256 Systematic Investor Series ft. Mark Rzepczynski – August 12th, 2023

[00:00:00] Intro

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

[00:00:23] Niels

Welcome and welcome back to this week’s edition of the Systematic Investor series with Mark Rzepczynski and I, Niels Kaastrup-Larsen, where each week, we take the pulse of the global market through the lens of a rules-based investor. Mark, wonderful to be back with you this week. How are you doing?

[00:00:38] Mark

Good. Always good to see you too. We’re in August, which is supposed to be a vacation month, but I don’t know if August has always turned out to be a vacation. It seems as though we always get hit with August surprises. I was having a discussion with some of my colleagues, and they said we don’t know whether August is going to be slow. Then someone else said some people say you should sell in May and wait until October, and then others have said you never want to trade October and you don’t want to trade December. The next thing you know, we came up with the reasons why we shouldn’t be trading for at least 9 to 10 months out of 12. So there never is a good or bad time for trading or vacations.

[00:01:24] Niels

Okay, fair enough. You know of course that historically, and this is very interesting, completely off topic but now that you bring it up, just speaking about trading (whether you should trade or whether you just invest and invest for the long term), I don’t know if you are aware of this and I’m not sure whether this is something that is also a rule in the US, but in Europe, at least, I think if you’re a broker, you have to disclose the percentage of your clients that lose money on their accounts and you have to disclose that on your website.

I came across this point listening to a podcast and reading a book recently, and sure enough, you can go to some of these brokers and find it. The percentage of clients that loses money in their brokerage account, it’s north of 70%, often it’s north of 80%. So maybe they should take your advice, not trade in October and not trade in any of the other months and just buy and hold. Who knows?

[00:02:28] Mark

But that opens up a bigger issue which maybe we’ll have time to talk about. Would you prefer a lot of small gains, for let’s say 10 out of 12 months, or would you prefer being let’s say down on 8 of 12 months, but then the four were very big months? This is fundamental to when we talk about trend following or even hedge fund trading. Which one is better? Are you worried about your total return at the end of the year or what is the process in which you’ve made that money? I think that most people always would prefer to make a little every month as opposed to losing more than 50% of the time but making bigger gains overall.

[00:03:19] Niels

Yeah.

It is quite educational, I would say. If you collect say 10 different track records of some of the managers that’s been around for 30, 40, or even like we at DUNN have for almost 50 years, and you just simply look at the monthly return table (because it is exactly like you said), there are probably six or seven losing months a year for sure in every single year, some years will have more. Then you have like one or two months where not a lot happens to the upside and then you have these one or two or three months per year where it’s a nice, healthy, good month, and that’s just how it is, not necessarily every single calendar year, but for the most part. But that is the challenge we are dealing with. That’s the challenge where we’re battling.

People can’t quite get their head around the fact that this is the “uncertainty” that they get, although they get certainty in terms of most of these strategies on a rolling 3-year basis, 5-year, 10-year for sure, are actually profitable.

[00:04:30] Mark

Right.

Are you interested in total return or are you interested in a smooth return that might be lower? I probably would say that the life we lead is an attempt to try to maximize total return. Albeit that means that you’re going to lose more hairs, lose more sleep, get more gray hairs because you might have more than 50% down months. In some sense, that’s part of a trading life.

[00:05:02] Niels

Yeah, very true.

Well, that was a little bit of a sidetrack, but I enjoyed it. We’ve got quite a good set of topics, as I always say, in my unbiased opinion of course, that we’re going to be tackling. But before that, I always love to hear what’s been on your radar, what’s been catching your attention in the last few weeks since we spoke last?

[00:05:23] Mark

Well, I guess what’s been catching my attention is that there’s not been anything that is really catching my attention. I think that there are a few things that really stand out; this China deflation slow growth. I think that you may have read in the newspaper that Jeff Currie, who is the analyst at Goldman Sachs, recently quit. He was a big advocate for the first supercycle and he’s been arguing for another supercycle. While most trend followers in commodities… You can’t say that you can “exploit” the supercycle but generally, if we have a strong upturn across most commodity market’s yield, it’s a little bit easier to make money from the long side than the short side. So a supercycle is a good thing for a lot of commodity traders.

I don’t think that that’s going to happen. You look at the Fed, we’re still not sure what they’re going to do. So that’s become a non-issue. You look at the soft landing story, at the beginning of the year, we thought that for sure we’d have a recession for the second half or near the end of 2023. Now it’s being pushed to 2024. In fiscal issues, we got a downgrade by Fitch from AAA to AA-plus. There should be no surprise there.

At the same time, the Treasury is going to issue close to $1 trillion dollars in this quarter and close to another $800 billion in the next quarter. So fiscal issues are a problem, but these are all creeping problems. When you look at shorter-term trends, that really doesn’t manifest in these creeping issues, supercycle, China deflation, fiscal issues, soft landing.

So we’re in an environment where there are non-issues which are driving trades. You see that in performance after the March bond debacle for a lot of trend followers. Markets have been fairly stable. We’ve been moving slightly higher, but there hasn’t been strong performance in the last three to four months.

[00:07:42] Niels

Yeah, very true. However, I did notice that Jeff Currie had left Goldman Sachs. Jeff was on the podcast only a few months ago, I would say. I have to say I like the way he explains his thesis. I know it hasn’t played out in the major commodities just yet, like energies, like a lot of people are focusing on. Not to make that as part of his thesis, but I do notice that there are some commodity markets, the softs, where some of them actually have made multi-year highs and are still making multi-year highs at the moment. But it doesn’t feel like a commodity supercycle because we need the metals, we need the energy markets, and maybe we need the greens as well to participate. But I still wouldn’t rule it out. We’ll see. Time will tell. Interesting stuff.

Now, let’s quickly just deal with the trend following part like we normally do. This week, as you alluded to in your comments, not a lot is going on and performance is treading water at the moment. My trend barometer is weak, down at 34 as of last night. That indicates a negative month for the CTA industry. That’s exactly what we’re seeing at the moment. BTOP 50 down about 1% for the month and 1% for the year. SocGen CTA is down about 75 basis points for the month, down about 1.8% for the year. The trend index down 1.4%, for the month down 2.85% for the year, and the Short-Term Traders Index losing about 80 basis points this month and down 3.8% this year. The backdrop of that is that MSCI World equity index is down 2.6% as of last night, up 14.6.

Bonds still struggling, down another 48 basis points in August, and the S&P 500 is down 2.6% as of last night and up 16.4% this year. But you’re right, there’s not a lot really going on in the markets enough for trend followers to have big swings in their performance.

Now, that actually leads me to the next topic, the first one I just want to throw at you, Mark. Because when nothing happens, performance is relatively quiet, which in some ways actually is beneficial to investors because people like certainty, they don’t like the uncertainty and all of that. So they’re certainly at the moment not getting a lot of uncertainty in their portfolios. Where I’m going with this is that investors sometimes change behavior. I think one of the times where we can think back of this happening was actually during the period where interest rates went to zero.

I think a lot of investors who used to be very happy with being bond holders changed the way they invest and they were “forced” to move into other investments like equities just to get some kind of return. Now interest rates are at 5% again, or more actually, in many countries. How do you see this? Do you think investors will change behavior again? If you do, in what way do you think they might change?

[00:11:10] Mark

Well, this is probably one of the top research topics for anyone involved in global macro and it should also be for those in trend following. When you think about the post-financial crisis period, that was the period, we’ll call it the reach for yield. So the reach for yield was that I’m not getting any money on my cash so I’m going to move out the curve, I’m going to buy risky assets, I’m going to choose riskier fixed income securities relative to treasuries. I’m going to constantly reach for new products and reach for yield. That period is over.

So now we’ve gone back to a completely different environment where not only is nominal yields higher, but the real yields which were negative for an extremely long period of time (depending on how you look at or measure inflation). Let’s just look at one-year inflationary expectations or just current inflation.

The real yield is now positive, which changes completely the behavior; that we should change the behavior we are seeing. So instead of the reach for yield, we should see the reach for cash. Why should you hold risky assets if you could already get a real return on your money, your cash money? Why would you want to hold a risky asset when you can get over 5% in the US and T bills? Not only is that a massive change in how people will perceive their investment universe, but also, you have an inverted yield curve.

So why should you actually go out the yield curve when you could hold cash and your return on cash is actually pretty good versus even moving out the Treasury curve? So there’s no need to do this. When you think about it, this should create a tremendous amount of caution in investors to say you’re going to have to prove to me that I’m going to get a higher return for me to move out of cash.

Now, if you told me that we’re going to be at 5% cash, you couldn’t tell me you’re going to have an inverted curve. If I asked you, what would happen to equities in this environment?

Niels, I’ll assume that you’re every man. As the representative of every man in Europe what do you think you would do or what would you think if I told you this information at the beginning of the year?

[00:13:57] Niels

Well, you probably wouldn’t have guessed that equities were doing so well so far this year.

[00:14:03] Mark

Right. Exactly.

You’d expect a change in behavior, you’d expect caution, and yet what we have is that the QQQs, if you look at that, is up significantly. If anything, it looks like it’s an overcrowded trade. You look at large cap, which is more tech, which should be more interest-sensitive because it has more equity duration, that’s done really well. We’ll say for the rest of the S&P, so if you look at the RSP which is the equally weighted S&P 500, it’s underperformed versus cap-weighted. So in some sense, there is a little bit of caution but you would not expect that we would have a strong equity market so far in 2023 given those facts.

So now what you say is, how quickly does behavior change? I would say we haven’t seen the change in behavior associated with higher interest rates. But when you think about it from futures trading, bond trading, FX trading, why would you want to take extra risk if you could make 5% nominal on your money? So now what happens is that should create caution. What caution means is that there’s going to be less speculative behavior. At some level, we always talk about that you need speculators to offset hedgers. You need more flow into risky assets to be able to balance off the other demands of other traders in the market. We probably won’t have that.

So, consequently, we should have a slower speed of adjustment in prices which might actually be good for trends, but we haven’t seen that either yet. So right now, I think that this is one of the key questions that we’re going to have to face. But as of yet, we haven’t seen the expected change in behavior given the higher yields and the inversion of the yield curve. A perfect example is that given that an inverted yield curve always (I’ll use the term in quotes), “always” predicts a recession, and now we don’t even know sure if that’s the case.

Can you imagine if we said here’s the inversion of the curve, not only in the front end, but along the entire length of the curve, we’ve been weighting this for almost well over a year, we’ll still say that there’s no idea exactly when that recession will occur. Now, maybe we’re trying to fit a story that just doesn’t exist and it will come, but this market environment doesn’t seem consistent with what we’ve seen in the past when we’ve seen inverted curves or higher real rates.

[00:17:05] Niels

Yeah, I would agree with that. I like the way you discussed the issue. I was kind of thinking in a slightly different way. I was thinking that, well, if you can get 5% risk free, you would certainly want to expect more from your other types of investments. Let’s look at that. What does it mean for trend following?

Well, as I’ve discussed with some of our other co-hosts in the last couple of weeks, well, actually, that 5% on cash, you can add that to your trend following return because a trend following fund will mostly have 70% to 80% of the money in cash. So during the 2010s, we were earning nothing. Now we’re earning say 4% on the total portfolio from interest rate income, so that’s already a boost to the return so those returns should be higher. That’s fine.

But the other thing I was thinking of (and I don’t want to make this too long a point, I want to get to your topics), but the other thing I was thinking of is if you lift risk-free returns and people start to expect or think that they want more than just 1% or 2% (which I think people got kind of got used to or satisfied with only a few years ago), maybe they’re also willing to take a little bit more; maybe they’re willing to accept a little bit more volatility to get to 12%, 13%, 14%, 15% returns, which is obviously what a lot of these trend following funds can produce over the long run, albeit with higher volatility for sure.

[00:18:49] Mark

Well, you’ve opened up a big Pandora’s box here. The big Pandora’s box is that when people look at track records for CTAs and managed futures and you look at what the benefit of cash has been on track records over the long run, it’s been fairly significant. So in the ’80s and ’90s when we had higher cash rates and higher real rates, you added that into your returns. It gave all the funds a good boost. So when you look at everyone let’s say post the financial crisis, there were a lot of skeptics who said that trend following is dead, no one’s making money, this is horrible.

At the same time, let’s assume that you were charging 2% and 20% and let’s say that the nominal rates were 4%. Well, that means you started out the year with a positive 2% return on your cash, we’ll just make adjustments for how much margin you use. So when rates were zero, then basically you had a 2% management fee. You’re 2% in the hole from day one because that was going to be that huge drag that you had nothing to balance it off so you had to trade out of that hole to begin with.

Now, at some level, that’s not huge. But at the same time, that would put a big drag on the performance of all of the types of futures traders that hold high cash balances relative to the amount that they use in margin. So this does have a big impact on track records. You always have to think about, what is the trader’s return ex of all of cash (because that tells you what their real skill is).

[00:20:49] Niels

Yeah. Actually, on that point, at the firm I work, we started removing that income from our composite track record back in 2007 because it wasn’t really our return. So when we show our numbers, we don’t include that extra in the program returns. You have to do it for a fund of course, so those returns will be higher. But you’re right, it’s an important point.

I want to pivot to something else, but it’s linked to this. Because you and I often discussed that investors don’t like uncertainty. We know that. As humans, we don’t like uncertainty. Back when interest rates were about this level or higher (this is some time ago), and certainly when you and I were working in the industry, a lot of funds grew their AUM by creating guaranteed products. Some of the biggest firms in our industry were masters of structuring these guaranteed products where they would take $100 and they would put $70 or $75 into a zero coupon bond yielding 5%, 7%, whatever, and then that free cash, because they knew they had that free cash for another five, seven years, that was pretty much enough to trade the program or the trend following strategy at full speed, almost at least.

I think that was a very popular product even though you could argue there’s no magic in it, you can do it yourself, you just put 70% of your money in a seven-year bond and you do the rest and you put it in trend following. But do you think because of the way investors behave and their preferences, now that we have real returns, would you expect these products to make a comeback?

[00:22:36] Mark

Well, there’s two answers. One is that we know that investors hate uncertainty. A lot of the drive for those structured products was when you have higher nominal interest rates because then what you could do is you could create the guarantee at a relatively cheap price. Because, in some sense, you could buy a zero-coupon bond and if let’s say it was 70 cents on the dollar, then you could take the other 30 cents and use that at margin and then be able to trade on this. So you could create products that give you the guarantee.

We still can’t be able to do that because I’ve looked at whether we can do constant proportion portfolio insurance or structured product. I was looking to do something like this for an insurance company. We’re still not at the levels, and it’s hard to structure it. People want that product. It’s also the problem of structuring it and whether you can actually deliver the product. Now, the key, when you think about uncertainty, is that uncertainty is not the same as volatility. Going back to the environment, when we talk about all of the things that are going on in 2023, you’d say, where would you think volatility would be at?

You look at the VIX, it’s actually on a fairly low level. It’s not at the lows and it has a very positive skew. But we’ll say that volatility in the marketplace is not very high on a measurable basis versus history. But then if you look at uncertainty, there are different ways to look at macro uncertainty; there’s policy uncertainty indices. Those are actually much higher. So there’s actually a discrepancy between the uncertainty in the market and the volatility in the market. Those two can’t be maintained because if you have policy uncertainty, it’s because we don’t know what choices are going to be made. It shows that there’s uncertainty in the newspaper.

Well, that uncertainty is going to have to be resolved and that’s going to cause markets to move. So people like to deal with the uncertainty. Can we offer products? Hard to structure. On the other hand, what you find is that when people do face uncertainty, actually, that they slow their behavior. So the adage is always that if there’s high uncertainty and if you’re an investor, well, then dollar cost average. So if I dollar cost average, then I’m going to get the average price over a period of time as opposed to saying, oh, gee, I bought in at the high and I sold in at the low.

Now, what happens is that if more people are dollar cost averaging across a number of different asset classes, what do you think will happen to the speed of adjustment? Probably say it will slow. If the speed of adjustment slows, what does that mean for finding trends or the existence of trends? We’d probably say that there will be a greater chance of an existence of trends. So high uncertainty, low volatility, or the perception that there’s high uncertainty will actually cause people to behave differently than if we had low uncertainty and similar volatility.

[00:26:18] Niels

Okay, cool. All right. Well, with that backdrop, let’s dive into some of the topics that you kindly brought along. As usual, I’m not entirely sure where we’re going to go on these, but I have some bullet points that I can share and then we’ll see where we go. The first thing we’re going to talk about is the macro-micro, not sure if it’s the environment or certain issues, and you mentioned a few of them, but I’m going to give you the mic and see where this leads.

[00:26:48] Mark

You always have to think about markets as being forward-looking as opposed to backward-looking. We’ve been in a period of relative stability on the macro side. But at the same time, we still have a lot of issues that still haven’t been resolved, whether it’s the China deflation story, the Fed story, maybe we’ll get some better idea after Jackson Hole, the soft landing story. All of those when you think about when we come back from Labor Day, those would be the three issues that are really going to play to a lot of the fears or concerns of investors. Along with that is going to be the fiscal issue because there’s just so much debt that’s being issued. At the same time, we’re still being engaged in quantitative tightening.

That being said, what we’re really also facing is a micro environment, which in some senses, we see that can have a big impact when the macro environment is more unclear. A perfect example of this is at the end of the second quarter. There was a lot of short covering by hedge funds and a lot of de-risking. We’ll probably say that that didn’t play out in market averages. But if you’re involved in individual stock trading, a lot of those structural issues of what other hedge funds are doing, the flows that they’re putting into the market or taking out of the market has actually had a much bigger impact on performance than maybe an investor would think about when he only thinks about top of the market.

What really that leads to is an area of research that I’m trying to focus on, and that is what we’ll call dispersion trading, and not dispersion as a trade, but the dispersion of markets and how that will influence your return potential. Now, we’ll probably say that post-financial crisis was the period of crisis alpha. I’m a little early to say this, but we might want to think about now dispersion alpha. What do I mean by that? Dispersion is the distance between individual markets relative to an index. So in some sense, it’s how diverse the return profile is for a set of markets we trade.

When you think about it, if you’re trading a set of markets, what you really want to have is you want to have more dispersion because then that’ll give you more opportunities both on the long and short side relative to a less dispersed environment. So we constantly want to think about if we can measure dispersion, can we then say something about what will be the ability for us to generate extra returns. Now, dispersion is different than correlation. Markets could be correlated, but that doesn’t mean that they’re going to be more or less dispersed because we’re looking at the actual return as opposed to the covariance to the two variants.

We do know that if correlations go higher, then dispersion is going to go down. But dispersion is what we actually want to trade, and that dispersion is what is going to generate us return for a portfolio.

[00:30:31] Niels

Yeah. I guess it goes back to this divergence-convergence discussion to some extent.

[00:30:39] Mark

So this is sort of like another level of thinking about convergence and divergence. What we really want to have is that if we trade a lot of markets, what you want to have is you want to have divergences on a more localized level. When we started our discussion, we talked about supercycle for commodities. We’ll say that softs are having some very big gains or highs. You look the at corn market on the other hand, corn has had a significant decline over the last couple of months. We see oil is actually going higher.

So in some sense, we have a lot of different dispersion that’s going on, and I think that that’s going to have an influence and what our return potential is. Another way to think of this is that I could have the overall market relatively stable, but if I have more dispersion, I could still be able to make money as a trend follower because it could be a matter that I’m short, such as in corn and I’m long crude. I’m long the softs. Overall, if you look at a commodity index, it doesn’t seem like there’s a lot going on, but under the surface there might be a lot of activities.

[00:32:06] Niels

All right. Well, let’s move on to the next challenge, so to speak, because you were talking about, in your comments to me, you talked about ex-ante and ex-post problem for quant research. So I would love to know where we are heading with this particular topic.

[00:32:24] Mark.

Well, I’m going to switch a little bit because the one thing that I found that was sort of an interesting piece of research is this whole idea of motivated reasoning and biases. That’s what we’re going to talk about, this ex-ante and ex-post. It’s interesting that this is more from the psychology research. They find that it takes more information to make you believe something you don’t want to believe than something you do believe.

So, if you believe something, then you’re always going to be looking for confirming evidence. You’ll say that evidence makes perfect sense. Then we’ll say that the stuff that doesn’t fit with your belief patterns or what you would like to believe, well, then you’re going to say, “Wow, okay.” You’re going to come up with a number of questions for why that doesn’t apply. I think that actually has a big impact on how we look and how we actually do research.

On the one hand, if you believe in discretionary, fundamental trading, it’s going to be pretty hard to convince someone that a quant strategy is going to be a better approach. So he’s going to constantly looking for evidence or that piece of evidence, let’s say the one quant that doesn’t do well. He’s like, of course, that proves my point why we want to be a discretionary trader. Similarly, when you think about it, if you’re a trend follower, then of course you might want to say, well, I’m always getting believe in trend following, I’m going to have to be pretty strongly motivated to switch my point of view.

Now, that doesn’t mean that you’re wrong. It just means that the level of information you’re going to need necessary to change your mind is going to be higher. In some sense, that’s why we want to be quants because in some sense, you can be able to use numbers to be able to take out (we’ll call it) this motivated reasoning bias. Because if you say I look at one strategy versus another strategy, I can say, well, what do the numbers tell me? If one makes more money, then you can say that it’s going to be hard to refute. That doesn’t mean motivated reasoning won’t try, but then it’s going to make a big difference.

Now, that being said, someone did comment when I started talking about this that there could be motivated reasoning bias for modelers themselves, and I think that that’s true. But I think that what we want to try to do as researchers is how can we be agnostic or as agnostic as possible? So that means that you say like, how can you say there are objective measures to say whether one model is better than another? Now, how does this all fit in with ex-ante or non ex-ante trading?

What I found is that what people do. and how they approach research, is very different and it’s good to understand how people approach a real research problem because that will determine exactly what kind of answer you’re going to get or how you pursue research. So one we’ll say is an ex-ante. You hear about a story, you see some phenomena, and then you say, can I form a testable hypothesis? So let’s take something that we did just talking a little bit earlier.

We talked about the inversion of the yield curve and we said that this should have a big impact on performance, performance meaning it should change people’s behavior. In some sense we can say that that’s very testable. We can be able to say let’s look at all the periods when there has been inversion of the yield curve. Let’s then see whether there has been a change in return patterns across a number of set of markets. If the answer is yes there have, then we could say this is a factor or feature that we want to try to add into a model.

If the answer is no, then we can say, well, we tested it, we found out whether it’s countable, and then we move forward and then we might say let’s just use our existing models. So from a trend following perspective, the hypothesis would be inversion would have some type of impact on trendiness. Then we could say, can we be able to test that? If the answer is no, we may say there’s no reason to change our existing behavior. If the answer is yes, then the question comes in, how do we adjust our behavior?

Now, the trend follower would probably have a different view. He’d say what I’m going to do is to say I don’t know whether these features or factors can have an influence and it really doesn’t matter because I’ll just follow my trend models and if there’s more trendiness during the inversion, well, I should pick it up and I’ll see higher performance. So I don’t need to test any alternative hypotheses. Ex-post, I could say, well, I made money, let’s look back and form a narrative on why I made money. But I’m not going to test ex-ante some type of feature that we think that might be influential.

So in some senses, the key to research is, can you come up with something that’s measurable, can you count it, and is there enough count that you could have that it could be meaningful? If you can’t count it or it’s a non-countable type event, then the whole idea would be to say, well, then we’ll just say things happen and you should stick with what you have.

So there are different ways of doing research of saying that I’m going to test features and look at where they’re accountable, or I say that I can’t discern what those features are going to mean for markets, so therefore, I’ll follow my existing strategy and then maybe use that as a narrative after the fact.

[00:39:27] Niels

This is obviously a very familiar topic also to this podcast because as you know, some of our best friends in this industry have decided to stay with what they’ve done for the past three or four decades and not deviate away from that philosophy that they were taught. What you’re saying here, I hear it, is that actually, as researchers, we should continue to be open-minded. I think the challenge, and I think it’s a fair challenge, that is if you have something that has worked for you for a long period of time and you therefore build a trust and belief in the process, it can be as a human being, I guess it can be very hard to want to change.

People in general don’t like change and so I think it does pose a question. Therefore, I wonder, deep down, if some of the changes we’ve seen in our industry in terms of research, in terms of how people do things, is basically based on the fact that they are bringing in other people with different backgrounds and they have maybe more the freedom of trying new things and then maybe potentially coming up with something that can convince or compel the approach to be changed.

[00:41:01] Mark

Right. In some sense, if you know your limitations, you might say I may need to have other people join my firm who have a different point of view, that can actually have a different perspective. It doesn’t mean that you have to change. It just means how can you be able to reduce or eliminate the potential for motivated reasoning and become more agnostic on what are the choices you’re going to make. Now, ultimately, this comes down to an issue, I call it the Goldilocks model or Goldilocks modeling issue. If you remember Goldilocks and her porridge, some were too hot, some were too cold and you’re always looking for just right.

The same applies to models. Some might have a lot of features, some might have few features. So how do you find the set of features that are just right? In general, there’s probably two schools of thought. One, we’ll call it the keep it as simple as possible. In the vernacular sometimes, people will say that that’s the KISS method, keep it simple stupid. I like to say you want to keep it sophisticatedly simple. Either way, the more features you put in or the more conditions you put in, then the greater likelihood that there could be a particular failure or there could be the potential for overfitting.

So we’ll say a traditional trend follower is that I am truly worried about overfitting issues. Given my fear of overfitting, I want to keep it as simple as possible. The other view would be to say I think that there are a lot of conditions that I could add that could potentially add value because there are a lot of exceptions that occur often enough that I could adjust my behavior accordingly. So now that becomes an issue. Well, which one is right? So no different than Goldilocks. Sometimes you like hotter porridge and some people like colder porridge. So that’s a choice that you have to make. But at the same time, you want to try to keep things as simple as possible.

[00:43:47] Niels

Now, the big issue that you really wanted to discuss, I’ve kind of kept you talking about other things so far, but there’s something that I think that was more on your mind this week, and that is something that may sound boring, but it’s incredibly important in our industry, and that’s transaction costs, something we don’t actually discuss very often on the podcast. So I am very interested in where you want to go with that and what are your thoughts and findings on this topic.

[00:44:19] Mark

Well, next to accounting, transactions cost is probably the topic that’s the most boring and people are least interested in wanting to have a discussion of, yet I’m going to venture into the area of the mundane and the boring with transactions cost. The reason why is I’ve been obviously working with my new partners. We’ve been building some new models and we’ve been talking to investors, and so we’re showing backtested results. When we show backtested results, Niels, you’ve seen those discussions. How do you think those usually go?

[00:45:00] Niels

Exactly, yes.

[00:45:02] Mark

So there’s a certain amount of cynicism. Most investors will say, I’ve never seen a bad backtest in my life. Then what they’ll do is they’ll say, whatever backtest you show me, my rule of thumb is that I’m actually going to cut it by 50%. So if you show me a Sharpe ratio of 1.5, I can assume in reality, it’s going to be 0.75. If you show me 10% returns on an annualized basis, I assume, if you get a little bit above 5%, then I’m going to say, yeah, that’s what probably I’m going to expect. So you go in there and you have the argument. You say, no, we’ve actually done a really good job of doing our backtests and we’ve accounted for all the costs associated with trading.

It’s almost as though they don’t really believe you. You still have to get at this issue of how do you get people more comfortable with backtesting or how do you actually account for that properly in your models. The issue is actually not trivial, because in some sense, when you look at the difference between live performance and theoretical performance, you could say that there’s a number of reasons for why the two could differ. One, it could be that you didn’t really account for transactions costs, so your live performance is going to be lower than your theoretical because you didn’t properly account for that. We’ll talk about what that means in just a second.

Second, it could be structural, so what you think you can trade. It’s not possible. That could be that if you’re shorting stocks, can you actually be able to short it, is going to be hard tomorrow, is going to be on the restricted list. So there are issues of structure. That could also include, for example, you want to trade a certain size, but some markets that have very little liquidity it’s not possible to be able to do it. The third would be the difference between when you trained your model in one period versus what the live period is going to look like.

Let’s say if you trained your model when interest rates were zero, and then you’re going to trade it during a period when interest rates are high, like 5%, would you expect that there would be a difference in performance between those two periods, between the live and we’ll call it theory or your training period? The answer is yes. So in some sense, that’s another way that you could affect your backtesting. But in the near term, that one’s a little bit harder to deal with because you have to then look at what history might provide and what the future might look like. But the one thing that you can control is the transactions cost.

There are two ways to do it for your backtesting. One is get a good estimate of what your true cost would be. That’s one thing. But then the other thing is, how do you trade less? If you trade less, what happens to your transactions costs or your trading costs? They’re going to go down. That’s a choice you could make inside models. So, the simplistic view of transactions costs is you say, well, I just assumed that I have to pay brokerage and then a bid-ask spread and so what I’m going to do is I’m going to nick every price on the entry and exit based on that fixed cost assumption. That’s going to lower my expected return, which will then influence my model performance. But that doesn’t really get at the heart of the issue, which is what are the total cost of trading a given strategy?

So what I did was to try to come up with what I call a checklist of what things investors should look at and what managers should look at to try to account for or deal with transactions cost when you build models. A universal view is that these are things that you should be aware of when you start to think about the cost structure and what the impact on what live versus theoretical would be. One would be, we’ll call it the strategy rationale itself. So that could be is it behavioral or structural what I’m going to actually generate or what I’m going to be able to profit from?

That also means if we call it to strategy rationale, are you trying to exploit long-term or short-term returns? So in general, if it’s a short-term model, it’s going to have higher costs and the costs are going to be more important than a longer-term model. So consequently, you have to say I’ve got to be more transaction conscious if I have a shorter-term model than a longer-term model.

Second is, what is the target universe? It could be that if I trade large-cap stocks, my transactions costs are going to be relatively low. If I trade an equity universe which is includes more small-cap stocks, well, then that trading universe or target universe is going to have a big impact on what my cost structure would be.

In the futures world, you can think about if you’re trading only the most liquid futures markets, transactions costs aren’t going to be as important. But we see that there are a number of managers who want to trade the less liquid names. If you trade the less liquid names, by definition, you’re going to have to be more cost conscious, the transactions costs are going to be more problematic, and the potential for structural issues like limited volume and your ability to be able to access the liquidity when you need it is going to be a much bigger issue.

Next, we’ll call it data input assessment. Whenever you think about transactions cost, you don’t really think about data. But data is becoming more expensive, especially if let’s say that you’re going to use alternative datasets. It also becomes much more of an issue if let’s say you’re a macro trader or trader, and data is subject to revisions. That’s a cost that you’re going to have to take into account or look at. Next when you think of this checklist is you have to think about the testing or the sample sizes that you use. If you have a shorter-term model, that means it’s going to have more trades.

You may not have to go back in history or your sample size may not have to be as large as if you have a longer-term model because you want to have a large number of trades to actually trade to look at to try to determine whether your model has been trained properly. You want to also have as longer-term backtests as possible because then you’re going to cover a lot more environments. At the same time, if you cover more environments, then you could see that your profit could be driven by periods that may not exist currently or may not exist in the future or are less likely to occur in the future. So more is not always better than less.

Now, the next issue, when you think about a checklist for backtesting and transactions cost, is how do you limit your transactions? Some models, they focus in on how do I reduce the amount of rebalancing? If I rebalance less, then I’m going to cut my transactions. So my overall cost structure is going to be less. Others are going to say, well, one way I could cut my trades is to have bands or thresholds. I won’t trade unless I cover the cost of trading on my trades.

One way that I’d say a lot of trend followers would look at it is let’s say a price goes above the moving average (on a very simple case), well, does it have to exceed by a certain amount? Is there a band before I actually trade that I need to get through before I would actually create a transaction? So others are looking at liquidity filters; this is that I filter out those names that I don’t think are going to be very liquid because the cost of trading is going to be greater.

On transactions costs itself, when you look at backtests, a lot of people if there’s a longer-term model will say, we always trade on the close. In reality, you find out that you may not actually trade that in production on the close. You may want to trade on the open, but then you have opening auctions. There’s sort of a cost associated with trading on the close. So how you set up the backtest is going to have an impact on what your live returns would be if, let’s say, you don’t trade them the same way.

Also, I think instead of trading at a certain price, more and more are trading on a TWAP or a VWAP basis where they’re looking at the weighted average price that they can get. If you’re going to trade that in a live version, then you got to do that in your backtest, otherwise, you’re going to get a lot of slippage.

Finally, I think that my little checklist is what is the production processing. I think that a lot of people when they build models, they think of themselves as being in a laboratory. That’s great. But when you trade a system, then you have to think like an engineer, because in some sense, you’re making a production process.

How does your production process vary from what you’ve done in a backtest? If the two of them don’t match up, well, then of course, you’re going to have let’s say tracking error between live and theoretical. So all of these things are important.

I guess I’d say that I’m on a mission to try to spend more time talking with investors to say, how do you add more realism in backtests, so that then the cynicism that we see in backtesting views by investors, or we’ll call it the haircut that you see can actually be reduced? Now, that doesn’t mean that a backtest is going to be representative of what you could do in the future, but you want to be able to say at least we’ve taken into account all of the factors and our production process is going to match what we’ve done in backtesting, so that people could be able to say, I can’t guarantee what the future might hold, but I do know it’s not going to be driven by transactions cost or production processing.

[00:56:55] Niels

Yeah. I think that’s actually a really important point. First of all, I know this may sound not as one of the most sexy things we do in trend following, but it is very, very important because every model that goes live comes from some kind of backtest, and therefore, we need to make sure that what we do is realistic and that we can replicate it in real life. But of course, what you also pointed out is that this is just one of those parts that goes into developing a model that actually you do believe will do well in the future, which is another can of worms we won’t go into today.

But it comes down to again these things about how much are these being influenced by historical data and so on and so forth. But yeah, it’s fascinating. It’s a fascinating journey. It is still fascinating to me that some of these concepts were developed so many years ago, decades ago, and that they still work. Because some of the questions we often get on the podcast, and that I also see discussed on other podcasts that I follow, and that is this thing about, well, the markets keep changing, how can these models continue to work, you would have to go and change them all the time? Actually, that’s not true. The concepts are very, very robust and adapt, so to speak, which is obviously one of the reasons why I have so much conviction in this strategy even after quite a long time in doing this.

[00:58:30] Mark

Well, that’s actually a good topic for next time we talk. How do you make a robust model? We’ll say that, if you’re using for example a rule base that’s sort of digital, okay, price is above a moving average (I’m simplifying), that’s going to be fairly robust because you’re not measuring any parameters, you’re just setting an on/off switch. So we’ll say it’s a digital decision, let’s say different models that actually look at factor risks, where they sort of sort based on some factor, and you say, I’m going to buy the top decile and sell the bottom decile, those are fairly robust. Now, that doesn’t mean they work, but they’re actually robust.

When you look at a regression model, then you say, well, there is the parameter of uncertainty because that has to be estimated. So what is the period that you estimate and will that estimated period then match the future? Traditional people who look at digital decisions in general are going to be more robust. That doesn’t mean that they’re right, but we’re going to say that they’re not going to be as influenced by parameter uncertainty.

On the other hand, the people who do regression, they might get better nuance and behavior, but then you constantly have to worry about, okay, is my estimation of the reaction function of people’s behavior correct today relative to what it was in the past? So it’s a different type of risk. But what we want to try to do at least with our transactions cost checklist is to say, how do we take that part of the backtest off the table and then have the discussion and focus on other things?

[01:00:35] Niels

No, no, absolutely. Well said.

Mark, as always, this was very, very good. Nice to catch up with you and thank you so much for bringing these topics to our audience. Next week, Cem will be back on the podcast so I’m sure we’re going to be tackling some topics from the world of volatility of which there seems to be very little off at the moment, but we’ll see. Of course, if you have any questions, as usual, you can email them to info@toptradersunplugged.com.

Make sure you go and follow not just us here, in terms of the hosts on Twitter and LinkedIn and wherever, but actually also check out the website. We actually post quite a lot of stuff now in written form as well. So why don’t you go and check that out before the week is over?

Anyways, that’s it for now. From Mark and me, thanks ever so much for listening. We look forward to being back with you next week. Until next time, take care of yourself and take care of each other.

[01:01:42] 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.