The AI Hype within Investing... a Step Too Far? 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.

[00:00:24] 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 markets through the lens of a rules-based investor. Rob, wonderful to be back with you this week. Hope you’re doing well. How are things in the UK in particular after the Battle of Denmark versus England in European football last night?

[00:00:49] Rob

I suppose on the upside, if you guys had beaten us, then you would have had a very grumpy English person on your podcast this morning. On the other hand, if we’d beaten you like five nil, then you would have been grumpy. On paper, we’re a better team. We certainly didn’t play like it last night, and that’s for sure. I still think we’ll qualify top of the group. But I doubt we’re going to go any further the way we’re playing, frankly.

So there you go. That’s my very detailed analysis of the football situation. But you’re right. The right result for us because it means that you’re probably a bit happier than me, but neither of us is upset, which is good.

[00:01:27] Niels

It’s funny you say that. Objectively, as objective as I can be, I certainly thought that Denmark should have won last night. They played really well, up against, on paper I should say, a better team for sure. To be frank, it was quite surprising. I actually didn’t think Denmark could play like that because they have been so poor for the last three or four years. So a nice surprise, but also a good result.

[00:01:54] Rob

Putting my quant hat on, I think you’re right. I think they did have more chances to score which they perhaps were unlucky with. So yeah, arguably, they deserved to win. But I’m glad they didn’t.

[00:02:05] Niels

Yeah, good. It’s a fair result, a diplomatic one. All right. Okay. Enough about the football, at least for now.

Before we start, always like to ask you what’s caught your attention, what’s been on your radar since we last spoke, nothing to do with the topics we’re going to be talking about later. We’ll also talk about your portfolio later on. But just something that you found interesting, something that’s had an impact on you.

[00:02:37] Rob

Yeah, the markets obviously; the equity markets are what most people follow. I think a lot of hype is coming out of the market a bit. I think Nvidia is now the world’s most valuable company, which I found quite incredible because I reckon if you’d gone back maybe even five years, then only like really hardcore gamers and maybe some people in our industry because the use of GPUs for massive quant computing has been around for a while.

So only a tiny percentage of people would have heard of the company, and most people wouldn’t have a clue. Now they’re the world’s most valuable company, and all on the back of, first crypto, and now artificial intelligence. So yeah, I think that’s quite incredible actually, personally. It surprises me, I would say. Because the thing about chips is it’s one of those businesses that has gradually got more commodified over time.

For example, if you’re talking about like a bog-standard memory chip, those are massively commodified, and basically, there’s only a small number of places that make them and not in huge volumes. The margins are very microscopic. But then even if you move up a level to the core CPU business where you’ve got people like Intel and AMD, their margins are better.

But it’s still the case that for the vast majority of the devices out there, like smartphones, for example, most of the chips are pretty commodified and the margins have come down. So kudos to Nvidia for being a player in a market that’s commodified but still making very decent money, massive growth. Although, objectively, the stock is quite highly rated I would say.

One of the reasons they’re one of the most valuable companies is that they have got a pretty incredible multiple, although not as incredible as Tesla, to pick another kind of hyped company. So that was the fact of the week that really made me sit up and go, wow, that’s actually quite something.

[00:04:41] Niels

It does tell you, or it does show you how unpredictable things are. Who would have talked about Nvidia, in this way at least, two or three years ago? So yeah, interesting.

Well, as I shared with you on my radar this week, it had nothing to do with finance or investing. It was the tough goodbye we had to say yesterday to our wonderful, beautiful golden retriever, Bella, who has been part of our family for the past 12.5 years.

So if I don’t sound as perky as usual or even a bit emotional, there’s a good reason for that. So I appreciate you taking the lead on all the topics we’re going to be talking about today, Rob, since I have been focusing elsewhere.

But nevertheless, from a trend following perspective, what’s going on in your portfolio? I know last time we spoke, you had decided to increase the trend following part of your overall portfolio as far as I recall.

[00:05:42] Rob

Yes, and I remember you expressing surprise. You said, well, why are you getting in at a high watermark? This is an example of me not practicing what I preach because there is a weak mean reversion effect in trend following returns. So if you wait for a modest drawdown, that’s probably a good time to get in. I think what happened was that I increased my allocation, and I was very pleased with myself because then everything went up, and I was patting myself on the back.

But of course, guess what’s happened since? I’ve gone down. So yeah, in late April, I actually reached a high watermark, which would have put year to date would have been 16%, which is pretty good for that. Now with pretty much half the year over, I’m down to about 7.8% I think for the year, which is in line with the indices. I think probably you’ll discuss that in a minute.

But obviously, in retrospect, I should have waited a few months to put the extra allocation in. But market timing is very difficult as we always say, which is why we leave it to our models to do that. So there we go. So just breaking down that P&L a little bit just to get a little bit of color on it. In terms of asset classes, the markets that have done well have been, interestingly, FX, particularly yen has done very well.

Also, yen, both against the dollar and the euro, has done well. Also, Swissy has made a bit of money. I also made a bit of money in equity markets, particularly Asian equities. So my best performing Asian markets are Taiwan and the Nikkei, which obviously is a Japanese market. Agricultural markets are still in the black. Cocoa is still one of my most profitable markets for the year, even though, obviously, I have given up a lot of the really outstanding performance I had earlier in the year, but it’s still in the black for the year for me.

My worst markets are energies and metals where I’ve lost a bit of money. Gas has been particularly bad for me, as has gold and platinum. So they’ve not been especially favorable markets. But there’s no massively clear pattern there. The other thing I like to do is take a forward-looking view and a backward-looking view and say, well, where am I currently positioned, because I think people might find that interesting.

So perhaps unsurprisingly, given that I’m in a drawdown and trends have reversed or weakened a bit, my risk is running at just under 10% annualized standard deviation. So that’s about just under half the long-term target. In terms of positioning within that, I’m long equities and FX as we discussed. I’m short bonds and I’m also short volatility.

My biggest short position actually is VIX. So that’s very much a pro-risk on position, obviously, because if the VIX comes down, then that means that people think that equity markets are getting less risky. So it’s not a surprise that my biggest longs are all in equities as well. So I’m long MIB which is the Italian stock index. I’m long AX, which is Dutch. I’m long the Nikkei, I’m long the NASDAQ, I’m long the S&P, so, a bunch of equity longs there.

I’m still long yen-euro and I’m long cattle. I don’t currently have a position in cocoa anymore. That’s long gone because of the reversal; that got me out of that position. So I guess it’s a reasonably positive beta, pro equities going up portfolio with a vol position in the same direction as well and then a few other positions that are not a strong pattern. So I’m certainly priced in for growth in equities.

[00:09:39] Niels

Yeah, absolutely. Sounds good. As you mentioned, it has been a couple of difficult months for trend followers after a super strong Q1. This month, obviously there’s still another 10 days or so left, but I think fixed income has probably been one of the more challenging sectors so far, as far as I can tell, based on the latest inflation data and of course also the continued hope for rate cuts now that we are seeing a few of them in Europe.

But the big one of course is still the Fed and how long they’re going to hold out for. Of course, we had the European elections, which meant the equity markets are somewhat divided. European equities are struggling, whilst the US markets just keep heading higher. Currency, I think you’re right. I think there has been a big mix, and certainly the metals seem to have been a challenging sector for trend followers.

My own trend barometer is super weak, so in line with performance numbers. It’s down to 25, which is certainly close to the weakest levels we normally get to. Although yesterday, this was at the end of yesterday, but I think yesterday was a pretty strong day for trend followers. Pretty broad in terms of what I saw this morning in terms of trends that are probably in the right direction from a longer-term point of view.

Nevertheless, performance numbers without the benefit of yesterday, BTOP 50 down 2.795% for the month of June, up 6.41% for the year. SocGen CTA Index down almost 3%, up 6.3% so far this year. The Trend Index down 3.57% in June, up 7.69%, and the Short-Term Traders Index down 1.93% and up 78 basis points so far this year.

On the other hand, traditional markets are doing well. MSCI up 2.16% for the month of June, up 11.06% for the year, and the S&P 500 up 3.80% and up 15.53% so far this year. Unfortunately, the World Government Bond Index seemed to have stopped updating at least where I look. So if anyone listening knows where I can get that data, then feel free to send me an email.

All right. We’ve got a couple of questions for you, Rob. Also, we’ve got the topics that you brought along, so it’s going to be fun. The first one is a bit long. I hope I can read it out without making too many mistakes. It’s from Steve. Steve writes, “I hope you’re well. A longtime listener. TTU remains my favorite investing podcast. Do please keep up the great work. Good luck in the Euros, but not this afternoon when England plays Denmark.” So he was hedging his bets there. But one-one, as I said. A diplomatic result.

“I have a question for Rob. I’m not sure we’ve heard Rob talk, in any depth, about systematic buying of put options as a tail risk strategy. Having read Scott Patterson’s entertaining book “Chaos Kings”, which is about Universa and their tail risk fund, I followed up by reading Mark Spitznagel’s book “Safe Haven”. It sets out how a safe haven with an insurance (in quotation 10 back) a payout, will have a negative expected return over the long run because the insurer makes a profit. And yet a portfolio with a 2% allocation to insurance and a 98% allocation to S&P 500 can theoretically achieve a higher geometric return, a CAGR, than the S&P 500 alone, so the insured makes a profit too.

I would be interested in Rob’s insights as to whether this is achievable in practice, as it would appear to be dependent upon a systematic program of buying far out of the money put options because you don’t know in advance when the fast equity crash will occur, and then a monetization strategy when the crash and associated volatility spike do happen. Finally, what factors would Rob consider when comparing and contrasting trend following and put buying strategies? For example, trend is less reliable than put buying in a fast equity downturn. What are likely to do better during prolonged bear markets? Put strategies do have a negative risk premium whereas trend has a positive return. All the best, from Steve.” All right, over to you, Rob.

[00:14:24] Rob

The first thing to say is that out of all the people you have on your podcast, I’m probably in the bottom decile for people who understand tail risk hedging. There’s lots of people who are on the podcast who could do a much better job on this. So I’m going to talk to this from a combination of very ancient experience.

Over 20 years ago now, I was an options trader, combined with some kind of theoretical handwaving. The experience that burned in my brain was when I’d been a junior trader on a desk for about six months, the guy who sat two tables down from me, watching him collect his belongings in a black bag and be escorted off the trading floor by security because he bought some short-dated options rather too expensively and there hadn’t been enough volatility.

So, he just, basically, bled theta and bled quite a lot of theta to the point where he was basically sacked. So that was like, oh, my goodness, be very careful. Buying optionality is, in the long run, a losing strategy because of the variance premium; the insurance premium if you like. But on the other hand, I was also in the situation where I’d sold volatility just before non-farm payroll and was certainly praying that the thing would move. So I’ve been on both sides of the fence there. Selling and buying volatility can both be very dangerous in different ways. Let’s just leave it at that.

Anyway, to move to the theoretical instructions. To be really theoretical, a way of thinking about this is, if you just think about portfolio optimization, then it’s actually possible to improve the expected either Sharpe ratio or geometric return. Both can be improved by adding an asset to your portfolio that has a negative return but is negatively correlated with the rest of your portfolio.

You won’t improve the actual return clearly, because you’re adding something with a negative return. So your actual return will reduce, but your Sharpe ratio will increase, which means you can apply more leverage, which means your geometric return will increase, which is basically what you want. Because geometric return will be the ultimate, I write about it a lot, but it’s the ultimate performance measure. It’s what everything comes back to essentially. So that’s a theoretical way of describing it.

A non-theoretical way of describing it is this, basically. Let’s say that you want to invest in a portfolio, and you can earn returns from a typical negatively skewed trading strategy. That could be something like FX carry, emerging markets carry, but it could also just be buying stocks, because stocks, especially individual stocks, have negative skew. They’re prone to very sharp, potentially strong negative returns.

There was once a 25% down day in the S&P 500 in 1987. There has never been a 25% up day. So it’s a negatively skewed asset, especially in the tails. Let’s say you want to invest in S&P with, don’t I know, four times leverage, which is spicy, and you think well, that’s going to be a profitable strategy. But I’m really concerned about this 25% drawdown day happening. The drawdown is so fast that I can’t derisk my portfolio and I’ll basically lose all my money.

So what you can do is basically buy an out-of-the-money put option that will protect you from a 20% movement in one day. That option is going to be cheap because it’s extremely likely that will happen. But it will also be systematically overpriced because, apart from the fact that very out-of-the-money options have very wide bid-ask spreads on them as well, even abstracting away from that, people over price volatility, but they overprice downside volatility even more because that’s what they’re really worried about.

There are not many people out there in the market hedging themselves against the market going 20% up in one day, but there are a lot of people out there who want to hedge themselves against the market going 20% down in one day. So that market will be systematically overpriced. Even so, although that option side of that trade will be losing money persistently, you will probably make more money than someone who just did that trade outright but without the leverage.

You might end up making more geometric return for the reasons I’ve discussed because you know you’ll be safe, with a massive drawdown happening, because you bought the protection. That gives you the comfort to leverage up, and that’s adding your leverage. Even though your net returns would go down without leverage, you add leverage and you’re back in same place.

So that’s a non-theoretical way of thinking about the effects going on here. This weird, counterintuitive thing where you’re buying something that you know is going to lose money. But because it’s giving you the comfort and protection of knowing you won’t get a massive drawdown, it’s allowing you to get a higher return. I’m a bit more skeptical of people who advocate buying tail-protect funds for non-leveraged investors.

Let’s say you’re an investor with a long-time horizon like a university endowment, it’s a classic example, and you’re not using leverage. I’m a bit skeptical as to why you need to buy tail protection, to be honest, because you’re not only going to see the costs and not the benefits because you don’t really need to protect your portfolio against drawdowns because you’ve got that long horizon.

Because you’re not leveraged, you’ve got no risk of margin calls. Why buy the tail protection? It’s just persistently losing money. It will reduce your returns, but it won’t improve your outright returns or your geometric returns, and that’s what matters.

Now, there are some people who say that they can do a better job of this tail hedging than just buying outright out-of-the-money put options the whole time because they can buy them when they’re, perhaps, systematically mispriced. They can use some conditional factors to try and time when they should buy them.

Yeah, it sounds plausible. Again, there are people who’ve been on the podcast who can talk about this stuff with a lot more knowledge than I can. But at the end of the day, any tail protect manager that tells you they’re going to give you their insurance policy for free, I’d be a bit skeptical. They’re either, A, a snake oil salesman, or, B, the protection you’re buying isn’t worth the money because they only selectively buy put options, but you may end up being unprotected just when you need it. Given the whole point of buying the insurance policy is to be protected, what’s the point of that?

So, yeah, that’s my take on it. I think it’s something that makes sense for leveraged investors in a modest way, clearly. If it’s going to cost you 2% of your portfolio to defend yourself against a 50% drawdown, that seems to me to be quite reasonable, because with additional leverage, you can definitely improve geometric return. But if it’s going to cost you 30%, then that’s a different story.

[00:21:53] Niels

I guess the challenge, Rob, is that these 50% drawdowns, for example, if that was really what people were worried about, they don’t happen very often, not even every 10 years. So if you’re paying 2% every year for 10 years, it becomes meaningful. But then as Steve was also alluding to, then there is this other option, which is of course the trend following part, which is a positive return strategy. But of course, it doesn’t have the same reaction pattern than just buying a put.

I remember a paper that Alan and I discussed a while back. I think it was Meketa that had written the paper where they described options strategies as first responders in a crisis, and then trend following was like the second responder, which is kind of what Steve alludes to, because if there’s a prolonged drawdown, then trend following tends to do well, but it may not do well in the initial phase. That’s more random. We can’t tell.

You mentioned maybe put options is not a bad idea if you have leverage in your portfolio. What about trend following? If the specific goal was to get some “protection”, is there a specific type of investor that fits better than others?

[00:23:13] Rob

I’m a little bit uncomfortable with the idea of putting trend following and these out-of-the-money put option buying strategies in the same bucket as protection because they’re doing very different things. The main difference of course is if you add, using historic data of course, blah, blah, blah, if you add trend following to a long-only portfolio, you can improve your geometric return without adding extra leverage because it has the positive expected return there and it’s uncorrelated, or very low correlated I’d say, to traditional markets.

So you can stick 10%, 20%, 25%, we talked about these numbers before, 30% of your portfolio in trend following, and your Sharpe ratio will improve, but your geometric mean will also improve. You don’t need to apply additional leverage. For me, it’s a completely different use case in a sense.

What I would then be wary oof is, let’s say you start off with a 60/40 risk parity portfolio that probably is already leveraged because it’s a risk parity portfolio, and then you add some trend following into that, and you add a big slug of trend following say 30%, it’s quite plausible that your risk is going to fall. I can’t do the math in my head, but your risk could easily fall from doing that, and you then say, oh, my risk is even lower now, I’m going to add even more leverage.

So now you’ve got this thing where it has got loads and loads leverage in it, and then a crisis comes, and your 60/40 does badly, but your trend following also does badly because it’s just the wrong kind of crisis. These things happen, and there’s no guarantee of protection unlike a tail protect situation. So now all of a sudden, you’re sitting on a bunch of assets that are all going down at the same time with massive leverage. Well, you don’t need me to tell you how this particular story ends, and it’s not pretty. So that would be my concern.

In the same way that I don’t think it makes sense to buy tail protection if you’re an unleveraged investor, on the other hand, it makes perfect sense to buy trend following if you’re an unleveraged investor. You see what I mean? They are different cases. Leveraged investors should buy tail protection but should be wary of buying trend following if they’re going to then use that as an excuse to apply even more leverage.

So that’s kind of almost the two ideal use cases for, on the one hand, tail protection, and on the other hand, trend following. A person who is leveraged should probably think about tail protection. A person who isn’t leveraged should probably think about trend following because they’re doing very different jobs.

As I said, I don’t think you should buy the tail protection if you’re unleveraged. It doesn’t make sense to me, because you’re not really getting the benefit. Conversely, if you are leveraged, yeah, by all means, buy trend following, of course. What else am I going to say on this show? But be just very careful of using that as an excuse to whack up your leverage even more. So that kind of summarizes it for me.

[00:26:22] Niels

I think that’s very well put and some really important points. It’s also why certainly over all these years, as much as I love the idea of crisis alpha, I’ve always said it’s dangerous to describe trend following as a crisis alpha strategy because no crisis is the same and we don’t know how we’re going to react. So I like the idea of just a non-correlated return strategy. Of course, we know, theoretically, that’s a good thing to blend with say equities and bonds.

All right, let’s move on. There’s a question in from Luis. It is regarding how to use carry in your strategies when you roll from a contract to the next one. “Imagine you’re trading a contract where you hold a contract that is not the nearest, and there is enough data to compute carry for different maturities. When you roll, should the roll carry of the previous held contract be used in the computation of the current forecast? I mean, the values that you put into your Exponential Moving Average smoothing should only refer to the currently held contract when it’s possible to compute. Depending on the shape of the curve, carrys of different contracts can be very different between them, and the EMA takes time to adapt. I’ve been playing with introducing carry rollovers in the time series of some markets to avoid this effect. I think this is a different issue from the ones discussed on the chapters Accurate Carry of AFTS.” That’s your book.

[00:26:22] Rob

Advanced Futures Trading Strategies.

[00:27:20] Niels

Exactly. Written by Rob Carver. Go and get the book everyone listening to this podcast today. So I did the plug for you, Rob.

[00:27:44] Rob

Yeah, thank you. As the rule, every time I’m on the podcast, there should be at least one plug for my book. So it’s good to get it out of the way early. Yeah, that’s a quite detailed and technical question. I’m going to try and explain to probably 8% of the listeners who haven’t read that chapter of my book who are head-scratching their heads, I guess. Let’s take a very concrete example.

So, I just quickly pulled up, because it’s always quite fun to look at often the term structure of the VIX futures contract, and lo and behold, it’s doing something weird. So if we look at the curve between now and the October delivery for that, it’s upward sloping. So the spot VIX is about 13.4, the NEX futures at 15, and then it goes up to 16.5 and then October delivery is just under 19.

That’s one of the reasons why I’m short VIX because of this roll down effect, this carry effect if you like. This curve is basically predicting that if you buy any of these futures, it’s predicting they’re going to go down in price, which is another way of saying, as we discussed in the previous question, that generally speaking, selling volatility is a profitable strategy because of the difference in expected and realized vol.

So that’s kind of normal if you like. But then something interesting happens. If you look at the November expiry, actually the December expiry as well, they’re actually lower than the October. Not a lot. I’m seeing November 18.1 and December 18.

[00:29:45] Niels

We know why of course.

[00:29:47] Rob

We may well discuss that later in the podcast, so let’s not get ahead of ourselves. But let’s just assume that we’re pure systematic traders who don’t know anything about what’s going in the world outside. Apart from watching football on TV, we never watch Bloomberg or CNN or anything like that, and don’t follow any of the news. Now, what that means is that if, let’s say that this weird pattern can persist, which, given what we know, it probably will, and you’re in a position where let’s say that it’s late August and you’re holding the October VIX contract.

If you look at the carry on that, it’s still very dramatically telling you should have a short position because basically vol is expected to decline. But once you roll your position, so you move forward to holding the next contract, which is the November contract, the carry on that is very different. It’s telling you the price is going to go up. This effectively is what Luis is talking about.

It’s actually related to another similar issue, which is that some people say, and I don’t think this is generally done anymore, but if you go back 25, 30 years, most people when calculating say momentum on a price would look at the actual contract they were trading. They would pull up the chart of say April, say of June VIX or, say September VIX, and they’d look at how September VIX had behaved, and based on that, they would draw their trend lines or calculate the moving averages or what have you.

What that means of course is when you roll from holding the September VIX to holding the October VIX, at that point, you need to essentially recalculate what you’ve done and say, well, now I’m looking at how the October VIX has behaved because it’s a different thing. We’re treating it as a different thing. I’m not sure many people do that now. I think it’s more common to do what I do, which is form a single back-adjusted price series of continuous contracts and then do your calculations on that.

Because one obvious implication of that is when that roll happens, you’re suddenly changing your mind on what’s going to happen, which seems a bit weird. It doesn’t seem logical that you should have a different position on for September VIX that you do for October VIX. They’re basically the same thing. So using that back-adjusted price also means that when you do that roll, it won’t actually affect your forecast, and therefore your positions to a first order of approximation anyway.

So what Luis is essentially saying is well, the carry, you should look at the individual contracts. Personally, I think this is one of those things I put in a bucket of we’re making an approximation here, but it’s going to make almost no difference to my position. Do I really care? Implementing what Luis suggested would make my system a lot more complex because it would have to have knowledge of individual contracts.

So in practice, what my system actually does, as Luis kind of alluded to, is essentially it would have quite a strong negative carry forecast going into that role, and then that negative carry forecast as it’s being smoothed out, my moving average will effectively reduce and then would turn positive, only modestly positive because the steepness is not that big. This doesn’t really affect too many things. VIX is exceptional.

It is more of a problem potentially in things with a very strong seasonality component like crude and gas for example, because obviously, the seasonality means that potentially you get different carry at different times of the year as well. But yeah, I’m relatively relaxed about this. I personally don’t think it’s worth worrying about too much. Of all the things to worry about, this is pretty low on my list, I would say.

[00:33:42] Niels

Good stuff, Rob. With that out of the way, and nice to hear you’re not so concerned about this part, maybe you are a little bit concerned about the next part, which is to do with elections and systematic traders. I can sense the energy coming into you already. So where are we going to go with this topic?

[00:34:02] Rob

Well, I know. The thing is of course there is an election going on in the UK at the moment, as I think everybody knows. Also, I read somewhere that I think this year, the highest proportion of the global population is taking part in some kind of election because obviously there were the European elections, the European Parliament, there’s the US election, which is why of course the VIX is spiking in October because of uncertainty around that, and there’s an election in India or has been election in India, and we’ve got an election in the UK.

[00:34:37] Niels

There was one in Russia as well, but I’m not so sure it was very uncertain.

[00:34:41] Rob

I think the uncertainty there may be quite small, I would suggest.

So, elections are kind of interesting as a trader because they do lead to uncertainty. The uncertainty could be for different reasons. One thing that’s quite depressing about the UK election, for me at least, is that a lot of the dialogue is around how much money each party is going to raise in tax and then spend and whether these numbers balance.

It’s a very kind of primitive economic view of things. It’s like John Maynard Keynes never lived really. The fact that often governments are spending more than they earn can be a good thing, for example. I’m one of those people who believe that if the UK had a more procyclical fiscal view since 2010, then we wouldn’t be in the situation we are in now, where, obviously, Brexit has not helped, but where a lot of economic indicators point to that our country has done pretty poorly.

So I think it’s quite a depressing reductive way of expressing government spending. One of the reasons why it’s happening is because of course of our previous but previous Prime Minister, a lady called Liz Truss, who did a pretty terrible job of communicating her economic plans and as a result nearly caused the entire UK financial system to disintegrate overnight, which is quite impressive.

So everyone’s now trying to model themselves. The Tories, the Conservative Party, the ruling party is saying, “Well, we’re definitely not like the previous Prime Minister. I know we’re the same party, but we’re definitely not like her.” Whereas of course the other side, the Labor Party is saying, “You guys are definitely like her.” So they’re both desperately trying to come across as the most fiscally prudent, which is depressing, reductive, and ultimately I think wrong.

Anyway, I think our election, from a trading perspective, is quite boring. Like the Russian election, we kind of know who’s going to win. It’s going to be the Labor Party and they’re going to have a pretty substantial majority probably. It’s not as quite nailed on as the Russian result I expect, but the polls are pretty convincing.

The other reason why there’s less market uncertainty around this is because I think the markets generally, because of this reductionism in the dialogue, generally think that the Labor Party and the conservatives, there’s not going to be that much difference as far as the markets are concerned, at least initially. So it’s not like some party is going to come in with a very radical plan to change everything, at which point everyone will be slightly panicking.

To get theoretical for a second, there are essentially two things that can affect markets in terms of political risk. One is the effect it has on what we would call the equity risk premium, so how much investors expect to be paid to own risky assets. Obviously, if things are going to change drastically, irrespective of whether that change is good or bad, then that may make markets potentially fall.

But there’s a slightly more subtle point as well, which is that given that essentially when you’re buying an equity, what you’re actually buying is a call option on the enterprise value of the firm. So I do remember some of my economics degree. It’s buried back in there somewhere in the back of my monkey brain. That could mean that volatility could actually be good, potentially. And then, of course, there’s the fact that various firms would expect to do well or badly out of a win by a particular party.

But again, because the policies here are so similar… So perhaps more interesting then is to talk about the US election, where we have a combination; a perfect storm of A, unlike Russia, unlike the UK, the result is very finely balanced. Every kind of poll predictor website, everything I’ve looked at, that’s independent at least, has said this thing is on a knife edge, it could go either way.

I think that’s partly because like the UK, the US has a system where you can lose the popular vote or do very badly in the popular vote and still become president, as happened in 2016 and could well happen again this time. I think it’s still quite likely that Joe Biden will win the popular vote, but I think it’s still 50/50 as to whether he will win the presidency. So we’ve got this kind of knife-edge situation.

But the other thing we have is that, unlike here, where this new party are going to come in, and everyone’s like, wow, it’d be pretty much the same as the last lot. It’s not going to make a massive difference. We can be quite relaxed. If the other guy wins, he could do some crazy stuff. In 2016 when he was campaigning, a lot of people were like, “Well, yeah, he sounds like a crazy guy. But once he gets in, he’ll be like a technocrat and he’ll be surrounded by these sensible technocratic type people. He’s a businessman and it’s going to be fine.” No one believes that anymore.

I think everyone believes that if the crazy guy does get in, he’s going to do crazy things. If there are any adults in the room around him, he’ll either not listen to them or he’ll fire them. There are lots of people individually in the finance industry in the US who probably would like to see Trump elected because they probably believe that he’d be better for their personal bank balances, which is fair enough if you’re going to vote on that basis.

But if you look at how the markets are pricing the potential of a Trump victory, I think they’re pretty scared. I think that VIX chart shows it. So that’s how to think about elections as a trader. Now we come to how to think about it as a systematic trader. Because the question you potentially get asked a lot, I know I do, is, well, what are you going to do with your system with regard to risk X? Whether it’d be the Brexit referendum, or election, or even down to next week’s non-farm payrolls, what adjustments are you going to make to your system? The correct answer is, Niels?

[00:41:20] Niels

Nothing.

[00:41:21] Rob

Nothing. Exactly. Yeah, nothing at all. I think it’s particularly hard. The more these things prey on the brain, the more you think about it, and you can tell I’ve been thinking about them a lot, the harder it is to step away from it and say, well, yeah, but that doesn’t mean I should do anything. It’s all very well to be interested in this and read the news, but it doesn’t mean I should actually do anything, because it’s really, really hard, really hard to be a political macro risk trader, shall we say, because you’ve got to do a lot of things.

Firstly, you’ve got to try and predict the outcome. Okay, well, that’s easy in the UK, it’s easy in Russia, very hard in the US to predict the outcome at the moment. It’s very, very evenly balanced. The second thing you’ve got to then do is to work out what the implications of that particular result will be. That’s a bit easier.

You can potentially look at the manifestos, look at the track record and say, well, it looks like this will be very good for this industry, very bad for this industry, looks like this would be good or bad for interest rates because we think that policy will be inflationary or disinflationary. Even if you don’t want to do that work, believe me, there’s lots of people working in investment banks who are pushing this stuff out every day telling you what the economic implications of Macron losing, or Sunak winning, or Trump winning will be. Don’t worry, there’s stuff out there you can read.

The third thing you’ve got to do, which is perhaps even the highest of all, is actually predict what the market reaction will be to the news. That may have nothing to do with the kind of theoretical whether it will be a good thing or not. So if I think back to Brexit, for example, I expected the pound to depreciate against the euro when the vote result came out, and that did happen. What I hadn’t really expected was the euro/dollar also moved quite a lot.

You can argue, well, isn’t this neutral for Europe? The UK is not a big part of Europe. Economically, I think one of the mistakes we made when we made the vote was we thought we were indispensable to the Europeans and they would give us this great trade deal. Of course, sadly, that’s not true, as it turns out. Actually, the euro/dollar also moved quite a lot, and that was unexpected for me at least. I hadn’t seen that coming, and I think a lot of people had.

So you’ve got to do these three things or two of them that are really, really hard. But you’ve also got to do them better than everybody else does because it’s no good predicting a foregone conclusion. No one’s going to give you any P&L for that. That’s not going to happen. So, personally, I think if you do think you can predict political outcomes, then there are websites, certainly in the UK, where you can bet directly on the political outcome. I’m not sure whether they’re legal in all countries so I’ll put in that disclaimer.

So yeah, if you want to do a little bit of political gambling on the side, and I have got a few bets on the UK election, I don’t mind telling you. I actually bet on Biden winning last time around as well quite early on. That was quite a good bet. But overriding your systematic trading system because you think you’ve got a gut feeling about the way the election is going to fly, maybe not.

[00:44:45] Niels

Actually, just for the record, I know you said that in the last election, when Trump got into the Oval Office, that a lot of things changed dramatically and policies and so on and so forth. But actually, for the record, a lot of policies have been maintained under Biden, so you never know what you get with these elections.

All right. So from a political, I wouldn’t call it rant, but from a political topic to something that also seems to be something that you have, say, a passionate interest in.

[00:45:25] Rob

A passionate disinterest more like, an anti-interest.

[00:45:29] Niels

An anti-interest, I think that’s a good point, and that’s the AI hype. We already touched a little bit on Nvidia, but this is not about Nvidia. But you came across a paper that Global Asset Management did where they talked about the growth of the asset management industry.

They talk about how it reached $120 trillion in 2023. Despite that growth, the industry is facing some challenges such as fee compression, rising costs, increased competition from passive funds, and so on. Then it shifts to the topic of AI. As I said, I didn’t have much time to prepare for this week, so tell me a little bit about what “excites” you about this paper.

[00:46:23] Rob

Well, actually, fee compression is another topic hopefully we’ll have time to talk about briefly as well because there’s been a more interesting paper about that. No, I didn’t mean to pick on the good people of Goldman Sachs specifically. There’s a lot of this stuff around. Actually, my apology, it’s not Goldman Sachs. It’s Boston, was it? It was Boston.

[00:46:48] Niels

I had written down Global Asset Management report, but of course it doesn’t say who actually wrote it.

[00:46:54] Rob

This particular report I think was done by BCG, so it’s a risk management consultant. I don’t want to pick on BCG in particular. But in the same way that I had a rant about crypto, I really feel I’m overdue a rant about AI. I think my timing is quite good because I am starting to see a bit of skepticism creep in to my Twitter feed about AI.

Also, if we look at the share prices not of Nvidia, but if we look at other companies that have done a very strong pivot to AI, they’re not doing so well actually. This particular bandwagon hasn’t proven to be quite the road to riches that they perhaps hoped. I guess I’m probably most well-qualified to talk about AI in the context of systematic trading I would say, and there’s been an awful lot of hype about it there as well.

Of course, the interesting thing is that when people talk about AI, they are talking most often about GPT, about large language models specifically. That’s because I would suggest that the average person on the street doesn’t have access to, for example, a neural network or a TensorFlow model or something like that on their desktop, and they have no idea what those things are or how they work.

But a lot of people have access to these large language models on their desktop, and indeed, they’re being put into all kinds of software and products often without people’s permission or liking necessarily, or knowledge. Yeah, exactly. So people have access to them, but they still don’t understand or know how they work. So they’re just using them purely as consumers.

In some cases, they seem like magic frankly because they can produce stuff that’s quite remarkable at times. Often, they don’t. Some of you may know I do a bit of part-time university lecturing, and as part of the course I teach, I set an assignment to students and I say to students, “Do not use chat GPT to answer this assignment because you will get a very low mark.” And they say, “Well, is that because you’ll know we’ve cheated?” I’m like, “No, it’s because it won’t do a very good job because this assignment, for what it’s worth involves analyzing some returns data from a strategy and comparing it to a couple of benchmarks, is kind of AI proof.”

It’s not a kind of open-ended essay question where you can just waffle. Sure enough, one student did clearly use an LLM to write their essay, and yes, I did give them zero because they didn’t actually answer the question. They just trotted out a lot of meaningless platitudes, basically. So let me be clear. I do think that there are, I would call, weird, unusual nonlinear linear patterns of financial data which are very hard to detect using classical statistical techniques.

Indeed, there is a very successful business that has done better than any other hedge fund on the planet in this field, and that is Renaissance Technologies. No one really knows what they do, but it’s probably that. Because they aren’t doing what everyone else does. If you look at the sort of people they hire, people who for example have worked in AI related fields like speech recognition, then that’s clearly where they are.

But that is very different from being able to use an off-the-shelf large language model to comb through some financial data and find some predictive effect that isn’t there, particularly as someone who does not really frankly understand the technology that they’re using. The chances of you overfitting and producing an effect that doesn’t really exist are just so high, and you’ve got no way of knowing that you’re doing it.

It’s not like a technique where you can very clearly see, okay, this is the fitting process I’ve done, these are the parameters that have come out, here is the statistical uncertainty on those parameters, here is where I’ve used in-sample data, out-of-sample data, and so on and so forth. It’s very unlikely that you’ll get that kind of rigor using the LLM; that you’ll get that kind of understanding of what you’re doing.

Now, I didn’t use an LLM for this, but I recently did an exercise to demonstrate to myself, I’ve not written a blog post about this, because it’s actually something I’m doing in Jason’s consultancy work. So I can’t share the exact details of what I did because it’s covered by an NDA. But very in rough terms, what I did was use a fairly simple way of essentially classifying time periods into whether the price went up or down.

So I ended up with a series of up or down price movements, and then what I did was essentially use some fairly basic fitting techniques to look for patterns in those up and downs. So a pattern might be something like if you get six ups and a down and then an up, that’s bullish for the price, for example. What I found quite interesting was the effect that happens if you basically increase the granularity of what you’re doing.

So you’re looking over a bigger and bigger set of potential patterns, and then you compare your results in sample and out of sample. You get this very, very striking effect. If I could work out how to, at some point, I’ll probably animate it as a JPEG. Basically, you can imagine, so the simplest possible case is you just look to see whether the previous period, the price has gone up or down, which is trend following.

Assuming the effect comes out as yes, after an up move, you should buy, after a down move, you should sell, that’s a trend following model. And you look at the out-of-sample and in-sample performance of that, and they’re very similar because the model is very simple. Also, it’s not bad. It does pretty well, actually, because it’s a simple trend following model. And as we both know, Niels, simple trend following models tend to do pretty well.

Then you increase that to 2, 3, 4, 5, 6, 7, 8, 9 levels of granularity, so now the number of patterns you get increases exponentially. So you start off with two possible patterns and you get two squared, which is four, two cubed, which is eight, and so on and so forth. You get up to 10 and you’ve got 1024 possible patterns. Two things happen.

Firstly, the in-sample performance improves to the point where roughly at about say seven, eight periods, it looks like a straight line account code. It never loses money. Simultaneously, and rather unfortunately, the out-of-sample performance gets worse and worse.

It’s a really nice way of illustrating the effects of overfitting. So I’m definitely going to find a way of sharing that with people once I can get out from under the cover of this NDA.

So that’s what’s going on. But the good thing is, of course, you can see it happening. It’s really clear and obvious. That’s what’s happening. What happens with any more sophisticated fitting technique, whether it be a large Langen model or something else, is you could be doing this bad thing but not even realize it. That’s the very specific issue I have with the use of LLMs for systematic trading.

We’ve given people get very sophisticated tool, and 99.9999% of them have no idea what they’re doing with it, frankly. That sort of adds to my general ‘grumpy old man’ level of grumpiness about people pivoting to AI and sticking AI here, there, and everywhere. when I go to Google and ask it for search results, I don’t want it to try and guess what I’m trying to work out and give me a made-up answer. I want it to just tell me what the best website is, for example.

[00:55:16] Niels

Yeah, it is a really important debate and it’s a very important debate for our industry. I think you kind of touched on it a little bit at the very end. The thing that I’ve thought about is, we know that it’s talked about as some kind of next big step, technology that can make things better in our industry, better performance because now we have all this wonderful technology and it’s getting smarter and smarter, and so on and so forth.

But what I was thinking about is, what about the liability? What about the liability of a firm using AI in its model and then the result comes out as being not what they hoped for or expected? Who’s liable here? So I think it’s certainly something to be used with caution. Maybe there are some operational efficiencies that can be had for sure.

But to allow AI to take an active part in code, and certainly without complete supervision and transparency so that whoever runs that system or model knows exactly what it’s doing and why it’s doing it, which has always been the beautiful part of trend following is that we put in some rules so we know exactly what’s coming up. We just don’t need to compute it every time ourselves. But the liability issue with using AI, I think we’re going to hear more about that.

[00:56:54] Rob

Definitely. If your Tesla runs someone over and kills them, who goes to jail? Is it you or Elon Musk? It’s someone working for Elon Musk more likely. Yeah, definitely.

[00:57:06] Niels

True. All right. Maybe very shortly before we wrap up, Ted Seides, a previous guest on the show, he has of course his own podcast, Capital Allocators. He comes from the allocator side, and he wrote a little blog post, or at least I received it by email, where he talked about this debate concerning performance fees and whether managers should be paying incentive fees on certain parts of the return such as interest income, for example, and so on and so forth. What are your thoughts on this?

[00:57:45] Rob

Obviously, my thoughts come from the other side of the fence, so to speak, because I used to be on the other side of the fence as someone who was allocated into rather than an allocator, so to speak. The conversations about fees are the most difficult in that kind of two-way relationship, that’s for sure. I think the main issue about Ted’s blog was I think he was kind of saying that, in a sense, the allocator’s bark is worse than their bite on this subject.

He’s basically saying that you guys are talking a big game, but actually, when you look at who you’re investing in, a lot of the people you’re allocating to are still charging quite hefty incentive fees and incentive fees that don’t have a hurdle rate. So it’s not like you’re only paying performance over, I keep wanting to say LIBOR. It always used to be like fees were calculated.

[00:58:42] Niels

Risk-free rates, let’s call it that.

[00:58:43] Rob

Risk-free rate. Exactly. Yeah. Some of them reference interest rate as we no longer have LIBOR of course. So essentially, he’s saying, well, you guys are talking tough, but actually, you’re not really willing to take your money out of these managers who are still charging these high incentive fees. So where’s the negotiation?

There’s no negotiation there. It’s like saying to your kids, “If you don’t clean your room, I’m not going to give you any pocket money,” and then saying, “There’s your pocket money.” The kids are like, “Well, okay. Fine. I’m not going to clean my room.” I speak as the father of an 18-year-old whose room frankly should be declared a super fund site. It’s so messy. Fortunately, he’s moving out soon to go to university and it’ll be somebody else’s problem.

The other point he makes, of course, is very valid. No one really cares what fees they’re paying, they care what net performance they get. So if you are lucky to be an investor in a fund that’s long since closed and is charging massive, massive fees because they can get away with it, and I’ll refer you back to our example of course, Renaissance Technology, then you’re going to just keep and paying those fees, and you’ll pay them very happily because your net return is still fantastic, irrespective of whether the risk-free rate is 50 basis points or 500 basis points. It’s still extremely good what you’re getting out of them.

All allocators essentially say, “Oh, yeah, yeah, we’re really tough on fees.” But if you look under the hood, actually, they’re quite happy to carry on investing in managers whose fees are, on paper, much higher than they would actually be willing to pay in practice. I guess where that makes things interesting or difficult is when it’s a new manager and the allocator says, “We’d like to invest in you but you’re a bit expensive.” The new manager can say, “Well, but you’re investing in these other guys who are much more expensive.” But of course, they’ve got the track record.

I always say that costs are known and returns are unknown, or costs are known with very high precision, returns are estimated based on a backtest with high uncertainty. So given the choice between two brand new managers, well, of course, you’re going to pick the one with the more competitive fees in the absence of any other information because you know for sure you’ll be getting a higher net return, assuming they have the same performance.

If you’ve got no information about what the performance will be, then of course that’s a decision you will make. I guess what we are going to end up with is a two-tier system where anyone who started a fund in the last few years is going to have much lower fees than the more established players, until or unless they can end up with a track record where they can potentially then try and raise fees, which is tricky because raising fees for existing investors, particularly their seed investors, is another one of those very, very awkward conversations.

So yeah, it was a nice little blog article. It’s unusual to see under the hood of an allocator’s decision. So that’s what I found quite interesting and quite fun with a little bit of empirical research as well. So, nice blog.

[01:02:07] Niels

I agree with that. I think in general, certainly in the last 35 years that I’ve been part of the CTA industry, let’s face it, our fees have come down quite a lot. I don’t think they’re unreasonable. Obviously, some people are trying to focus on super low fees nowadays to be their way of getting attention and raising assets. I’m not in that camp. I don’t believe that to be the right thing to do for our industry because it does cost money to run these strategies.

As you said, it’s probably not a bad thing to look at net returns at the end of the day. But even the larger managers who have long track records and great returns, fees have come down to I think reasonable levels.

[01:02:56] Rob

Yeah. When I was working at AHL, some of the fees on some of the legacy funds we had were just eye-popping, things that have been launched in the early 90s when interest rates were a lot higher and were the business model, in terms of paying trial fees to advisors and things like that, was still a big part of things and some of the fees were very, very high.

Even while I was there, as those funds aged out or were replaced, the fee margin was coming down, and it’s come down since. I think the other point, I make this point every time we talk about fees, you should always look at fees adjusting for volatility because you may think that a manager with say a 25% vol target which is quite punchy, you may grumble about paying them say, I don’t know, 2 and 20 off the top of my head.

But remember, that’s basically the same as paying 1 and 10 to a manager with half the vol target. Given that often you’re comparing CTAs with relatively high vol targets with say equity long-short funds in another part of your portfolio which often have very low vol targets you know, as an allocator, you should always compare them on a like-for-like basis after adjusting vol. Once you’ve done that, the CTAs often end up looking quite cheap.

[01:04:13] Niels

Good stuff. Well, if those are our parting words, CTAs end up looking quite cheap, I think that’s a good place to stop. Rob, I appreciate all the hard work to get some topics this week in particular.

If you like these conversations, feel free to go over to iTunes, Spotify, Amazon, and tell how great Rob is and leave a rating and review.

A little bit of an update, in the next couple of weeks I will be away, but in my absence, Alan will be hosting Nick Baltas and Cem Karsan on the podcast, so you can look forward to that. If you have questions that Alan should pose, then send them to the usual email address info@toptradersunplugged.com and I’ll make sure that he gets them in time for the recordings. I think the first one up is Nick next week and then followed by Cem the following week.

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

[01:05:21] 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.