Trend Following or Mean Reversion: What Works Best When? ft. Rob Carver

[00:00:02] 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.

[00:00:25] 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.

Rob, it is great to have you back this week. I think actually it’s the first time in 2025. So how are things on your side? How are things in the UK?

[00:00:45] Rob

Things are fine. It’s a bit cold and damp here and I’ve actually had, I’ve got a cold, which I’ve had for several weeks and isn’t going away.

So, listeners should be aware that if there’s any weird gaps in the conversation, it’s because the editors had to take out about five minutes of me coughing. My voice sounds even sort of lower and gravelier than usual as well.

[00:01:05] Niels

Fair enough. I’m sure we’ll work our way through that.

We are, however, going to keep you pretty busy talking today because we’ve got a ton of questions in for you, which is great, so we much appreciate that. But before we even get to that, let me just ask you the usual question and that is, since we last spoke, lots of things have happened. Anything in particular that stuck on your radar the last few weeks?

[00:01:31] Rob

Yeah, I mean, something that’s come up quite recently actually is what on earth is going on with gold? Right?

I mean, so, gold’s gone up, which, you know, is one of these things that happens. And the causes of it, we could argue about – instability and uncertainty politically, which is interesting. But the thing that I find interesting about this specific thing is that there’s some kind of weird technical stuff going on in the background.

So, according to market reports, what’s happening is that gold is basically being moved from London to the US. And I’m not sure whether that’s a physical movement or the gold bars are staying in the same place, but the kind of legal right to it is moving. I’m not completely familiar with what’s going on there. And what’s happening as a result is that…

So, when gold is a futures contract, like a lot of futures contracts, the futures price will depend on the spot price plus any kind of yield that you earn on it, less the interest rates for funding the position. But because gold doesn’t earn a yield, you actually have to effectively put in essentially a borrowing cost and a storage cost.

So, the storage cost is, you know, if you’ve got a lot of gold in a warehouse, you’ve got to hire people with guns and thick walls and stuff to keep it safe, I guess, keep it underground somewhere.

But actually, amortized over a large amount of gold, the storage cost isn’t very much. So, what really drives the difference in the futures and the spot price is the borrowing cost. And borrowing costs have just exploded. I mean they’re up something like they normally follow pretty closely the sort of “risk free rates”. You’d expect them to be about kind of 4.50%, 5%.

Like roughly the kind of sort of Fed dollar rate, borrowing rates, because gold is priced in dollars of course. But actually, they’ve jumped up to like 10%, 11%, 12% which is just crazy because of this weird imbalance in inventories across warehouses.

And if I look at the futures price at the moment, so, for example, gold for delivery in say December is 150 points or something like that, which it’s up to about 10% annualized over the spot price, which is just weird. So, we have this interesting situation where, as a futures trader, gold is going up. I want to bet on gold going up.

And actually, if I look at my own forecasts, I’ve got a long position on gold and on silver incidentally, and on Bitcoin (which, you know, it’s digital gold, isn’t it?). But the cost of carry on that position is negative because the future is well above the spot price.

So, it’s one of those weird situations where you are kind of getting mixed signals from the price movement and the carry movement. And I love this sort of weird technical stuff that goes on underneath futures markets. And this is an interesting example of it. So, we’ll see what happens over the next few weeks.

[00:04:21] Niels

Yeah, I had not picked up on that. Well, I will say I have been traveling for about a month, so, I guess that slipped my radar. So, I’m glad you brought it up.

Does it say anything about, about who’s moving their gold back to New York?

[00:04:33] Rob

I’m looking at the article. So, there’s been a few articles. Some of them are in the less kind of accurate end of the financial press, shall we say.

But I’m looking at the Financial Times, which is normally pretty accurate and, and it doesn’t say so. So yeah, it’s a mystery to me exactly what’s going on. I’m sure that you can read all kinds of conspiracy theories on the Internet, but fair enough. But for the time being, yeah, it’s definitely causing some issues.

[00:04:57] Niels

Yeah, very interesting. Thanks for bringing that up.

For me, what hit my desk this week was an interesting, but maybe not sort of surprising in some ways, article that Bloomberg had about fees in the “hedge fund world”. And both you and I are old enough to remember when the traditional model 2 and 20 was the norm. Then, over the years, it was seen as being very rich and way too high for most investors. I think a lot of institutional investors certainly also helped push fees down in our industry.

And interestingly enough, of course now, actually the 2 and 20 model can be seen as pretty cheap and that probably needs to be explained somewhat. And it’s this article on Bloomberg that basically compares the 2 and 20 model to the new multi strat/pod shop pass-through model.

I mean I have to say it’s pretty scary reading if you’re an investor paying those fees. Although I do accept that the net return has been, for the most part, very good.

But there are some examples, and I’m not going to go through all of them. But there is, for example, one quote where they estimate clients are effectively paying something like 7 and 20 or even up to 15 and 20 – compare that to the 2 and 20 that hedge funds was known for.

And it all starts out with a comparison of how much was left by investors or for investors, I should say, from the gain of around, was it 15.2% gain that the Balyyasny Asset Enhanced Offshore Fund delivered in 2023. Before fees it delivered 15.2%. After fees, what the client got was 2.8%.

Now, I have argued before that, of course, the net return is the most important thing to some extent. What surprises me, really, and I’m not sure it’s covered by the article as such, is that we’ve seen (as many know) an enormous amount of interest and growth and money being allocated to this space. It’s kind of the new thing in our world.

And that, you know, leads me to believe that this must be large institutions that can allocate this amount of capital. Otherwise, it just wouldn’t be these numbers that we are talking about.

And so, if that is the case, then I will say I am surprised that some of these pension funds, insurance companies, et cetera, et cetera, are accepting the level of fees being put on these investments, at least compared to what I have seen in my career in terms of pushback from large investors, even in the low, relatively low fee world that I’ve been operating in. So, that actually is something that caught my eye. I know I sent the link to you. I don’t know if you had a chance to look at it or had any thoughts.

[00:08:14] Rob

I was sort of aware of this discussion, and actually I think it’s interesting because I think it comes down to transparency. I think, for right or for wrong, the old model where we were like, “this is our management fee, this is our performance fee”, was very clear and transparent. Whereas now it’s like, well, we have these management fee performance fees, but they’re quite low. And then there are these other fees kind of falling out of the back door of the fund that you can’t necessarily see because everything’s been charged effectively to the client’s account.

So, I think the issue might be that institutions just look at, as you say, look at the net performance, look at the kind of headline fees, and think, well, this seems fine without realizing that there’s all this money kind of disappearing out the back door almost invisibly. So, yeah, I mean, it’s not a new problem in the sense that if you think about a kind of fund of funds model.

So, you know, before Mr. Madoff came along, the fund of funds business was the way that people tended to get exposure to lots of different hedge fund strategies at the same time. The sort of multi strategy pod shop was less common.

But in that model, you had the issue where, for example, if you had managers that were doing really well, but the overall portfolio was doing badly, you’d have to pay performance fees to the managers that were doing well. So that’s another issue with the pods.

I mean, if you’re sitting in your pod and the whole strat fund, as a whole, is down, you’re still going to want to get paid. And the pod guys want to keep these guys sitting there in their seats, so they’re still going to pay them their bonuses even if the fund overall is losing money. And that’s another thing that kind of ratchets up the overall expenses.

So, that specific issue is not new, but I think the issue of the transparency of costs, I mean, I feel like we’re going backwards. Because in the retail world, transparency costs actually improved a lot. If you look at things like UCITS, the transparency costs is much better for the retail investors. But it seems like these multi strategy pods are taking a step backwards in terms of transparency, which I don’t think is a good thing, frankly.

[00:10:09] Niels

No, neither do I actually.

Although I will say someone mentioned to me that even the UCITS space, you can now find examples of people, if you read the perspectives close enough, where you have the official fee. So, everybody says, oh yeah, that’s great, and they may even state a certain expense ratio. But then when you drill down, there are some other costs, like research costs, et cetera, et cetera, that crop up. And so that’s a little bit worrisome if we start seeing that in the UCITS space because it really should be crystal clear, from the expense ratio, what people are paying and what people are not paying for.

Okay, so let’s leave that aside because I do want to just very briefly mention one other thing because it was on my radar when I saw it. It was just this picture of Elon Musk with one of his many children on his shoulders in the Oval Office. I don’t want to make this political, but I thought it was very telling of the times we live in. And then people have to make up their own mind as to what they think of it. I know we’re going to come back to some of this a little bit later, but from an economic point, of course.

Anyways, it’s too bad people can’t see your face right now, Rob, because you really want to say something. But I will now gently move on to the trend following update that has also been very interesting. I’m really curious to hear your thoughts on the first six, seven weeks of the year.

Now, as far as I can tell from looking at the indices, it’s been a mixed start across the industry. Different managers doing, you know, well, not so well in terms of performance. The CT indices are not moving a lot, frankly, away from zero. Some above, some, some below.

Obviously when you think about the market moves we’ve had so far, you would think things like equities have done well for trend followers, coffee, even some of the metals. You mentioned gold, for sure. And frankly, also, at least if you have a longer term horizon, I would have thought that fixed income had also done okay, despite the recent rally we saw in bonds. But now it’s selling off again with the latest inflation figures.

The only thing I can kind of see, from my vantage point, that has been a little bit tricky this year has been the currency sector, and that’s mostly been in February actually. So, does this resonate with what you’re seeing in your different models?

[00:12:39] Rob

To be honest, I’ve not looked at my performance. So, I’m actually just going to do that now. From memory, my gut feeling is that I think I’m sort of up a little bit this year. But, if you give me a moment, I’ll be able to tell you for sure.

[00:12:55] Niels

Yeah, no, I’m just curious to see because obviously all managers are different so these could also just be general observations even though I’m sure you don’t follow…

[00:13:06] Rob

I’m up like 1% for the year, so, basically noise, to be honest. And that consists of being down about 1 1/2% in January and then, so far, in February being up like 2 1/2%.

So, not very meaningful to be honest.

[00:13:27] Niels

No.

My own trend barometer finished yesterday at 30 which is actually a weak reading. But, again, it’s a different time frame for what I use for calculating that to what we see in the indices will also play a role. I think yesterday, which was Wednesday, probably was an up day for most people.

Anyways, in terms of numbers, BTOP50 is up 46 basis points as of Tuesday, up 1.68% so far this year. So that’s actually doing the best of all of the indices. SocGen CTA index up 15 basis points in February, up 77 basis points for the year. The Trend Index up 42 basis points so far in Feb, and only up 57 basis points this year. And the Short Term Traders Index down 18 basis points in Feb, and down 12 basis points in this year so far, and continues to struggle, frankly. I talked a little bit with Tom about that a few weeks ago. And so, I’ll probably bring that up with him next time he’s on the podcast.

MSCI World up 30 basis points in Feb, and up 3.79 so far this year. And the 20+ Year S&P Treasury Bond index is down 32 basis points (it obviously was hit a bit by these new inflation numbers), but still up 15 basis points so far this year. And the S&P 500 Total Return is pretty flat, up about a quarter percent this month, and up 3% so far this year.

All right, as I mentioned, we have a ton of questions in which is great. Now, first of all, some of them are long to read and I’m going to stumble across them, but I’m going to do my best. Some of it is also a little bit technical, although I have tried to weed out at least one that I thought was just maybe too narrow because we want something that is something that many people can benefit from. So, we’ll do our best.

We obviously take the questions as they come, but just bear with us and then we’ll move on to your topics which are truly very, very interesting, Rob. So, let’s do it.

So. the first question is from David, all the way from Spain. “Thank you both for creating such a high quality content for retail investors. I’ve been studying Rob’s book and working on putting the concepts into practice both for a long only portfolio and a managed futures portfolio. Question for Rob, I’ve been studying smart portfolios and am in the process of designing my own portfolio.

Since the book was published, several multi asset leverage ETFs have become available such as the WisdomTree Efficient Core, series such as the WisdomTree EF, and a one and a half times leveraged 60/40 US Equity Bond ETF. And there are some return stack portfolios, as he mentions. Anyways, the question is, do you think these products have a place in a long term portfolio? If so, what kind of allocation would you consider reasonable?”

Now, I want to preface this, David, and to all the other questions. Of course, we do not provide investment advice on the podcast, and of course each of us will just voice our own opinion. So, it’ll be as much as we can say. But don’t take it as investment advice.

[00:16:40] Rob

Definitely not. Because I’m actually not regulated to give investment advice anywhere. So, I used to be, but not anymore.

So yeah, this is an interesting one because actually if you do read smart portfolios and this is a kind of good general piece of advice, leveraged ETFs are generally a little bit dangerous, especially for holding for long periods of time. What happens is, if they go down a lot and then go up by the same amount, if they go down 10% and they go up 10%, you actually end up down. So, you’re not back where you started and then that’s leverage. So, instead of going down 10%, you go down 20% and then up 20% and again you’re even further back from where you started.

So, what will happen over a long period of time, with very volatile assets, is the value of these will tend to drift down. So, if you’re underlying is something that’s already quite volatile, like say the S&P 500, or let’s get really silly and look at, say, MicroStrategy, it’s called strategy now isn’t it?
The strategy company, which is basically just a bag of Bitcoin which you can buy at twice the value of the Bitcoin plus a small software business. I would definitely not, in a million years, buy a leveraged ETF on that because the underlying is very volatile and the value of that’s going to end up getting sucked down to zero over time with these large volatile movements.

Now to get technical for a second, the appropriate level of leverage and risk depends on something called the Kelly criteria, which depends on the expected performance of the thing you’re investing in. And that’s true for ETF, it’s true for someone targeting a futures trend following strategy or anything like that. And so, as a rule of thumb, if you kind of say, well, if the risk you’re getting on something is more than about 20%, 25%, 30%, it’s potentially quite likely that that’s going to be more than the amount of risk you should actually be taking because it’s unlikely that your performance will end up being high enough to justify that.

So that’s why, for example, I wouldn’t invest in say a two-times leveraged S&P 500 ETF because that’s going to have volatility of 30%, 40% a year, which I think is too high. I’m certainly not investing in a strategy times two ETF because that’s going to have a volatility of hundreds of percent, probably.

Now these particular products though, so, if you look at say 60/40 leveraged by times 1.5, that’s probably going to have (I’ve not looked at the product documentation), just off the top of my head, I would imagine that’s going to have a volatility of somewhere around the 12%, 13%, 14% level, something like that, 15% maybe. So, on that basis, I’d say that that’s probably okay, that’s probably a reasonably safe thing to invest in, just purely from whether the leverage is appropriate or not.

Not with any respect as to whether 60/40 is a good investment, or whether that particular product is a good investment, or whether the fees on that particular product are a reasonable level because I haven’t looked at any of that stuff. The return stack stuff, again, so, it’s two times leveraged S&P plus managed futures. That’s a little bit, sounding a little bit scarier.

I do know and have a great deal of respect for the people that actually launched this product. So, you know, they’re very sensible people who think very carefully about what they’re doing. So, for that I’m not going to just say, oh, it’s probably fine. I’m saying, okay, I’d want to have a close look at the documentation, look at the volatility of that product and look at how that’s come out.

And I would be a little bit skeptical and a little bit concerned because it’s probably relying on the fact that, if you look at the risk of that thing, if the correlation of managed futures and S&P stays relatively low, then it’s going to have a lowish risk, and applying some leverage to it is going to be fairly safe. The risk is potentially, of course, if the correlation of those two things increases and stays increased for a long period of time, then the volatility is going to be higher and it may potentially then be beyond the level which I’d consider a safe level of leverage.

So, I’m reasonably comfortable with 1 1/2 times 60/40. I’d need to think quite carefully about 2 times S&P plus anything, never mind managed futures. And as to what allocation, you’d have those in your portfolio. Well, I mean, you know, that’s an impossible question to answer in a short period of time because it’s very much going to depend on what’s in the rest of your portfolio, to be honest.

[00:21:19] Niels

Next question is from Carlos and with some of the questions that I can sort of quickly overstate oversee here, I’m going to rephrase them and make it shorter just so we have more time actually.

But Carlos brings up an interesting question I thought actually, and that is, if you start out with a trading account where you are able to trade 10 markets but you’re just using one model for that, you know, could be, you know, one approach, call it that. If you then suddenly have more money, would you then rather split the money and trade, you know, equal amounts of money but using more systems (so, say, a system 2 and trading the same markets), or would you add more markets to your model that you’re already running?

I know this is of course completely impossible to answer without lots of research, but philosophically I guess the question is, do you gain more from diversifying on models than you do on markets?

[00:22:25] Rob

I love the way you give me all these impossible questions, Niels. I really appreciate that.

[00:22:28] Niels

Well, Carlos actually gave it to us.

[00:22:30] Rob

Oh Carlos. Anyway, thanks Carlos.

Okay, so the answer is it depends, right?

So, if for example, your trend following system was relatively undiversified and just consisted of a single trading speed and then you were thinking about adding something to that, well, it’s quite likely you’ll get more diversification from adding more markets than by adding further trend following systems which are fairly similar. Because it comes purely under correlation.

So, the extra markets going in are probably going to have a correlation of 0.4, 0.5 with the ones that are in there, something like that. Another trend following system might have a correlation of 0.8 0.9 because there are only so many ways you can do trend following, even if you’re doing it at different speeds, it’s going to be fairly similar. So, I probably instinctively go towards more markets with my first answer.

When would be a case when you wouldn’t do that? Well, if you’ve already got quite a lot of markets, for example, then the additional markets going in are going to have a very small marginal benefit to the existing portfolio.

And if you’re then adding not just under the trend flowing system, but something that’s a bit different, like say carry, which we talked about briefly when we talk about gold earlier, then that potentially has got a correlation of maybe only about 0.7 with the existing system. So, at that point the pendulum swings from more markets being better to a different system being better.

And the other advantage of adding systems is at least if you do it the way that I do it, you don’t actually need more capital to do that. So, adding systems is virtually free as far as capital goes, whereas adding markets isn’t.

So, my answer is yes, markets, definitely. But given that adding systems is sort of “free” if you’re fully automated, it’s just a matter of writing some code. You know, obviously you lose a bit in terms of intuitively and complexity of your system. I wouldn’t, you know, rule out completely. I wouldn’t just add a thousand different signals to my model just because they all might produce a tiny marginal increase. I think there’s a point at which that’s not really adding any real value.

But yeah, markets first is my normal instinctive answer to that question.

[00:24:48] Niels

Yeah, that makes perfect sense.

All right, we’re going to jump to a quick question from Chris again. I’m going to try and summarize it.

Essentially Chris is asking you, Rob, whether using ETFs to backtest trend following strategies, you know, will give an accurate representation of performance. Of course, Chris is aware of the challenges with rolling inside an ETF if it’s based on futures, but also compared to obviously having to roll yourself if you’re using futures contracts in your backtest. Any thoughts on this particular issue?

[00:25:27] Rob

Well, the first question I have is what are you actually going to trade, Chris? I mean if you’re going to trade futures, then you really probably should be using futures to actually do your backtesting with. If, on the other hand, you are trading ETFs then it would probably be better, if you can, to use ETFs to do your backtesting with.

So, with that in mind, what are the differences between, say, holding an ETF which has underlying it some contracts like, say, the Bitcoin ETFs that have futures underneath them and holding the actual future itself? So, what are the differences between doing it one way or the other?

Well, one difference is fees. So, there’ll be fees applied to the ETF product and costs. And as we’ve discussed already, some of those costs may be obvious, some not be obvious, but what costs are those people going to have to pay? I mean, obviously they’re going to have to pay some administrative costs, they want some profits.

And there’ll also be trading costs from rolling from one contract to the next. And of all of those costs, the only one that you’d have to pay in the futures space is the actual rolling costs. So, you know, you should be able to get a rough idea of how much it’s going to cost you to roll and then compare that with the total annual expense ratio of the ETF and then check that does include everything that you think it includes and there’s no hidden stuff coming out of the back, and that’ll give you a fair comparison.

And ultimately, you’re probably going to end up paying a bit more for the ETF, I would imagine, because although, in principle, a big asset manager has got economies of scale and can actually probably end up getting lower costs than you can potentially, because they’re big they’re going to have more slippage, so they’ll end up with higher costs. And secondly, because they’ve got to make a profit and support all of these, you know, various functions, they’re going to have higher costs coming in there. So, all the things being equal, I would expect the ETF to cost more money.

[00:27:24] Niels

Yeah, and one final thing I just want to add to that, Chris, and that is just be aware also of liquidity. A lot of ETFs have been issued, but they don’t all have very good liquidity, frankly. So, you know, just be aware of that.

[00:27:36] Rob

Yeah, and the other difference, of course, between them is that if you’re looking at the futures price, then you’re basically, you have to sort of effectively add on the risk-free rate to that because the margin that you’re holding against that futures contract, you will actually earn interest on it. If you just look at your backtest, you won’t actually see that money coming in.

Whereas the ETF will actually include that interest within the price of the ETF, because the ETF is actually earning that interest on the capital, it’s got the exchange and it can return that to the investor as well. And that might be in the form of, you know, an outright dividend yield or it might be imputed into the price.

If it’s a dividend yield, then, again, you’ve got to kind of add it back in. So, essentially you want to be computing what I’d call a true total return series. So, for the ETFs, that’s going to include any dividend yields and it’s going to be less any costs that you’re going to have to pay, either implicit costs that are hidden or explicit costs in terms of a management fee. And then you can compare that to the futures price, back adjusted price, and that’s effectively, again, a total return series. But you need to add in the risk-free rate or deduct it from the ETF to get a fair comparison. So, this is why it’s much simpler if you can, if you’re trading ETFs to use ETFs in your backtest, if you’re trading futures to use futures in your backtest.

And then a second question is, what is better? Well, as you say, I think costs and liquidity are the two main points definitely to consider. But the reason why you would want to go down the ETF route would potentially be market access and contract size.

So, if the contracts are really big in the world of futures and you need a lot of capital to diversify, well, you may be better off going down the ETF route where the share prices are smaller and potentially even you can buy fractional shares. So, as far as the decision between ETFs and futures go, it’s not straightforward.

All of the things being equal, I’d say generally speaking, if you’ve got enough capital, futures are better. But not everyone’s in that position, of course.

[00:29:34] Niels

So, we can summarize it to test what you trade and trade what you test.

[00:29:40] Rob

That is a good thing to have. Definitely, always.

[00:29:43] Niels

All right, next question that came in is from Steve, and Steve writes, “In AFTs, (which, of course, I had to ask you, what exactly is AFTs? Of course it’s a good way to plug one of your many books, Advanced Futures Trading Strategies), all forecasting techniques are rules based. Any pointers on how to use predictive modeling techniques like linear regression etc. and how could we combine it with your forecast scaling framework? Also, can you comment on potential objective functions?”

I think again, let’s keep it broad so that most people can get some use for it and just allow for the rest of the questions too.

[00:30:19] Rob

Yeah, so this is kind of a general thing which is how do we get from what, in machine learning, they called a feature to a forecast of a price. But, in general terms, you’ve done some analysis, you’ve come up with something you think predicts futures prices. How do you get from say that wiggly line on the graph to a thing saying, right, this means we should buy X many futures contracts in say gold, which we’ve already talked about in the episode.

And the sort of simplest way of doing that, which is what I do, is literally to say, well I’m going to treat that wiggly line as something that has some kind of distribution. I’m going to construct in such a way that if it’s positive then I’m bullish, if it’s negative I’m bearish. And then I’m going to kind of calculate some scaling around it. So, I’ve got some way of saying is it high, is it low?

And that comes down to quite simply just dividing it by a number and producing something like, if you’re familiar with the terminology, something a bit like a Z score. Now, that process could equally be done by, say, a linear regression. And with a linear regression what you’d say is well, I’m trying to predict prices.

So, on the left-hand side of my regression equation I’ve got the price, or probably you want a normalized return, actually, a volatility normalized return on the left-hand side. And on the right-hand side of regression is the thing that you’re trying to predict it with. Well, that will be the wiggly line on the graph. And then the alpha and the beta of that regression will effectively be, the beta’s going to be (we won’t go to the details of calculations), it’s going to be very much the same thing in the sense that the coefficient on the regression is going to be something that tells you how big the wiggly line is. You know, is this a big forecast or a small forecast? And then the alpha, the insert on the regression, well that’s just a way of essentially removing any systematic bias from forecasts that are systematic long or systematically short, which you may not want to do, by the way. And that’s a whole big debate we can have on another podcast.

So, actually, there’s not really any fundamental difference between using say linear regression and doing what I do, with the possible exception of the fact that I don’t, generally speaking, remove systematic biases because (and we can have a big discussion about that) I just prefer not to. But in principle I could.

So, in answer to the question about the objective function, which just means in plain English, what is it we’re trying to forecast? Well, I would always be trying to forecast risk adjusted returns. I think that’s the most appropriate thing because we then want to size our positions according to risk.

[00:32:54] Niels

Yeah, cool. Good question. Next question that came in is from Vic.

Vic writes, “I’m curious about limits of research in finding new or improving systematic trading rules in the liquid mid low frequency space. Once you’ve included established risk premier rules like trend, carry, and fundamental valuations, do most research efforts by experienced teams in big and small firms amount to just fancy branding exercises? In a competitive environment where everyone is working with more or less the same data, is it possible to meaningfully move the needle? Would love to hear your views and thanks and all the best.”

What are your thoughts?

This is obviously super difficult because we don’t know what goes on inside the research teams, but we know they have some very clever people working there. What are your thoughts, actually? I have my thoughts, but what are your thoughts?

[00:33:46] Rob

Yeah, I mean, this is an interesting one and it’s quite a cynical view, isn’t it, to say that, well, everyone’s just doing the same thing. It’s just fancy branding and all this sort of stuff. So, you know, there are Indeed some CTAs that have not changed their model for years and not done any research and are just plugging along quite happily and that may be a very valid way of working as well, to be honest.

So, what are they doing inside these big shops with hundreds of PhDs? Well, they could be doing things like, for example, implementing new markets, some of which have issues with pricing. So, certainly when I worked at AHL, that was something that we were pushing to do in a big way, and don’t mind me plugging it, their very successful Evolution Fund was a result of that. And of course, there are other funds out there like Florin Court that have also pushed big into alt markets. And this is something we’ve talked about in the podcast before. So, that’s a big job.

And going back to the earlier question in terms of whether you should be adding markets or systems? Well, actually, adding markets can often give you the biggest bang for your buck. So maybe that’s what you should be doing.

You can be looking at things like improving execution as well. The bigger that you are, the more important execution is. So, for me, I can do a pretty decent job of execution with an algorithm that’s a few lines of code long. But if you’re a big fund trading hundreds of millions or even billions of dollars of notional a day, then execution is something that you should definitely be thinking about.

Then the other alt, of course, out there is alt data. So, there are people looking at alternative sources of data. And that’s also quite a big growth area.

I think where there’s probably less research effort than you might expect is in using, let’s say, alt methodologies. So, we had all the alts in this question. Alt methodologies, so that’s your neural networks, your machine learning, your artificial intelligence. So, basically working with existing data, but doing it in kind of fancier ways. That’s an area where I think you’re less likely to get much value, although undoubtedly people are doing it. But I’d be very wary of any sort of team of researchers that were purely focusing exclusively on that area of improvement, because I think the lower hanging fruit is quite high up in the tree there. And I think there aren’t many places, with the obvious exception of Renaissance Technologies, that are really good at that kind of stuff.

[00:36:15] Niels

Yes. At least for their proprietary fund, I might add. But there we are. I completely agree with what you just mentioned.

It’s not just the kind of data that I think firms are looking at. It’s actually also what to do with the data before they stick it into their algorithms that I think is an area of interest for these firms.

But I tend to agree. I don’t think necessarily that, as an industry, we’re coming up with many new ways of doing trend following. Although I don’t necessarily think it’s a bad thing that you use more than one approach to trend following. Instead of saying, “Oh, I’m wedded to moving average crossover,” well, okay, maybe you can combine that with something else and actually get a better result. So that’s kind of one small thing.

But the other thing I was going to say is that I think where I would suspect we see the most evolution still, and where there’s still room to improve, is probably risk management. I think that, at least what I see, is that better ways of dealing with risk, forecasting risk and all of that stuff I think is pretty interesting. And I think, as an industry, I think we’ve always been risk managers, first and foremost, and I think we’ve done a pretty good job. It’s rare that you hear about a trend follower blowing up unless it’s specifically because they were running like a 5x leverage version of their strategy. That’s obviously something I have seen in the past, which is crazy.

In one of the conversations we had when we did the SocGen CTA Index series with all the managers, I think some of the ones, maybe was AHL where they talked about that probably of their research budget, 35%, 40% of that goes to course execution – improving execution to not lose out when they get more inflows and manage bigger amounts of money.

So, I do think that is true and that’s obviously where managers have to be careful that they could still improve enough to increase the capacity of the strategy. But thanks for the question.

The next question is from Andrew, and Andrew writes, “Thank you very much, Rob, for your books and your transparency in your trading. Question, approximately about a year and a half ago or more you published on X that you were making a discretionary trade increasing your bond position. I’m just curious how that trade worked out and if you think, in retrospect, that discretionary call was correct. And are there any other learnings for the rest of us about when to know if a discretionary call makes sense?”

[00:38:54] Rob

Yeah, I have to say I really didn’t like this question because…

[00:39:01] Niels

Well, when you asked for it on X…

[00:39:03] Rob

I know, I know. Well, I’m a very good systematic trader. So, if you ask me how a particular trade works out, I can tell you with precision because it’s all in a big database.

But, the small number of discretionary trades I make, and the last one I made was during Covid, I’m not very good at kind of keeping records of them and sort of saying how they did in terms of P&L.

So, I did do that for my Covid trading because there was a lot of it in quite a short period, and I did work out that I had actually made some money. So, you know, that was nice.

But this particular one I actually just had to quickly check while you were talking, and have a look, and I did quite well in catching the bottom of the bond market, the top of the market in terms of yield terms. But I didn’t do a very good job of sort of closing the position. So, I think I actually closed the position basically flat.

So, I made a good entry decision but a poor exit decision. I should have had a, I mean this is ridiculous because I literally have written books about this, but I didn’t have a predefined sort of stop loss or exit criteria for my trade which is just crazy. And this is why I’m not a discretionary trader because I’m rubbish at it. Absolutely rubbish.

So, the learnings from this is don’t do it, I think at least as far as I’m concerned.

[00:40:30] Niels

All right, all right, good question. I’m glad we got that straightened out.

The next question is from Paul. Paul writes, “I have a question about incorporating value/long term mean reversion strategies. In Advanced Futures Trading Strategies, Rob introduces a mean reversion strategy based on past five-year performance relative to each instrument asset class. The strategy has a negative Sharpe ratio but improves the performance of his baseline trend plus carry strategy. I was wondering what the benefits/drawbacks of having an absolute strategy that just looked at if the post returns were positive or negative rather than relative to the performance of the asset class. In the academic paper Time Series Momentum, Moskowitz, al, in 2012, the authors show that returns years 2 through 5 are negatively related to subsequent returns. Given this result, it seems like applying a value approach on an absolute basis could increase the Sharpe on the standalone value measure while still maintaining the strategies negative correlation to trend.”

I know you have this in writing so don’t rely on my reading this question.

[00:41:47] Rob

I’m trying to, I’m really trying to dig through my mind and I can’t remember if I’ve ever tested an absolute momentum, an absolute long term mean reversion, rather, which is just negative momentum.

So, this should definitely work, and actually one of the things I want to talk about later is a paper that talks about momentum and mean reversion behavior across different time periods. So, this is a nice kind of preview of that. So, it should work in principle.

I don’t think I’ve tested it in like the last 10 years because I’m quite good at blogging about things that I’ve researched and I’m pretty sure I haven’t blogged about it. So yeah, I’ll have a look at it. I mean it’s in terms of Occam’s razor, you should always go for the simplest possible version of something.

And obviously this is simpler than the relative mean reversion. And even if it’s sort of similar in performance, it’s probably diversifying. It’s probably going to give you something a bit different.

So yeah. Now, as the questioner says, there’s a lot of research in it, particularly in equities. I mean there’s papers by people like Richard Thaler and stuff on mean reversion and you know, it’s sort of related to value effect and equities. So yeah, I’m a fan of the idea of it.

Of course, as a long term signal it’s going to be quite hard to get statistical significance. So, you know, and it’s never going to be that good in terms of Sharpe ratio because of that. And it may even be negative in the backtest. But yeah, I’ll make a note of that and have a look at it.

[00:43:13] Niels

Okay. All right. The next question is from Samuel. It’s a long one which I’ll probably butcher a few places, but I’ll try and do my best.

He starts out by saying, “I’m a big fan of TTU for a number of years now, but a few concepts have made their way into my head that would apply to the trend following universe and yet haven’t been covered on the show. (Well, there we are. Good that you bring them up.) Namely, what does the research say, if any, of trend following strategies that don’t rely on lagging indicators?

If I recall correctly, EMA – exponential moving average (that’s what I was just about to say) crossovers versus Donchian breakout strategies, if applied systematically, don’t change backtests all that much on a diversified basket. As Rob Smith highlighted (I’m not sure who Rob Smith is but), as Rob Smith highlighted in his May 2022 presentation, price doesn’t have mass. So, using the term momentum with stocks is more like describing a sports team that has momentum. It’s not literally applicable to the thing being described.

One simply needs to look at any duration candlestick chart to recognize that price often turns on a dime. Bright green on one candle, bright red on the next one, changing without any hint of a transition. To your knowledge, has anyone done any studies using the current state of monthly, quarterly, yearly candles for a trend following system, say reducing volatility at the beginning of those time periods rather than on a rolling basis. Same thing with adding to positions in addition to or in lieu of the various channel breakouts of EMA crossovers. Why not look at the current state of high time frame candles to increase exposure progressively? Same thing on reducing exposure, should something that was doing great one quarter turn around immediately be the next?”

All right. Oh, hopefully I know you have it in writing. So…

[00:45:19] Rob

So first a few caveats. Things I do not understand in this question or don’t know about Donchian breakouts. I’m not familiar with the work of Rob Smith and I don’t tend to look at candles.

With all that in mind. ultimately, all indicators are lagging because they look at the past, right? How can we reduce lag?

Well, we can reduce it by using less of the past and more recent periods. So, for example, we can speed up a moving average by using shorter numbers in the moving average. Exponential moving averages weight more recent periods more than periods longer ago. Okay, so that’s another way of doing it.

So basically, to get technical for a second, both the moving average and exponential moving average, and indeed any indicator that takes a series of past returns, is a weighting function over those past returns. So, a simple moving average is literally just the last, say 20 returns equally weighted so that the response function for that would be flat. The exponential weighting response function, obviously, is exponential. So, it’s high for recent periods and then goes down.

So, I think, if I understand the question correctly, it seems that he’s talking about doing something weird with the most, perhaps most recent observations, and weighting those. Either weighting them more highly or changing your response in a more nonlinear way to that.

So, for example, to paraphrase it might be something like, well, the moving average says we should be long, but because the last week or so is negative, we should actually be short or change our position. Something like that. I’m not generally a fan of sort of nonlinear stuff because it’s not very intuitive. And also, it’s highly, potentially can be highly overfit because you need additional parameters to do it.

So, you know, to implement the kind of thing I’ve discussed, you’d need to have a parameter saying, well, how far back do we look, what do we actually do when this thing reverses? I mean, there’s quite a few extra parameters potentially there. It’s making the system more complicated and potentially more overfitted.

It might be better. A simpler way of doing that is to say something like, well, I’m not saying this probably isn’t true, and I’ll discuss why in a bit when I get to my part of the podcast. But if you think, for example, that prices trend over six months but then tend to mean revert, if they have been trending for six months and they start to mean revert suddenly, then you should go short.

Well, a better way of doing that is to have a separate mean reversion one week signal, or to fit some kind of response function, as we were talking about earlier with the question about regression, between how prices move depending on how strong your forecast is. And again, I’ve done that and there are some effects there, but I’ve judged that the complexity they add is not worth the tiny, tiny, insignificant performance that they add.

So yeah, I think this is one of those things that kind of sounds like a good idea. Let’s get rid of lagging indicators and use indicators that don’t lag. Well, actually it’s impossible to do that.

[00:48:29] Niels

Actually, it would be better to have future indicators, right? So, we would always know.

[00:48:33] Rob

I mean, I would prefer to have future indicators. Unfortunately, I’ve not been able to find any because, you know, time travel is not possible.

[00:48:41] Niels

Not yet.

Anyways, last question and then we get to your topics. I have to preface here. First of all, it came from Crypto Captain. Now Crypto Captain is, I think, a longtime listener, so I really appreciate that. And Crypto Captain has also asked questions before, as far as I recall. I do think, however, I did mention last time, Crypto Captain, that you really should use your own name or at least tell us who you really are because we don’t really appreciate people being anonymous on this. I will never mention your last name, but let’s make it more direct instead of using these different names.

Anyways, you asked two questions, Crypto Captain. We will answer question two because the first question was simply, in our opinion, too narrow for our audience. And so, I’m sure you will understand that. However, your second question is something that we both felt was relevant. So here goes.

You asked, “How to handle missing data when contracts get delisted and then relisted. In my case, many contracts in some commodities got delisted in June 2020 and then got relisted in February 2023. ChatGPT suggested I use co-integration and error correction models to fill the missing data because the larger contract data is available. What are other things I can try out?”

So, Rob, over to you.

[00:50:03] Rob

Well, the easiest thing to do is to ignore any data before February 2023. So, basically ignore the period it was trading earlier and obviously ignore the gap.

The next thing to do that’s still kind of okay, but more complicated is to create your trading system so it can actually deal with missing data. So, then what would happen is that in your backtest you’d be trading this thing for June 2020, and then you’d go to a position of zero until the prices started coming in again. And then once there was enough prices to form an opinion about what the forecast should be and what the volatility should be, et cetera, et cetera, then you’d go back to having a position.

I would really, really not interpolate data, price data, and then, then use that in a backtest and say, oh yes, look at, this is great. I think it’s a fundamentally stupid thing to do, to be honest, and I’m not surprised that ChatGPT has suggested it because, you know, I’m not a big fan of AI, as you know. I really, really wouldn’t do that, to be honest.

Now, there are some limited cases in which it might make sense to do this. So, for example, if you are, say, estimating a volatility and you’ve got hourly data, but obviously you’ve got a period where when markets are closed, then it’s probably a reasonable thing to do to get a better estimate of volatility to actually interpolate those overnight hours. I’ve seen people do that. It’s a reasonable thing to do.

In terms of techniques, I wouldn’t use co-integration or an ECM. I’d use a Brownian bridge. If you don’t know what one of those is, you shouldn’t be doing this, frankly, because, you know, it’s quite complicated stuff and you need to be very careful with it. But I would use it in that specific instance and if I think hard, I can probably think of a few more. But 99.9% of the time, interpolating missing prices is a fundamentally stupid thing to do, that only an AI would suggest.

[00:52:02] Niels

All right, let’s move on to your topics. Now we’re going to talk about your most recent blog post, which is on a very interesting topic, which has been discussed in different shapes and forms over the years now. However, actually, there is a very nice sort of bridge into you into this topic from the most recent, which is Q4 2024, paper from Quantica, our friends here in Switzerland, who write some excellent stuff. People should go and check it out.

Now, I think, and I can’t remember if I did the discussion on this paper or maybe Alan did with Katy. I’m not entirely sure. Anyways, maybe you could just quickly summarize what they concluded about dynamic position sizing and so on, and so forth, and then gently take us into your blog post and guide us through that.

[00:52:54] Rob

Yeah, so, the Quantica paper is a really nice paper and I definitely encourage people to read it. I’m not going to summarize it in great detail here because that’s not the main point of the conversation, but it’s about evaluating three different kinds of position sizing framework.

One, where you enter a trade and you take a certain number of contracts and you hold that fixed number of contracts.

The second method is fixed notional, where you would say, all right, I want to get say $100,000 of exposure to this particular future. On day one, that might be five contracts. On day two, maybe the price has gone up a bit, therefore you might lower it to four contracts. Obviously with a very small amount of capital it would be quite hard to get an exact notional, but with large enough capital you can obviously get pretty close to the notional you want to target.

And then the third methodology is the methodology I use, which is to say I want to get a certain amount of risk on my contract. So, you’d say I want to have $25,000 of annualized risk on that contract. What does that correspond to? And then that will change if the price changes, but it will also change if the volatility changes. So, most notably, if the volatility goes up a lot, then you’ll reduce the number of contracts that you hold. And they look at this specific example of cocoa, because obviously cocoa was the poster child trade of 2024.

And they then sort of evaluate these different techniques. And I’ve done a similar work myself, and they come to the conclusion that the volatility adjustment has the highest Sharpe ratio. Okay.

What they don’t do, however, and I have done in my own work, is look at skew. So, you know, trend followers reduce positive skew. And it turns out that the closer to your sort of fixed position sizing the system you’re running, or even the notional position sizing, the greater the skew you’ll get.

And the reason for that intuitively is, well, what’s happening is that if you have something like cocoa that explodes in price and goes up a lot, and you’re just holding a fixed number of contracts, then that’s going to produce an outsize effect on your P&L and an outsize effect, positive outlier on the upside of your P&L. And the same thing doesn’t happen on the downside because obviously, when things move against us, we close our positions.

So that’s the kind of intuitive logic behind that. So that’s the Quantica paper. Go away and read it. It’s very interesting.

But this comes down to essentially a question we should ask whenever evaluating any kind of strategy or asset in finance, which is what should we be paying for risk? And we use Sharpe ratios because as futures traders we can use leverage. And that means essentially if by risk, if you mean volatility, well we can get any level of volatility we like. We just need to change our leverage. And that’s not going to change our Sharpe ratio,

So, effectively, the price of risk is basically zero for a leveraged trader. We can get any amount of risk that we want to get. That’s not true of skew though, necessarily.

So often when we’re evaluating different options or, say, different hedge fund strategies, we might have a choice between something that has a really good Sharpe ratio but negative skew. And an example of that would be something like… An extreme example of it would be something like an option selling strategy. A less extreme example of that would be something like an equity market neutral strategy. They tend to have negative skew as well.

And then you might be comparing that with something that has positive skew, like, say, a trend following strategy. And you could also be comparing different kinds of trend following strategies, so ones that are closer to mine, where you’ve got good Sharpe ratios but the skew maybe isn’t so good and then you’ve got other funds that have lower Sharpe ratios but very, very high positive skew.

So, what I wanted to do was, in a sort of intuitive way, kind of say, well, if I’m comparing two different assets, whether they be funds or strategies or underlying instruments, and they’ve got different Sharpe ratios and different skews, what should the kind of trade off between those two things be, at least in theory?

And I say in theory because in practice people have preferences for this sort of thing. So, some people really like positive skew and they’ll, you know, happily give up more of them, their Sharpe ratio to get it. Other people won’t. So, this sort of is like a risk neutral approach, if you like, as far as skew goes.

Anyway, my conclusions were quite interesting because I was surprised to find that trade off wasn’t actually that substantial. So, in other words, the amount of Sharpe ratio you should be giving up to “buy” positive skew was actually be very small.

To put it another way, if you have two strategies, one with a very good Sharpe ratio and one with a slightly worse Sharpe ratio, but with very good positive skew, generally speaking, you want to go for the higher Sharpe ratio strategy because the geometric return of the product is going to be better. And the geometric return, sometimes called the CAGR, the Compound Annual Growth Rate, maximizing that basically maximizes the amount of money that you have at the end of your investment horizon. That’s, I believe, the kind of main fundamental metric that everyone should be using when they’re evaluating anything.

Sharpe ratio only works if everything has the same skew. And here we’re looking at a specific example where things have different skews.

So yeah, it was interesting, and I guess for me it was another nail in the coffin, if you like, of the idea of using something like a constant contract or a constant notional, as is in the Quantica paper, because they do have a lower Sharpe ratio. I found that. Quantica showed that as well.

But any improvement in skew… There’s no conceivable amount of improvement in skew that would justify that lower Sharpe ratio and sort of pay for that lower Sharpe ratio if you like.

[00:59:02] Niels

So first of all, people should go and read this full blog post on your website and we’ll put a link to that in the show notes, of course. And again, because we’re starting to run out of time a little bit, I just have one general question that I think some people might think and sit with, hearing your thoughts on this.

And that is, well, on many of these episodes we’ve had in the past decade or so, I’m sure many people, including myself, would have said, well, hang on, Sharpe is not really great to optimize for when it comes to trend following falling because it penalizes upside volatility. How should people think about that when you say, well actually we should still optimize for Sharpe?

[00:59:43] Rob

It penalizes upside volatility, sure.

But the point is that if an investment has a high Sharpe ratio, you can sort of leverage it up so that the benefits of getting the upside and the downside…Yeah, this is quite a hard question to answer actually.

[01:00:03] Niels

That’s fine. It was on the fly, so don’t feel like…

[01:00:06] Rob

So, I’m trying to think of an intuitive way of explaining it, but basically what I did was sort of simulate the effect of holding different investments with different levels of Sharpe ratio and skew. And I said, well, the only metric I care about is how much money I have at the end of time.

[01:00:21] Niels

Right.

[01:00:21] Rob

So that simulation accounts the fact that the high skew, positive skew, lower Sharpe ratio investments, their pattern of returns is going to be getting all of this extra upside volatility.

The point is that, in this framework, you don’t really think about volatility. Volatility only matters in as much as it will reduce how much money you have at the end of time if it moves against you.

So, the point was basically that the additional benefits of having a higher Sharpe ratio massively more than compensate for the fact that we’re not getting those big upside volatility moments. So, I think it’s quite a good framework thinking about things, because you don’t need to say, well, okay, yes, upside volatility should be valued more than downside volatility, which Sharpe ratio doesn’t account for, but skew does.

But actually, combining those two things together, combining a measure of symmetry, essentially, in your performance judgment, which is what skew does, it still tells you that you should generally be hunting for higher Sharpe ratio investments. Because, you know, the benefits of positive skew are, when you actually look at how much money you’re going to end up with, you know, they’re limited.

[01:01:36] Niels

Yeah. And of course, always a warning that some very high Sharpe strategies, I can think of one like Bernie Madoff, may not always turn out to be that great of an investment at the end of the day.

[01:01:50] Rob

Absolutely. Yeah.

[01:01:51] Niels

All right.

[01:01:51] Rob

Especially if they’ve got a lot of, you know… Ignoring like outright frauds like Bernie Madoff, I mean, we should always be careful of high Sharpe ratio strategies that require a lot of leverage because even if they haven’t got negative skew risk in, in the backtest during the historic returns, it’s something you should always be concerned about.

[01:02:07] Niels

Yeah, and are opaque at the same time in some cases.

Okay, all right, the next one, we’ll keep the best for last, of course. So, we will get through this one first because you mentioned that this is actually an interesting paper and I simply hadn’t got the time, when I came back last night from my travels, to dive into it in any great details. But you already mentioned that it’s somewhat relevant to our previous discussion today.

So, I’d love for you to take us through this paper that is very recent. It came out, I think only a few days ago. I think it’s called Trends and Reversion in Financial Markets on Timescales from Minutes to Decades.

And I should of course have mentioned the authors. I don’t have it in front of me here. You may have it, just to be full credit.

[01:02:52] Rob

Yes, Sara Safari and Christof (and I’m probably going to mangle this) Schmidhuber, both of whom are not far from you, Niels.

[01:03:02] Niels

Exactly. That’s exactly why we want to definitely give a plug for Zurich University, which I think this is where they relate from.

Anyway, I’m going to turn it over to you, Rob. You read it much more carefully than I did.

[01:03:14] Rob

Yeah, I mean this is a really interesting paper.

So, we’ve mentioned my previous book already, but in my previous book I say, well one thing that’s interesting is that at different timescales mean reversion and momentum tend to do better or worse. So, as we discussed with one of the earlier questions, if your time period is multiple years, then generally speaking you’re probably looking at mean reversion. Momentum seems to work well, empirically, certainly in futures, across multiple asset classes for time periods of say a month up to a year.

And we also know that if we go right, right down to kind of really small time increments, mean reversion tends to work well because that’s where the high frequency traders are operating and their strategy is very simple. It’s buying on the bid, selling on the ask and they’re relying on the prices kind of bouncing between those two points.

And in my book I say, well, there’s a sort of a gap between this high frequency trading and this one week, one month time horizon where momentum or mean reversion may be working. And I kind of, unfortunately I didn’t have the data to do an analysis and say what actually happened in those time periods.

I kind of waved my hands around and came up with some suppositions that actually this paper says are false. So that’s kind of, I don’t mind having my vague guesses refuted. I’m much happier to see hard evidence because, apart from anything else, it’s a really good guide to if you’re thinking about sort of going into faster trading, whether that faster trading should be mean reversion or momentum. I think it’s really useful to have that as a starting point.

But anyway, what they do is they, they look at probably the widest range of time frequencies I’ve seen in any paper ever, which is fantastic. I won’t go into the technical details of what they’re doing, but basically what they do is for different time horizons, time frequencies, they basically say, is this a time frequency where we see momentum or is this time frequency where we see mean reversion? That’s kind of what the paper boils down to.

And if you do nothing else, go to page 28, figure 10 and that’s the figure I’m now going to describe to you. And that basically summarizes the paper beautifully.

Now what complicates things slightly is that the way that they analyze trends is a bit weird. They fit a cubic polynomial, which is a slightly unusual way of doing it. And to get technical for a second, it allows you to model both the sort of relationship between trend strength and mean reversion and also the general trend. But we’ll not talk about trend strength because there is some interesting stuff in there but I think it takes away from the key idea in the paper I want to bring out, which is the relationship between, as I said, at a given horizon, do we see trends or do we see mean reversion?

So, they go right down to sort of minute level data, and they basically find that, let’s say for time periods of less than an hour, mean reversion occurs. Okay. And I think the most mean reversion occurs at roughly a five minute time window. So,that’s kind of the area where, if you’re going to be a mean reversion trader, you want to be playing in.

And a huge caveat here, you know, trading at those kinds of frequencies is a massive engineering and backtesting exercise and it’s not something that you should be casually doing. You don’t just now sit at your computer and look at charts and every five minutes do mean reversion trades. Do not do that, whatever you do. But empirically that seems to be what’s going on.

Now, if you look at trend horizons of more than an hour, they find momentum occurring. And this is where this sort of fills in the gaps in my previous knowledge because I wasn’t sure what would be happening at these time horizons. But basically, if you’re trading for holding positions for an hour, or two hours, or four hours, or a day, you should probably be trading momentum. And again, big caveats about trading that quickly.

[01:07:19] Niels

Sure.

[01:07:19] Rob

Trading costs, in particular, are going to be very hard to overcome if you were trend following in short time frequencies, so be very careful there.

And then they go on to sort of two days, three days, four days, five days, ten days. It’s still momentum. You know, three weeks, six weeks, three months, six months, one year, it’s still momentum. So that, you know, it’s momentum all the way.

This is a great paper for our industry because it’s basically saying that as long as you’re not really a really fast trader, you should probably be a momentum trader, which of course is what most CTOs do. And then is when the switch happens.

Then is when the switch happens. So, anything longer than a year is when mean reversion kicks in.

And as I said, they do look at ridiculous amounts of data because they go up to 16 years. They look at data out 16 years and they’re still finding mean reversion out there. And to do that they’re looking at data from 1692. So, they’re looking at, you know, 330 years of data to do this analysis.

So, it’s an incredibly thorough job and very, very, very impressive. But yes, the bottom line is, so we talked earlier about looking at absolute mean reversion over multiple years. This paper supports the idea that if you’re trading, is it really trading if it’s multiple years or is it just investing? I don’t know.

Yeah, but if your forecast horizon is, you know, two, three, four years, definitely a mean reversion strategy is more likely to make sense. If your time horizon is anywhere between one hour and one year, you should be a momentum trader.

And if you’re able to trade at sub one-hour frequencies, then yeah, you could look at mean reversion. So, it’s a beautiful empirical survey of everything from right down to the tiny, tiny subatomic structure of high frequency trading, zooming out to the giant galactic views of multiple year holding periods.

[01:09:18] Niels

I’m surprised, actually, that it cuts off at one year, a little bit, because I do think that many trend followers use lookback periods that are somewhat longer than one year.

[01:09:26] Rob

Yeah, well actually, the cutoff point is two years. One year has the strongest, has a very strong trend flowing performance. Two years is pretty much flat. So, you might get away with 18 months.

[01:09:39] Niels

Yeah, that’s actually what I would have thought.

[01:09:41] Rob

Yeah.

[01:09:41] Niels

Without doing all the research, of course. There we are.

Okay, we’ve come to the last topic brought to you or brought by you, I should say. And it’s about one of your favorite persons to talk about, Trump, but not in a political way. It is from a economic way.

[01:10:00] Rob

Yeah.

[01:10:01] Niels

What does it mean?

[01:10:02] Rob

What does it mean? What does it all mean? Yes, what is the point? What are the economic consequences of Donald Trump?

[01:10:08] Niels

Correct, absolutely. In your view.

[01:10:09] Rob

There’s a paper written, by John Maynard Keynes about Winston Churchill, almost exactly 100 years ago. Yeah, so, I’ve been told I’m not allowed to be political on the podcast. It’s not a politics podcast and I might offend some of the people listening who are fans of the man. So, this is not political at all.

This is a pure hardheaded macroeconomic analysis of the likely consequences of Donald Trump and, of course, the implications for any investments that you might care to make over the next four years. So, we’ll start with the big one, tariffs, of course.

By the way, I should preface this by saying that I’m going to assume that he is successful in his endeavors so that he’s going to actually do the things that, A, he said he’s going to do and, B, he appears to be trying to do. So, you know, there are some instances already of pushback from the courts, potentially some Republican politicians. And it’s going to be quite interesting to see how the sort of conflicts between the different branches of the US Government resolve themselves. Because there are going to be conflicts and there are going to be arguments and discussions, that’s for sure.

I think a lot will depend on how much he gets done in the next two years because I can’t really see the midterms going that well. And midterms generally don’t go well. Like, for presidents it’s sort of a stop light.

It’s pretty usual that if you start a presidential term with a majority in the House and the Senate and the presidency, it’s pretty likely you’ll end up losing one of those majorities in the midterms. That happens nearly all the time, mainly because people just don’t like sitting, you know, they don’t like sitting governments. So, the midterms are almost a bit of a protest vote.
And we see a similar thing in the UK with sort of local council elections, but those are far less important than the midterms, clearly.

So, yeah, it’s going to come down a lot to what he manages to get done in the next two years before he loses, I think he’ll probably lose the Legislature.

Anyway, having said all that, let’s start with the big one which is tariffs. The tariffs are interesting because it’s probably the one of his policies that there’s the most pushback by people who, actually, he’s going to listen to to. Because most Republicans think that increasing tariffs is a terrible idea. Trump uniquely seems to think they’re a good idea. But it’s generally accepted that tariffs will increase inflation and just generally be a bad thing.

And I don’t think I need to talk about that in a lot of detail because a lot of ink’s been spilt on why tariffs are a terrible thing, and almost no mainstream economist thinks that they’re a good thing. So, they’re going to increase inflation, but of course they won’t just increase inflation in the US, they will increase inflation globally, I think, for sure, due to retaliation and just generally. So, let’s put that one aside and look at other things that he’s up to.

So, he’s planning to deport a lot of people, and send them back to where they came from. What effect will that have? Okay, well, simple supply and demand. If you reduce the amount of labor in the market, then that will probably increase wage costs, I would imagine, which is more inflation. Now, there’ll be an effect on the demand side as well, but I think it’ll be less substantial. But the other thing that really worries me is the likely effect that this will have on supply chains.

I think what Covid really showed us is that the sort of network of supply chains in the world is a very delicate thing, and anything that causes damage to it can have consequences which are very problematic. And you end up with stuff in the wrong place, and stuff not being manufactured, and issues with that.

And I think there are also potentially supply chain consequences from the tariffs as well, because, for example, I know that US cars, bits of cars, go backwards and forwards between Canada and the US across the border. Think about where Detroit is actually physically located for a start. So that’s again, potentially going to lead to inflation. Again, I think a lot of these things are inflationary, I really do.

Then we get into something a bit more esoteric, which is regulation. So, I think it’s fair to say that Trump doesn’t like regulation. And there’s a sort of naive view that all regulation is a negative cost for businesses. So, therefore, less regulation should be positive for share prices because businesses will make more profits. Obviously, there is some truth in that to an extent. But actually, what businesses want and like is things like certainty, and the rule of law, and a set of rules and regulations that they can kind of rely on. And if you start messing around with things like that, then what that’s probably going to do is actually increase what economists call, the risk premium.

So, people will demand to be paid more to hold risky assets, because everything’s getting riskier, everything’s changing, everything’s all over the place. So, I think potentially, actually things like regulation and things like tariff policies that change every five minutes even if they don’t end up going in the wrong direction, that’s going to increase the risk premium, which would be bad for equities.

I think that zooming out a bit more, and looking at the fact that he seems to be, how can I put this politely, making some fairly radical changes to the way that the sort of US Government operates, and potentially even doing things like literally, metaphorically putting his finger (or perhaps it should be Elon Musk’s finger) on various spigots of money that are flowing and keeping the US Economy moving and going. Just putting a finger on and saying, what happens if I just stop this payment?

Again, what’s that going to do? Well, potentially it’s going to make people unemployed, it’s going to cause supply shocks, it’s going to cause demand shocks, it’s going to cause uncertainty.

And so, I think the fact that he’s sort of breaking the contracts that the American government has with its people, and also that the American government has with other governments, it’s going to increase uncertainty, it’s going to increase risk premium, it’s going to be bad for equities. I think there’s going to be inflation, which is going to be bad for bonds. And of course, the conclusion of this is we should just all buy CTAs, lock ourselves in our bunkers with our shotguns and our baked beans and hope for the best.

[01:16:42] Niels

Well, I mean, there’s also a little bit of a nuanced view on this. I don’t, I don’t disagree with some of the stuff you’ve said.

And actually, you tricked me a little bit, Rob, because you sent me a link to an article by the FT and that that was a slightly different version of what will happen under Trump. So, you know, kudos for me to agree to this topic.

[01:17:05] Rob

You can still cut it out at the edit, now.

[01:17:07] Niels

No, absolutely not. That’s not how we do things here.

But I think there are a couple of interesting observations in the paper or in the article in the FT, because I agree with you that there are certainly a lot of risks in doing what’s likely to happen. But there’s also this conundrum that we see the risks showing up in only parts of the financial markets at the moment. Right?

So fixed income is probably showing a little bit more concern about what’s going on while equities…

[01:17:40] Rob

So is gold of course.

[01:17:41] Niels

And gold, as we talked about, yes. Whilst equities are not really showing a lot of angst at the moment, if we just measure angst by the price level on many of these indices. So, it is an interesting time.

I’ve obviously alluded to it in my previous conversations and I will dig a little bit deeper with a very special guest in a couple of months time, because I think what you’re saying and what I’m saying, in a slightly different way, is I think that not just what happens right now in the White House, but actually what’s happened in the last couple of decades, is an erosion of trust, erosion of trust in institutions. I think that’s probably also why I mentioned the picture from the Oval Office earlier in our conversation. I do think we are losing respect and trust in a lot of these institutions.

And that to me is a serious issue. And in a world where there is definitely a disconnect also happening between what is value and what’s the price. I do agree with you that, actually, a price-based strategy that doesn’t care about ‘is it the right value or not’, but just follows the price. Of course, I would at all times say that that’s a pretty good strategy to have in your portfolio. And trend following is certainly one of very few that I can think of.

[01:19:08] Rob

So actually, if I think back to 2007, 2008, the equity markets for a long time thought everything was fine and it was in the bond markets, in the CDS markets and so on, the corporate bond markets and the mortgage backed security markets that the initial pain was and the initial foresight was.

And I do think that I’m reluctant to say that market X always leads market Y, but I do think there is an argument for the fact that most of the people trading equities are naturally, how should we say this, optimistic people who might be slow to kind of make a judgment about market news. And that’s probably particularly true now that I think equity trading now has got a much bigger percentage of retail traders than it ever used to have.

The bond market however, is still, I think, dominated by more professional traders. And I think bond investors also are naturally grumpier and more conservative than equity investors. They must be to accept that kind of 4% or 5% yield.

So, I do think that potentially this could be a situation where the bond market could be a bit ahead of the curve and maybe even the gold market in saying, well, look at, there’s some scary stuff going on here.

And obviously there are different drivers because the bond market’s probably more concerned about inflation rather than, say, the risks of a recession, whereas the equity market. Is inflation good or bad for equities? There is not an obvious answer to that question.

So it may be that it’s just a more direct thing, that Trump’s policies are clearly inflationary, therefore bonds will probably react, equities, not so sure. But I do think that as some of these other effects start, I mean, he’s not been in office that long, Right?

You think about the amount of stuff he’s done already, but, you know, a lot of the things that he’s doing, there’ll be quite a lag before they have an effect on the real economy and start showing up in things like jobs numbers and even bigger lag before they show up in equities. So, I’d say watch this space.

[01:21:04] Niels

Yes. And I’ll finish with one thing which actually I do think might be also a little bit of the signs that we’re seeing now. Many, many years ago, I came across someone who talked about this idea of cycles between public and private, Where sometimes the public trust is high, sometimes it’s very low, and it’s the private…

And I will say I have been thinking about this concept a little more recently, and I would not be surprised if what people think is safe, i.e. government bonds, will turn out to be not so safe. And actually, what we think of, maybe more risky normally, such as equities, might actually turn out to be more of a safe harbor.

This is not a market forecast, but I just think we need to revisit or even take out of the archives some of these concepts, some of these cycles that come across so rare that we don’t think about them day to day. And always, at least in my mind, I always think about the conversation we had with Neil Howe and the books that he or the book he wrote back in the early 90s, The Fourth Turning.

I think that is a concept that we should not ignore at this point in time. And I fully, firmly believe that this is what we’re seeing right now. And it will turn more ugly and more surprising before it’s over. So, it will be interesting times and there’ll be lots of things for us to talk about every week on the podcast.

Rob, thank you ever so much for doing such a thorough job without coughing, despite having to bite your tongue at times when we discuss certain elements on the podcast today. Great stuff and I hope people appreciate all the preparation that Rob put into this. If you did, by all means go and leave a rating and review on your favorite podcast platform to show your appreciation.

Next week I have another interesting, super insightful guest that used to work actually with Rob, namely Graham Robertson from AHL. So, that’s going to be another fun and very insightful conversation.

If you have some questions for Graham, something that you might want to challenge him about, then by all means send your questions to info@toptradersunplugged.com and I’ll do my best to get them in front of him.

And of course, as you can tell from my dyslexic way of pronouncing some of these words, by all means make them short and easy for me to put forward to him. Anyways, this is it from Rob and me. Thanks ever so much for listening. We do look forward to being back with you next week. And in the meantime, as usual, take care of yourself and take care of each other.

[01:23:44] Ending

Thanks for listening to the Systematic Investor podcast series. If you enjoy 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.

And 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 is 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.