They said we’re going to add a trend following strategy, but rather than picking a particular fund, they chose the SG CTA Index, and indices generally have a lower volatility than any one of their components. Now the SG CTA Index has a much lower standard deviation and returns a much lower risk than a typical CTA hedge fund will have. A CTA fund, a typical CTA ETF. However, you’re getting your trend following exposure. That means that including this very low volatility asset into the next would automatically dampen down the geometric return and make the thing look fat.

I just thought I’d throw that in, because the relationship between geometric return and leverage is something I’ve talked about before. Yeah, it is briefly in the year end podcast, but people might not understand the technical details. We don’t really go into them. So that there is a bit of a teaser. Perhaps you can listen to the podcast and then come back and listen to my explanation if that will make sense.

[01:05:40] Niels

Absolutely appreciate that very much, Rob. By the way, if you haven’t left a rating and review so far, we would love if you would take five minutes of your time to do it right after listening to this episode. It would help more people find the podcast and we would appreciate that very much.

On that note, we’re going to wrap up this conversation. We hope you enjoyed it. Next week, I’m going to be joined by Cem for another fun, insightful conversation. It is of course our picks next week, so make sure you send in your questions. As always, you can email them to info@toptradersunplugged.com, and we will do our best to make sure we address them.

From Rob and I, thanks ever so much for listening. We look forward to being back with you next week. In the meantime, take care of yourself and take care of each other.

[01:06:30] 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.