213 Systematic Investor Series ft. Richard Brennan – October 8th, 2022

[00:00:00] Richard

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 Investors Series with Richard Brennan and I, Niels Kaastrup-Larsen, where each week we take the pulse of the global market through the lens of a rules based investor. If you’re new to the show, I hope that today’s episode will trigger your curiosity enough to check out the back catalog and listen to the past episodes that you may have missed like my conversation with Alan, last week, where we discussed what causes trends, and how trend following may handle a soft, hard, or even no landing, and why we may still be at the beginning of a commodity supercycle. 

I would also encourage you to listen to the midweek episode where Hari spoke to Michael Gayed about his approach to the markets based on indicators like the lumber to gold ratio. It also turned out to be a very raw account of what it has been like for him going through 2022, which he describes as hell. I’m sure he’s not the only investor who feels this way. 

If I can plug next week’s episode, on Wednesday with Annie Duke, this is an amazing episode, particularly if you like systematic investing strategies as there are so many lessons from her new book Quit that we can use in our world. So, head over and check out these episodes when you have done listening to Rich and I today. 

Anyway, Rich, great to be back with you as always. How are you doing? How are things downunder? 

[00:01:45] Richard

Really good, Niels. It is slowly getting warmer here, but we’ve had a sort of an extended winter. Yep, it’s starting to pick up over here, so it’s getting more and more humid as time progresses. But, it’s all good. I’m keeping my eyes out on what’s happening over in Ukraine. I’m a bit nervous about what Russia’s plans are. Yeah, it’s still an uncertain environment, and the trends are doing very well at the moment. 

[00:02:11] Niels

Yeah, definitely. We’ll get back to that in a second.

Now, in terms of a market wrap for this week, there’s no doubt that the headline is higher rates for longer. That was the concise message out of the Federal Reserve this week after an attempt at rallying on Monday, both stocks and bond prices rose with quarterly rebalancing and short covering. Markets, again, were overwhelmed by the Fed’s message by the end of this week. Now, the week started out pretty strong for the S&P 500, up about 5.5% at the end of Tuesday. The yield to maturity on the two-year treasury notes finished at a low point around 17 basis points lower at 4.09%, in terms of interest rate. 

Now, these rallies were mainly driven, in part, by the third attempt (essentially) at the narrative of hopes that the U.S. Central Bank is on the cusp of slowing the pace of rate hikes, even now with the UN weighing in this week to put pressure on Powell and company, maybe not so fast as all the Fed members are on the same page and advertise the same resolute message, namely higher rates, and higher rates for longer to squash inflation. 

By close of business Friday the S&P was still positive for the week, although Friday’s ferocious selloff, it raised about 3%, and the two-year note was back to its quarter end high of 4.3%. Now the catalysts were many, but U.S. non-farm payroll data, released on Friday morning, will get the most attention for sure. Payroll employment figures came in slightly above expectations and the labor force contracted a touch.

The result was that the unemployment rate was 3.5% for the month; an improvement from 3.7% reported in August. Although employment gains have slowed from the blistering pace of 2021 and early 2022, a number of 263,000 payroll is still way above trend growth. So, we conclude that despite the tightening of financial conditions since the beginning of 2022, employment remains firm. 

Now, for bond investors, this year has been the worst drawdown, so far, in a broad fixed income market, on record. At the end of September the aggregate index now is down 15%. Now, perhaps there is more to come of that because the Reserve Bank credit dipped by a healthy $45 billion from seven days ago, leaving the Feds holding of interest bearing assets at $8.73 trillion.

Now that’s down $ 190 billion since late March where it peaked. So $45 billion in a week is not bad. All eyes, next week, will be focused on inflation and retail sales data as well as the first of the third quarter earnings that are being released. So, the question is, of course, has the financial tightening been enough or will the Fed need to use both feet on the brakes? Only time will tell.

Anyway, Rich since you were last on, what has been your focus either in terms of market moves, maybe in terms of some of the performance that we’ve seen coming through the grapevine, and of course, The Battleship. I hope it’s cruising along well. 

[00:05:37] Richard

Certainly is.

So, September was a great month for The Battleship. We managed to pay some very good gains in that month. Notably because my portfolio is focused more towards currencies and has less exposure in bonds. The things that were standouts for me was this continuing trend in the yen pairs such as the U.S. dollar/Japanese yen and the Swiss/Yen pairs.

The Euro/U.S. dollar, of course, we’ve got a nice kick with the collapse in the pound with my currency. Yeah, it’s a magnificent time. At the commencement of this month (which we’ve only had about eight or so days of), I did get a bit of volatility there, but fortunately because of my loose pants models they weren’t thrown out of the trends.

We’ve found that they’ve reverted back now, back into those trends that have been preexisting for quite a few years now, for many of them. So, everything is still going along as per normal. I just noticed how a lot of the trend following programs are posting some of their record periods year to date, I think historically. I’ve got a number of large trend following funds posting magnificent returns for this calendar year, which is really an extension of the last two to three years. We’ve had this magnificent trending regime, and it’s particularly pronounced this year.

I don’t know how long this is going to go. None of us know how long this is going to continue, but we certainly make hay while the sun shines, as they say. 

[00:07:21] Niels

It’s interesting, isn’t it, because, there’s obviously been a lot of talk about whether these simple strategies still work or has market structure changed? Are markets moving more quickly now because information is disseminated so quickly and everyone has access to it, and so on, and so forth? 

In that sense, I actually find it quite interesting to see that we are witnessing, perhaps, the strongest trend following year, maybe even on record, but certainly among the top three (I would say), probably throughout trend following history, more or less (top five, top three doesn’t really matter), but certainly up there. So, what does that tell you, in a sense, if you think about the fact that, okay, from the seventies, to the eighties, to the nineties, so much has changed in the past 20 years. It is so different, really, compared to back then yet performance is obviously as good as, or maybe even better than, what we saw back then. What does that tell you? 

[00:08:38] Richard

It’s very interesting to me because I noticed that those trend following programs, with a very long term track record, haven’t necessarily adjusted their strategies significantly over that 30 year period.But now we are witnessing this explosive growth. 

It’s almost saying to me to place less importance on the back test, or what history has presented, provided your models provide this upside, and you cut losses short, you will enter regimes we’ve never seen before which can certainly produce a bonanza in these days. You’ve got to praise these trend following programs that have had the persistence and the patients over 25, 30 years of not tampering with their models. You see the benefit of this, maybe 30 years down the track, of not tampering with their models where they suddenly achieve the best returns they’ve ever achieved in historical terms.

So, it’s very impressive to me, and it just shows how robust the method is. Despite the desire to want to change things (you might think that markets are changing), I think there are these golden rules that you need to stick with and be very patient with over the very long term period.

And it’s the market that delivers your bounty. It’s not your finesse in changing your system rules, or the intricacies of system development. It’s more the patients and persistence to wait until the market decides when to deliver its bounty. 

[00:10:20] Niels

The other thing that I think is interesting, because, as you mentioned, there are many firms doing really well this year. But we’re not the same. We’re all a little bit different. We trade a different number of markets. We trade different markets. Some people don’t change the position size, some people do change the position size. It’s all of these things we’ve talked about on the podcast that makes us different.

Yet, as we can see, it is a pretty broad-based rally (let’s call it that) in terms of performance of these strategies. To me that goes to this term, that you mentioned there and that we talked about a lot, but it is elusive (a little bit), and that is this robustness. Meaning that as long as we stick to the golden rules of trend following, whether we do it a little bit this way or whether we do it a little bit that way, it doesn’t actually spoil the strategy and its ability to continue to adapt and (really more importantly in my view), to continue to navigate an environment that it has not seen before. I think that’s the critical part. 

I do believe that a lot of strategies rely on things repeating themselves, but in our case it’s really not the case. It’s really about completely new [developments]. Just like we saw oil go negative; we’ve never seen that before, but no CTAs blew up because of that. There are all of these completely crazy things we’ve seen in the markets, yet, we’re all still here, for the most part, and doing well. As you rightly said, many of these firms and strategies have been around for 30, 40, even 50 years. I think that is impressive and I’ve never quite understood why that is not being appreciated more by the people looking from the outside in. 

[00:12:33] Richard

No, definitely. Yeah, it’s always surprising to me. I believe that trend followers take opportunity under uncertainty. This is unlike other methods that are relying on predictive stability. 

I think our methods blossom under uncertainty, and it’s exactly because we haven’t seen these environments before. Our models tend to capitalize on these opportunities. So, if we imagine this market is a zero sum game. For those traders that are working on the premise that history repeats, or the patterns repeat, when their models break down, I think this is where there’s a mass shift in transference of wealth to those models that capitalize on the opposite side of the spectrum, which is capitalize on uncertainty as opposed to predictability. As we go through these different regimes, I think there are certain models that do well under predictive certainty, and then there are certain models that do well under uncertainty .

[00:13:39] Niels

Of course, that is true. It is true that trend following has done well during uncertain times. What I always want to stress is that we don’t need uncertainty per se. We’ve done perfectly well during times where things were more stable; stable in the sense not that the prices don’t move, but stable in the sense that we don’t have a war in our backyard, or we don’t have inflation at 10% and all of that stuff.

This is what still fascinates me even after all these decades of doing it. It fascinates me that you can have a strategy that actually is able to handle, probably, both environments pretty well. Of course we know that the transition period is always going to be somewhat painful because markets change direction and therefore we have to adapt our positions accordingly.

Anyway, I’m sure we’ll come back to it, but it’s just still fascinating. 

This week I think that most trend followers had a pretty soft start to Q4. There were some short lived corrections in the markets. As I mentioned earlier, this was driven to some extent by the UN suddenly weighing in on what the Federal Reserve should do with its monetary policy. Which strikes me as being somewhat out of order that an organization like that should put pressure on a central bank and question their independence to that extent.

Having said that, after a couple of days, the focus shifted (to some extent) because we started looking at some of the challenges that we are still facing, not least the high energy prices. That came in the wake of the OPEC+ making this very blunt statement saying that energy security comes at a cost.

This was in the wake of them announcing that they’re going to cut 2 million barrels per day of production. It’s a very interesting time to do this, and I’m sure there’ll be plenty to talk about going forward with that. It could actually, from a trend following point of view, put a little bit of a floor under energy prices, which have been under pressure for the last four months, maybe five months, since early March, really when they peaked.

So, this will be quite interesting to follow. You and I were talking off-air, before we pressed record, a little bit about that energies is an interesting sector right now because are trend followers short? Are they long? Are they neutral? It’s hard to tell. It depends on what kind of methodology and timeframe they’re applying, of course. But if you are longer-term, I think you would probably still be long. Then this week would’ve been very productive from a performance point of view in the energy markets. Fixed income is, of course, our friend that keeps paying dividends, (so to speak) did well this week. Currencies, maybe even grains contributed positively. I’m not so sure about metal, softs, and equities, again, it’s a little bit depending on timeframe and lookback and all of that. 

My own trend barometer finished the week at 48, so it’s a little towards the higher end of neutral. So, it could go either way. 

Performance for the week so far, and for the month so far, and for the quarter so far, for the indices we follow (this is as of Thursday); Friday was definitely a positive day for trend followers. But I’m not sure it makes it all the way to positive territory. But the BTOP50 Index was down 46 basis points as of Thursday up 18.88% for the year.

CTA Index down 73 basis points, up 25.19%. SocGen Trend Index down 1.46% for the month up 33.6% for the year. The SocGen Short Term Traders Index was down 53 basis points, up 12.4% for the year. Comparing that with the MSCI World, which actually did finish the week positive up 1.64%, but down 25.19% for the year. The World Government Bond Index had another losing month again, down 49 basis points so far in October.

We have one question, which is not going to take that long to answer. So, let’s do that first and then we have some quite interesting topics about trend following, funnily enough, that we want to dive into, but maybe from a slightly different perspective than what we normally talk about. So, let me just read the question and then you can give your view, Rich. 

So this is from Josh. Josh writes in, “A question for you and your guest, when opening a trend following position (here I’m talking about equity, and long, and fractional in my small retail Robin Hood account), and I calculate a 14 or 15 day average true range based on (as a stop-loss) say two ATRs. Do you enter the stop and have it fire intra-day, or do you keep the stop-loss price as a mental price and take it based on the close? Or do you just watch it all day?”

[00:19:10] Richard

Yeah, good question from Josh. Look, the first thing I’d just like to say is that when we are talking about trend following, and we’re talking about the returns delivered by trend following, we’re talking about these significant outliers that we target. Now, because of the magnitude of these outliers, it means that the materiality of things like entry, and the things like when we stop-out of our position, really is quite immaterial over the long-term with our statistics. 

The benefit of trading the medium to long-term timeframes is that the impact of trading costs (associated with the finesse of how we enter, how we exit, et cetera), is far less material than if we’re trading more aggressive strategies that we’re, say, trading intra-day, where suddenly things such as trading costs (when we enter all of these things), become exceedingly important because the costs of those trading decisions are material to your PnL.

But in relation to the materiality to my PnL, the questions that Josh proposes are quite immaterial, really. The way I do it is that I use algorithms that base their calculations 30 minutes after the open of a market. I’m not actually doing my calculations on the open, or at the close of the market. I’m waiting 30 minutes. I do this for all of my systems, just to wait until the frenetic activity of the open starts dissipating a bit. Then my algorithms execute 30 minutes after the open. 

At that point in time they do a calculation. If that calculation says, “Enter the trade,” I adopt the market order and go straight into that trade, I don’t adopt stop orders or I don’t adopt limit orders. One of the reasons I do that is that I don’t like brokers knowing where my order entries are. I adopt a market entry position, which does occasionally give me a bit of slippage. But, as I said before, the materiality of that is insignificant over the long term because of the nature of the outliers that I’m targeting through my method.

Yep, it’s on the open, about 30 minutes after the open, a market order for entry, and my trailing stops however, they do operate at all times during the day. Once that trail is hit, I’m out of that position. I just do that as a matter of principle because a lot can happen in the course of a day. I like to execute my exit as soon as that price is touched on my trailing stop rather than, for instance, wait until the end of the day or the open of the next day to exit my position. 

[00:2:06] Niels

Yeah, I agree with that. You might as well leave the stop sitting in the market. That seems logical. I’ve even actually heard about managers who execute their signals a couple of days after they are generated because, as you rightly say, long-term trend following is not that sensitive to that. 

I’m a little bit surprised to hear that you need (and maybe I misunderstood it), that you need a little bit of data to run certain algorithms at the open. Because for a long-term system, you would think that one day’s worth of data, or half an hour’s worth of data, it doesn’t really matter.

[00:2:50] Richard

What happens in my world, with the markets that open (according to my brokers), they don’t all open at the same time. They open in tranches. So, over the course of a half hour period, different markets are opening consecutively. That’s why I wait until 30 minutes after the open, effectively, because they have a tranche opening, as opposed to all open at the same time. 

[00:23:14] Niels

Another question that just comes to mind when we talk about this and that is, I’m not sure you and I have talked about it, I know I’ve talked about it some with some of the other guys, and that is, do you think that the exit may be more important than the entry? What I mean by that is, of course, the entry is important because we need to get in. This is not the question. What I mean by it is that whether you apply moving average, crossover, breakout, time series, momentum, whatever it might be, I would be so bold to say that I think the entry points are not going to be vastly different.

Of course, they’re going to be different if you’re super short-term to long-term and all of that. Generally speaking, I think we can almost see a trend emerging, or when it starts, just by looking at a chart. You can almost see it with your eyes. I’ve always had this slight feeling that we all get in, more or less, at the same time (let’s say within the same week for example). But, where we get out can be more defining for how much of the profit (if it’s a profitable trade) we can capture. Also, I think that there are many more exit methodologies or rules that people apply, therefore they are probably somewhat more different in terms of where they are in the markets. Is that something you thought about? 

[00:25:02] Richard

I hear a lot of people say the entry is not important… 

[00:25:06] Niels

No, no, no, I didn’t say that. I didn’t say that it’s not important. What I’m just saying is that I think that we, as an industry, are more likely to be getting in (more or less) in the same area, not the same time necessary, more or less in the same area. But I’m not so sure that I could say the same about where we get out. That’s all. 

[00:25:31] Richard

Yeah, for me entry is important to me because, for instance, if I randomly entered the market at any point in time irrespective of when trends had materialized, if I did that then clearly I’d have a hell of a lot of whipsaws and be unprofitable. So, clearly I’m waiting for some price event to occur before I even consider participating in that event. 

To me, the more I delay that entry event, the more material the price move has become before I enter that trend. Now, what that does to me is that it, therefore, pushes my entries out towards the tails of the distribution. I think that’s a good thing for me because I like to think that [I’m targeting] a particular area of that market distribution at the left and right tails. Now, no one knows where they begin or end, but we do know that as we wait to enter trends (that become more and more material in nature), you do increase the probability that you are entering an outlier, by restricting your trade entry you avoid what I call “the noise and mean reversion” that tends to be associated with a lot of trending environments. So, once you do wait for this extreme entry, I think this does reduce the probability landscape to have a great chance of capturing outliers in your distribution of trade returns. To me, that’s why entry is important to me. 

As you say, exit is very important because the method of exit ultimately determines the level of PnL you obtain; how profitable you are. I’m a big believer in trailing my stop. I probably have less loose pants than some other trend followers in relation to how I do my trailing technique. Clearly I don’t set profit targets. A profit target to me is not a correct way to validly trade a trend. If you set a profit target, that means that you have some idea or understanding of where that trend is going to end. Now, because we are targeting these outliers, which are anomalies (by definition), and unpredictable in nature, we need to let them run their course because no one knows how long they’re going to go for.

So a trailing method is the ideal technique to capitalize on this propensity for some trends to become extreme in magnitude, which means that you don’t lose all of your profit off the table by waiting for it to come back to your original end (your stop-loss), your trailing stop is continually progressively (with beneficial price moves) ratcheting up behind your price. It means that you do give maximum freedom for those trends to be of significant magnitude. But you do ultimately have to take that profit once that trailing stop hits. 

[00:28:44] Niels

Yeah, I think that makes sense that you can, to some extent, if you are really targeting outliers, you can tilt your entry rules as well to increase the probability of that. I think that is fair. As we’ve talked about before, you would rather make more money from the 5% or 10% of trades, maybe compared to someone who might say, “I may not make as much on money on those, but I’m going to make more money on the 50% other profitable, or 40% other profitable trades I’m going to have.” So, I think that’s fair. 

All right, Rich, we are going to dive into some trend identification and certain things we have to avoid and so on and so forth. So, it’s going to be another masterclass, I have a feeling . Let’s dive into all of this and I’ll do my best to keep up with you.

[00:29:40] Richard

Thanks, Niels. Look, we are going back to the basics here. What we are doing is we are talking about trends and we’re talking about visual methods of identifying trends versus systematic rules based methods of identifying trends. We’re comparing and contrasting the different methods.

When most people begin their trend following journey they tend to be drawn to the idea that we need to be prescriptive in the way we classify trends. You might go back to Dow theory, which says that for a valid trend you need three consecutive highs and three consecutive lows, et cetera. Or there might be other definitions of trend where they’re using moving averages to define what the trend is. There is this plethora of different ways of identifying trends. 

The problem with that plethora of methods of identifying trends is that no one has a common understanding of what the trend is. They all have a personal interpretation of what they view a trend is, dependent on the interpretation they’re using to identify what that trend is. So, this is different to, for instance, the diversified systematic trend followers who use more quantitative methods for trend identifications.

They’re using the data as a basis to give them clues as to whether there is a trending environment or not, as opposed to the visual form of a trend. One of the problems we face is that if, for instance, I went to my technical analysis class where I looked at identifying trends, and we put up on the blackboard all of these different trend configurations. Everyone will be saying look, that’s a trend, that’s a trend, no, that’s a trend, no, that one’s a trend ,that’s a trend, et cetera. There is just no consistency. Having a systematic rules based approach means that everyone will agree, if we use the same systematic rules based approach, everyone will agree on that interpretation of a trend.

It differs, of course, when you’re applying discretionary judgment, and using visual clues, and the way your brain works to identify what is a trending condition. We often find, when we revisit something that previously we disregarded as a visual trend, when we come back to it, we say to ourselves, “Oh, how did I disregard that as a trend? Clearly that is a trend now.” 

So, your brain and its pattern recognition changes over time as well. There is no “stationary” to how your brain perceives trends. It’s a moving feast of emerging patterns that change over the course of time. That is why, one of the problems is that, if you are very prescriptive in determining what is a trend, you will find that you potentially miss out on a vast array of different visual forms that also are trends. And you will find that you potentially miss the outlier because you are demanding some prescriptive form of visual trend before you entertain the prospects of trading that trend. 

So, what we say in our quantitative world is, It’s better to be less prescriptive in how we define a trend and simply recognize a trend, over the course of time, as a directionally extended series of price data as opposed to getting into all of the nitty gritty about the exact visual form that trend should take.

What we find is (if we do get prescriptive in how we define a trend), we can have these, what we call, type two errors. For a trend follower, a type two error is perhaps the biggest sin. We, of course, have lots of small losses but they are only small losses to us. A type two error is where you miss the outlier. Because those outliers are such large contributions to our PnL, if you miss one of them, that can be quite drastic for our fortunes over the long term. 

We avoid this prescriptive tendency to classify trends according to what specific visual form it has. We use quantitative methods to basically give a degree of loose pants in our definition of how we describe a trend. So, the looser we define that trend, the better the chance we have of capturing a vast array of different possible forms of trend as opposed to being prescriptive. What do you think about that? 

[00:34:37] Niels

Yeah, first of all, by the way, so you talk about a type two error. What’s the type one error in your book?

[00:34:36] Richard

Now you’ve got me, I’d have to go back to my book.

[00:34:39] Niels

That’s fair, that’s fair. It’s just that I hear these terms being talked about a lot, and I’m not always sure what people mean by it, but that’s fine. It doesn’t really matter. 

[00:34:251] Richard

A type one error is a mistake caused by some programmatic error.

A type two area is more like an opportunity cost; missing out on something, I think. 

[00:35:03] Niels

Anyways, I’m going to try and visualize what you just described. We’ll talk about this, I’m sure, a little bit more. One thing I noticed, in our own data, is that the best performing market (for us), in the last five years, is a market that did not make any money for quite a while. I’m actually talking about short-term interest rates. 

So, here we have a group of markets that, if you looked at the price from, say, 2000 (and this is just ballpark), say, 2018 through 2022, it’s like a flat straight line because they were all pegged at zero. So, there’s nothing going on. You could still, because of roll yields and so on and so forth, you could still get a signal. Then earlier this year (or towards the end of last year, I can’t remember exactly when), you started getting this little breakout, which looked like nothing. Given the fact that you’ve had like a four year period of flat lining, you think, “Ah, this must be a mistake.” You’re not going to fall for that. It’s like the JDB trade, like the widow trade, where every time you try to short it, you’re going to be stopped-out, right?

But it just goes to show that a market that essentially had given you no opportunities for quite a while suddenly can give you amazing opportunities. As you rightly say, it didn’t look anything like what a normal trend would look like visually, because it was just a flat line and then a little blip, and then suddenly it became a real trend.

So, that was what I was thinking about when you were talking about different types of trends. Our gut reaction when we think about trends is some kind of, “Oh yeah, here’s a range of a market. It’s consolidating. Then it breaks out and then it takes off. Okay, that’s trend.”

But there are these weird other situations where, It just looks completely different. 

[00:37:25] Richard

Yeah, exactly. One of the problems we are facing is that if we are using our brains to interpret what trends are… I’m fascinated with some of these recent research findings by some of the cognitive neuroscientists like Anil Seth and Donald Hoffman. They’re very interesting. So, what they’re proposing is that our brains are effectively these prediction engines. 

So, if we could imagine the brain itself is encased in a skull. It’s reliant on all of the sensory data being received by our senses: through our eyes, through our taste, through our nose, through our touch, et cetera. These are just sensory impulses. I love sometimes thinking to myself, “What happens if we don’t have any ears?” So, for a person who has no ears, there is no sound. That’s because what we are really picking up with what we refer to as sound (which we believe is out there in the environment), is just these pressure points of molecules vibrating. That’s effectively what it is. So it’s the human, the human ear that is interpreting this as sound. But ask another creature that might not have an ear, how they’re interpreting that same sensory input, and they might come up with a totally different conclusion. It’s very vivid when I think of sound.

If I don’t have ears, I don’t hear any sound. This is something that is familiar on this planet earth, but if we go out into space, there is no sound out in space because there are no molecules to vibrate in the spatial arena outside of our planet earth. So, in space, no one hears you scream. 

We are so familiar with these concepts of sound, sight, smell, et cetera, but really they are our methods of interpreting our external environment. That external environment may be significantly different. The reality of that external environment may be completely different to how we are interpreting it. Effectively, as a species, as a human species, we have evolved these senses to give us a survival edge.

It has given us a competitive advantage, so we hear sound because it’s given us a competitive advantage. There might not actually be this concept called sound out there, but how our bodies, over time, have evolved, and how we have adapted the sensory inputs into how we interpret it has given us a survival advantage.

People like Anil Seth and Donald Hoffman are now recognizing that the world in our external environment may be completely different to what we perceive as human beings with a brain. One thing we are certain about is the brain’s ability to predict. Anil Seth refers to our perception of experience around us, in our environment, as a controlled hallucination. In other words, he calls it a controlled hallucination because at least everybody agrees on that same hallucination. This is different to the hallucination you might receive under a certain drug intakes or whatever. So, everyone’s receiving the same control hallucination simply because we’ve all evolved and we have the same ears or similar ears, similar design features, so we were interpreting things in a similar way (but it could be considerably different). It’s given us a competitive advantage in this natural environment. 

As traders trying to trade this market, the markets, the financial market is something we haven’t evolved in. So, what the brain is doing might be contra, or different to the way we should be interpreting these financial markets.

We haven’t evolved in those financial markets. We don’t have sensory apparatus that has honed in on what the financial markets are doing. The brain is a pretty bad guide to how to trade financial markets, which is why I always adopt or prefer a systematic rules based approach as opposed to letting my brain override my rules based approach. This is simply because that brain has evolved in a different context to the financial markets which I’m trying to trade.

Does that make sense? 

[00:42:19] Niels

Yeah. So do you think that (if we’re going to stay with your analogy here), do you think the fact that for those of us who only really need an open, high, low, close, on a daily basis (in terms of price), is that kind of the same as not having any ears? Meaning we don’t hear the noise intra-today, we just say, “Oh yeah, it was this high, this low. This is where it opened. This is where it closed. That’s all I need to know. I don’t need all the noise that took place.” 

[00:42:52] Richard

Yeah, it gets worse than that though, because a chart pattern is a visual representation. It’s not actually the way the market operates.

When we see how the market operates, with the orders coming in, and the matching of orders, and then the transactions taking place, that is a different representation to what we visually see on a chart. We like to simplify things on a visual chart and say, “There was an open, there was a high, there was a low, there was a close.” But that’s not actually the way the market really operates. It’s just our visual representation of that. You can use different visual charts (as opposed to the open, high, low, closed chart), to represent different interpretations of those transactions that are occurring in the financial markets.

So, those people that are hell-bent on using charts as the definitive guide to identify a trend, they’re really restricting their abilities to a degree. That’s not necessarily the way the market actually works. I think quantitative methods… 

So in other words, one of the key reasons we use these extensive backtests is because we want to reduce the ability of our brains to override our decisions and ensure that we’ve got a systematic rules based process that everybody can agree on (because we use the same system rules). So, if I pass my system rules onto you, Niels, you will get exactly the same definition of trend as I’m getting because we are using the same rules based process. Those rules based processes also have been forged over very large data samples. So, it has reduced the fickle nature of our minds; one moment we might see a trend and when we revisit it at a later stage, we’ll say, “That was never a trend.”

This sort of eliminates that discretionary override in the way our brains work to a much more systematic rule based process that is more definitive, can be agreed on by everybody, and is a much more rigorous way of validating empirically the nature of a trend or a bias price series.

[00:45:10] Niels

Before we move on to the next point I did find my notes from when we recorded with Annie Duke and there was this one quote that I wanted to share with you, even though it’s going to come out in a few days. So, this is a little sneak preview, Rich, but I think you’re going to like it. Annie says, at some point she says, “The more you can systematize the decision, the better off you will be.”

Which of course is something that just plays so well into our world. Of course she comes at this as an incredibly successful poker player. Now she’s someone who coaches people about making decisions during uncertain times. We just think (many investors and certainly trend followers) that’s exactly what we’ve been dealing with for many decades. 

[00:46:01] Richard

That’s very exciting. I can’t wait for that podcast, Niels. I spent a lot of my youth in the casinos. I do love looking into the details of the different games of chance and looking where the edge resides, or if there is an edge present, or if there is simply a House edge present, et cetera.

I think these things I learned at the casino were quite foundational for my move into trend following. One of the reasons I moved into trend following was that clearly I saw this definitive edge in trend following, particularly in the nature of markets to be non-normal in nature and have these tail properties.

When I started seeing these tail properties of markets… Every liquid market I assessed, I saw these tail features of these liquid markets. I thought, “There’s my opportunity that lies outside the normal distribution. This is something I can capitalize on.” Thanks to the seedy days of the casino, I think it has taught me a lot about how to identify where an edge exists.

[00:47:14] Niels

Yeah, absolutely. By the way, she’s also a wonderful storyteller. She has some amazing stories that she shares which actually helps explain… We know how important narrative is, right? But it’s given me a couple of new ways to talk about things like stop-losses, and stuff like that. But I’m not going to steal her thunder. Let’s wait until at least next week. So anyways, where are we going to go next with this trend thing? 

[00:47:45] Richard

Okay. So, we’ve talked about the trends in the market data, but there are two things that define trends (as far as my understanding). Firstly, there needs to be a trend in the price data, which is this directional series of events. Not only does there need to be a trend in the market price data, but also our ability to capture trends is dependent on our system configuration. In other words, what I mean is that, to be able to extract that edge from the market data, we need to have systems that are capable of extracting that edge.

The way we do this is through these very simple designs that limit the degree of price freedom that is allowed when we enter a trade. So, when we enter a trade, we can configure the initial stop, where it needs to be located. We can configure the trailing stop. Effectively, they are like system constraints. If you can consider a system like a container, it limits the adverse price excursion that the price is allowed to go before you exit that trade. So that, therefore, minimizes our adverse risk exposure. We don’t set any profit target. We are unbridled with our directional extent; the future potential of that directional extent. So, we use this trailing stop mechanism. If you can imagine, yes, we do mitigate all adverse risk exposure with our small bet size, our trailing stop, and our initial stop. We allow for this unlimited possibility in relation to how we want price to move. Our ability to capture these trends is both a component of the trends that exist in the market data and our system configuration.

What this means is that, as we said before, because there are these many different various forms of visual trend we need a very loose pants definition of how we capture that trending price series to give maximum degrees of freedom for price to play out, because there is a large component of random gyrations in any price series. But there is a much smaller component of bias in that price series that ultimately manifests in this momentum, that exists in that trending price series, to enable us to capture this trending price series. 

Effectively, I just want to make listeners aware that, when we do talk about trends, it is… People say, “I looked at trends in 2012 and 2014,” where people were declaring that trend opportunities don’t exist. When I looked at the monthly price data of all of these liquid markets, in those periods of time where people were saying that trend following no longer works, I thought to myself why is that? Visually, I can see these trends existing, but why aren’t trend followers capitalizing on those trends?

What we find is that with these adaptive markets, over the course of time, trends do change in form and they change in their degree of volatility. That means that, over the course of time, we do need to adapt our systems to take into account this growing variability of trend form.

Whilst we might have developed trend following systems back in the 1970s, the variables associated with those simply designed trend following systems might have changed over the past. The trends I saw in 2012, 2014, they were valid trends, but they were very difficult to trade. The reason was not because trends in the market data didn’t exist. They did. Our system’s ability to capture those trends was more difficult because these trends were much less defined, much more volatile. They had a greater variety of different forms than what we’d train those trend following systems on in the past.

So, there does need to be this degree of adaptability and our trend following models to continuously refresh, in light of new data we receive; new trend form information that we receive. We do need to continually include this progressive additional data so that we can retain this very loose pants system design.

As we’ve talked about, quite regularly, over our podcast series. The more prescriptive your design elements are… So, in other words, the more parameters and variables that you have in your system design, the more the system constraints restrict the degrees of freedom of price. What that means is that the more variables that you adapt, you are more prescriptively defining how that trend needs to perform in the future based on a backtest series.

So, for instance, if I decided to have tight pants, I could develop trend following systems that mirrored how they’d behaved in the past, exactly. That would restrict my degrees of freedom in relation to new opportunities, new emerging trends that we hadn’t seen before. I wouldn’t be able to necessarily capture those trends if I overly restrict my designs to suit what has occurred in the past. So, I need this freedom to allow trends to adapt to change in nature, to have these loose pants. 

[00:53:36] Niels

Okay, so let me ask a question. We look at, essentially, how different lookback periods would have performed, every calendar year, going back in time to give us an idea of whether, a 40 day lookback or 240 day lookback, what seems to be better (let’s use that word). Now, when you look at the evidence (meaning if you look at the data), there is a clear gravitation that the better lookback periods are not the shorter ones. 

So, my question to you, here, is do you actually think (from what you just said), that this is the reason why shorter-term lookback periods are just not profitable to the same degree as longer-term simply for what you just said, namely that we need (in lack of a better word) looser pants; we need to allow room for many different types of trends? If you, even with that single parameter as a lookback period, if you restricted too much, it just doesn’t work as well. Is it both that complicated but also that simple?

[00:55:03] Richard

Yeah, I have looked into this quite extensively. How I do that is that I research how short-term models that are configured exactly the same way (apart from the lookback), compared to my medium to long-term models. What I find is that all three (let’s say I have short-term models, medium-term long models, and long-term models), they’re using a similar ATR in their initial stop, a similar ATR in their trailing stop, and similar multipliers. But the things that do change is simply the lookback. 

What I find is that the short-term models, the medium-term models and the long term models, yes, they all capture outliers. There’s no question about that. But what the short-term models do, more than the medium-term models and the long-term models, is they also catch a lot more noise and mean reversion. Even though they capture these outliers, they’re capturing a lot more noise and mean reversion that compromise the PnL.

So, when I look at the… Let’s say I have a short-term model, and let’s say that short-term model generates a thousand trades over a 30 year period. Then I’ve got a medium-term model. The only thing I’ve changed is my lookback and my long term model. The medium-term model might only have 300 trades. The long-term model might only have 100 trades simply because of the change in the lookback. What I find is that all three have a representative inclusion of outliers in their distribution; their trade distribution. What is much less prevalent in the medium and long-term models is this degree of noise and mean reversion; these linear inconsequential wins or losses.

What I’m finding is there is more frictional drag in my short-term models that compromise my results. What this tells me is that we always talk about how we are targeting these outliers and these significant anomalies tha always are far larger than what we can predict, according to a normal distribution. What I think we find is that, as we step out to these medium to long-term models, what we’re inevitably doing is we are playing more in the tail region of the distribution than we are in the bulk of the distribution. So, when we play with our models in the bulk of the distribution, we are capturing a lot more convergence and noise than what we do when we step out to the tail regions. I think the longer-term lookbacks naturally migrate us further out to those tail regions.

[00:57:43] Niels

Yeah, I think that makes sense. Anything else in terms of the points you wanted to bring up today? 

[00:57:51] Richard

Let’s have a look. Okay, what I’d like to do is quite a contentious thing here and suggest that, in reality, nearly all traders or investors are actually trend followers. Now, what I mean by this is that apart from some restricted strategies like options selling, nearly everybody, as a trader, wants a different entry price to their exit price. They’re looking at a difference in originating price and exit price to generate their profit, or potentially their loss. So, they’re looking for a trending price series between their entry and their exit. 

In broadest possible terms, everyone: the gambler, the convergent trader and the divergent trader, are all effectively trend followers. You know how I said we all tend to interpret trends differently? A segment of a mean reverting cycle, which goes out from a maximum favorable excursion back to an equilibrium, has a trending component. If you look at that on a small scale, it looks like a classic trend. A lot of people that trade counter trend are trading those mean reverting components of a bigger mean reverting cycle. The divergent trader is also capitalizing on trends. 

So, what sets apart the divergent trader from the convergent trader and the gambler, if they’re all trading trends? What we are looking at as divergent traders is this enduring directional persistence in that trend. We’ve talked about visual trends and how difficult it is to do it, but really, the reality is that they’re more academic topics. When we are trading on that right hand side of the chart, and making our entry, we don’t know what price is going to do in the future. So, we are staring into a black void. 

We know what’s occurred in the recent past because we’ve got this trending price series. Our systematic rules have identified that there is a trend that has occurred but now, at my point of entry, I’m now staring into the black abyss. So what a divergent trend follower is relying on is some form of enduring persistence in that directional move, at the time of entry, that continues on in the same direction or a similar direction in the future.

That enduring persistence, or what we often refer to as momentum, is what we refer to as serial correlation. In this particular example of a divergent trend trader, we are looking for positive serial correlation. Now, this is different to negative serial correlation. Negative serial correlation is something that a convergent trader, or a mean reverting trader is looking for. A negative serial correlation means that a high is followed by a low, is followed by a high, is followed by a low. You can see that it oscillates. That is the symptom of convergence associated with negative correlation. 

Then there is positive correlation where a high is followed by a high, and another high, and a future high, and a future high. There is a directional persistence in the same trajectory; as well as a low is followed by another low, is followed by another low, et cetera. That is positive serial correlation. There is something amplifying that directional price move in the same direction as opposed to switching direction, which is what negative serial correlation does.

Now, of course, sitting in the middle, between negative serial correlation and positive serial correlation, is when there is no serial correlation in the data. Now what that means is that price is independent at all times. So, that is effectively dealing with a random series. When you plot a random series of price, it does neatly fit within a normal distribution, which is no surprise as a normal distribution is a definition of an independent random price series.

When we are talking about the tail regions, we are talking about looking at this positive serial correlation. When we’re talking about a convergence series, we are looking at the peak of the distribution about an equilibrium where we’re referring to this negative serial correlation. What the difference between the convergent trader, the gambler, and the divergent trader effectively boils down to is the degree or what form of serial correlation, if it exists, what form does it take: positive or negative, or is there no serial correlation?

So, a quantitative trader, as opposed to a visual trader, will be looking at this serial correlation. This is very interesting to me because I often do studies where I create random distributions of price series. Now, they’re quite easy to create. What I potentially will do is I will say, here is an existing price series of data. We’ve got a 30 year price series of data. Now, what I want to do is randomly reshuffle that data to break up any correlation (, or clusters of correlation), that exist in that data. So, by randomizing that price series, I can basically make them less correlated. So, the price at each interval is more independent of each other.

When I put my trend following models on that price series, low and behold they fail. Also I can create, from scratch, a randomly generated price series where I allow the price series just to stochastically vary in relation to the immediate close of the prior price, how it varies to the next price, how it varies to the next price. I make them independent and I can create a random price series. 

Now, when I plot (with the visual graph) a random price series, visually, I simply cannot tell whether it is a trending price series, a random price series, or whatever. It has all of the features and forms that we normally associate with data that we assume is non-random. What this tells me is that in these efficient markets this random price series, or the visual form of the random price series, is what tricks so many of us traders into thinking there is something of causative substance or endurance in that price series. Ultimately, It is random.

So what I do then is I say, let’s take a random price series. Now, what I do is I say, on top of that random price series, let’s insert clusters of positive serial correlation in that random price series at discrete intervals across that price series. What I’m doing there is I’m creating a small bias in the random price series so that visually it still looks the same as the random price series. The only difference is, taken from the origin, the random price series and the price series with positive serial correlation in it, they diverge away from each other. It’s this divergence which is this momentum that we are looking for, which we are targeting with our trend following models.

It’s not a visual thing, because we cannot tell, visually, what the difference is between a correlated price series and a random price series. There is a divergence in the price series itself, which allows for our systematic rules based process to capitalize on these trending conditions.

It’s far less a game of visual identification, and it’s more a game of how do we know if we’ve got this degree of positive serial correlation in that price series? I’ve talked about how I do it, creating these random price series, but really that is another academic exercise. When I’m trading the right hand side of the chart, I’m not going to be able to statistically immediately analyze that prior price series to see if there’s correlation in it.

What I am doing is delaying my entry into that trend to wait until there is a greater chance of reaching this tail region. One of the reasons that the tail regions are so important is that we know there are clusters of positive serial correlation in tail data. In other words, that is what causes the extreme magnitude of the tails on the left and right tails of the distribution. There is this momentum baked in the price series that is amplifying the price series, in this directional way, well beyond what a normal distribution implies. These power laws, that reside in these tails, are a consequence of these clusters of serial correlation or these positive serial correlation.

Whilst I can’t definitively tell whether I’m riding a trend which has this positive serial correlation in it, what I am doing, with my entry signal, is delaying my entry until price becomes of sufficient material extent that there is a great chance it doesn’t fit neatly within that normal distribution, and there is a greater chance that it lies within those tail regions where we know positive serial correlation exists.

[01:07:21] Niels

So, in other words, this exercise of manipulating the data which, of course, you could say that’s completely irrelevant because the data is the data; the actual market data is the data. We can’t manipulate that. 

What you’re saying, if I understand correctly, is that by doing this exercise, it has helped you refine (a little bit) how you want your models, or when you want your models to enter. Once that exercise is done, and it’s proven the point you want to use, this is not an exercise you need to do again and again, I imagine.

[01:08:04] Richard

Yeah, exactly.

[01:08:07] Niels

I think this is one thing, Rich, where there was a Twitter comment this week that I commented on. There was someone who wrote something about trend following, that we are over (not complicating), but we are overdoing the automation. I didn’t quite understand the point of how we are doing that. Yes, we do use math, because if it was so simple, then clearly everyone would be doing it. We make it sound easy, but it doesn’t mean that it’s simple. So of course there is some math in it. What you’ve just described is some of the math you are using, so to speak, to prove a point that then can help you in designing your algo, so to speak.

I’m also aware that we don’t want to overcomplicate. There is this fine balance where we definitely don’t want to overcomplicate matters too much. I also think we have to be transparent and say there is a reason why we have highly educated and experience researchers, like yourself, Rich, and of course people who work for these trend following firms who spend a lot of time thinking about small tweaks, small details that actually may make a decent enhancement if you can get it right. I’m sure that 90%, or maybe 95% of the stuff they work on will be dismissed, because it just turns out not to be robust enough. 

Let’s leave the topics for today like that.

This was great, Rich, as always. Hopefully it’s given people something to think about in terms of trends, and how they may form, and the fact that there are many ways of looking at it. Also, what I really like about the fact, because I’ve seen some of your data, and that is of course, you’re absolutely right, to the naked eye you really can’t tell the difference: whether a time series is random or whether it has a bias. It’s very hard to tell the difference. The devil is definitely in the detail sometimes, and so that I appreciate a lot. 

Of course, everyone listening to our conversation today, if you love these conversations, then please do go and to Spotify, iTunes, Amazon, wherever you listen to these conversations and leave a rating and review. That would be great. 

Which reminds me, actually, about a rating I got this week. It’s rare we get it, but we did get a low rating from someone. I have no idea who it is, who basically said it’s great that you talk about how wonderful trend following is, but can’t you just tell us how. I don’t think that the person who left that rating has actually listened carefully because I do think we spent quite a lot of time, and not least by you, Rich, being very transparent as to how we do this. But this is not a podcast that’s going to just give you a cheat sheet and say, “Oh yeah, this is the five things you need to put into your system, and then you’re off to the races.” Some people do that, some people want to make it sound like, “Oh, you can just get it off the shelf, or you can just get it from the internet, and then you’re a trend follower.” That’s just not how it works in reality. 

Hopefully these conversations will help people, encourage people to either do it themselves, do the work. This is about doing the work. You can’t just rely on other people to do the work. 

Maybe, to the listener who left that rating of, “Just give me the how,” well the how is then go and invest with an experienced manager. That is the how. That’s the easiest way you can do [the how], because there are plenty of good funds out there. So, that is, of course, one way to solve that, I would say. 

Anyways, next week we will have a fun conversation, I think. It’s a little bit out of the usual format because we are going to be joined by Andrew Beer, or I’m going to be joined by Andrew Beer to discuss trend replication; a topic that has come up a few weeks ago. There are already questions that have come in, but I would like some more to choose from. So, do send your questions to info@toptradersunplugged.com. I’ll try and have a meaningful conversation with Andrew. Of course I have my own questions and concerns. But I think it’ll be a fun conversation. Andrew is great at explaining these things. So yes, next week will be a little bit different but hopefully lots of fun as well. 

From Rich 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.

Thanks for listening to the Systematic Investor Podcast series. If you enjoy this series, go on over to iTunes and leave an honest rating and review and be sure to listen to all the other episodes from Top Traders Unplugged. If you have questions about systematic investing, send us an email with the word “question” in the subject line to info@toptradersunplugged.com and we’ll try to get it on the show.

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.