Thanks for tuning back into our show, we know it has been a while since we’ve published an episode. However, you are in for a real treat: this is the first episode where we have a roundtable of guests, representing some of the most prominent firms in the trend following space. Our guests have a combined 100 years of track record, and we’ll explore it all in this episode.
We discuss the history of trend following, what it takes to practice trend following, and the human biases that effect our trading.
Thanks for listening and please welcome Katy Kaminski, Alex Greyserman and Roberto Osorio.
In This Episode, You’ll Learn:
Resources & Links Mentioned in this Episode:
Follow Katy Kaminski on Linkedin
Follow Alex Greyserman on Linkedin
Follow Roberto Osorio on Linkedin
IT’s TRUE 👀 – most CIO’s read 50+ books each year – get your FREE copy of the Ultimate Guide to the Best Investment Books ever written here.
And you can get a free copy of my latest book “The Many Flavors of Trend Following” here.
Learn more about the Trend Barometer here.
Send your questions to email@example.com
And please share this episode with a like-minded friend and leave an honest rating & review on iTunes so more people can discover the podcast.
It’s been awhile since I published my last episode. First of all, thanks for being patient and sticking with me and the podcast. However, I hope you will agree with me that it’s been worth the wait because what I have in store for you today I think is really unique, and something that I’ve never done before. So far I have focused on very detailed one-on-one conversations with some really interesting and successful traders. I’ve been overwhelmed with your positive feedback.
I have published what I think are some useful eBooks written by experts that I hope will highlight some important issues that can help you regardless of whether you are an investor, a trader, or just interested in the markets. Every day I publish my trend barometer that explains the environment for a typical trend following strategy, and the daily market score that gives you full insight and transparency to how a classical trend following system will be positioned today and how strong the market trends are. But there is one thing I have not done until now which I really wanted to try, and that is to bring some of these successful traders together on the same episode to discuss important topics about trading, and initially focusing on systematic trading.
I wanted to bring you a unique opportunity to hear some the best minds in our industry share their most valuable insights and debate their views in a new and exciting format. But as we are all busy people, it took some time to find an opportunity to sit down in the same room and have this discussion. The good news is we managed to do it and with the help and support of our good friends at BarclayHedge, and Sol Waksman in particular. Today you’ll be able to hear part 1 of this conversation.
Of course when you try something new you always face challenges you did not expect, and this was no different. For example, we had the cleaning lady who decided to vacuum right outside the room we were in during our conversation. Added to that the fact that we had to turn off the air conditioning to avoid background noise, which made it a very hot event, to say the least. Despite these challenges I think the result is as good as I could have hoped for when it comes to production quality. As for the conversation, I think my guests were amazing. So let me reveal who is on today’s episode and how you’ll be able to distinguish one from the other since I, as the untrained moderator, did not always mention their name before they spoke.
The three guests represent three of the legendary firms in the CTA industry with a combined track record and experience of more than 100 years. Campbell and Company, ISAM, and Dunn Capital Management have all had a huge impact when it comes to systematic trading and trend following in particular. I was honored and privileged to sit down with Katy Kaminski, who previous had joined me on the podcast before she had joined Campbell; Alex Greyserman, who is the chief scientist at ISAM and whom you will recognize due to his deep New York accent; and Roberto Osorio, my Brazilian colleague at Dunn Capital Management where he heads up our research and who has been responsible for some of the newest and most innovative improvements to our trading model in the last five years.
As you can hear, the lineup is pretty amazing. Since this is a new format for me, do let me know what you think, and if you’d like to hear more of these kinds of episodes. Leave me a comment, send me an email, give the podcast a rating and review in iTunes. It’s the only way for me to know if I’m adding value to your day. Now let’s get started with part 1 of my conversation. I hope you will enjoy it.
Katy, Alex, and Roberto, thank so much for being with us today. I really appreciate your time. Now today’s conversation will be very different as I have all of you in the same room, and you all represent some of the most legendary firms in the CTA industry with a combined experience of more than 100 years. So I think it’s safe to say that we’ll have some very unique discussions in store for our listeners. Let’s jump right into the first topic of the conversation.
Many years ago Lady Thatcher said, “If you don’t understand history, you might be condemned to repeat its mistakes.” Let’s start with another lady, so you, Katy, why don’t you set the stage for our conversation today by telling the history of trend following based on some of the research and studies that you and Alex did in preparation for writing your latest book, Trend Following with Managed Futures: The Search for Crisis Alpha, which by the way, I would recommend that anyone interested in this strategy should get a copy of.
Thank you, Niels. I think Alex would agree with me on this, that in our book we decided to start with history because history, although it’s not as precise in terms of the mathematical… whether or not you can agree that the exact numbers of your analysis, but history really provides some context for us to understand why momentum, as a strategy or trend following as a strategy has persisted throughout the ages. What we found, we started off our book by examining 800 years of data on trend following and demonstrated that trend following strategies, or momentum approaches work over the centuries. I’m sure Alex can also add a little bit more here as well. We also looked at Crisis Alpha. We looked at how momentum strategies worked during periods of difficulty for equity markets over a three-hundred-year period. What we found is that trends have always existed. Momentum has always been a phenomenon in financial markets. Whether it was in the Middle Ages or it’s today, we’ve always seen herding effects. We’ve always seen that there are opportunities to follow trends in markets. So maybe I’ll turn that over to you, Alex, to add some other comments as well if you want.
Alexdates before, I think it’s:
When do you think, by the way, that trend following as a concept that we know today, when does that date back to? What’s the first evidence of trend following managers? I know that the three of you represent some of them, but is there anyone that you can think of that was even earlier with trend following as a concept?
Katyt they’ve been around since:
I think the key, if I can interject, is the social psychology. People show herd behavior. People like to follow each other, and this is not going to go away. It’s inside our brain hardware. That’s why I believe that trend following is not going to go away soon. It’s also… This first chapter of Alex and Katy’s book is a very impressive piece of evidence of how long it has been around and the fact that it has been around for, what is it, twelve-hundred years at least, is a good sign that it’s not going to go away in a decade or so.
Nielslso a quote in the book, from:
“Cut short your losses and let your profits run on.” That’s the first quote of the book and it was from the legendary political economist David…
Alexagic, though. Actually in the:
Robertoas also been around since the:
I think if we turn a little bit more as well… One of the things that Alex and I looked at as well was thinking about this concept of what does trend following capture? What we discuss in our book is that trend following is a divergent trading strategy. So what you’re doing then is you’re cutting your losses and your following your winners. What are you trying to capture? Well, you’re trying to capture divergence – movements across markets. So you can see why a lot of people think, well, there are no… Markets are efficient. Technology, everything is getting more convergent to be more efficient. So Alex and I actually tested that hypothesis. We looked at the measure of divergence across markets over a long history.Wasn’t back to:
Before we jump to the next topic, I want to ask sort of slightly personal question about your own history with trend following. So let’s start with you, Alex. If you could share your first experience with trend following. I seem to remember a story when you met Larry Hite for the first time, so maybe that’s a starting point. I’ll let you share what you remember.
I came very close to staying in engineering. I went to a job interview with Larry Hite in the late ‘80s. The book called The Market Wizards came out right around that time and I remember him handing me a copy and asking me to read his chapter. There was a question by Jack Schwager, (the former best person who asked questions, but now you’re doing a better job… interviewer), and he asked Larry something like, “What makes MINT?” Which was the first fund that Larry started. “What makes MINT good, or what’s the edge?” Something like that. Coming out of science I expected an answer like, “We have better PhDs.” I was expecting a scientific answer. The answer in the book, and this is a quote now, “Because we know that we don’t know.” I was like, “You have to be kidding me!” I was working at RCA doing high definition television signal processing. It was geeky cute stuff. So he did offer me $4,000 a year more, so I did take the job, but I was kind of very hesitant because it all seemed like a psychology thing. That actually is the essence… It’s one thing for us to put this on paper and say this works. It’s another thing to actually have the discipline to follow it. So that was my first experience. Basically knowing that we don’t know. Truly, to this day, if you ask me what the edge is, that’s the edge. The edge is knowing that we don’t know. I’m not going to get all Donald Rumsfeld on you with whole known knowns, and known unknowns. Don’t try that question on me.
I do want to follow up on that, before we hear what Roberto and Katy have to say about it. One thing is to be exposed to trend following, another thing is to really believe and start trusting trend following. How long did it take for you to get to that stage? That’s obviously not something that you get to just by reading a chapter.
It took me years, probably 5, 10 years. The comfort level with having more losing trades than winning trades, we’re constantly losing money. The comfort level with the risk aversion and the stops and, OK we take a loss and we move on. There’s just no mental reaction, there’s no reaction to try to data mind parameters, rerun our systems. The discipline to do this is actually easier said than done. I’m going to tell you, because it’s kind of like a parallel story because I’m involved in academics and I teach, and I find that the combination that, at least we look for in researchers… I’ll tell you that the combination is extremely rare, it’s extremely rare to find very smart people who are very humble. I have two next to me, but…
I’m extremely humble, and I’m proud of it. (laughter)
And it takes a set of people, and we talk about that in our organization, we’re just wrong most of the time in the trades. It’s OK in having the humbleness together with the ability to actually run these systems and have discipline. It’s actually very unique. I’ll tell you that 99% of the students that I teach at Columbia are not humble. They think they can walk on water. (laughter) This is maybe why (and this is getting a little advanced), but maybe this is why the strategy works. There’s not that many people who can actually follow it.
Most of the world and most of the participants in the investment management industry tend to be mean reversion type players. We actually provide the liquidity and I don’t mean to get too technical, but this is… It’s almost not possible for this to go away. Maybe that’s why it’s worked for so many hundreds of years. It’s not easy for the majority of people to actually have the discipline to do it. They’re going to want to override. They’re going to want to do this, they’re going to want to do that. It took me quite a while. It’s not just something that you can read in a book.
Sure, absolutely. Now Roberto, we of course as colleagues, know a little bit about your background which includes another strategy, namely statistical arbitrage, but I actually don’t know how you got your first exposure to trend following. So, why don’t you share that.
Robertok for the kids. That, in late:
I guess I got convinced pretty quickly that it worked, because I started doing some simulations and in a few months it looked pretty real to me that this is really a very strong signal and we can’t ignore it. There was a track record of decades going on there.
I liked Alex’s comment on the fact that humility is very important because that’s one of the main psychological biases is, coming from this research in behavioral psychology – overconfidence, in some circles that’s called the Lake Wobegon syndrome. There’s this program on NPR, by Garrison Keillor where there is this mythical place called Lake Wobegon, where every child is above average. That’s how we all feel, right?
If you do research of what people think, when they behave as drivers, for instance, 80% to 90% are going to think they’re better drivers than average. So it’s a universal phenomenon. That’s why I agree it’s better to let the machines decide. You take out human emotion out of the decision process.
Once you build an algorithm, you believe if you really have a good theoretical foundation to believe in that algorithm, and you’re sure you didn’t fall into data mining traps, and all those considerations that are very important to do, you should let the algorithm do your trades. You can improve it. You can add new signals to it, but you should not be afraid because you have a losing month. That’s part of the process.
We’ll get into that. What about you, Katy, how did you get started with trend following?
I actually started with trend following back in my PhD thesis. So I was working with Andrew Lo, at the Sloan School, and, to give a little more context, I grew up with a father who was a clinical neurologist. When I was a kid, he was always telling me about how synapses were working to create fast and slow type approaches, and how the brain was lazy. He used to tell me that all the time at dinner time, “The brain is lazy. You’re going to try and find a simple approach to do this and you have to think about how the brain is reacting to make decisions.”
So when I was doing my PhD I decided to do my topic on heuristics. I was fascinated by the fact that the academic literature treated the heuristics universe, which would be trend following technical analysis, as “voodoo finance.” Andrew posed a very good question to me, he said, “People in practice use these rules all the time. There must be a reason why they do.” Because of this, I said that there has to be. I combed the utility functions. I tried to find a way to explain it, but what I came to realize is just how important and how useful heuristics are in investment, and how pretty much everything that we do in practice is about heuristics.
I want to add one other point to what Roberto was saying as well. At 50/50, behavioral finance literature says that that’s your highest point of overconfidence. So if you’re working in a world where you have a 50/50 chance, that’s when your relative overconfidence is at its maximum. So that’s when we have to be a little bit more worried in what we do, and why it’s good to have humility and to try and think about how those effects – the behavioral effects -impact us.
Another thing I’d also point out that I was interested in when I was a Doctoral student, is behavioral connections to heuristics was the disposition effect. The disposition effect says that we tend to hold on to the losers and we tend to sell the winners too short. So if you think about what that does, when people are selling the winners with interesting information, you end up with some sort of momentum or trend in prices. What we do is cut the losers and buy the winners. In some sense we’re really systematizing a behavior that is hardwired into the brain by allowing our systems to make decisions that typical investors really have trouble doing.
I think that is why trend following works, and that is where I got interested in the strategy at the beginning, is trying to understand why does a heuristic work? What are the behavioral reasons that create a scenario that might make a heuristic help you to make a better choice? So, why would you use a stop loss rule? Why would you decide to go into a casino with X amount of dollars in your pocket so that you don’t come home with even less?
People who are following the disposition effect, or are victims of this disposition effect, the other side - people against who trend following makes money.
Did it take a long time for you to be convinced that this really is solid, robust?
I was actually puzzled, back when I was a PhD student, I had a colleague named Jasmina Hasanhodzic and she’s written a couple of books on technical analysis. I was really fascinated by the fact that most of the technical analysis strategies and heuristics were applying the FX. I as very shocked by that because I thought, as a student at that point, we preached the efficient market hypothesis and I thought, I can understand that maybe some sort of rule might work in some liquid market or something like that, because there’s inefficiency, and so you can figure out how to take advantage of that. But, why is it that the case that these really diligent rules, using technical information, work better in really liquid markets? That particular idea stuck in my head until I went further into doing more research on trend following, and that’s where it all started to come together. It’s about being able to be liquid and to take those loses, follow the winners and being able to do that dynamically over time, is why those type of heuristics work. They depend on that liquidity.
Sure, sure. Why don’t we spend a little bit of time, and I don’t know whether we can do that in terms of defining trend following. What I mean by that is, in the old days there was a very clear link between being a CTA and being a trend follower as almost the same thing, but it’s not so anymore. So, I wonder if you, Katy, if we can stay with you for a little while longer, is there a way to define trend following so that we don’t think that all CTAs are trend followers?
That’s a very good question. I tend to think of trend following as one of the predominant strategies in CTAs. If you really thing about it, what we’re trying to do is find any sort of indication of whether or not something is going up or down. So, you can easily widen that definition to include other data sources, cross information. You can think about all the different ways that you can use information to try and estimate is this going up, or down? If it’s not, you cut it. If it is, you continue. So really a lot of different strategies classify by that and the typical trend following definition is use an individual market’s past price, but these days we’ve gotten creative. What we can do is we can use other market’s past price, or we can create indices of markets and follow those prices; or we can even look at forward looking information in prices to try and determine what the trend is.
What about timeframe? Is there some level of timeframe where you say, “If you’re below that timeframe, you’re not trend following anymore?” You’re some sort of short term trader, but we wouldn’t define it as trend following.
I think, and I look forward to hearing what Roberto and Alex say as well, is that ultimately, in this game, it’s a numbers game. As Alex explained, you’re losing a lot and winning bigger sometimes. So what we have to do is diversify. So define as many atomic units of opportunities across markets to play the numbers game to reduce the volatility in our portfolios and try and find the best risk adjusted opportunity sets.
If I can jump in with that, we talked, in the book, about convergent/divergent and all these things. So to me it’s actually the way you follow the trade, after you get in, which defines whether you’re a trend follower or not. It’s not (this may be confusing), it’s not whether you get in after a trend. It so happens that you actually have to separate those two things.
So you could get in after a market has actually gone down and go long. I’m not saying we do this, I’m saying conceptually. But then when you go long, if you do it with some sort of a trailing stop, you are actually a trend follower in your trade. You have what we call a divergent payoff profile. So actually this is one of the… After I read the chapter, We Know What We Don’t Know, the next thing that I did (this is 25 years ago) was this kind of a piece of research which opened my eyes to a lot of these things, which is what we call a random entry system, which is where you flip a coin. This is the simplest way a practitioner can do this, the problem is they may not hire us and we don’t make any fees, but this is the simplest way that your listener or listeners can do this.
So get a coin, a proper coin with a heads and tales, flip it and let’s just say if it turns up heads, you take a long position. Let’s just say that we’re talking about one market and no big portfolios. If it ends up tails, you take a short position. If it goes heads and you take a long position and you set what we call a trailing stop. Some kind of simple calculation of maybe one standard deviation, or something like that. Let’s say the price is at $100, you flip a coin and it tells you to go long, you set a stop at let’s just say 10 points because that’s maybe some measure of volatility and off you go. So if it goes to $101, the trailing stop goes to $91 and at some point you obviously get stopped out because it will turn around by at least $10, in this case.
When that happens you flip the coin again. OK, and if you get heads again, then you go long again with another $10 stop just to keep it simple and off you go like this. Obviously if you run this in the computer you’re going to get all sorts of results and you’re going to have to average them out, etc. in some kind of simulation. You will get maybe half of the results of a professional trend following system because you will sort of (my eyes were opened to this in the early ‘90s) you actually stumble onto trends.
Now this is a philosophical argument we can have between three PhDs here, is it trend following or not? Because, I am not using any past prices, at all. I’m just flipping a coin to get in. I am only money managing… In the early ‘90s when I worked with Larry Hite he didn’t know any buzz words. He went to NYU and majored in film studies. So we didn’t know alpha, beta, optimization, even trend following wasn’t a common word. It was really just a form of risk managing your portfolio. In fact, all these meetings now, all day long people ask me about these fancy questions because they read all these text books. You’d be much better to go back 25 years ago and talk about what it is that we actually do, rather than what’s your alpha and beta, and maybe half the people don’t know what that means to begin with. Anyway, back to the coin flipping. Is this trend following or not? Let me just ask Katy or Roberto. The think I just described, yes or no, is that trend following?
Yes, it’s trend following, but you are still using the past history, once you are in a position that was decided by a random event, a coin flip, you are going to keep a positive position if the price had gone up in the recent past since you started the position. So in this sense, it is trend following. You’re following momentum to decide if you are going to exit or not. So I view the position that you make at each point in time as really more important than the act of getting to the position.
I’ll tell you quickly. I teach a class at Columbia. Before the first class I ask people a question. These are all smart students, maybe not as smart as MIT, but their pretty good (cut another one out). So I asked them, there’s three things to any trade. People make it complicated. We’re doing this taping during a major conference for the industry, and there’s all sorts of smart people, and everybody tries to make everything very complicated, maybe us included. But by the end of the day there’s only three decisions to any trade. Period: When you get in…or any investment… when you get in; when you get out; and what’s the size of your position.
So I ask in my class, “What do you think, class, is the most important of those three in and in some kind of order: getting in, getting out, or sizing the position? Invariably most people think that getting is the most important thing. Now grant you, in some investment categories it probably is the most important thing. You probably should buy real estate, if you buy some crap and you can’t get out, you have to actually understand getting in is very important. But in liquid instruments, which is more or less what we’re talking about, it’s actually the least important. So all this CNBC talk and all this talk about forecasting and everything is actually the least important. Then I make them do this exercise, so yeah, I think we all agree this is trend following and we’re obviously using history to keep the position, but we’re not using any history whatsoever to decide whether to get in.
Alex, I’d say that if you look at that example, I like that example because it divides the two apart. But going back to the example I was talking about, well trend following is really two pieces. One is how you determine your position, and then the other is that you’re consistent with the trading approach: that you cut the losses and you follow the winners. So it depends which lever you’re considering more of the trend following. I think that the decision to get out is really more of what makes a trend following position.
In this example I think the transaction cost is an important component. Like Alex mentioned, real estate, of course getting in is important because it’s not the liquid instrument that you’re getting. If liquidity is high, I really see every moment as a decision point. If you hold the position you make a decision at every moment if you’re going to continue to hold that position or if you’re going to decrease, or increase, or if you’re going to get out completely, or even if you’re going to take a reverse position. So there are different components of trend following that you can divide into getting in or getting out, or exit points, stop loss strategy, whatever. Really, I like to see every point in time as a decision moment, if you want. Of course there will be some granularity in this. Most people will use every day as a decision point.
Absolutely. A lot of people… And we’ve touch a little bit upon this, but a lot of people say that trend following is easy and there aren’t that many ways you can do it. On the other hand, I don’t think that firms like yours would spend so much money and time on developing research systems, infrastructure, if it was that easy. So what’s fact, what’s fiction about how difficult it is to be a trend follower?
That’s an excellent question, Niels. I often like to turn again to another behavioral effect. We, as individuals, tend to suffer from something called conjunction fallacy, or representative bias. What this mean is that we tend to think that two things that we think are representative of each other are highly correlated. So good companies are good investments, for example. A good company actually should be a properly priced company so therefore is not necessarily a good investment.
In our space this occurs in the following sense: investors tend to think that high correlation means similar returns. What I mean is that it’s very, very easy to replicate, just as Alex’s example showed, a strategy that has high correlation to trend. But in practice, to have a high risk adjusted return, or good performance in the space really requires acute attention to detail, making the day to day decisions in risk management, and I often am quoted in saying, “the devil is in the details.”
Yeah, absolutely. You can have two time series that have perfect correlation and they could be simulating a return series where one has a positive drift and another has a negative drift. So the correlations go back to style analysis or factor analysis. This would be reflected in the styles, effects, or principal components if you do it in a statistical basis. But there’s always an alpha left and that will distinguish different managers that may look very correlated but are really doing different things. Because the alpha is whatever is not captured by your styles, your factors, or your principal components. These are things that may be very unique, and I agree completely with Katy that Hell is in the details, the implementation – or the devil is in the details.
The details of the implementation and efficient software, making sure that there are no bugs, no interruptions due to defective software. All this type of detail is very important. This is like having a CTA based on systematic methods, just basically going to have software make your decisions. It’s very important, like in any machine shop, that all the gears and levers are in the right place. That if there is not the very good management of those details you can have catastrophic events like a wrong trade because of a bad line of code. Some event that might happen once a year, or once a decade may kick in and trigger that line of code. It’s very important to have very careful code review, testing platforms, testing protocols.
Let me… Roberto as a physicist and talks about gears and levers, let me maybe to your listeners, explain this in a little bit more plain English. Most people invest in equities. Most people will think, OK so what am I going to do here? I’m going to get that coin out and I’m going to apply a trend following system, somehow on the S&P. So you have to realize that the sharpe ratio, or the information ratio, or the reward to risk, or however you want to measure it. One market, on any trend following system is extremely low. Let’s just say, for the sake of argument it’s .1 or .2, whatever the exact number is if we subtract risk free from the sharpe ratio or not; or let’s just refer to it as return over risk, more or less, and let’s just say it’s something like .1. That is extremely low. Too low for any investor to actually rely on, too low to use, by the way, it’s too low for any energy company or McDonalds or anybody to actually not use the actual markets for hedging and rely on moving averages not to hedge future prices of oil, which creates opportunities for us.
Anyway, backing up, one market the sharpe ratio is extremely low. So anybody who’s going to try to say, OK let me use some kind of a simple trend following system to get me out of my equities, or to go short is doomed to fail. The information ratio is way too low. What makes it professional, but that’s probably what somebody is going to try to do – take a few markets and just apply it. If you take a few markets the information ratio will obviously go up, it’ll go up to .2, .3. It’s too low for any practical purposes.
So what we as professionals do is we actually aggregate these so called trend following risk premiums, if you want to use buzz words, or just simple P&L from different markets and we aggregate them. Aggregate how? By trading in many markets, a few hundred markets or such, markets that most people wouldn’t even follow or know that they even exist. All the bond markets, all the currency markets, also things around the world, and also applying different speeds, which we haven’t really talked about, but there’s different speeds of trend following. So what we do, when we say collectively as an industry, we try to take that information ratio from .1 to .5, .8, 1.0, something competitive to the information ratio of other asset classes, which is somewhere in that high 0. some high number, .9, .8, .7, it depends what you look at, and that makes it then a compelling asset to consider. But it has to be managed across a wide portfolio of markets, sectors, speeds, and not just one thing and one market. That little random entry coin flipping thing will only produce a sharpe ratio of maybe .1 for one market.
Sure, so when we hear people say that trend following is easy, it’s not that easy for sure.
No, and that’s part of the cost of entry: the fact that you need a good number of markets to be properly diversified. Futures contracts are not cheap. A futures contract typically are a few 10s of thousands of dollars, and of course, when you enter into a futures contract you only need to put the margin, but still you’re talking about you need a few million dollars after you have a proper model for trend following. So there is a cost of entry.
Sure. Another myth, so to speak, that we often hear managers mention when they’re asked how they build their trend following systems, is that often they say that the most robust systems are the simplest ones. Is that really true? What do you think Alex?
Who said it? I don’t want Roberto to tell me that I miss quoted somebody. We now have to re-tape this whole thing because now he has to go first instead of Katy. (laugh) But who said that? Was it Einstein? “Keep everything simple, but not any simpler.”
It’s something along those lines.
So yes, simple but it depends on (I don’t want to be a Bill Clinton here) but it depends on what the definition of simple is. There has to be… we use these words like robust, and all those kinds of statistical words. It has to be solid and rigorous and there’s scientists we have on staff, all of us collectively, I guess for a reason. It’s relatively speaking, simple. The very, very simple version? In my mind that’s probably a range of about 100%. So what I mean is like a factor of 2 to 1, maybe 3 to 1 between a very simple approach, a textbook system from, what is it called, The Handbook of Futures Markets, or something, on a small number of markets that you implement in. Trade Station or something like that you’ll get a result of X and a professional version across all markets implemented with market access and connectivity and low transaction cost and diversification and time frames, whatever, is probably 2 or 3 times X. You’re not going to get a factor of 100, but you’ll probably range, at the end of the day in your kind of information ratio, between .3, which you can get from that random entry in a couple of markets, up to maybe 1.1, but that’s a meaningful difference in terms of your expectations for performance going forward. Sot that’s probably about the range you’re going to get.
I would also say that we have to go back to the 50/50 situation here, again. And Alex’s points about being humble is that really, in the business we’re in where it’s very difficult to forecast and predict things, you need to be substantially proven that there exists evidence contrary to one over N, to the heuristic that we all have engrained, these simple heuristics. To give an example that one of my favorite papers in the literature was one that looked at the one over N diversification effect.
Some researchers from London Business School looking at all these really fancy asset allocations in the academic literature. What they found was that the one over N heuristic, over long time horizons, often tended to beat them. This kind of gives you some intuition about how to think about why we think so much about trying to use a very simple version. Because even if you take another simple example of the mean variance analysis… I spent several years teaching MBA courses and we sort of beat mean variance analysis to death.
In this process we do mean variance optimization and what you see is, depending on… it’s very unstable, very, very unstable. Yet we spent a lot of the time sort of teaching that. But even the research on that shows that the minimum variance portfolio, out of sample, significantly outperforms mean variance. So what this tells us is that the simpler and the more robust and simple the solution is, the better.
So Alex and I had done some work on this as well, over long time horizons, the typical market based trend following approach tends to outperform. So one over N. So you allocate equal risk to every market. That particular strategy over long time horizons does very well because it basically has no view which market is going to do better than another. If you’re going to have a view, you need to have proof as to the contrary, and proof is hard to ascertain in a world where you’re stuck at 50/50, or 51/49.
It’s the role of uncertainty that really makes mean variance not to do the job that some people expect that theoretical financial economists would expect it to do. Like Katy said, we don’t know what the expected returns are. At the most you make an informed bet that things are going to go up or down based on some statistical evidence of how they behave with respect to their past history, but it’s very hard to be any more precise than that. When you put that in a mean variance optimization engine small changes in expected returns yield very different answers.
Going back to the problem of simplicity of complexity, I think in my opinion, you should not introduce complexity gratuitously. I think Einstein said that we should be as simple as possible, but not simpler than granted by the problem – I’m paraphrasing now. Probably worse than he put it.
So the idea is that the components should be simple. The components of a model should be simple, but once you have new signals, you should introduce those signals in a way that’s not a Rube Goldberg machine, if you know what I’m talking about – those famous old comics where you have those very complicated problems to just draw a ball from point A to point B. We don’t want to do that. If you went to grab some effect you have to try first the simplest way possible, then you can introduce a bit of overlay of, let’s say granted complexity, that is granted that’s ensured by the evidence that you have, by the data evidence that this added complexity yields better results. This is always to be done very carefully because, again, you don’t want complexity for the sake…
Ready to learn more about the world's top traders? Go to TOPTRADERSUNPLUGGED.COM and sign up to receive the full transcripts of the first 10 episodes of the show and visit the show notes where you can find useful links to other amazing resources. Thanks for listening and we'll see you on the next episode of Top Traders Unplugged.