In our continued conversation with Robert Sinnott, Katy Kaminski and I dive into why AlphaSimplex has been so successful, why Robert values transparency in his firm so highly, and what he advises for new and rising-star investors. He goes deep into what he is excited about and afraid of with managed futures, and what he’s looking forward to accomplishing with AlphaSimplex in the future.
Thanks for listening and please welcome back Robert Sinnott.
In This Episode, You’ll Learn:
Resources & Links Mentioned in this Episode:
Follow Rob 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 firstname.lastname@example.org
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.
I have a few questions/observations which you may or may not agree with. So first was just a question, do you use machine learning throughout the program? Would you say it’s a machine learning, trend following strategy?
No, I wouldn’t. We have a portion of our portfolio, about forty percent of the risk of the portfolio, with partially informed machine learning. To be specific, what I’m saying is that forty percent of our portfolio is composed of trend signals that are then differentially weighted based on the outputs of these machine learning tools.
I should, again, say that these machine learning tools are not neural networks. They are things like decision trees. They are things like kernel regression, where, as a research scientist I can go in, and I can tell you exactly why this market and these trends are being weighted the way that they are. I can go in and I can tell you, “Oh, it’s because in these periods it did particularly well, or in these periods it didn’t do particularly well, and that that’s the source of the entropy that is causing this decision tree to make the decision that it is – to come out with that bifurcation, with that decision, with that rule.” So, it’s not a black box in any way, shape, or form to us and that’s critical in terms of how we develop these models.
Moreover, we’re really sensitive to the idea that any mechanism, any source of value-add is going to break. George Box, the statistician, once said (and again it’s been quoted by a whole bunch of people, but I like to say that George said this one, “All models are wrong, but some are useful.” So, when we compose our portfolio we build that portfolio with the assumption that some fraction of the models that we’re using are wrong; that some fraction of the value adds that we’re using are wrong (especially at least some of the time) like any other model.
By taking a diversified approach, by combining about thirty percent of our portfolio in very, very academically sourced classic trend following signals; about a third of our portfolio and what we consider our specialized short horizon models that don’t use machine learning but do try to add value through asset selection; and about forty percent of our portfolio is in these algorithmically informed trend models. That allows us to have more confidence that, regardless of what happens in the future, we have a fighting chance of being successful capturing the next trend or the next allocation of risk.
That’s interesting. In a sense, I’m glad about the answer that you gave because, of course, trend following comes from a fairly simple way of trading the markets and a lot of people would argue that the reason why it is successful, still, is that it’s not too complex at its core. It doesn’t mean you shouldn’t innovate. It doesn’t mean you shouldn’t evolve, but at its core, it’s not too complicated. Therefore, personally, there’s always this worry that if you over complicate something that is relatively simple that, at some point, it stops working.
Absolutely, Occam’s razor holds in trend following like in anything else. If you go into even our most complicated model, you go into it, and you look at its individual pieces. What you realize is that it’s a combination of really simple transparent steps that are just being systematically applied. It sounds like what we’re doing is really fancy, but what it comes down to is we’re just being very efficient and very systematic about how we learn those simple decision rules and how we constrain what can be learned by those simple decision rules.
Every element, every sub, sub, sub signal in our portfolio is either a binary or ternary positive, flat or negative; or it has got a continuous signal. Some of them should be somewhat positive, somewhat flat, and somewhat negative. It’s just a combination of those elements. They are all trend following. There is nothing in there that says I should be short this trend.
There are some things that say this trend is very, very strong and has gone a long way and that on a forward going basis my expected return is lower than what it would have been if I hadn’t gone through that extremely strong period. But we’re never going to fight the trend. It’s always asking the question, based on what we’ve seen in momentum, and assuming that that momentum continues, at least to some extent, what is the best portfolio that I can construct? It’s simple things layered on top of each other, tightly constrained.
So Rob, actually this is a big debate right now, as well, is that pure trend versus non-pure trend. So, it sounds like you’re more on the pure trend side, so focusing explicitly on trend. Is there a strategic reason behind this, or what is your thinking on the difference between those two, including other strategies into a trend program?
So, I’m going to break the trend (so to speak) of managers. We are a pure trend program. And, this is even crazier; our flagship program is a pure trend product. Why do we do that?
If I were to say that our flagship program was a pure trend product fifteen years ago, or ten years ago, no one would bat an eye – that would be unsurprising. That’s where the industry was.
Today that’s not the case, and there are a few reasons for that. One has to do with the business economics of a flagship program for many managers, especially those with large legacy infrastructures. Another part of it is that trend following is a rough ride, and many managers will add in other elements that will smooth that ride over time.
But, let’s think a little bit about what that means “to smooth the ride” because when we think about trend following (and I’m going to steal your words here, but I think they’re very, very good) trend following is fundamentally a divergence bet. You’re saying that markets have moved a certain degree, and they’re going to keep moving in that direction. They’re not going to revert to where they used to be. They’re not going to revert to the historical range. They are going to break out, and they will continue that move for as long as they can.really popular right now, in:
So, if you think about trend following among dislocation based strategies, it has, over the long term, provided a positive return; not just the excess return of the cash collateral but typically a bit more, especially during these periods of crisis. So, that’s, we think, the fundamental…
In the words of economics – if you were to think about the Arrow-Debreu security that is trend following, it’s a small payout (in most environments) in expectation. Hopefully, if everything works correctly (but not in a guaranteed sense), there is a big payout if things hit the fan in either direction. Things could go into a mania or a crisis, but managed futures should do well.
Other than options there are no other investments, at least that I know of, that have that same payout characteristic. So, if you were to start adding classic risk premia, if you were to add a long bias to equities, if you were even to add a long bias to commodities, those bets break down in a crisis. They are things that will muddle or cancel out part of that crisis return.
So, when we designed this product, it was for the view of let’s be a component of a portfolio that is complimentary to stocks and bonds and that traditional allocation, and if we were to add in those convergence bets (that carry, that value), we would be going against that goal. The benefit of being a young manager is that we were able to build our infrastructure, research team and everything else consistent with the program that we have. As a result, we were able to create a flagship product that was just trend following; that met that institutional need.
Where we have been able to differentiate ourselves, and where we’ve been able to (knock on wood) continue our relative degree of performance is in these value adds, in these adaptive measures, in how we follow short horizon trends, in these particular edges that we’ve created, but that is all in the context of trend following rather than striking out. I like to say that investing in any premia, like momentum, follows the 80/20 rule. You can get eighty percent of the return, especially on a correlation basis, really cheaply, with a small number of assets, and with a really simple signal.created some of them back in:
So, if you think about trend following as the core of what you do, to some degree maybe you can think that there must be a limit to how much juice we can squeeze out of trend following. What are some of the hard problems that are left to be solved in this space, in your opinion? What are the things that you really want to achieve or overcome in that particular space?
Oh my goodness, there are so many. The challenge, of course, is to know how confident you should be in any particular solution to them. Like any other manager of size, when we have someone come in and they show us their program and they want to be hired, or they want to be bought out, and they’ll show us these Sharpe two strategies because they think they’ve solved one of the critical problems in trend following and of course you look at that… Let’s put it this way probably nine hundred and ninety-nine thousand nine hundred and ninety-nine out of a million of them are probably going to be wrong.
So, just these are, literally, billion dollar questions. If you could answer any of these questions to any degree of satisfaction you could put the rest of us out of business, at least eventually.
One of the largest challenges is, when is the trend going to stop? What is the proper way of identifying when reversion is going to start? When the trend has gone from not having incorporated…
So, basic model of trend following: you started out at price level one hundred, new information comes in, prices start to rise as some market participants discover the new information, then the trend followers come in because the markets started moving, and they’ll start repricing. At some point, suppose the underlying information says that the price was at a hundred, now it should be at two hundred. At some point in the classic version, your trend is going to pass two hundred. Then it’s going to keel over, as some people really know that it is suppose to be a hundred, and they’re going to start shorting the market and it will flat-line and become range bound at two hundred.
Being able to identify when the trend has become fully valued, consistently, over time, especially in periods of crisis, would be immensely valuable and it’s an incredibly hard challenge. Because, for any bit of information, chances are you don’t know. Even if you have the bit of information that happens to be driving that market move, you don’t know that that’s the only bit of information that’s driving that market move.
That incompleteness of understanding makes the timing of reversion really tricky. It gets even worse when you go through a crisis where everyone may know that at an S&P value in the six hundreds, things are oversold, but because of liquidity needs, some people are still selling, and you don’t know when that’s going to stop. So, trying to time reversals, in a generic sense, is absolutely a source of significant value-add to a trend following strategy because that would allow you to reduce risk, at the very minimum, if not remove risk.
Another great one, how do you identify the impetus? How do you identify when a trend has begun that is going to persist rather than revert? We’ve spent a lot of time in our specialized short horizon models trying to identify which short-term signals are going to persist and which ones are going to revert.
I would say if we’re lucky, and if we assume that our in-sample results are in anyway consistent with the out of sample results, we’ve improved our success rate by a couple of percent. To do so, even at a five or ten percent margin would be huge. It would be revolutionary in this space.
Another one would be, let’s finally solve the execution question. One of the big challenges, as a trend follower in today’s day and age, is how do you execute your trades in a very, very cost-effective way? I’m not talking about clearing costs because once you get to a certain size if you can negotiate down your costs… Well, if you don’t negotiate down your costs, you’re not doing your job.
In terms of your actual execution – your market slippage, that is a very competitive game that, even if you invest a huge amount of capital in it, depending on really the supply and demand dynamic of the market at that point in time, having the best algorithms in the world won’t help you if every other CTA is trading at the same time. So, those are just a short example of some of the challenges, some of the complications that we struggle with every day, that we try to add value incrementally to, but are still illusive and, more importantly, we think that even if we get some edges, those edges are going to disappear over time.
So, maybe just a quick question about the recent growth of the firm. I thought that maybe… So, in the last few years you’ve definitely been growing, increasing your investor base. What do you think some of the success factors have been for AlphaSimplex and what has led to helping to increase your footprint in the space?
So, AlphaSimplex has been successful for a number of reasons. The biggest has been the consistency of our program, our sticking to our mandate, and our focus on transparency.
So, if you’re an investor in AlphaSimplex… Gone are the old days where you don’t know what the holdings are and doubly gone are the old days where you didn’t know why we were holding the positions that we are. If you invest in, really, any of our programs, and you want to see the holdings of your portfolio at the end of the month, that’s publicly available. If you want to understand about the positions that we hold, well, we’re entirely a trend follower.
If you look at the price chart of a typical asset, unless we’re in a tight range, you will know at least what direction we’re holding. You may not know the magnitude, but you’ll know the direction. We go through a lot of effort to try to make it very accessible as to what we do as a manager, what kinds of algorithms we use, and our goals are (let’s call them) our constructive philosophy about how we build the portfolio.
When you combine that with a very, very consistent investment process, and I’m not just talking about being a pure trend follower, though that’s helped, but also being clean and clear about our research process, about how we take ideas from either blue sky research or academic papers, build them, test them, put them in out of sample testing and then finally deploy them. That consistency is also something that has given a lot of investors, be they full discretion probably less so for your average retail investor, but certainly on the institutional side, or the investment advisor side. They want that consistency of process. They want that consistency of mandate. They want that consistency of understanding.
Then, finally, I think that we do a very, very, again, consistent job of risk management – where we look, when we think about risk. We don’t just think about volatility like many managers over the short term. We look at volatility over lots of different horizons with the view that what has just happened in the last few days may not actually be representative of what happens in the next few days.
Our deputy CIO, a guy named Alex Healy, who is one of the more intelligent people that I have ever met in my life has a saying. That saying is, “if your risk model thinks that a two percent drop in the S&P is a four sigma or five sigma move (which, in fact, it would have been last year based on short-term windows), then your risk model is wrong.” That is wholly unacceptable because the chance of a large upset like that is completely well within the realm of reason.
So, I don’t do this anymore but used to, as a kind of side gig, was an advisor to some students at Harvard College, helping them with statistics, academic advising and the like. I was aligned with one of the houses, a house called Dunster House. On the night of the election, I was running the projector, we had CNN, we had Fox News, we had MSNBC (you know, covered the full bases), and then we had the Mexican peso (my particular addition to that set).
So if you were tracking markets on the night of the election, you would remember that the S&P dropped more than five percent, the NASDAQ hit its overnight drop limit. It couldn’t trade lower. The stopping mechanisms, the fail-safes kicked in. If that happened in November and we’ve had similar instances that were very short lived more recently, it’s totally reasonable that that could happen during the day, but your short term volatility estimate might not capture that.
So, when we build our risk models, we want to make sure that they are aware of that jump risk. Whether they take into account that longer history, that we don’t just look at volatility, but we look at other moments, that we look at your value at risk, that we look at your net asset exposures, and your gross asset exposures, and things like your dv01 and your concentration in your portfolio. Again, that comprehensive approach to risk management has been another strong selling point for our clients.of:
We had a very, very negative correlation during that time and it served as a proof point. So, investors, as they become more educated about managed futures, have realized that if everything in their portfolio is all green and is always all green at the same time, well maybe they’re not well diversified and that therefore, something like managed futures (and especially one that is designed to be a compliment to a traditional portfolio, like ours) is probably going to receive a certain amount of good attention.
Sure, absolutely. Speaking of receiving good attention, we’ve also seen that some strategies have grown very quickly, in particular, some of these low-cost products, of course. Is there a limit to how much… Because you also talked about the risk of everyone doing trading at the same time and all of those things, is there a limit to how big you can be (if we talk about you specifically) before you feel that this is actually going to impact what we do? Because the range of managers, we have the hundred million dollar managers, we have the thirty-five billion dollar manager. The range is huge, and there’s always this debate about does size kill performance, or at what point does it kill performance?
So, the tiny little compliance angel on my shoulder is going to be kicking me when I say this. I would say that there is a tradeoff and it’s partially a tradeoff of how you trade. The more short horizon momentum, the more reactive your portfolio is, and the more biased you are towards an equal risk allocation to all of the assets that you trade. So, that is to say, if you want to hold the same potential exposure to soy meal as you do to the S&P 500 you’re going to have constraints as to the size. That’s going to be a function of both market liquidity as well as some of the regulatory constraints of exchanges, or of the CFTC or any number of other bodies.
So, what I would say is that if you want to have a reactive portfolio that doesn’t have a long bias, then there will be constraints as to how big your portfolio can be. I will say that our fund cannot be, at least in its current form, anything close to the size of the really large firms in this space.
We will be constrained at that point. We’re not constrained yet, but that it’s our intention to constrain ourselves before we see that degradation. Because, in fact, we do have that short horizon emphasis and we do have that bias towards trading in size some of those less liquid futures.
In terms of the industry as a whole, that’s a much trickier question, and it’s going to be something that varies over time. I think a good example of this might be in base metals. So, there are some times where base metal volumes have been very low, and there have been times where base metal volumes have been very high based on growth expectation out of China, or out of supply considerations out of South America, or any number of things. Then, based on where we are in that range of volumes, there are going to be some times where CTAs are a hefty fraction of the volume of some markets, not necessarily base markets, but of some markets, and there are going to be some times where we’re very, very little.ewEdge did a study on that in:
I think that once you account for some of the CFTC limits and whatnot, it’s not quite as bleak as they paint it, not that they paint it particularly bleak, but it’s not as bleak as they paint it, but I think it is a consideration. So, what does that mean? It means that when we think about portfolio liquidity, and asset liquidity from a risk management standpoint, yes, we care about that, but the larger concern is not so much about what fraction of total volume is being driven by CTAs, it’s what fraction of today’s volume or the current period in today, so this hour’s volume or this fifteen minute period’s volume is being driven by CTAs. In so much that you can avoid those points of concentration when you can avoid the crowding, you should, all else equal, be more successful, at least if you’re going in the same direction.
Yeah, that’s true. I think there’s the whole debate about how big is this industry? This industry has grown, but mainly because we’ve got one big company that was suddenly included as a CTA, which has a hundred and seventy billion under management. You take them out and suddenly the CTA industry, it hasn’t grown that much. So, it’s more about the participants whatever we call them. You’re right; it’s clearly do you get the liquidity that you need at the time that you need it? So, I take that on.and age, now we’re entering:
Rob: So the compelling things left in this space, fortunately, are always changing. I think we’ve had a few interesting possibilities; interesting opportunities come up in the recent past. There’s been a shift towards, what you might call, extended markets, which will be substantially less liquid but do trade different instruments.
So while it wouldn’t necessarily… It could potentially, but wouldn’t necessarily fit that crisis alpha niche. If you don’t have liquidity, it’s very hard to tactically trade in a crisis, just from first principles, but it could be very attractive. So, I think that’s something that the managed futures industry as a whole is looking at, even if it’s starting to bend the definition of a futures contract.
I think that managed futures continues to go through this constant evolution where new edges, new ideas, new behaviors of markets come out. That is the whole game, that is what wakes you up in the morning, that is what’s challenging, that is what gives you ulcers at night. It’s how are the market participants causing that aggregate macro behavior to change? That ever newness is what makes this industry interesting.
So, maybe turning that around too, Rob, what are the things that keep you up at night? What are the things that worry you?
That’s the DIY work that keeps him up at night.
Fortunately, not so much anymore. That was a very short period but a very enjoyable period of my past; now it’s infants. Other than my small son, the things that keep me up at night are the big unexpected, unknown macro shocks.
Whenever you construct a portfolio, especially managed futures, we in the US would say you’d probably use leverage. In the UK they would probably would say that you geared it to a certain volatility. But, anytime you do that you’re increasing the risk of your portfolio in order to achieve some amount of desired volatility; if you’re using the Kelly criterion to scale that, great, if you’re using a static risk allocation, my humble opinion less great, but it still works. All of that is based off of some assumption of volatility and some assumption of correlation.
As we all know, market distributions are not normal. Those fat tails exist, and they can surprise you. I think there are many instances where, in managed futures, we have all been surprised.rd,:
Those are the things that you worry about. Those changes in correlation structure, those changes in trend because for better or for worse, if you’re a trend follower, your correlation structure and your trend signals often change at the same time, and while we try to spend a lot of effort trying to mitigate the cost of that, you can’t get rid of it. It’s kind of a central weakness, a central pivot of the systems. So that’s something that you always worry about.
The second thing that you always worry about is just simple model decay. I think trend following has a lot. Momentum as a factor, as a future, has a lot of reasons why it should persist for a long period of time. Yes, it has these behavioral features, but it also has this jack-of-all-trades aspect where it can detect mispricing and capture shifts that may be informed by who knows what, but it will capture some fraction of.
That should persist in so much that markets are not perfectly informed and that price does not perfectly incorporate information over time. But, anything we do is going to degrade over time. So, the constant worry is that one, will our models decay? And then two, in our fight against that decay, in that fight for an edge have we added in some accidental overfit? Regardless of how careful you are, there is going to be some possibility there.
So, that’s what I would say; it’s big macro economic shocks that we don’t have time to capture and change direction for and model overfit. Those are the two boogie men that keep me up at night.
What about maybe not a final thought, but close to a final thought. If you’re going to share something that you feel is important right now for investors, or even a rising star, someone who wants to do what you do one day, what would that be? It could be a piece of good advice or something.
I would say four things, one take up ballroom dancing, but more seriously, I would say the first thing you should realize is that any model that you construct, any model that you read about is going to be wrong. It might be effective, it might be useful, but it’s going to be wrong.
The EMH… There are a number of critiques out there that say that if the EMH was true, it would all of a sudden fail. The joke about if you have a behavioral economist walking down the street along with an efficient markets practitioner and they look down, and they see a hundred dollar bill. The behavioral economist asks the EMH practitioner why didn’t he pick up the hundred dollar bill? He’d say well you can’t; it’s not actually there. If it was there, somebody would have already picked it up.
So, the models that you learn, the theories that you learn, none of them are going to be perfectly correct and that there’s so much different between theory and practice. That being said, models are useful. Models can give you information: can inform your views, can inform your intuitions and can inform your process. If you combine a disciplined constructed model and hopefully, a disciplined constructed view on probability, you can be, over the long term with a fair probability, successful in markets.
It’s not going to be sexy. If you find that you’ve returned two hundred percent overnight: either you’re immensely lucky, or you’re a crook, but you’re certainly not following a disciplined process – one of those two, but certainly not the third. If you’re going to have that kind of return and following a disciplined process, it should take you a very long time to develop and to hone that skill.
Then finally, if you want to go into this space, actually the best book I’ve actually ever read, the one that I love more than anything else, if you want to learn probability, I would suggest picking up Joe Blitzstein’s book on Introduction to Probability. If you want to learn about CTAs, I love Robert Carver’s book on Systematic Trading.
It doesn’t go to quite the same depth of what I think anyone really does in production, but in terms of building out the concepts, building out the intuition, building out the understanding of how you think about these programs, I think Robert Carver’s book does an excellent job, and if you want to learn about managed futures, of course, Katy Kaminski, here, sitting across the table – I can’t recommend her book enough.
I couldn’t agree more and just reminding the listeners that Robert Carver has been on the podcast so that they can go back and listen to him as has Katy, of course. From my point of view, I would just say that if you think about an investor, right, in this day and age, and with all the things that are changing, with all the things that you’ve been talking about, what’s the most important question they can be asking themselves right now?
If I knew that I could make a lot of money. I would say that the thing you have to remember more than anything else is that trying to find out the right question to ask is often as important if not more important than the right answer. If you could tell me with great confidence that you knew the right question to ask today to be successful in markets? Then you, sir, would have a very, very valuable quantity on your hands.
I would say, be humble. I would say, and I said this at the beginning, everyone thinks that there’s a lot of noise in markets, all that noise is, is a lot information and a lot of action of people who are probably behaving fairly rationally, using different objectives and different information sets than you are.
It reminds you that markets are incredibly dynamic, incredibly complex, and that noise, or interventions by the ECB, or anything else are things that we have to deal with. You can complain about them all you want, but if you want to be successful you have to be humble, you have to take what the market gives you, and you have to learn and adapt and adjust to that state of affairs, not the one that you want to happen to be the case.
Perfect, excellent, on that note let’s wrap up this fascinating conversation recorded live, here in Miami.
Rob, thank you so much for being on the podcast today and sharing your thoughts and experiences with Katy and me. It is so important that practitioners, like you, come and share these ideas because when ideas become conversations that lead to action, that’s when real change happens.
To our listeners around the world, let me finish by saying I hope you got a lot of value from today’s conversation and that if you could, share these episodes with your friends and colleagues so that the conversation can continue.
From me, Niels Kaastrup-Larsen and Katy Kaminski, thanks for listening and we look forward to being back with you on the next episode of Top Traders Unplugged. In the meantime, go check out all of the amazing free resources that you can find on the website.
Ready to learn more about the world’s Top Traders? Go to TOPTRADERSUNPLUGGED.COM and signup to receive the full transcripts of the first ten 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.