Episode 51: Exploring Dual Momentum with Gary Antonacci
59:11 Share

Welcome to Fill the Gap, the official podcast series of the CMT Association, hosted by David Lundgren and Tyler Wood. This monthly podcast will bring veteran market analysts and money managers into conversations that will explore the interviewee's investment philosophy, their process, and decision-making tools.

By learning more about their key mentors, early influences, and their long careers in financial services, Fill the Gap will highlight lessons our guests have learned over many decades and multiple market cycles. Join us in conversation with the men and women of Wall Street who discovered, engineered, and refined the discipline of technical market analysis. ♪

Fill the Gap is brought to you with support from Optima, a professional charting and data analytics platform. Whether you're a professional analyst, portfolio manager, or trader, Optima provides advanced technical and quantitative software to help you discover financial opportunities. Candidates in the CMT program gain free access to these powerful tools during the course of their study. Learn more at Optima.com. ♪

Hello and welcome to episode 51 of Fill the Gap, the official podcast of CMT Association. My name is Tyler Wood and I'm joined as always by Dave Lundgren, although he's a little tired today. How was your flight home from France, my friend? Oh my goodness. Yeah, lots of travel.

Lots of things going on business-wise. I apologize if I fall asleep in the middle of this discussion, but all good. It's all good stuff. Super exciting. And I'm really happy to talk about this podcast because this was a really, for me, it was a highlight. Absolutely.

Absolutely. For those of you who joined us in San Francisco, California, in early April, you got firsthand views, front row seats to an incredible lineup of speakers, Gina Martin-Adams, Jeff DeGraff. We had Chris Kane. And this session was filmed live at the conference between Dave Lundgren and Gary Antonacci. And

My goodness, I've known Gary at least for 10 years, but it was an eye-opener to see how much new insight he has gleaned from the markets, all available in his forthcoming book. But Dave, talk to us a little bit about the conversation you had with Gary. Obviously, most of our listeners are going to be familiar with his dual momentum concept, looking at relative and absolute. But where did the conversation jump to with that as the launchpad?

Well, I mean, first of all, I would just say that if you haven't read his book, Dual Momentum, you should. And when his next book comes out, it's going to be sort of an update of that whole thing, which is something I want to discuss as well. But to me, I mean, just the idea that he literally in his early career worked with and allocated to what is essentially the sluggish row of trend followers back in the day. So he worked with Paul Tudor Jones, Richard Dennis, Louis Bacon, Monroe Trout, and others,

Of course, at the time, we could not have known who they would become, but he clearly had a skill in identifying strong future trend followers. And I mean, that just must have been...

such an amazing experience to be able to do that, to have conversations and track their performance and just learn from them and what they do and how they do it, see the differences and cross asset classes, timeframes, et cetera. So I really wanted to spend a little bit of time on that just because that's such a unique circumstance. We should definitely pull on that thread, right? But the other thing I thought was really interesting was his book, Dual Momentum, was really a

Almost like a foundational book, because it's not like it's rocket science. It's just really straightforward as all momentum in trend following strategy should be. They should be just straightforward. You almost don't want to say plain vanilla, but that's really what the more simple they are, the better. And so his book was was foundational in many ways in the sense that it just shed light on just how powerful these concepts are.

But but what he has spoke about is that is how much he's learned since then. And the book wouldn't be written the same way as he originally wrote it if he were to write it again. And we'll find out why in this in his next book. Right. How different will it be? But the idea is that when you start tweaking with things at work based on experience, there's always the risk that you're just making. It's almost like you're you're adjusting things because of discomfort or something.

You know, it's almost, you know, messing with your back tests and things like that. So I wanted to dig a little bit on that and find out.

What processes what lessons he learned and then what processes he went through to ensure that what these adjustments he was making were still grounded in statistical significance and and you know proper modeling and things like that and You know, it just came away with I can't wait to get this book I mean, it's yeah, it was a great discussion But it really got me excited about what what a guy like this has been working on for the past. What I

15 years is that how long the book was 10 15 years what's he been working on and he showed me some of the models we talked about them a little bit in the in the in the uh fill the gap but

Really, really innovative, but again, simple. Yes. Yes. So great stuff. And I feel like the the lessons he was sharing, you know, learning from the pandemic crash in 2020 as a moment of awakening for how dramatic some whipsaws can be if you're, you know, completely wedded to a monthly rebalancing process.

Rather than tweak his models, as you say, he was thinking about strategy diversification, that you could run in parallel things that are shorter time-frame,

even volatility capture for mean reversion against a trend following strategy. And it was just the perfect session in the day, which included John Lewis, noted trend follower with the Dorsey Wright team, but also some Boston-based friends of yours who run a similar thing. How do you control for the volatility in a high momentum strategy? And I think for the audience, it was incredible.

It was a really great session to tie a lot of the content from the conference all together. So with that, I'll invite our listeners to have a rouge cochon or whatever they call Red Bull in France. Dave can inform us. And enjoy this episode with Gary Antonacci here on Fill the Gap. So it is my distinct pleasure to introduce to you Gary Antonacci. Hello.

A dear friend of the organization, George Shade, wrote to me a few months ago and he said, Tyler, I see that you're going to do something in San Francisco. You have got to reach out to my friend Gary. I said, oh, of course I know Gary. Bill Kelleher and I had the pleasure of having him at the 2017 New York Symposium. He said, yeah, but San Francisco is a lot closer to where Gary lives. This one's a layup, so have him there. And I actually brought his bio with me because I want to get this right.

In 2017, he presented some of the ideas in Dual Momentum. Raise your hand. Do you all own that book? I got at least two copies on my shelves. That was perhaps the elementary education in Momentum. And since that time, Gary describes having gone through a PhD program in understanding Momentum. He is the founder of Optimal Momentum, has over 50 years of experience as an investment professional, focusing on underexploited investment opportunities.

After receiving his MBA from Harvard Business School, Gary concentrated on developing innovative investment strategies based on academic research. His research on momentum investing was the first place winner in 2012 and the second place winner in 2011 of the Founders Award from the National Association of Active Investment Managers. Gary introduced the world to dual momentum, combining relative strength performance and

Relative Strength Price Momentum with Trend Following Absolute Momentum. He's the author of the award-winning book, Dual Momentum Investing, an innovative approach for higher returns with lower risk. And Gary now focuses on constructing diverse multi-model portfolios utilizing dual momentum and trend. You can find out more about Gary and his work at OptimalMomentum.com. It's my distinct pleasure to welcome Gary and my dear friend Dave Lundgren to share some insights with us. Please help me in welcoming them to the stage. Thank you.

Okay, testing? Yeah, you can hear me. I can hear me. So we're good? Good. Everybody good? Okay, so I had a bunch of questions planned for you, as you know. We spoke earlier, and then John Bollinger came up and gave me the advice to give you hell and make it really hard for you. So I'm just going to throw this away, and we're going to go a different route, if that's okay. Sure. Now, we will start with...

I mean, we're such a fortunate place to be able to have somebody like you as a member of our community contributing your thoughts and your research and models and innovation. And it's especially pertinent because you started in the 70s in this business. And so from that starting point, you've worked with some incredible people, some of the legends of the business.

You're a legend in the business now yourself. You may not believe it, but you are. By the way, not enough hands went up when it was asked if you bought this book. You should definitely get this book. I've read it two or three times, and I just read it again just to prepare for this discussion. And so I really highly recommend it. But let's get a quick understanding of the beginning of your career. What got you into the business? And then perhaps most importantly, what got you tilting towards technical versus what other things you could have chosen?

I started in the investment business in 1974 with Merrill Lynch as a broker because nobody else wanted to at that time.

And the markets were terrible. So I tried to look for other things going on. So I focused on stock options and gold and gold stocks. And I worked hard. I was in the office late hours. And the best producer in the office was there late hours as well. So we developed some rapport. And one day he calls me over and he says, hey, kid, you want to know the secret of why I do so well? I said, sure.

So he reaches down into his desk where he had hidden away the Mansfield stock charts. And when you open it up, there are all the Dow stocks and there's a 39 week moving average. And he explained to me how he used that

And that got me hooked on technical analysis and I tried to learn everything I could about it. There wasn't a lot out then. There was some books on point and figure. There was Edwards and McGee. There was Dow theory. So I read Dow and Hamilton and Rhea and Schaefer and Richard Russell. In fact, when Richard Russell passed away, I was one of the writers on his letter for a while.

And most of the technical stuff was being done in the commodities area. So there was this person named Dick Donchin, who you should have heard of. Sure. Being in CMT. And he put out a weekly letter. It actually revolved around his moving average crossover system. I believe it was five and 25 days. But he also had a four-week system, which is channel breakout one.

And that's what people seem to remember now. Some years later, Chester Keltner developed his channel breakout system as well.

So later on, after I went to Harvard Business School, I wanted to be a money manager, so I went there. And one of my professors there was building econometric forecasting models for futures, and he asked me to handle the technical side. So I would look at the charts and other indicators, and watching his signals, I learned the truth of Keynes' theory

saying years ago where he said the markets could be wrong longer than you could solve them. So I stopped doing that and figured I would just have my own hedge funds to trade technically. And I knew there were some really great traders out there like Paul Tudor Jones, Richard Dennis,

Louis Bacon, Monroe Trout, John Henry. So I used them. I put them all together into several hedge funds. And I handled the portfolio optimization part of it.

Back then nobody was using Markowitz mean variance optimization. So we wrote our own code and we learned quickly why nobody was using it. It doesn't work very well in the real world because the inputs tend to be unstable, especially the returns. And when you do the matrix inversions, the errors get multiplied. So the allocations were going all over the place. And so we developed a Bayesian approach

using equal weights, some MVO and the idea that Paul Tudor Jones is better than everyone else. And then later on that became Jim Simons was better than everyone else. - Who you allocated to as well? - Yeah, the hedge funds did. My business was taken over by a brokerage firm

So that's when I just took some time off, but I kept an eye on everything, all the research that I could find, especially on the academic side, because that tended to be very thorough and would go back as far as possible. And so... Before we go on from there, because we can't...

just let pass the fact that you allocated to all these legendary investors. I mean, what was it like at the time you had this opportunity to invest in what was basically the future Hall of Famers in the business all at once? I mean, did you know that that's what you had before you at the time, or is it just... Well, you knew they were good. Or were you just a really good manager selection? I mean, part of my job was due diligence, so there were a lot of them that I passed on. But it was part of my... Were there any you passed on that ended up being really good? No.

They're around, some of them. I don't want to mention names, but they're still around, a few of them.

It was sort of an education for me too because I'd see all their trades every day and I could go and visit with them and sit there and watch them trade. I mean watching Paul Tudor Jones trade was really an incredible experience. Everything everybody says it was? Yeah, it was. He was really good. Probably the best trader that I've ever seen. So then...

I needed to know what to do with my own money because we had done very well. We never had a losing year and some brokerage room bought me out. So I kept an eye on different types of research, especially the academic side and

About 10 to 15 years ago, I stumbled across the Momentum research that had been done. Now Momentum was actually written about first in 1937 by Cowles and Jones, but it was ignored until the early 90s when Jagadish and Tippmann did an academic type study on Momentum. And what they did was they segmented the stock market into deciles

and they compared how they would do in the future to how they did in the past, and they found going back three, six, nine, and 12 months, stocks that outperformed tended to continue to outperform in the future.

So this created quite a stir in the academic world because it was a slap in the face to efficient market. The greatest embarrassment to the... Yeah, it was. So academics started researching it as much as they could to try to disprove it, basically. Exactly, yeah. So they were looking at different...

things like industries, other countries, whatever. And they found that momentum worked pretty much everywhere on everything. It was universal. Can I ask you, in your experience, especially working and reading so much of the academic research, why do you think it is that you could fill this room with all the papers that have been written from academia in support of various forms of technicals, whether it's trend following, momentum, mean reversion even?

And yet in the classrooms, in my own classroom, I was literally told that we spent five minutes on it and the professor said it was BS. And then we just dismissed the whole chapter on it. And then years later, I was teaching technical analysis at Brandeis, which I took over from Dave Keller. And I was in the middle of the break of the class. And my son, at the time, my oldest, called me, just happened to, he sent me a text saying, hey, we're in the middle of,

finance class and we just covered technical analysis and the professor says it's BS and spent two seconds on it. So, and that's with, that's almost how many years that is, but it's a long time of research papers being written, actual managers managing money with momentum and trend and beating the market and they still deny it. Why do you think that is?

Because that's how they were taught. Basically when efficient markets... Well, they're scientists. Aren't they taught to be questioning data and accepting the truth? The evidence wasn't really strong. Basically, Kuttner came out with his book and basically he started this whole efficient market thing. In fact, they thought it was random walk because they weren't looking at it the right way. They were looking at simple runs analysis and...

serial correlation using the wrong time frames. So that's the problem when you look at things in having incomplete information. Now since then, like you say, there's been academic validation of all forms of friend following.

Except there wasn't for channel breakouts, so I did that myself with a researcher in Switzerland. We just did that last year. We went back through the French data, took 100 years of industry data and also sector ETFs, and we showed that using the Keltner and Donchian channels gave

gave a big advantage over buy and hold and reduced the downside exposure considerably. But there's been over 100 years of research in both types of momentum, relative strength momentum, Giecksky and Samenov went back to the year 1800. And absolute momentum, which is the other form of dual momentum, that's

Kaminsky and Grazerman went back to the beginning of every market. That's the 1600s in stocks, the 1400s in bonds, the 1200s in the Japanese rice market, and they found it beats buy and hold and reduces downside. There's been comparable studies on moving averages the same way. Also on Dow theory.

And I put references to all of this in some of my writings. - Right, and I know you've done a lot of work on general breakouts and doubt theory and what not, incorporating it into your new kind of like, everything you've learned about momentum since dual momentum, and I wanna talk about that, but let's just first level set everybody, just quickly describe what you're,

what dual momentum is, what your findings were, how it definitely changed the way people think about combining trend and momentum, relative performance in trend. And then we can kind of jump off into the other things that you've learned since then. - Okay, well, there are two types of momentum. The first I described already, which is relative strength. Academics call it cross-sectional because it started out

where you segment the stock market into 10 deciles. But you can also look at it across different assets. So it's really relative strength, which we're all familiar with. So I call that relative momentum. You know, it makes sense. - But we're in a room full of CMT, so when we think relative strength, we do think cross-sectional momentum comparing, you know, deciling the performance of the constituents according to how they did over the past,

three months, six months, whatever. But we also think relative performance, which is stock price divided by index. Have you done work on relative performance as well in terms of incorporating this into your work? Yeah, it's the same thing. If you were to compare, say, one sector to all the other sectors, that's similar to comparing one sector. Using a relative performance line, though, like putting moving averages on it and all that? That's one way to do it. That is a form of momentum. Now,

The charting packages also have absolute momentum, which is a basic trend following rate of change over time. They'll call it momentum or they'll call it rate of change, but that's the other type of momentum where you're looking at an asset's performance itself over time. Momentum means persistence in performance. So an asset, let's say, that's done well over the

the past year should continue to do well into the future. That's what absolute momentum is. Now my contribution, which is in my book and was in my papers before my book, is combining the two together. So I have a simple model in my book, which is meant for the public. It's not the best investment model, but it will

hopefully keep them from having horrendous drawdowns and put them ahead over the long run. So what that does is it will determine whether the trend in stocks is positive or negative. That's absolute momentum and it does that by focusing on the S&P because that tends to lead other markets. And if it determines the trend is not positive, then it will shift over into bonds or cash.

If it is positive, then you want to know what part of the stock market should you be in. And my simple model just used US stocks and non-US stocks. It would use relative momentum or relative strength to determine which of those two to be in. And when you, if I remember correctly, when you calculate the absolute return of the index to determine if you should be in it or not, you would calculate a net of treasury bills.

Yes, you're looking at excess return. Right, and why would you do that? Because it almost feels to me like you're double counting the returns of T-bills in the comparison. Well, that's why you take T-bill return, you subtract it away. See, the idea, and academics usually look at excess returns. The idea being that if you can earn a comparable or higher return in T-bills, why would you want to assume any risk?

So that's a hurdle rate. In order to be in a certain asset, you want to be earning more than what you could get in T-bills. Right, right. Okay. And then you also spoke about within the equity side, you also talked about buying the top, was it five sectors?

as well on the equity side? - I mentioned that in my book, but we don't do sector rotation. We found better ways of going about-- - Did you find that it didn't work as well? 'Cause that's kind of what I thought. - Yeah, I did, I did. - Why do you think that is? That's kind of interesting. - Well, I only had 20-some years of data at the time, and I'm very data intensive in everything I do. So when I got more data, I found it didn't hold up as well, and there are better ways of doing things is what I found.

Okay, so the book came out, it was a splash, won awards, fantastic book. What do you wish was in the book that wasn't in the book? Well, that'll be in my next book. Perfect. As all books should be, right? But it's an evolution process, just like the history of investing is evolutionary. My history is evolutionary, too. Fortunately, I have some information to draw on since the 1970s, looking at so many different things.

So 2020 was a bit of a wake-up call early part of that year because what was happening?

Just kidding. Well, you had a mini crash, and so models that adjust on a monthly basis had some problems there. Whipsaw, right? Yeah, we got out, so if the market had continued to go down 50%, we would have looked great. But then it came right back up, the V spike. So there was a whipsaw, and that got me to think, well, maybe I should be looking at shorter time frames.

So that's what I did on my next model. I incorporated daily data into it and by doing that I could determine an optimal rebalancing period which instead of being one month for this type of approach was actually three weeks. So you're able to get out sooner? We get out or in sooner. Also I wanted to incorporate other information in terms of approaches because I believe we're not just in diversification with assets.

but diversification with approaches that reduces your specification error and so forth. So I thought well what can I apply

in addition to dual momentum. So I segmented the model into two parts. One was using dual momentum with different stock ETFs and the other was using dual momentum mostly with bond ETFs. And I thought, well what's gonna tell me what the primary trend of the market is so I'll know which to be in? So anyone wanna guess what that was? Gave you a clue, I said primary trend.

Okay, well, Dow Theory, you know, and that's gone through some evolution. The fellow named Jack Schaunup wrote a book, Dow Theory for the 21st Century, that kind of updated things. And there's a research paper that went back and showed that Dow Theory has held up well over that time. And have you, can you maybe talk a little bit about how it's changed? Because if you look at the Dow today, it's

specifically Dow industrials, it's actually not industrial at all. So do you use the Dow industrials or do you use like the S&P capital goods index? - We use the Dow industrials, the transports, which were the rails back then and also the S&P 500. - Right, okay. - Now that's a little bit passe now too, because I found that it was good to have that additional sensitivity and to have that additional information from Dow theory.

But I thought, well, let's take this a little further. Maybe I could get something that was entirely based on daily data and I wouldn't have to have a rebalancing period. In addition, I wanted to incorporate more diverse assets. And I'd always been attracted to gold because that's all, pretty much all I had. - Started in the 70s. - Yeah, back then. So as you probably know if you've done any research, gold is very hard to model.

I mean, Hussman and Friedberg had some models, but they weren't great. And then using the typical rate of change stuff, like moving averages and absolute, it would give some benefit, but it wasn't terrific. So I thought, well, let's go back to something that's been around a while and that has worked, and let's take a look. And anyone want to guess what that was? I'd be cheating if I said so. Okay. Leave it to everybody else. You'd be cheating. You know already. Okay.

Well, let's look at channel breakouts because Jack Dreyfuss became a billionaire using something similar like that. He was buying stocks making new highs. Richard Dennis taught it to his reptile traders. Reptile traders.

Nicholas Darvis. Yeah, Darvis. Darvis was one of my inspirations when I got started in the business in the 70s. Everybody read that book? How I Made $2 Million? How I Made $2 Million. $2 Million in the Stock Market? Fantastic book. I read it every year. Have you read it yet? Yeah. Fantastic. Yeah, he was traveling around. He was a dancer. By the way, if you get a chance, you should meet this guy, Russ Evan. I mean, he's not here very often, but man, he's one of the smartest investors I know. I've known him for years. He was a guest on the podcast.

Is he a good man? After you. So I found that channel breakouts worked really well in the gold market. I added some price action filter to it, but it was basically a channel breakout approach, which was great because during times of market stress,

A lot of people think that diversifying with bonds is the way to go. You hear about these 60/40 type things. But what people don't realize is that stocks and bonds have been correlated as much as they've been uncorrelated. And during times of market stress, the only thing that goes up is correlation. Yeah. Everything- Correlations go to one. Yeah, so people are looking for a flight to safety. So what they'll do is they'll go to cash or they'll go into gold, except for last Friday.

So I really wanted to see if I could have something that worked and this worked. Now I thought, well, if it's working on a market that's so hard to model like gold, maybe it'll work on other things too.

And as part of the robustness testing I always do with models, I apply models to everything anyway. And I found that it worked pretty much on all markets. And I also use the S&P going back to 1962. It held up there and it held up everywhere else. So I built channel based models for stocks, bonds,

I had gold and then digital based assets, which was an interesting. So that's a question I want to ask you about. But I know you've built a lot of models around each of these things. And one of the things I have to credit you with is I'm pretty good at coming up with acronyms to try to capture the theme of what I. And I think you are the master of acronyms. I mean, he has a he has a model called Badass.

And I got to imagine that you came up with the name first and then you fit the words to come up with it. But that's your digital asset one, right? Yeah, it's Bitcoin and Digital Asset Synergistic System. That sounds totally like you.

What can I squeeze into badass? Okay, but then you've also got... For gold, it's gold long. Yeah, that's the one. You were sleeping on that one. That's the GLTR. I mean, you can do better than that. Glitter, yeah. And then for stocks, it's stock market upside reversal factor, SMRF. SMRF. And then for bonds, bonds had included some other things. So it was bond and equities,

systematic bonds, equity, anomalistic systematic trading, that was BEAST. - BEAST. - Yeah. - Right. - So those are my acronyms. - Okay, so my question for you on the digital asset one though is because you are very adamant about having lots of history before you test something and certainly before you invest in the findings of the model. This obviously doesn't have that

that history that you would need? Is this just more of a function of you believe the model is so robust that really it works on everything, so testing something back 100 or 200 years is good enough to say that it'll work on digital assets? - Well, that was part of it, but you want to make sure that it also works on whatever data you do have. So I tested on Bitcoin going back to 2012 when it became viable and it worked beautifully.

You can't really use Bitcoin because it's a 24/7 market and I don't want to do stuff in the middle of the night on a Saturday or whatever. Plus I always match up with other things so you couldn't do that. So my research went to the Greystone Trust which went back to 2014. It held up really well there.

And there still wasn't any way to do it on Bitcoin, so I used blockchain technology ETFs. BLOK goes back the furthest to 2018, and it worked beautifully there. When SpotBitcoins came along last year in 2018, I incorporated that into the model. But even without that until October, that was my best performing model last year. It was up 87%.

and it is up nicely this year. One of the reasons for that is I don't just develop a model and say, okay, when we're not risk on, we go into cash. That's what a lot of people do. I said, well, there's probably other things that the model could go into. That's my idea of relative strength now is let's look at applying the model to alternative assets or switching into another model.

So the way that works is my badass model is bifurcated. I have part of it in blockchain and part of it in Bitcoin. And the part in Bitcoin will go into the gold model when it's risk-off because those are both stores of value. And the part that's in blockchain will go into the Smurf model, the stock one, when it's risk-off because they're both stock market things. And then those two models glitter and...

When they go risk off, they'll go into the bond model beast. Now, bond has nowhere to go, so I just layered within that model. I'll start with tips and then use senior loan,

and then I'll go down to T-bill equivalent. So we can capture a lot of the short and intermediate sectors of that market. - So if most of the bonds, 'cause the correlations can go to one, even over cyclical periods, not just necessarily during bear markets. So if you walk all the way down that decision tree and you get all the way down to bonds and everything, there isn't a downtrend, you just go to cash.

- Yeah, we're in cash, T-bill equivalents, floating rate notes tied to T-bills. Right now, what we're positioned in through all the models, we still have gold, we have a really nice profit in that, but we're getting close to taking the profit. If this craziness continues,

But it may not. And then we're in short-term tips right now. That's it. We got out of the stocks at the end of February, first part of March. Right. And there is no, I'm assuming that there is no regime identification, so you have no idea when you're going to get back in again. You don't really think technically about it. It's more just momentum has me in...

It's whatever the models tell me. I stick to my models religiously. I think that's a problem some people have. They'll override...

their models or delay or whatever it is. So you have to really be disciplined. The models are simple, but implementing them can sometimes be hard. - Yeah, that's the whole idea, right? We were talking about that earlier, that one of the great appeals of even the book you wrote was it's so incredibly simple.

But it's not easy. And that's where into the conversation comes all these things we've been talking about, which is the behavioral biases and the things where you... It's kind of like Adam's been talking about it and Zoe and others talking about educating their clients to make sure that Pamela's well... That when...

when you have the opportunity to educate them when things are going well that this is what it will look like when things are going bad is you should have those conversations ahead of time right and i think it's particularly important for if you have somebody that's that's invested in a momentum-based model they love it when it's working but when they don't when it's not working like wait a minute what's momentum you just bought it because it went up that doesn't make any sense now they want out of the whole strategy when in fact it's one of the best strategies to be in over time so when you're working with clients you spend a lot of time with them on

helping them manage their behavioral biases before they come home to roost? - Yeah, I don't deal with the public per se. - Good for you. - I've determined that I wouldn't,

I developed these models for my own use. I have 97% of my liquid assets are invested in my own models and I thought since I'm doing the work anyway, I can release signals to people who understand and appreciate what I'm doing. So I have a foundation, I have family offices, and I have a handful of advisors that I give the model signals to and hopefully they understand. Not all the

All of them do, so there is some turnover. It's usually because they can't follow the models or they just don't have a good enough understanding. It's a continuing education process. Every model, no matter how good it is, will have periods where it may underperform its benchmarks. You have to educate people to that so that they keep the big picture in mind.

We don't lose people because of drawdowns. We have very modest ones so far. But we lose them because of impatience, because people are myopic and they tend to want to move from one thing to another. So like you'll have a period of inactivity and they'll want to move on to something else? Well, they'll see something else. Shiny objects? I had a licensee one time.

who his wife was invested in REITs, which were hot. She was making all this money, so he pulled his money away and went into those REITs and promptly lost 30% of his capital. So just thinking about how you have evolved since Dual Momentum, your original book, it sounds like you've more...

kind of evolved into a strategy allocator. So you have all based on the same core principles. John would refer to them as first principles, all trend following momentum, but you're just doing it across different asset classes. And so what's the end game there? Is that enable you to create an opportunity

an appealing return stream over time, but with less volatility. Ray Dalio said if you can have low correlated return streams, like four of them or more, that's the holy grail of investing. So that's kind of what we're doing. It's modeled along like Harry Brown's permanent portfolio. He had stocks, bonds, gold, and T-bills. We have stocks, bonds, gold, and digital. But I have some other things going on there too because I didn't want to be entirely channel-based

although I do have some different filters in there. So one of my newer models, I replaced an older stock one with this, with the Smurf model. And what it does is it looks at moving averages, moving average crossovers, long-term moving average filter,

volatility and kind of breakout stuff, some mean reversion. And that gives a different type of return stream than the others. It's only in the stock market two thirds of the time.

So the other third of the time it'll be in the bond model. So whereas people will have a 60-40 where they're continually, and the bonds are a drag on long-term performance and wealth accumulation because they have a lower return. Well, we do it a little differently. We're in stocks when it makes the most sense, and then when it doesn't, and bonds are in a trend, we'll be there. Otherwise, we're cash equivalents. The other thing we have is

- Yeah, we have to talk about this because that's like the antithesis of, like I think momentum is like growth investing and mean reversion is value investing. They're antithetical to each other in the fundamental side and I kind of feel the same way with trend following and mean reversion so this should be interesting. - Yeah, momentum tends to work on an intermediate term basis, usually from three to 12 months, you could extend that a little bit. But mean reversion,

Mean reversion works on a longer term basis, three to five years, which gets you into value investing, or on a short term basis, which would be a week or less. So we incorporate that in most of the… Both of the mean reversions? No. Just the short term ones? Only the short ones. The other ones, you have to have a strong stomach and a lot of patience. Yeah, right. So…

What we do is, we don't do this with Bitcoin because it can rip your face off. It just goes crazy. So we wouldn't want to do mean reversion there. But for the other ones, if the markets get ahead of themselves, we have a way of standing aside. And oftentimes, there'll be a reversal. So we're able to get out early and capture more of the gains. Or there'll at least be a consolidation, in which case we get back in.

We'll also do it when we're out of a market and it gets clobbered, you know, it gets knocked down extremely. We'll step in for, you know, one to three or four days. And those trades are very successful. Very high hit rate, right? Yeah. The one for blockchain, we had a trade a month ago where we got in at 40 and got out the next day at 44. Yeah.

When you do the trades, the mean reversion, do you make them trend aware so that to be down

10% in two days or however you're measuring mean reversion is arguably something different if it's above the 200-day versus below the 200-day. It just doesn't matter. It doesn't matter what the longer-term trend is. Because oftentimes, even in a bull market, stocks will take a dip, a strong dip. And actually, the odds are in our favor there. But they're also in our favor when the markets are going down. They don't go down all the time. They have bounces along the way. Unless you're

investing last week. They just don't stop. So when you're doing the mean reversion, in the back testing that you did, did you find the profitability statistics for buying mean reversion in a trend versus in a downtrend? Did you find them to be both profitable, but the drivers of the profitability are different? Maybe there's a

higher gain per win if it's trending positively, but there's a higher frequency of wins, but a smaller gain if it's in a downtrend? No, it's really, really indifferent to what the trend of the market is. On a short-term basis, the markets will respond the same. They get whacked and then they bounce.

That happens whether in an uptrend or a downtrend. Now, there is research on mean reversion. I cite three different mean reversion papers. Everything I do has to be founded in academic research and have 100 years of back testing.

Let me mention that, that's important too, because it's so easy to fit models to data now. There's lots of data, there's lots of platforms you can use, and you don't even have to program, you can use AI to write code for you. So it's very easy to think you have something when you may not, because regimes change every

15 or 20 years you could be in a totally different regime. So if you're looking at something over say a 20 year period of data, there's a good chance it won't hold up. Plus there's a danger of overfitting. So I try to keep my model simple. I try to make sure there's academic validation and there's lots of data supporting it and lots of robustness across different assets and different time frames. - And on the breakouts which you've introduced,

That to me has always been a signal in the context of trend. And you're finding that there's improvement. It's one thing to identify an uptrend just because it's above the 200-day or the 200-day is rising. You're finding that if you actually wait for the breakout, that's an even better signal than... Because obviously if it's breaking out, it's in an uptrend. It's probably above the 200-day. So are you finding that you're better off waiting for that actual breakout?

Yeah, with the channel breakout one that's what they're based on now in the Smurf model the stock one we actually will I think this is ideal if you can do it we'll buy dips within the long-term uptrend right and we'll sell rallies right with when we need to as well so we try to we that's the ideal way if you could buy a dip within a long-term uptrend so that's built into that model and in the on the on the channel breakouts is is

I know you use a 12-month look back on your momentum. For channel breakouts? No, for your momentum scores, your momentum ranks. Is it a 12-month look back or 352 days? We use different look backs. That's exactly my question on breakouts. Do you use different...

time frames of breakout or is it always the same look back? Is it different over asset classes as well? - Yeah, it tends to be shorter term on the breakouts. We're not looking at months, we're looking more at weeks or days. The study that Carlo and I did used 20 day channels of Keltner and-- - Which is the old Donchian, right? - Yeah, and then we used 40 days for the exit to give it some room.

So we're more along those lines.

it makes these models more sensitive and at the same time they don't have as many whipsaws as moving averages or absolute momentum would. Now if you're interested in moving averages, Valery Zakomulin, he's a professor in Europe, he's coming out with a book later this year, I just read it,

And he explores moving averages more than anybody ever has. He's looked at every possible moving average there is and he ranks them against each other. And the best one, I hope I'm not giving away too much, is just a simple two moving average crossover.

So that's incorporated in one of our models. Yeah, fantastic. By the way, all of the research that you're mentioning and all the other studies that you've done, that's all available on your website? Not all on my website, but I have fact sheets on all my models people can request. And I give links to academic and other studies on those fact sheets.

Okay, we should definitely save some time for questions. What time is it, Pamela? Do you have the time? 12:19. All right, perfect. Before we go to questions, then maybe just wrap up with what's coming next? What's the next frontier for you? What are you most excited to work on next?

I pretty well finished with the model development. There are a couple of areas that we haven't touched on that are important to me. One is how do you evaluate a model? And there'll be a talk this afternoon from the Optima guys about that.

but they're gonna show the limitations of a chart ratio and I believe that, I believe in that too. So I have other things I look at. But I think an ensemble approach to that makes sense. So I use things like the,

- Ulcer performance index, which in place of the standard deviation of the Sharpe ratio, it uses the ulcer index. - Is that drawdown? - It's the magnitude and duration of drawdowns. And I use a skewness adjusted Sharpe ratio. I use CAGR. - Skewness adjusted Sharpe ratios? 'Cause the problem with Sharpe ratio is that it doesn't distinguish between up and down volatility. - Yeah, exactly. - So this gives credit for upside volatility?

You can adjust for skewness. I write about that in my book, and Lopez de Pardo has a whole more complicated way of doing that. Look at CAGR. Believe it or not, that's a risk-adjustment measure, although a very aggressive one. When you do a regression, they say after you look at the numbers, you should look at the...

plot of the residuals. So it's important to look at the equity graph of your model, make sure it's smooth and consistent. And then the other thing that's important to what I do is portfolio optimization.

My last blog post goes into that, but basically we start with equal weight and then we adjust the weights to optimize everything I just mentioned. And I'll have a number of choices that people can pick and choose from based on risk preferences. That's in the book. That'll be in the new book. Right, exactly. Fantastic. All right, last thing for maybe the advisors in the room. I'm sure there's a lot of advisors that use either what you've

published in the book or something similar. The name of my company is Motor, which is momentum and trend. So it's a very well accepted idea. So when you're thinking about advisors or portfolio managers that might be thinking of allocating to a strategy like this, what advice would you give them? What are the things that you think they should know before they get into something like this to make them better investors in it? Well, first of all, they should be aware of the dangers of

model overfitting, not having enough data. Lopez de Pardo has some papers on that which are good to read. So keep that in mind. That's why I try to have academic validation for everything I do and 100 years of data supporting it. The other thing too is don't just diversify with respect to assets but diversify with respect to your methods because they

They all have different ways of adding value. - So this would be like mean reversion versus trend following versus the workouts. - Yeah, mean reversion, different means of trend following. Now being in CMT, you know a lot of these, so don't be shy. Don't just be satisfied with one particular approach like I did back in 10 years ago. Branch out and look at other things. And the third thing is think about portfolio optimization.

There are ways of making things better than just equal weight, but equal weight is probably better than mean variance optimization. Yeah, that's a dirty little secret. Fantastic. I'd like to open it up for questions. I'm sure there are many. So let's get started with... I've got one right here. John has a question for you. You're in trouble. Thank you, Gary.

So I get this question from clients a lot, so I'm going to ask you. So formulating a momentum factor, whether it be absolute or relative, when you're doing it with individual stocks, you typically are skipping recent weeks or skipping the most recent month. Like you mentioned Jagadish and Tittman, they did 12-1, skipping the most recent month. But then when you do an index, typically you're not skipping anything. So customers ask me, why is this?

What's your answer to that? Yeah, it's because stocks tend to mean revert a lot and indices that all get smoothed out so you don't need to. The other thing I should point out too is that research shows, my research and also by Gecksky and Samanov, they show that there are better ways to use momentum than applied to individual stocks.

Because individual stocks have what they tend to be more volatile. They have wide bid-ask spreads and you have idiosyncratic volatility that creates these jump processes that make it difficult to model. So I've always used stock indices and the Getsky and Samanov study shows that that outperforms momentum applied to a lot of different things. Commodities, fixed income, individual stocks or whatever.

Yes, John. You're going to take me to task here. First of all, I'm a big fan. Love your book. Love your work. Really, really very good. And Mr. Lundgren, excellent interviewer. Another great job. Thank you. When can we expect your book and what will it be called? Do you have any suggestions? LAUGHTER

I haven't had time to work on the book recently because I have all these models and there's been a lot of interest in them. But I intend to start working seriously on it within the next month or two. So I expect I'll have it done by the end of the year and probably be the middle of next year before it's published. Any other questions from the audience? We got one right here from Mike Jones.

You had mentioned some advisors who worked for you and who said that you lose people who won't follow the system or the signals. I'm wondering if you have any insights as to why people aren't ready, willing, and able to carry it out when the time comes. Dave could answer that, too. It's behavioral biases. Simple but not easy, right? Yeah. People are motivated by fear and greed and myopia.

So I'll give you an example. I have a model that I don't tell the public about. It's called Snapback. I've had it for over five years now. And I use almost 100 different ETFs and mega cap stocks. And it does the mean reversion, both long and short, that I was telling you about. And I started to give the signals out to some people who are interested, and they just couldn't handle it psychologically. For instance, I'm going to get--

I'll have some buy orders in on Monday. You know, I mean who's going to want to take those trades? And it's because of the hit rate, right? There's not a high hit rate or it? It has a high hit rate. Yeah. It's still difficult to do. Yeah, with the indices, the hit rate's over 75. When you're dealing with individual ETFs for sectors or whatever, it's down around 70. But 70% wins when wins are greater than losses. I mean, why would you not take those? Exactly.

Fantastic. We've got one more from John Lewis. Thanks, Lockery. I'm just curious, can you share a little bit about how you go about testing people

a mean reversion system like snapback, something like that. Um, the thing I'm most curious about is normally when I see those tests, people just will run it, run the system on all hundred things in there and then give you an average. But in my experience, like the reality is all those signals kind of come at the same time. So you can't buy everything at the same time. So like, how do you kind of go through that process? Uh,

Because I customize the signal based on the volatility of each thing that I'm trading. So although the models are the same, the parameters will differ. Also, things move at different rates and different times. So it is true that we might start off, let's say this Monday, there's a couple things that we'll go into. And then the next day, if the markets continue to tank, there'll be more.

So what I do is I split my capital up amongst them. I don't increase my investment size. But if we go two, three, four days, I don't think we've ever gone more than that, where we're still getting signals, the probability of success goes up each day. Even though it's harder to make those trades, they tend to be better trades, the more overbought or oversold the market becomes.

Now this is a minor part of what I do, don't get me wrong. This is just an enhancement and it's a minor kind of thing. It's nothing that I would go and bet the farm on. It's a process diversifier. Yeah, I mean you could end up, occasionally you'll end up taking a loss on it.

But overall, they're worth doing as long as you keep your capital size within reason.

Here at CMT Association, 2025 is a landmark year. For the first time ever, the official CMT program curriculum will be included with exam registration. That means if you register for an exam, the corresponding study materials are immediately made available. The CMT curriculum was specifically crafted to give candidates the skills needed to gain an edge in an intensely competitive

Thank you.

That's right. The CMT curriculum has gone digital so you can study anytime, anywhere, home, work, your mobile device, while you're binging every season of The Office, or when your kids make you watch the latest video they found on social media. Registration for the December 2025 exam cycle is open now at the early registration price. And with the curriculum now included in your registration, the CMT program is more accessible and inexpensive than it's ever been.

Head over to cmtassociation.org to learn more today. Fill the Gap is brought to you with support from Optima. In addition to candidate study of the official CMT curriculum, Optima provides a full video course on all of the material that candidates need to know for each level of the CMT exams. Each course is broken up into modules ranging from 15 to 45 minutes, depending on the complexity and length of the topics being covered. Learn more at Optima.com.

Bye.