Bill Gebhardt from 10Dynamics discusses the intriguing intersection of market efficiency and systematic trading strategies. He emphasizes the idea that while traditional views suggest markets are efficient, many hedge funds continue to find success using price-based strategies that should theoretically not work. This paradox leads to a fascinating exploration of the psychological aspects of trading, particularly how human behavior can create opportunities for systematic traders. Gebhardt shares insights on the operational complexities of running a nearly continuous trading process, utilizing intraday data, and employing a diverse range of signals. The conversation explores the risks of underfitting versus overfitting in trading models, ultimately advocating for a strategy that embraces empirical success without excessive optimization.
-----
50 YEARS OF TREND FOLLOWING BOOK AND BEHIND-THE-SCENES VIDEO FOR ACCREDITED INVESTORS - CLICK HERE
-----
Follow Niels on Twitter, LinkedIn, YouTube or via the TTU website.
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 “Ten Reasons to Add Trend Following to Your Portfolio” here.
Learn more about the Trend Barometer here.
Send your questions to info@toptradersunplugged.com
And please share this episode with a like-minded friend and leave an honest Rating & Review on iTunes or Spotify so more people can discover the podcast.
Follow Moritz on Linkedin.
Follow Bill on X.
Episode TimeStamps:
02:17 - Introduction to Bill Gebhardt
06:35 - No pain, no gain
08:40 - Why are they called 10Dynamics?
10:08 - What is their philosophy behind how markets work?
12:24 - Are the signals that they run based on trend following?
13:47 - How they handle the operational complexity of their system
18:15 - How they assess the validity of their data
20:39 - Exploring the technical details behind the execution of their system
22:39 - How they approach trading times
25:07 - Combining short and long term trading
26:52 - How they avoid overfitting and underfitting
30:58 - How Gebhardt implements his experience from the energy space
33:54 - How many markets do they trade?
35:50 - Accessing the domestic futures markets
36:56 - Why are single name equities so special to 10Dynamics?
41:08 - How they approach VIX futures
43:39 - Their perspective on replication strategies
Copyright © 2024 – CMC AG – All Rights Reserved
----
PLUS: Whenever you're ready... here are 3 ways I can help you in your investment Journey:
1. eBooks that cover key topics that you need to know about
In my eBooks, I put together some key discoveries and things I have learnt during the more than 3 decades I have worked in the Trend Following industry, which I hope you will find useful. Click Here
2. Daily Trend Barometer and Market Score
One of the things I’m really proud of, is the fact that I have managed to published the Trend Barometer and Market Score each day for more than a decade...as these tools are really good at describing the environment for trend following managers as well as giving insights into the general positioning of a trend following strategy! Click Here
3. Other Resources that can help you
And if you are hungry for more useful resources from the trend following world...check out some precious resources that I have found over the years to be really valuable. Click Here
What we ignore is the fact, and it is a fact, that markets have different characteristics for sure. So, we definitely miss opportunities to specialize for a given market. Right? So, we are underfitting, as you would say.
In my career, I've never seen anyone go bankrupt because they underfit, but I've seen loads of people go bankrupt because they overfit. So, you know, in terms of the risk/reward, I'll bet on underfitting all day long and, you know, stay away from overfitting as much as possible.
Intro:Imagine spending an hour with the world's greatest traders. Imagine learning from their experiences, their successes and their failures. Imagine no more. Welcome to Top Traders Unplugged, the place where you can learn from the best hedge fund managers in the world so you can take your manager due diligence or investment career to the next level.
Before we begin today's conversation, remember to keep two things in mind. All the discussion we'll 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 product before you make investment decisions. Here's your host, veteran hedge fund manager, Niels Kaastrup-Larsen.
Niels:Welcome to another episode in the Open Interest series on Top Traders Unplugged, hosted by Moritz Siebert. In life as well as in trading, maintaining a spirit of curiosity and open mindedness is key. And this is precisely what the Open Interest series is all about.
Join Moritz as he engages in candid conversations with seasoned professionals from around the globe to uncover their insights, successes and failures, offering you a unique perspective on the investment landscape. So, with no further ado, please enjoy the conversation.
Moritz:Hello everyone, it's Moritz here with episode number 13 of the Open Interest series on Top Traders Unplugged. Today I'll be speaking with Bill Gephardt from 10Dynamics, a London based systematic trading firm. Bill has a really interesting background which can be divided into two parts.
In the earlier stages of his career, and after Having completed a PhD in Finance at Cornell University, Bill was an Assistant professor of Finance at the University of California at Davis, focusing on applied trading strategies grounded in behavioral finance theories. He then left academia to work for Barclays Global Investors, being responsible for the research and development of trading signals for their long/short equity portfolio.
he European book. And then in:Now let me also tell you why I decided to invite Bill onto the Open Interest series here. And that's because first, the way they trade is different to most other CTAs in that they run an almost continuous trading process on several different timescales, day and night. And second, because of Bill's background in the energies arena and his deep experience with spread and arbitrage trades in the power and gas markets.
Okay, let me stop it here. Bill, welcome to Open Interest.
Bill:Thanks, Moritz. So, it's a great intro. You do it much quicker than I do. I appreciate that. You hit the highlights. It's good.
Moritz:There you go. There you go. I always like that.
Well, speaking about your background, and as I've just mentioned, Bill, you started as an academic, as an assistant professor. So, you were teaching under the assumption that markets are efficient or very close to efficient, and that the trading systems which you and I run really shouldn't work in practice, but they do.
Bill:Yep.
Moritz:And so, you ended up in this space and I ended up in this space. But how do you reconcile this? You know, you've crossed the border, so to say.
Bill:Well, for sure. I think it's really one of the most amazing, I don't know, it is all of what we're doing is this giant anomaly. And, you know, it's been a few years. I haven't read any of the academic literature recently, but certainly when I did my PhD, and everything I'm aware of has been like, well, at the very least the market should be efficient enough that price based strategies shouldn't work, which is, it's the most basic form of market efficiency. And yet you have, you know, countless hedge funds that seem to be doing it successfully and doing it differently.
You talk to lots of people like ourselves, and I think you see the breadth and strategies in different timeframes. So, there's some incredible knowledge gap that we have in our understanding of how markets actually function. I think about this all the time. I feel like I've run out of great ideas to solve the problem.
I do feel like what we're doing is purely empirical. We're doing something that works without any real grounding in why it works. And I really feel like if somebody could figure out why this works, they could come up with a pretty good strategy, potentially that's better than everybody else.
But it's amazing to me that if you look at what the market efficiency type people typically do when they have something they don't understand, they say, oh, it's a risk factor. It's some undefined risk factor that investors are exposed to and so therefore it has a return.
But it's pretty hard to argue what the risk factor is in trend or CTA type strategies that all seem to be producing money over time and correlated. But where's the risk? Particularly when you look at the return profile, it tends to do well in high volatility periods, which is when the true market risk factor, that everybody who focuses on, the beta, does poorly.
So, it's actually a hedge to the traditional risk factors. Which, again, goes counter to, you know… If anything, the risk premium on trend following should be negative, not positive, and yet it is. It's positive and consistent. It's really an odd puzzle. I don't know what the answer is.
Moritz:Look, I recently listened to an interesting conversation, with Cliff Asness, say that maybe one of the reasons or a source for the edge is the fact that these systems cause pain and it's a feature and not a bug. It's kind of like by design. And because there is pain involved in the way that we trade with drawdowns, and inconvenience, and you're buying highs and you're selling lows, you're kind of like doing the opposite of what our human brains want us to do. That's the edge that just doesn't go away.
Bill:Yeah, definitely. That's what I've always thought is that it's some psychological thing, right? That’s why it's repeatable is because humans aren't changing that much. So, we do the same things, we have the same tendencies. And so therefore it's exploiting some sort of psychological trait.
But the problem is, the market efficiency idea is that if it works it should be arbitraged away, right? That people should be out-competing each other to try to make their strategy quicker or whatever. And so, the return is sort of like a true arbitrage, that if enough people do it, it should go away.
And the ironic thing is we're all doing the systematic trading where we believe that we've figured out something that's predictable and we don't want to tell anyone else what we're doing. And why do we think we don't want to tell anyone else what we're doing? Because we're afraid it'll be arbitraged away if everyone does it right.
So, somehow we all are believing that market efficiency is true. That's why we all are hiding what we're doing. But at the same time, you know, what we're doing sort of proves that market efficiency isn't true. So, one of those two things is not compatible. I'm not sure which one it is, but I think that this is kind of crazy.
You know, people are so secretive about what they're doing, and what they're doing is magic. You know, I don't know if you've looked, if you go on Wikipedia and you look up technical analysis, it's listed as a pseudoscience. So, we're all pseudoscientists out here doing technical or systematic trading, whatever you want to, however you want to brand it, which is kind of funny.
Moritz:It is.
Now, Bill, back to your firm, 10Dynamics. Why is it called 10Dynamics? I'm sure there's a reason behind it. Spill the beans, tell us why.
Bill:You know, at least the process or the history that I've been through is a bit different, I think, than more of a quantitative approach to how a lot of people end up doing systematic trading. Really, because my trading background was on the proprietary side and the more fundamental based, bottom up, kind of typical commodity trading strategy. You know, I had built these tools over time to help me with my timing particularly.
And they were all tools based on, I have a very kind of, I guess, philosophical view on how the markets actually move and work. And I built tools specifically to capture that idea.
I changed was probably about:So, they've been around for a while, but there's 10 of them. And so, 10Dynamics came from the 10 signals that we use. And it seemed to be a trend to put a number in front of your name. So, we thought, well, 10Dynamics makes sense. It's trending.
Moritz:Yeah, it sounds good to me. You just mentioned you have a view or a philosophy about how markets work. What is that philosophy?
Bill:Well, this goes back to, I think I've told this story a few times, and it goes back to my PhD. I sat in a class with a professor who's quite a famous market efficiency guy, built one of the asset pricing models that I think a lot of people work with and he did the classic, he didn’t put up anything, he said, here's a chart. What stock is this? It was that kind of a class. And everybody made their guess. And he says, oh surprise, surprise, it's a random walk. You know, it's not Apple or anything like that.
And then I went from that class and thought, well that's kind of interesting. But I went on my own and generated about a hundred different charts of random walks. And I had had trading experience before I did my PhD, and when I looked at those charts I thought, well, these don't look anything like what actual charts look like.
And so, I just felt like, well, statistically maybe that description of prices makes a lot of sense, and those kind of Brownian motion type things. But I thought there was a lot more structure than what comes out of those kind of models. And I got exposed to… I think a lot of people in our industry look at sort of Mandelbrot, and the misbehavior markets, and the fractals kind of ideas, and I really do think there's something there.
I haven't been able to really pin it down specifically, like the mathematical side of it, but it definitely influences us philosophically in terms of how we build the models and how we weight things and combine our strategies, and, as you mentioned earlier in the introduction, how we look across multiple time frames. That all comes from this fractally idea that really prices should be predictable on lots of different timeframes and there should be some self-similarity there, and things like that.
And it's been the guiding light, I guess, if they are talking about what is it that markets are actually doing? That one thing seems to work pretty well for us. I think there are other people out there doing it. But I think it's an important perspective to have when you're looking at price behavior.
Moritz:Now speaking about the systems or tying that back to the philosophy, are the signals the 10 signals that your firm runs, are they all rooted and grounded in trend? Is trend following the only thing you do, or do you also trade other signals?
Bill:I would say that of these signals, 8 out of 10 are trend or trying to identify trends where 2 out of the 10 are more reversal in nature. But because we equal weight all of our signals, what that effectively means is the reversal signals tend to act like profit taking signals.
We would never find ourselves countertrend, like truly against, you know, against a trend on any time frame. We certainly do end up lightening up our position under certain situations.
We have these separate signals, and we talk about them a little bit, you know, publicly. Our reversal, we call it ‘the reversal probability’, which is just those two reversal signals by themselves. And you can have moments where markets are trending strongly and the reversal probability is very low. And then you can have times where, actually, the markets are trending strongly, the reversal probability is really high.
So, that feeds into our overall positioning that, as we scale are based on the different indicators and the different signals and how they offset each other.
Moritz:Got it.
Now, one of the interesting or one of the many interesting things that you do is that you run an almost continuous trading process. At least this is my understanding when I had a look at the materials that you shared with me and the pre-call that we had. You are using intraday data, you're pulling it in all the time, you are then resampling it (I think every 30 minutes if I remember that correctly), and then you trade on multiple different time frames from 30 minutes or like very short term to two-week lookbacks, all of that.
Now this comes with a lot of operational complexity because you need to have the engine and the machines running 24/7 or 24/5. How do you, I mean, just give us some background on that. I mean, how do you handle all this? This sounds like a big tech lift.
Bill:It is, it is a big tech lift. And probably we went in a bit naïve, or certainly I did, because I hadn't really implemented something, as you say, it as complex.
It's the whole self-driving car issue, right? It's, you know, if everything's going fine, it's great. As soon as a rabbit runs across the road, you know what happens? It's the same with what we're doing. And to be honest, we just got, I think a couple lucky things happened.
We built this system in the era of Python and object-oriented coding. Which immediately lent itself to a concept where we created this validation idea on all of our objects. And so, every object is continually being tested to see if it's valid or not.
And then you come up with, you basically have to walk through all your process and say for every step in the process. There's an input and an output. How can the inputs be wrong and how can the outputs be wrong?
And there are lots of different dimensions that you're looking at. Like how frequently do things change, how much do they change, are there missing values? It’s all that sort of stuff. But that's for every different type of object. And then all your objects interact with each other. So, you have this chaining that happens between everything.
So, you're constantly evaluating the validity of each of the objects inside your system and all the instances of those objects, and then catching when one of those objects is behaving abnormally in some way, and then you kind of pass it to… Well, it's a semi-automated system where it goes into another system that tries to decide if it can figure out what's wrong, and if it can, then it can fix it. If it doesn't, then it gets elevated for the humans to deal with.
So, we have a lot of different messaging, and that sort of stuff set up so that we're constantly in contact with the machine, if you will. It's telling us what's going on and if there's anything abnormal.
I think we got kind of lucky really, if I look at that we certainly didn't go into it with that design idea in mind. But what's happened is it's really a robust way to do it. Honestly, I don't know how other people do it, so I don't know if it's unique or not, but it really provides a huge amount of control and confidence to be able to do that.
And if you don't do that, I think it's so easy, if you're going to trade on an automated basis, as you say, throughout the day and on a relatively frequent interval, then the potential for problems is there. And you have to be quite careful about dealing with it. At the beginning, as we were figuring it out, it took quite a lot of testing to, and as you do, try to break the system in every way you can think of.
But even things like that we have a database. Everybody has a database. What happens if you make a query to the database and the data you get back is corrupted or unexpected in some way, or maybe for some reason you get something old in there. And all those… We're now up to like hundreds of checks that we run across all the hundreds of different instances of our objects. And so, it's quite robust in the way it handles that.
I mean, you can of probably tell I'm a bit proud of it. If I had to rank the system versus the control of the system, which one I'm more proud of? Actually, I don't know. It's a tossup. I think they're both super important when you're trading in an automated fashion.
Moritz:Look, I mean, I don't think it necessarily makes 10Dynamics unique. I guess your setup is unique because nobody else has built it in an identical way.
But you know, Most of the CTAs out there or most of the funds in our space, they would trade a daily bar. They wouldn't necessarily dissect the day into 30-minute intervals or use tick data. So, one of the complexities that comes with that is, you know, garbage in, garbage out. I mean, do you have a continuous process of checking the validity of the data, the cleanliness of the data? How does that work?
Bill:Yeah, that's all part of the ongoing checks that are happening. So, we have like for instance, we have a raw data object which is the data that comes directly from the exchange. That has its own validation and cleaning process that it goes through. And you have different checks around volatility and gaps, and all that sort of stuff, to make sure that things are coming correctly.
And, you're right. I think running everything on a 30-minute basis versus a daily kind of process is an order of magnitude more complex. But I don't know, I've never been involved in high frequency trading. I can only imagine that that's even another magnitude of order more complex in how you have to manage things. Because while we do run 30 minutes, we're still batching, we're not streaming.
So, I think if you're in a streaming data setup and you're doing all low-latency stuff, I think then, yeah, you better have some pretty robust systems to do it. So, I wouldn't even want to have to tackle that. I'm glad that we're… I think 30 minutes, honestly, is kind of the limit for what we do. We can see, we look at the profitability portrayed by time frame. It's a monotonic increasing function.
So, the shortest timeframes are the least profitable and the longest timeframes are the most profitable. So, I think you reach a limit where capturing breakouts or capturing whatever you want to call a trend, capturing movements in a particular direction gets hard. Once your frequency gets too low, then it becomes all about microstructure and bid ask stuff and all that kind of stuff that we're not really interested in or involved in.
Moritz:So, that's interesting. You just mentioned that you're not streaming but you're batching. You're batching 30-minute blocks. Right? Does that mean that the quickest your machine needs to run is kind of like a 30-minute time frame? Because you would get one 30-minute batch in and say, you now have a 30 minute open, high/low, close and maybe you're signals goes, okay, buy or sell on the next 30 minute close.
So, you would, in that example you would have 30 minutes for the machine to run the next set, come up with the new signals, and then put it into your execution framework, which I presume is the way you trade probably fixed based and automatic. So, you don't have humans pointing and clicking into the order book. You just, every 30 minutes, run the machine.
Bill:Yeah, we run it every 30 minutes. We do, like a lot of people do now, we do a huge amount of parallel processing. So, we kind of like run each asset as an independent thing. So, we run all the assets at the same time, every 30 minutes, and we can process everything in about five minutes. And that's the entire signal space plus all of the adjustments to all the positions and everything. So, we're pretty quick, I think, in how we digest the data.
So really, it's like you run every 30 minutes, you run for five minutes, execute orders, check to see what comes back. There is, as you said, if you're going to do this, you do need to be doing some sort of auto execution. We use algos, but you do have to monitor if things are not getting filled or whatever, depending on liquidity. So that you are doing stuff in the intervening 30 minutes with your orders; you're using limits and stuff and adjusting them based on market conditions. So, the process around the execution, I would say, is more of a continuous process. But the actual system itself is, yeah, a 30-minute batch.
Moritz:When we look at the data, certainly in equity markets, but also in other markets, throughout the day what we see is that there is liquidity clustering at the open and at the close or at settlement in the case of a futures contract. But there's also volatility clustering. Most of the noise happens at the open and at the close.
This is kind of like where the action is, and you have a reduction of volatility on average between these two points. Now, with your system, because you're not necessarily targeting the settlement or the open, you'd be whatever, trading at midday.
y, I'm settling at, whatever,: Bill:Well, that's, you know, it's actually an ongoing research area for us because, as you said, the close to open gap… So, if you're going to trade daily and then you wait for the close to get your signal and then you trade on the open, which everybody does, I think there are definite strategies to improve on that.
I think that is an easy way to give away some money on everybody trading at the open when the signals are coming in on the close from the prior day. So, we're testing things. There are some interesting results like daily bars, close to close. But you could do a daily bar from, as you said, midday to midday. It's a bar, and do we really think that running midday to midday is going to be a worse performance than close to close over time? You know, I don't know, but it might be better when you consider the close to open gap.
Bill:So, we are looking at things like definitely it looks like potentially delaying execution also can work better even though you do get a reduction in volume or in liquidity. Now the one thing I will say, we're small enough now, we don't have any liquidity issues anywhere in anything we trade. We're not really worried about liquidity. It's a future problem we hope to have someday. But for now, at our current size, the liquidity is fine because we trade such a diversified basket too. That's what helps.
Moritz:You also mentioned that, on average, per day, you have 0.7 to 0.8 trades per instrument. That's kind of like you're touching every instrument almost once every day, doing something with it. And then you have an average hold period of two weeks. So, I guess it's fair to put you more in the short-term trader’s bucket. Would you say that's fair? I mean it doesn't sound like a long-term trend following strategy.
Bill:It's a little bit… There's some nuance to it.
So, because what we're doing, you know, the way the model works and because we're trading so many time frames at once, we're effectively scaling in and out of our positions all the time. So, you're right. So, we're making adjustments to our positions of about 5% to maybe 10% on each day.
So, something like as you said, 70% of our assets will trade each day. The adjustment of the max position will be seen at 5% to 10%. So, we're kind of continually scaling in and out of things.
And when we say our turnover is two weeks, that is specifically that change from, say, we're 10% long an asset, and then we move to 15% long, and then back to 10%. That is a two-week holding period.
But if you look at our net position, so on a given asset, net-long or net-short, how long are we net-long or net-short? That's more like two and a half months.
So, when people ask about our holding period, it's like, well, if you're talking about our net position, that's more like one to two months, which is long-term, but our trading is actually shorter-term. So, our turnover looks more like a two-week type turnover.
Moritz:The other thing that kind of like jumped at me when I looked at some of the documents you shared with me is you're saying there's zero parameter optimization going on and no overfitting. And I was kind of like, well, yeah, that's nice, but it's also very difficult or maybe even impossible because as soon as you start working with historical data, I mean, you're working with that data set.
What's your view on this? I mean, how do you not over-optimize anything? How do you avoid overfitting? Also maybe speak, in that same context, about the risk of underfitting, which I think is also a challenge, you know, not fitting enough. Yeah, just share your thoughts on this, please.
Bill:Yeah, I think this is one of those areas where we're maybe a bit different. And you're right, I mean, our signals have been developed over time in real prop trading situations. So, for sure they've been influenced by the success or weaknesses of those signals over time.
But the signals, the way the signals work are they're not like moving averages that have a time frame, or breakouts that are looking at the last X number of bars. Those types of parameters aren't in any of the signals that we use because our signals are self-referential, if that makes any sense. So basically, we take the noise out of the prices we kind of create. We take a poly series, and we turn it into these linear segments of up and down moves and then all of our signals look at the last three segments.
So, if you're in an up move right now on a time frame, then you have the prior down move, whatever that was, and you have the up move that preceded it. And so, all of our signals are looking at those three segments and determining if just on a true/false basis, are they changing in some way?
So, you know, if we're looking at a simple thing as well, how long have we been going up? If we've been going up longer than we went down, then we'd consider that bullish, let's say. So, there's no parameter there. The only parameter comes in the process of creating the linear segments. And that's really just a volatility-based thing that we do.
Yeah, I guess you'd say, well you're picking some sort of number for that. We're just using sort of basic statistics. We're not doing anything fancy there. But we haven't optimized what the exact volatility parameter is that we should use in every market to create these charts. We use the same one in everything and it just comes from statistics. There's no back fitting in that sense.
So, when we say we don't optimize parameters, we don't really have parameters to optimize. You're right though. Like what we ignore is the fact, and it is a fact, that markets have different characteristics for sure. So, we definitely miss opportunities to specialize for a given market. So, we are underfitting, as you would say. But in my career I've never seen anyone go bankrupt because they underfit. But I've seen loads of people go bankrupt because they overfit.
So, you know, in terms of the risk reward, I'll bet on underfitting all day long and stay away from overfitting as much as possible.
Moritz:Roger that.
I think what you've just mentioned with that up, down, up, or zigzag type of move, ties it to the fractals or could possibly tie to the fractal theory and the Mandelbrot theories that you were mentioning earlier where you actually do not need a parameter such as a 100 day high that you need to observe. So, that parameter just doesn't exist.
Bill:Exactly. Yeah, that's exactly right.
Moritz:Great.
So, now, let's maybe tie it, or try to bring the conversation and combine it with your background in the energy space because I think that's super interesting. And you have just mentioned, Bill, that the systems that you have implemented and decided to implement, they have a background and a history in some of the prop stuff that you've done on the desk trading for Deutsche Bank or Trailstone.
There’s so much interesting stuff going on in these markets like in coal, and gas, and dirty spark spreads, and clean spark spreads, and emissions, and now LNG is becoming a bigger market. I mean do these markets, do you still trade them? Did you take, or did you bring what you've learned, your experiences from Trailstone and Deutsche over to 10Dynamics so that you're now trading these spread and arbitrage trades still?
Bill:Yeah, I mean, definitely my experience in energy has been hugely influential in the signals for sure, in how we use them, definitely. Also, just in my history of being involved on the fundamental side for a long time and then seeing how challenging that space has become and how the competition for fundamental information edge has gotten very, very intense. And I think it's much harder today than it was 15 years ago because everyone's using the same models now, everyone gets the same data.
So, it's quite hard to have a new idea where, when I look at the return on systematic strategies they have this kind of clear cycle of outperformance and underperformance over time; over kind of multi-year timeframes. I think every sector has this natural sort of trendy environment that lasts for a few years and then not for a while.
d say the Sharpe ratio around:So, for sure, that experience was very influential in my sort of belief that what we're doing is the right approach. And you know, I used to think, as I was developing the signals that I thought, well, there were times that I thought the signals (because we didn't have the full suite at the time or whatever) that there were times where, as a human, I could do a better job than a fully automated system. But I don't believe that anymore.
And I actually, I joke, I don't tie my shoes if the system doesn't say to do it. So, I really think the system does as good a job as I could do in estimating the probability for a given market move.
Moritz:I forgot to ask, how many markets do you guys trade? Do you have a specific focus on the energies complex and power and gas and all that, or are you completely diverse? I mean, how large is the portfolio?
Bill:Well, ideally, we want it to be as large as possible because again, the whole approach is to have something that's generic. And by taking the risk of, say, underfitting, we can make up for that through diversification. So, if the model is successful in lots of different markets, then the more markets we add, the more our Sharpe ratio improves.
What we're currently running on is, you know, our current managed accounts are trading in the energy space. That's where we started primarily, I think because that's my background. So, that's where it was the easiest place to get investors interested in what we were doing.
But we've also recently, we've added the Chinese futures markets to what we cover with the model. We have some interest there, which I think it's great. It's such a broad spectrum of markets. You can trade there, and the liquidity is pretty good. It’s just difficult, as everybody knows, it's difficult to access and that's the challenge. But yeah, we'd love to run it as broadly as we can.
We do all the commodities markets. We do the Treasuries and European bond futures. We do stock indices with it. We do FX futures. So, anywhere that we can find liquid futures, we tend to apply the model and continue to look for more markets.
Unfortunately, on the futures side, we're a little bit limited. China is kind of the last real liquid environment, I think, where we can extend things.
But yeah, in general we'd like to apply it to as many places as we can and bet on the consistency and that all markets should trend at some point, and the model should capture that when it happens.
Moritz:Are you trading the internationalized Chinese futures markets which are easier to get to because you can margin them in dollars? Or are you trading the really the kind of like onshore Chinese markets which are still kind of like blocked away from most users - most offshore users?
Bill:We're planning on using a WFOE to access the domestic futures markets which is, you know, again, it has all of its own challenges and risks and stuff. If somebody wanted to do the dollar funded ones, we can certainly do it. There's just not as many of them.
And the nice thing, I don't know how closely you've looked at the domestic ones, but I think there's 45 contracts that are liquid there. Interestingly, they have a lot of contracts that we don't have. You know, a lot of the chemicals contracts that you have available there are quite interesting. It's a shame because it is a really nice broad portfolio. It would be amazing if it was a little bit easier for everybody to trade. But we're going to make an effort and see if we can access it.
Moritz:Absolutely. What I also noticed, Bill, is that you included some single name large cap US equities and sector ETFs in your portfolio. And I just want to ask, is this in addition to the equity index futures or is it a replacement for equity index future risk? Or, maybe put differently, what interests you in the single name equities? What's special about them for you?
Bill:This is another one of the things that we do that, I guess, more of the traditional CTAs or systematics don't as much, but we can apply the model to single name equities and ETFs. And it works just as well in those markets as it does in the future space.
And the reason to include them, and you know this if you have ever done any portfolio construction, that equities tend to do very well in times when maybe the commodity markets aren't doing as well. And so there is a natural diversification. If you trade single name equities, the years that you do really well tend to be the years that the future side is not doing as well. I think there's real macro reasons for that. So, I think that will continue to be the case.
I think the best portfolio would be the combination of the two. And we trade large cap equities and the liquid ETFs that are out there for kind of sector ETFs. also the commodity ETFs, also the VIX related ETFs because we trade VIX futures as well.
The one thing that we do see, and I don't know if maybe other people will disagree with this, but for our model we want things to behave quite normally in the sense that things should trend and not be too jumpy. And so, we do apply it to certain spreads in the commodity market. We don't apply it to spreads that tend to be stable and then have some sort of dislocating event that moves the spread because I think that's not really what the model is designed to catch.
So, what's interesting in the equity space is the system works very well in really heavily traded, highly followed names where you get good volatility and you've got liquidity and everything. But they tend not to be super gappy. Even on earnings announcements they do a bit, but not so much. Where, when you get down into the mid cap space, the mid cap and the small caps become really just beta plus earnings announcements plus noise kind of thing. So that's a tougher space, I think, for trend following.
So, we limit ourselves to the, basically the S&P 500, and even in there we don't do all the industries. We stay away from things like biotech, and things that are, again, more jump related; news event related. But, in general, it works great with things like Nvidia and some of the big trending names we've talked about. It is a really nice complement to the futures side.
And we asked about stock indices versus the single names. What we see is, because we don't constrain the system on the equity side, we allow it to do whatever it wants to do, that means you essentially get factor timing in there. So, you trade a hundred stocks, stocks are going up, you're going to be 80% long, 20% short in that environment. So, you're going to have beta exposure.
When stocks are going down, when you're in a bear market, you're going to have the opposite. So, you're going to be 80% short, 20% long. So, you've got time varying factor exposure. So, you're definitely betting on the factors. And if you look at the return on the strategy, you make money betting on the factors, but you also make money on the individual stock selection.
So, it's a, it's a different approach and it's not, as you know, about the equity space, it tends to be like market neutral and everybody has to be market neutral at the end of the day. So, it doesn't really fit that kind of strategy.
So, I think it's a harder thing for most investors to get their head around, what we're doing there. But you know, from my perspective, it's just a great fit from the futures side that, ideally, we'll get to the point where that will be our flagship strategy that will be something like 50% equities, 50% futures, and run it globally. That would be the dream when we get there.
Moritz:Yes. Good.
Now, you just mentioned the VIX futures, and the question I always like to ask is, do you constrain the model on just trading them on the long side or would you just, as well, be short VIX futures? How do you do that?
Bill:Yeah, we do constrain it. We don't go short the VIX futures. And, in general, I have a very strong philosophy, or I guess preference, against negative convexity tail risk type strategies, which is why I don't like mean reversion in general because that tends to be a negative convexity strategy. So, if you look at our returns, we have positive skew in our return profile, which is great. I love that. That's where I want to be.
And so, for the VIX, we're able, on the VIX side, the system historically has performed on a net basis by buying VIX. So, that's basically saying you're buying insurance and getting paid to do it. And that to me is like a great addition to the system.
And it works as you would expect. When you don't have a big event, you tend to be sort of bleeding a little bit on the VIX, but hopefully it's a lot less than the time decay you would get by just being long VIX. That's the idea behind the strategy.
And then of course when it performs, you know, you're picking up a lot of return there. And I think it's a great addition. It's hugely popular. There's a bunch of people doing it now. I think, for us, it makes a lot of sense, and I just don’t want to be short VIX because, you know, we haven't really seen a complete out-of-the-blue VIX spike.
I mean it moves fast when it moves, but it's never been completely like you wake up one day and the VIX is at 60, right? It goes from 15 to 60 over a few days or weeks at least. But I just don't want to be short that.
You know, you don't want to walk in and lose 20% of your portfolio on a day just because you're short VIX. So, we’re happy to stay away from that, even though, historically. it looks pretty good. Short VIX looks pretty good historically, but yeah, that's for braver people than me.
Moritz:Yeah, look, I mean selling out of the money puts on the S&P 500 also looks good historically. The question is whether you can maintain the path. You look at that ensemble, but maybe at some point the pain is too large, and if you're the one who kind of like has an ankle point in a drawdown, you're no longer part of that average, and your return profile and experience will look very different, and maybe you will never make your losses back. So, it is a tough one.
Now, I was just thinking, with the way that you trade and the 30-minute resampling, and the big machine that you have running, I was just thinking, as I'm sure you know, there are several trend replicating ETFs that have come up in recent months, years. First, what do you think of them? Then second, do you think 10Dynamics could ever be replicated given the way that you trade?
Bill:Well, this comes a little bit down to what we were talking about at the very beginning. You know, is it a factor? Like, if it's a factor, can people figure out what the factor is and replicate it like we do with beta?
So, in, in theory, yes, I think that's possible. Although it seems to be a lot harder because the replicating guys, there's been some good strides lately to do that, but actually, the shelf ratio is relatively low.
So, we've been thinking about it because people are talking about, well, is trend a commodity? We thought well okay, well if trends a commodity then we're going to create a market neutral, systematic, where we're long our strategy and we're short one of these replicators. We've already looked at the return profiles and it drops our Sharpe a bit. Certainly, we still have completely uncorrelated Sharpe with the replicator. So, to the extent that they have captured whatever the trend is, then yeah, that's fine. I don't think that's an unreasonable approach for systematic strategies.
But the caveat to that is, you know, we've talked to a few allocators and stuff that see a lot of systematic strategies. They said, one of the characteristics of the systematic space is the huge dispersion in returns. So yes, there seems to be an average, call it trend, but there's a massive dispersion. Look at beta. With beta there's no dispersion. There was a very little dispersion around the beta strategy. You kind of know what it is. So, I don't know. Again, what are we actually doing with systematic? What are we capturing?
And so, any investor who says, oh well, you're just commoditized trend, I'll say, well, then let's pick your favorite trend replicating strategy and we'll go short that against us and let's see how that looks. I guess that would be our sort of way of dealing with that.
Moritz:Yeah, maybe this is a great point to end the conversation. Is trend following a factor? Is it beta? Is it alpha? Can it easily be replicated?
We've heard many times, or some people say that trend following is a commoditized beta return stream. I don't think it is true. That is my opinion because we're long and short markets at different points in time. We go in and out. So, we're kind of like actively decoupling from the beta return stream. And whatever it is that we're producing cannot just be beta given the fact that we are trading on both sides at different points in time.
But yeah, leave it there. Bill, it's been great chatting with you, good trading conversation and thank you for joining me today on the Open Interest series. And a quick note to our podcast listeners. As usual, I say this all the time. We'll include the most important points of today's discussion with Bill in our Show Notes. And please, if you have any questions, be that for Bill or I, just email us. You can reach us at info@toptradersunplugged.com and will always pick it up and respond.
ard to being back with you in: Ending:Thanks for listening to Top Traders Unplugged. If you feel you learned something of value from today's episode, the best way to stay updated is to go on over to iTunes and subscribe to the show so that you'll be sure to get all the new episodes as they're released. We have some amazing guests lined up for you, and to ensure our show continues to grow, please leave us an honest rating and review in iTunes. It only takes a minute and it's the best way to show us you love the podcast. We'll see you next time on Top Traders Unplugged.