The Derivative - Trend Following, Signal Deterioration, & Crypto Modelling With Quant Artur Sepp

Episode Date: April 28, 2022

This week, we're adding another stamp to our Derivative passport and traveling to Zurich, Switzerland, as we talk to the Head of Systematic Solutions and Portfolio Construction at Sygnum Bank's Asset Management, Artur Sepp @ArturSepp — who specializes in crypto-assets and decentralized finance. Artur has led quantitative research at systematic trend-following hedge fund Quantica Capital, focusing on data-driven investment strategies and asset allocation in global managed futures. In this episode, we dig into his background to discover what it's like being a quant (not as much like TNG character Data as Jeff would like…) and discuss; coding, mathematical modeling, and why statistics matter (testing simple, yet complicated models), the framework of trend following (be sure to download our trend-following guide here), pros and cons of risk premia strategies, quants trying to figure out the short data sets in Crypto and more! Plus, find out where Artur would invest 1K, 1 MM, and 100MM in Crypto. Chapters: 00:00-01:44 = Intro 01:45-10:47 = Coding, Math, Data, Stats - Testing simple, yet complicated models 10:38-38:12 = Framework of Trend following: The Carry, When factors are identified & Why Trend following could fall flat 38:13-42:37 =  Pros & Cons of Risk Premia strategies 42:38-53:38 = Dynamic Trends in Option volumes & Is the VIX dangerous? 53:39-01:13:25 = Quants on Crypto 01:13:26-01:23:44 = What would you invest in? ---- From the Episode: Check out our podcast with Roy Niederhoffer - Making Market Music Follow Artur on Twitter at @ArturSepp and check out his Blog on Quantitative Investment Strategies Don't forget to subscribe to The Derivative, and follow us on Twitter at @rcmAlts and our host Jeff at @AttainCap2, or LinkedIn , and Facebook, and sign-up for our blog digest. Disclaimer: This podcast is provided for informational purposes only and should not be relied upon as legal, business, or tax advice. All opinions expressed by podcast participants are solely their own opinions and do not necessarily reflect the opinions of RCM Alternatives, their affiliates, or companies featured. Due to industry regulations, participants on this podcast are instructed not to make specific trade recommendations, nor reference past or potential profits. And listeners are reminded that managed futures, commodity trading, and other alternative investments are complex and carry a risk of substantial losses. As such, they are not suitable for all investors. For more information, visit www.rcmalternatives.com/disclaimer

Welcome to the Derivative by RCM Alternatives, where we dive into what makes alternative investments go, analyze the strategies of unique hedge fund managers, and chat with interesting guests from across the investment world. Happy birthday to me. Well, tomorrow, but if you can't wish yourself happy birthday, what's the point of even having a podcast? Speaking of, we've got some good pods coming in May. We'll be riffing with Resolve about return stacking, talking prop trading and ball surfaces with Noel Smith, and getting into wine and entrepreneurship with Anthony Zhang.

So make sure to subscribe and check out those upcoming shows. This pod was super interesting, digging into the quirky quant details with Artur Sepp, head of systematic solutions and portfolio construction at Signum Bank in Zurich. We chat about his days at TrendFalling from Quantica and research into quicker and quicker information flow affecting returns, his work on volatility modeling, and why the data and testing and quant approach in crypto is both infuriating and fascinating all at once. Send it. This episode is brought to you by RCM's Outsource Trading Group, where their 24-6 trade desks do call-in, spreadsheet, FTP, direct-to-exchange, algos, and more.

Order types to help firms be cost-efficient and not have to staff their own trade desk. Check it out under the services slash trading firm slash 24 hour desk on the main navigation at rcmalternatives.com. Now back to the show. All right, everyone, we're here with Artur Sepp, live from Switzerland. What part of Switzerland, Artur? Hi, everyone. I'm based in Zurich. Zurich, all right. And I got to ask,

what's this little painting or drawing over your left shoulder there? Well, it's something from Chagall. It's a favorite painting of my wife. Perfect. I love it. You're saving up to buy the real one? Maybe. Talking nice, talking nice to share. One day. So let's start with telling us just a little bit of what it's like to be a quant. I think most people think of them as sort of the data character from star trek not sure if you're a star trek fan um but it seems like you're not all android there so what's what's the quant life like well it's a it's a hard question

to answer i think uh well it's uh for me it's more lifestyle, right? Relationships to much of my profession, right? I'm one by profession. What it means is that I'm tempted to look at things in a broader scale. For me, it's more interesting to build up a system of how and why things happen. And therefore, it's a subtraction. It's what I like in mathematics. I'm not married to math per se. It's just part of my profession, part of what I can deliver working in larger teams where some people tend to be more discretionary or more intuitive. And here, I stick to a model. For me, I think the key of being a plant is really to have a systematic way of looking at things and to be able to answer why. It seems just the modernized version of the word engineer, right?

Yes, exactly. Actually, for me it's the same. I actually have a degree in industrial engineering from Noroeste University. So, I studied several topics and for me it's always, it's indeed, it's creating a system that can work through different cycles, that can work through, say, process different data sources, different parameters, and in the end it produces some utility. Right, and I think that's the main engineering carryover is kind of to make it last too, right? For more than one purpose, to exist in more than one environment. Exactly, yes. Yes. Actually, with this profession, what I like is almost, I think I noticed already since I changed a few firms, a few companies, but more or less what I do is the same. It has different names between, say, crypto, volatility, CTAs, but more or less it's almost the same. Not almost the same, but it

can be attacked or approached with the same set of tools and set of approaches. And have you coded in the same language that whole time? Well, that's a good question. Actually, it's changed. So since when I just started, early 2000s, it was C++ was predominant. And to tell you this, C++ is almost like probably, I don't know, Latin language, right? It has everything. But if you process it, you understand. I think it's more about architecture, how to make things scalable, what is important.

And right now, of course, most of the time now I work with Python. It's simpler, but the idea... I think when we talk about programming, ability to put mathematical formulas and data into code is a skill that does not depend on language. Something that you have to learn, some old ones, especially the ones that once you work in industry, in asset management, but it's say language, it should not depend on language you choose. Yeah. And what came first for you, the interest in the statistics and probabilities in math

or the coding skill? Actually, it's also a good point. Actually, as an undergraduate, I studied finance and economics. It was fairly, it was like very quantitative, right? But then we studied modern portfolio theory like Markowitz and Black-Scholes. At that time, I started to actually I was very curious because those were mathematical models that you could implement in that time in Excel and VBA. But then over time, I thought that it's very interesting. It's why I went to graduate program in statistics, but in between I got, after that I got PhD and I got three master's degrees. And yes, indeed, through all time, I did coding. I think it's important, as say, more people with applied, say, background, you always...

Mathematics is one way you make a hypothesis, you can create nice theories, but in order to apply things in practice, you need to do some experiments. And the experiments, it's using data and running some outputs. And therefore, I think programming skills are important. Nowadays, everyone who plans to get some sort of

science degree should take enough time to learn this. It seems to me we're almost at a tipping point, right? Where you have some of these no-code platforms and things where the code can kind of create itself. Especially, like you said, for some of these math equations that seems like a simple plug-in, for lack of a better term, right?

Are we going to get to a point eventually where the coding is less important and the more important part is knowing how to apply and do the different testing and the different models on it? It's a valid point. I think yes and no. Nowadays what I see in candidates, the one that I interview, most of, especially this data science guys, they know how to use particular, say, Python package, right? As scikit-learn or TensorFlow. What I think they're missing this understanding of why actually under what conditions you expect something

to work, what are numerical representations, right? Say convergence, how would you analyze the convergence of your algorithm? So therefore, and especially also in, say, in production, right?

I can do something for experiment. I can take luck and play solution. But usually it's always, it's very limited. It's limited almost, it's limited to work in 95% of cases, right? All this is mathematics. It's a fairly complicated algorithm. Yes, it's very simple to use for some say

nice problems, but the ones that we have in finance you almost hit that 5% that you have to think yourself a little bit how to tweak it. Well if that five percent can put you out of business it's important. Moving on, can't have you on and not talk about trend given what's going on in the world right

now and given your time at Quantica. So part of me feels bad for you, right? You left Quantica in August of 21, which is right around the time trend started to go vertical. Do you look back and say, whoops, I shouldn't have left or do you seek and kick yourself or it's all good? All good. There are two points. So first of all, like CTAs in particular, trend phone CTAs cannot be timed.

Don't try. You can leave or you can enter them all the time. You cannot say, oh yes, I left at that time. Actually, when I exited this before that, and actually last year, what was interesting, last year was very similar to this year. So almost also was short trend, like rates rose, right? We had the correction rates. So all CTAs were short bonds, long commodities, agricultural especially, and flat-ish equities. But then around May, it's all corrected, high trend reversed. So

I wouldn't say that I went like almost say, I can say I left at the local peak. Got it. And the second point, I still have my own portfolio, more or less CTA-based. Oh, okay. So I still participate, and I like it. I think TrendFall, for me, is one of the best systematic strategies. It fits my mentality and I like, say, in terms of the justification, in terms of, say, return potential and, say, skewness. It's one of the best stuff out there and I think for me there is no way I can't exit this. Well, now it seems you have the best of both worlds.

You're not dependent on it for your paycheck and for your investments. Now you can have some diversity in your exposures, right? Yes. So what was some of the most interesting trend research you did there? You were head of research, right? Yes. So you got any good stories of some of the cool stuff you did? Yes, it's definitely interesting.

As I said, CTS is an interesting business because you have data, right? You conduct it like... Yeah, tons and tons. For example, for crypto, right, you have most cases it's like two, three years of data. And here we have data since even tradable data for commodities since 1960s, right?

And my research, one of the interesting parts, and also in line with our peers, is to see the deterioration of trend following since the 70s, 80s, 90s,

and how, what are implications? So, in say, in 70s, 80s, even you had and how what are implications so in 70s 80s even you had a transaction cost

very high you could run a short term trend following system meaning that your holding period on average

would be maybe two weeks less than one month right and you had stellar performance it was it's almost was like an alpha strategy. It traded very, it traded, turnover was high,

transaction costs were high, but the statistical nature of markets where you had this self-feeding cycles, they generated performance, it depends on specification, but say between two and three shafts, up to mid-90s. Then, of course, professional CTAs started.

The bigger guys, they started to enter the market around 90s. You had bigger players. Then, of course, it started to deteriorate. And the way to respond is, of course, to increase the trend, the scale. So your holding period would have to be around one month. Increase your holding period.

Exactly. Increase in some way, it's increasing the decay rate of your signal so the decay you always take some sort of a look back depends on what weighting you use but you always depend on some history of

say one week ideally would be one week then you can adapt fast if these conditions persist you can adapt fast but if these conditions persist you would adapt fast. that's interesting. real quick if you're thinking of it in like terms of bars right so if i'm trading daily bars and the signal deteriorates after 500 bars or something if i'm trading monthly bars that's 500 months I've greatly expanded the time till deterioration.

Yeah. And it's also a good point. So with bars, you can think also how bars, like say week over week, right? Ideal for trend following if bars kind of increase, right? You have a stir. Do they go up? Lower lows. Right? Yes. a stair. It will go up. Lower lows. Right? Yes.

Or down. Or vice versa. And what happens since 2000, and especially over the last decade, market changed, right? You have like, say, market takes elevator down, and then goes and steps up, right? So you have the negative, And if it goes too much

up, then it kind of, again, it's corrected. And this environment is more like mean reversion, right? You have mean reversion. And to avoid that on shorter scales, on say one week, right? So if one week is positive, next week is more likely to be negative. And so on. Across almost all classes. And this is a difficult condition

for trend following, because what you want is market persistency. The statistical persistency called positive autocorrelation. So this was one part that I found interesting. What was maybe unusual or…

Real quick on that, did you get to a reason of why that changed? Some people say the space got too big, there were too many assets chasing too few contracts. Others have argued against that, saying that the stats don't really show. The price action you could say, but the volume stats don't really show it. Well, I'm a mathematician, so I can answer in mathematical terms. The nature of the market changed. Before the market was more like this positive autocorrelation, what I call that. Yeah. Probably it was slow propagation of news, right?

It would take a time that market would converse to equilibrium through several steps, through several weeks or several days, several weeks. And therefore by say kind of discovery, early discovery, you would be able still to participate. And nowadays, especially over past decades, it's more like this kind of mean reversion where the news propagates very fast, market corrects very fast, and then the rest, you have this positive trend, but it's not because it's positive. There is QE programs or people getting into equities.

This is not something that you want. It's something resistant. It starts to become more binary. There's some new shift that just shifts the market from here to here, and then it's chomping along here. So, yeah, that's interesting. That's right around the time the internet, obviously, then social media, then computers.

So, right, everything was just getting faster in terms of the market moving news, propagating high-frequency trading firms. So for you, not because trend following got too big, the market moving news propagating high frequency trading firms. So for you, not because trend following got too big, but rather the rest of the players got too much information. Too efficient, too fast, too efficient. Trading, especially in commodity markets, there is informed traders, right?

There will be bigger trading houses that could adjust very fast, much faster than say traditionally it would be say more clusters of smaller traders that would take longer time to adjust. Nowadays you have few concentrated players that could move very fast in a way that is not tradable by like CTAs. But so then how do you explain the recent performance and how they keep going? So because again, it's a narrative, right? It's we waited for a long time and trends were very good. If you it's a perfect environment because I think rates-year UST rate, it was last August,

it was 1.2. And then it went almost like this straight line to now we have 2.7, right? So it was 1.5. It's really, and correction didn't last more than one month. So it's a perfect environment where you get like increase, increase. This is exactly what you want. For a trend forward, you increase shorts, shorts, and it goes your way. A little bit at a time, you can move your stock. Exactly.

Yeah. And also this oil was very good. Some kind of it was very good combination. Also oil. And oil also, I think the trend really strengthened mid-October. We were under 60. And end of February, it was 130. And actually, CTS, I think, were long, I would say, November, December.

You didn't wait until 1.30, right? After the correction, I think actually most of the guys must be flat. It was also good in terms of, say, portfolio verification. When the correction in oil happened, right, the rates went up. So actually, it was very smooth. Do you have any thoughts? I've long said on here that a lot of trend followers went long bias, added more equity beta to survive because it was such a long period of nothingness that they had to do something to survive and now those programs are kind of underperforming here.

Were you tempted in your research to say, hey, if we just made long bias, it would work a lot better? Well, so the part that I know from other sources, I can tell you. So part of Frisier, what also important framework I developed was carry. And I tell you why it's important. For futures, carry, say for bonds, is very important. Carry. Carry. The one that your return from backwardation and contact. So when commodities in backwardation, it means there are

economic benefits of holding the asset. So for example, for bonds, and also commodities are always success return. So if I buy, say, 10-year treasury bond on a margin, I get the coupon

that my rate carries. I get the coupon and i pay funding for me it's beneficial if funding rate is zero right and i buy 10 year coupon at one percent i get it the second important component is roll down because i am constantly changing my contract to 10-year future and if slopes are rising, if market expects the rate to rise, but actually nothing happens, say in one year, 10-year bond will be at 0.9. So I get 10 basis points multiplied by duration. But that's what was happening for the past 30 years almost. There was always expectation of higher and it never came to be.

But actually, this is the surprising part. A lot of people, you have to understand fixed income. Actually, with negative rates, German bonds were the best investment in your CTA portfolio. Yeah, in 19? Yes, even before that. Especially, say, 10 years. 10 years was... Or 5 years.

And then you pay excess return, right? And if your funding rate is minus 75, but your coupon is like 50 basis points, you get the spread of 25 basis points, right? Plus, if it's term structure, right? So you bought it, let's say, at minus 50, but then it moved to minus 75, right? You get this duration.

It declined. More negative. And plus, because we are all, almost everything is on risk basis, right? Volatility was very low. So the components, say, carry components of bonds, of short-dated,

say, both were more than one. So, sharp ratio. If I just buy bond at negative rate, I hold my expected return is sharp of one. On the risk, I just... There's a lot of money being spent on stuff more fancy than just buying.

Yes, but it means also in commodities. So, similar in commodities, a lot of stuff that say right now, it's always some sort of deterioration. Right now, short-term supply is affected. So short-term, people expect higher prices, but lower-term lawyers. So we have backwardation, right? What actually also most of the time is actually if you buy now,

say if nothing happens, right? So you get this correct company. And this is for many commodities. It's always economic thinking, things this that all producers once they see higher prices, they would like supply the market. Yeah, but it's never like this. And usually, but usually so curry also in commodities is important company. And then we come.

Sorry, go ahead. And then, so what brings us to trend following is that over time, say in 70s, 80s, if you make a breakdown of your P&L, how much you actually got because prices changed or how much you got from just getting the correct component, right? That since didn't change until you just get your coupon or your backwardation in commodities. Actually, 10-20%. In the early 70s, 80s, 90s, right now, over the past decade, especially last year, it came to 40-50%. Prior to the last year.

So, 70s, 80s, 20% of the profits were from Cary. And 90s and 2000s, 80% of the profits from Cary. No, no, no, no. Vice versa. 90s were still 20%. no. 90s were still 20%. 90s were 20%, 2030s, 2010s, 40%. Got it. And this is the reason exactly as you said, two companies that holding periods increased. And maybe not many CTAs even realize that you actually benefited a lot from past years,

most people were loan bonds and bonds didn't. Okay, there was, of course, we benefited from falling rates, but it's not as big as benefiting from this carry effect. More so benefiting from poor estimates on what the future would be, on that rates would be high. Exactly. And so what do you make of, have you seen Roy Niederhofer's research on that? If you flip the chart, right?

If bonds go on a 30-year rising interest rates, because of this carry, trend following might be flat to negative, right? It's not necessarily going to make the same money. Have you seen his piece? No, but I see the reasoning. And I tell you, actually, why I'm positive now on trend following, as I said, I did a lot of long-term studies. And late 70s, early 80s were the best time for trend following because they were both short equities and bonds. And the reason is right now actually the carry effect is not very strong because term structures are flat. It's actually the opposite. If, for example, so if I have say now slopes are negative, I'm actually benefiting of being short. Yeah. Right? So,

so, but being said that it's always, you have to be careful, right? With strength following with say, I prefer to keep it as pure as it is. And for me, carries a second order effect,

right? But it's important. It's important to maybe adjust position or understand what are, say, drivers of returns. To his point, that it costs, it would cost when, say, you short long-term bonds and your current rate is still very small, you would

lose on this coupon carry. There is not much of a roll down. Right now the term structure is flat. Yeah, which we could see in five years or something, maybe short terms at 2% and longer terms at 10% or something. But now, right, we're basically flat or twos are above 10, right? Because bonds are, futures are excess returns. So if say short-term rate goes to whatever, 5% and 30-year bond yields you 5% and term structure is flat, my carry is zero. Yeah. So then it's just the price action. Yes, exactly.

That's the difference, that I'm not shorting, say, cash bonds. In the future, you always have some excess funding rate. And so was part of your leaving Quantica that there wasn't enough to research in trend? Did you get bored with the trend research? It seems like there's only so many ways to skin the cat. You can do a bunch of moving averages, you can put a bunch of filters on, but there's

not like a blue ocean of new things to research there. Is that right or wrong? Well, I mean, I think CTA has survived, right? It's probably the oldest quant strategy out there. A lot of equity like this kind of mean reversion, it didn't survive. And I think partly because it's very conservative. But there is a lot of things to do. Some part I mentioned also, some what I think most interesting part of say traditional price-based trend following was sentiment data. So sentiment, you know that you heard that some people applied it for stocks. What I found interesting in my

application is that you can also trade, say, global futures or, say, US-based futures, equity, bond, commodities, based on news flow. And news flow, in my opinion, seems like a now question. It has been a new word since the corona crisis. The reason is that traditional models, they need to take some time. So, you need to always observe the phenomena, you need to train, you need to think how I translate these phenomena into actionable ideas, how I build my model, how I backtest it, and how I put it in production.

And so, of course, it takes time. With sentiment, what we observe, intuitively, say, each news, economic news, for example, unemployment numbers, right? If you expect that in certain regimes, a bad number is a good number in some sense like right now i would say bad number would mean that the fate is on pause so you actually market would benefit right before crisis say no one knew the extent, right? How deep the economy can contract,

and therefore everyone would expect that bad number, it's a bad number, next time it will be even worse. And this type of system where you have different news, right? So imagine you have different sentiment. You can track, say, news related to unemployment, news related to COVID, to oil news related to unemployment, news related to COVID, to oil supply, to wars,

to everything, and you can create a continuous time series that would say maybe you can normalize so it would be almost like a score. So negative score would mean that this news feed right now is very negative.

And then you study impact. So a little bit of machine learning. We try to select factors that actually drive right now the market. Right now, everything that relates to theater right now would drive the market. Anything that relates to war. And the trick there is, how do you know when it switches, right? So it's a short-term system where you constantly update your model, maybe week, every month.

You just try to select factors that matter most and also their logins. This is perfect. What I like is statistics. First of all, for me, it's a high frequency. It's not high frequency high turnover high turnover strategy it has very good defensive profile it's almost like being

low volatility it's very sensitive to transaction costs but what I found interesting is for me it's some sort of trend following on news on news sentiment and it can be made to work right and basically long and short For me, it's some sort of trend following on news, on news sentiment.

And it can be made to work. Right. And basically, long and short, the same piece of news at different environments. Yes. And this is also what people were calling now casting. I think it's become popular. But yes, I come up with some model at least that I like. And what do you make, I can't remember the authors of the paper, but there was a paper about a year or two ago talking about as soon as a factor is identified, it actually ceases

to be a factor, right? Like the very fact that you've identified it and put it into the market, other players have done the same work and identified it, put it in the market, and it ceases to be a value to drive the alpha that you thought it could drive. Have you seen that in real practice?

Fortunately, not probably in my case. I can tell you, it's always a question what kind of factor, right? So most of academic literature, especially factor, is very simplistic ways. You find some sort of fundamental or some sort of behavioral factor, you sort of build your rankings, then you put like some kind of

dark test, right? Whereas we go, say for example, and it's of course, I mean, first of all, it's obviously if there is something that you come up, like say, okay, it's academia, you would anyway expect that people profit. If it has value, it's natural that it will disappear because people would profit. In our systematic trading, I mean, everyone knows what CTA is.

I'd say, for example, CTAs. But there are so many different layers. You cannot actually tell that it's a risk premium. It's a dynamic strategy that is more, say, the values from rules, from applying the rules and from harvesting the beta. So for me, it's more like... With sticking to it, right? The values from enduring.

So you mentioned risk premia, you were at Julius Baird doing risk premia, right? Yes, I did. So dish the dirt for us on the pros and cons of risk premium strategy. To me, sometimes a little too simplistic. And like we just talked about, if everyone rushes into one on some bank platform, it's likely to stop working. What are your thoughts? Well, risk-free means because we have, say, fundamental factors like probably valuation. We have structural, like say short volatility, that say small demand imbalances. In my opinion, you have to be very clever how you implement it.

It's a strategy for good times, in my opinion. You can build a well-diversified portfolio of different risk premiums. What is good, probably, risk premiums, you have always expectations. Always say, like simple cash and carry arbitrage. You buy or short the future, depending on where the term structure is. And you buy and sell spot. And of course, problem is inversions.

So with carry strategies, or with say risk premium strategies, most of them you make money 80% of the time, 80, 90. The problem is that those remaining 10 percent, they actually can invalidate any return that you made. So therefore, I believe that a lot of this stuff exists, like say short volatility, is because it's difficult to over time, right? You have to be there through worse of the times, right? And therefore, yes, it exists. But they're like terminal

break-even, you're saying, basically. Exactly. If you access it, make sure you only do it for a certain period of time, but then you're under timing, which is difficult. But there's surely positive skew risk premium strategies or no? Or people don't select those because they lose, lose, lose, lose,

and then make money, right? Can I go on a platform and get the momentum risk premium or the trend premium, right? I think for me, it's more like, say, a classic trend phone is a dynamic strategy. So it's actually like say a classic trend following this dynamic strategy. So it's actually making you trying to replicate this data profile that works in certain situations. Risk premium for me, it's almost always like some sort of a term structure, right? When say in implied volatilities, if there is a normal period, say, of calmness,

but you still expect something bad to happen, so you're actually betting that these things are not going to happen, right? And you get your excess return. So to your point, and maybe in equities, if you say, okay, there's a value, but the problem is that it's not something that you have to be sufficiently long period of time, right? And if you're saying, okay, if strategy has positive skewness it means it must realize over short periods of time and the value factor so you have like a decay and suddenly you can have a very sharp return most of the risk premium is naturally short skewness. Okay interesting. Yeah to me it hasn't really built, it hasn't fulfilled

its promise that was coming out a few years ago like oh you're going to be able to do anything and pull some alpha off the shelf so to speak when it's by definition not really alpha it's some sort of beta right? Yes kind kind of delayed a bit. So let's move on to vol trading. You had your delta hedging paper way back on. We've had some debates on this podcast between different episodes of some people who are basically saying the whole the gamma, the Delta hedging, that's what the whole game is in the S and P. And if you're not cued into where those levels are from the dealers,

you're going to, you're not going to do so well. And other guys saying, yes, it's a factor, but it's very small, right? Maybe it's moving the S and P points at certain thing, but right. It's not causing it to move 50 handles or something. So take us through what you found in your paper and what your thoughts are on the whole dynamic, especially currently as options volume has exploded and whatnot. So I tell you my background with Zvol. So first, I was working for this, big size and liquid, liquid underlines. And where that idea came from that you really try to look at your risk and how much it would cost you to hedge. So that origin of that patent. Now, if, and of course, flows are like on OTC market,

flows are important, but there are one way. It's only the dealer who dictates the price. And you can always say, you always can say, put your spreads high. And, but moreover, to the to the point yes your positioning creates feedback loops and they are very strong if you and most of the time yes it depends how clients positioned and how that definitely would create a very strong feedback effect. Going to S&P, I doubt. I think the market is too deep and it's more positioning of market

overall that matters rather than position of several dealers. Like for example what we had in February 2018 when the Wall market was down. The market was everyone was short Wall. So I don't know what the dealers Yeah, well you could argue the S&P didn't really move all that much, even though vol exploded up, right? Yes, it was the position of the market. But some people believe in this level. For me, I think I thought this was more like definitely flows are important, especially on the loan side. Most of, for example, traditionally, especially private banks where clients are typically short volatility. So they sell through structured products to the bank

in long dated options or in return to some sort of kind of coupon,

right? The problem but it's almost everywhere. Even though you can buy it cheaper

than you would say go in listed markets, you are long volatility and you have this current decay. You need something bad to happen in markets. You need that flow.

So to be long volatility, somehow you need to be attached to a good flow of cheap volatility. Yeah, your client's flow that's short the other side. You cannot just buy on markets. For example, I would not assume that you would be able to profit systematically from being low in S&P options unless you incorporate some sort of other information. And what your VIX blow-up paper as well, what was the conclusion there on those products? We're having

VolShares who just launched the new VIX ETFs on either the week after or the week before you. But what are your thoughts on those products? Are they useful tools? Are they dangerous? I think they are very dangerous, especially for, I think, for retail. I mean, professional players, they would implement anyway through this kind of term structure. They would not be, of course, how they would have, you have to manage your margin call at the margins. But here it was product that was just wiped out.

You just wipe out of equity. And I think, yes, people discuss it. It was a feedback effect because it closed. I think the reason was that it closed the end of the day, right? And price declined and dealers, like well-known Swiss banks, they had to liquidate their short position in a 15-minute period when markets, like cash was closed, futures was open, and that really created some crazy liquidity effect that product blowout. Right. And I think it's so... Go ahead. blowout. Statistically, I think it was high convexity event. The S&P dropped like minus

four, minus 5% and VIX almost doubled. Yeah, right. Because of those flow effects and because in the prospectus it said they'd have to exit, right? If it was below a certain level that was seemed to be the biggest mistake but at the same time it was they wanted to stay solvent right so they want they had that language in there so they had an out instead of you know going bankrupt and then what have you done any quant any modeling on the VIX itself what do you see as the challenges on like doing quant work on the VIX itself? What do you see as the challenges on like doing quant work on the VIX? It seems like its own animal there. Yes indeed, I did some conservative and empirical work. I think the most interesting aspect is S&P puts, SKU and VIX, right? Yeah.

And VIX calls SKU, right? So it's more or less the same function. S&P tanks, right? You can be, say, long S&P puts, but also long VIX calls, right? So this is interesting. Another interesting effect is... So this is more like a skewness. And that ratio should be the same-ish, right? You're saying it gets dislocated, it gets out of way?

Yes, typically as a second-order premium, VIX skewness is steeper than say two. That opens up say a part like say arbitrage right you would be short VIX calls and long VIX puts, S&P puts. Yeah except Feb 18 that would have taken you to the cleaner right? So the skew is probably there because there's things that can happen where the VIX moves without the S&P moving. Yes. Another one, so this is more like, yes, this kind of skewness. Another one is term structure that is like the most interesting counterintuitive. So when term structure is in contiguous, you should sell.

Yeah. Because when it's actually in backwardation, you should buy, but usually it inverts when VIX is above 30. So this creates some sort of interesting strategies where if VIX is above 30, you would buy it. You would buy even though it's not cheap. But the expectation is again, if not the carry-based sort of... You buy the future because the cash is above the future. Exactly. yes. Even though it's in contango.

If nothing happens, you are on volatility and you get the card. This is ideal. It works a few times in power, but again, it's not tradable. It's interesting, but it's very hard because we're really looking at a handful of events where it would work. And all the time,

it's so much conditional on your rebalancing, what future expiry you got. But this is definitely interesting. I know that some people they do base strategies where

looking at the term structure of DIGs, you would trade one way or another with S&P, mixing S&P, S&P options, DIGs and probably VIX options. And then do you tie in a little bit of machine learning?

Like if there's so few of those big spikes, right? Like how do you do any valuable backtest or research of, right? Do you have enough data on those big spikes to make it statistically significant, your research. No, that's the point that you cannot make... It's really like I saw in back this kind of loan volatility, it's really, you make money two or three points, right? And therefore, it's always this strategy. You need to look at robustness, right? Somehow, if you want to do something like that,

as a standalone, it will not work. It should be part of more, say, broad, some kind of, how they're called hedge yeah yeah actually or just a long equity portfolio yeah yes okay let's move on finally to your new gig at Signum. Am I saying that correct? Signum? Yes, Signum. Signum. So tell us what you're doing there, why you joined and all the goods. Well, so I actually, part of my decision was a new book. So I have a bookshelf and there is some

mathematical methods, expected returns, risk pymia, active portfolio management, and then there is a new one that is blockchain and distributed ledger. That's your book?

It's not, unfortunately, not mine. Not yet, at least. It's Alex Lipton. He's is not unfortunately not mine not yet at least it's Alex Lipton he he's he

was previously he was global head of once at bank of

at Merlin Bank of America Merlin and we worked he was manager

of my manager but we had still very good relationships. And I saw he moved to blockchain like a few years back. And then I was following here and there his publications. Last year, we decided to work, start to work more closely like DeFi applications. So automated market makers, stuff like that. And so I kind of got interested. I still have, say, my reservation about where this all is

in go, but it's interesting, it's developing fast. And when I saw opportunity at Signum, I already had some feet in this area. I said, okay, this is it, I should go. And they're Zurich based there? Zurich, yes. Zurich is a digital bank. We do a lot of things like tokenization, custody, but also sort of traditional asset management so digital banks

all crypto or has some fiat has good old fashioned francs and dollars as well like you can hold cash but all activities are devoted to crypto, especially on the investment side. And so we just talked about the craziness of modeling VIX, now take us through the craziness of doing some quant modeling on crypto. Yes, that's perfect.

Yeah. Unfortunately, I cannot share my screen, but I tell you one, it's very interesting piece of figure that I want to show to everyone. Yeah. Part of my role, I created a database of all tokens, all protocol tokens or coins that were traded over time. And traded anywhere, any place?

Traded anywhere, yes. So we work with several data providers, so they all have this data it's just a lot of stuff change over time you have different names different tickers but I did work and I have more than 5000 different stuff that traded moreover the staff had volumes that volumes are

non-nones that market cap 5000 two and a half still active that volumes are non-none and market cap. 5,000, two and a half still active. Active in crypto meaning that there are still some activities. The protocol can be dead, right? So most of the time it goes up, it spikes, then it goes down to say whatever, zero, zero, zero,

something. But because in crypto per se there are no listing requirements, so since it can be data. So in my data set, there are 5,000 sort of debt stuff in five. So 5,000 total, however, it's kind of, it's just disappeared. It probably was not economical. Yeah. To keep track of them. Two in the 5000 are still even though some of them don't read frequently. Then what I did for each of them during the period that they were live, I computed the maximum drawdown and the PA return just.

Yeah. Right. And if I plot, so if I plot my X is a drawdown at my Y is PA return, right? So you understand that drawdown, usually if this goes minus 90 percent say drawdown minus 95 is no way to get back because you you need like yeah a thousand takes returns right doesn't happen and of course then your per annum return also is negative so actually you would be surprised how much since are in lower quadrant with drawdown

more than 90% and negative per annum return. Yeah. So you want to be in the top left quadrant, right? Yes. So 95% stuff it didn't make through. 95, right? That means that they were legit kind of projects with market cap, with volumes, but they didn't make it through. 95%. Our negative return and... Yes, negative return and effectively drawdown. So if you invested one dollar, maybe right now you would be less than one cent. Even though it's still, you would still have it in your wallet, right? You would not probably... What does that tell you? Trade these things?

It tells you that there is a case. Nevertheless, if we compute, so of course there are two biggest survivors, is Bitcoin and Ethereum. Moreover, if you build sort of theoretically equal weighted portfolio, right, that you invest in all tokens that have available market cap, you would actually have decent return. So, it would be of course risky, but say we would be talking of maybe 200% annual. Even with the 95% failure? Yes, because few of them, there were exceptional guys like like more than say 2000. Of course, I would not, there was exceptional return. But more, so that would be, for me, it's more like a venture capital. You would invest in smaller stuff. You would make money.

Also market cap. so instead of just taking equally weighted, we just go market cap, that would also make it through. So you invest in all investable universe, whatever is available at that time, just in proportion to the market cap. So you would also something even slightly better than Bitcoin and Ethereum, not accounting for transaction costs. But does that put too much faith in the past? Is most of that return in the 15 to 20 range? I think for me, yes, but also it's fighting them down. So what for me tells crypto is all about the recitation,

especially if you want to go, okay, you can invest in Ethereum, Bitcoin, probably those are already established. The rest, it's almost like lottery tickets, right? Don't buy, if you think that it's kind of everything will fall between, just go buy lottery tickets. It happens faster, right? And what about trading and what about building quant models to go in and out of them? Yes. So we actually, we're building right now, what I think most interesting, what we're

building is sectors. And in crypto, of course, it's very, there are trends. What everyone say last, like say beginning of last year was DeFi and mid of was Metaverse and FPEs Web3 and in my opinion now also it's much better

chances if you say if you select the right sector with proper diversification, you can generate high excess returns outperformance or Bitcoin and for me

what we are building it's we are building crypto sectors one of the projects that I think is very innovative and very interesting, what we're also doing NLP-based classification. Because unlike the traditional

industry, you have very well-defined sectors like financials, utilities, oil, pharma, and so on. For pharma, you have sub-sectors maybe. In crypto, there is nothing like this. So people know, say, DeFi sector they can define. Maybe Web3 you can define. But sub-sectors of Web3 are difficult to define.

And what we do is NLP. What's your take on Web3? There's a lot of hot takes that it's kind of a scam and just made up. Yes. Yes. Web3, I mean, well, Web3 layer zero, what is called layer zero is Polkadot, ICP, Atom. These are more almost like a foundational layers to create blockchain so to create a distributed chain of interconnected blockchain right and the important is that those are more almost like a programs where you could replace the protocol change the source code and it would preserve the existing application right as you know with blockchain, with Bitcoin, you have this forked version. When something happens that algorithm or protocol cannot develop or there are some

very big conflict where it cannot be resolved with say blocks, some kind of either fraudulent or some logical failure that you have to fork. You have to create a restart, a new copy of your blockchain with layer zero avoid it. And it's a code, it's a platform to create the apps, digital applications. So all kind of maybe DeFi, NFT, gaming, they can be built seamlessly on this platform. So therefore,

this part is definitely the top ones that they will survive. But I definitely don't think it's a... And from an investor standpoint, can I... I've heard before, right?

It's like getting paid to own the HTTPS protocol, right? Or all that stuff. So it's similar to that? We're saying you could own a piece of that protocol and get paid every time it gets used? Or is it more programming language

that we're just going to build things on? This one would be more programming language, right? And the protocols themselves that would generate the utility, like the apps, applications, they would generate utility through transaction costs. Yeah, but that seems to be the knock on all this too.

Nobody wants to pay all these gas fees and all these transaction costs in the new environment. What's the solution there? That is the point where I'm a bit skeptical on DeFi in the sense that there is not economic utility yet created for DeFi. in relationship to metaverse to gaming and if he is these things are like uh say it's a gaming right people play to to to to play yeah these are natural why i i personally actually i i believe in metaverse not not say more of this because it it doesn't have a clear competitor in traditional space, right? It's been developed, it's traditional players are coming there and it creates some need for, say, these DApps. That is some kind of entry point to your universe, to your game, to your collection. And what's an example of that? Like I could bring my virtual currency across games or something, right?

Spend it in this game, spend it in that game, use it there is like several. So Web3, as we talk, is interesting. Like as I said, we run NLP. So we have NLP being, again, so imagine in traditional company, you always have very big, when company like its first IPO, right? You have everything what they are doing. Once they are up and running, you have quarterly calls. You know exactly what they are doing.

In this space, in crypto space, zero, right? You have Bitcoin white paper is 10 pages or 11. This is what they do. Most of the protocols. But what we created, we created what is called natural language processing, where we would analyze actually textual content of protocols. And you can do different analysis, like say, what words do they most frequently use? And that information, once you have that information, you can say build clusters, right?

More or less, we can identify say Web3, and we can create some clusters. And so, for example, Web3, and we can create some clusters. And so, for example, Web3, so layer zero, the one that we said that used to build application layers, the most, the three words that make it different from layer one, from like say Ethereum and DeFi, like say from Uniswap with all this, is data security and provider. So which makes sense. So Web3 is about providing secure data,

is a storage or exchange for usage, right? For them, we go into this like chain link, for example. The one that actually use for them specific words are data and Oracle. That being said, they connect different applications and especially where there's a price information or there's some sort of reward information. This type of stuff that kind of connects you, allows you to connect different metaverses or exchange your money through different metaverses.

I'm lying that I got it, but it sort of makes sense. It's a nice, it's interesting stuff, right? And the best within Web3 was the best last year was best was performing actually a business to customer. The one that kind of distribute either videos, right? Or some sort of information. But that's a company or a token or a coin? It's a sector. I call it more of a sub-sector. So those sectors are made up of the coins, not the companies?

Yes. Yes. Got it. Okay. Strictly speaking, it's DAO, Decentralized Autonomous Organization. But for me, it's almost like some sort of maybe startup. There are some things that they have valid applications, they have valid white papers, developers. Right, so it's your whole point of coming back to the VC world, the digital VC world.

Who knows who the winners are going to be? Choose a sector. Choose a sector, diversify market cap, volume market cap, a lot of, especially biggest challenge for building a systematic solution

is really data. I spend much more than I imagined that I want to. Yeah. Data on cleaning, on understanding differences between different providers. And there's no standard for their APIs to pull the data. Yeah. So you're in the Wild West there. And then how many exchanges are there that you've identified?

Trading application is another thing. Yeah. Usually I would always... Fundamental data is important. At least this kind of, say, screening, that at least there's a white paper, you can compare it, you can identify it. You have market cap that some protocols actually,

they may overstate, understate. So you need to have different ways of comparing actually outliers, picking outliers, then accounting for volumes, especially once we build up a scalable asset management solution. I'm going to let you go here, but I need to know first what's on your shirt. Looks like some math. Ah, yes. It's actually, I want it to be for this. I like it it's from Stern it's

just some formulas. I thought it said Jeff for me I see a lot of F's on the top line. okay I think it's from physics so it's a force or something. The force mass times volume or mass times velocity is force, right? We'll show our lack of physics knowledge here. And then I was going to close with our, what would you invest in? But you sort of answered. So if you got a thousand dollars crypto addition,,000, where are you putting it?

So $1,000, it does go with Ethereum. Ethereum, all right. I mean, with Bitcoin, you probably want max for your buck. So I think Ethereum will outperform Bitcoin. And now if I 100x that, I'm 100,000 then I would go a little bit more like say sector based or growth market but a little bit

with some fundamental ways even as simple as market cap you don't need to rebalance frequently but it's definitely worse to have more deletification. And then so at a million same thing just larger numbers?

At a million yes I would do the same. I would try to build up a diversified portfolio. And then a hundred million in crypto. Is that scary? That scares me. So actually, I would say this. For example, at Signum, we actually have different pockets of investment. So we have venture capital, so first part of venture capital is the highest risk, highest return potential. Then we have, say, fund of funds. We are also building up systematic strategies. But most of the stuff, say, to invest in crypto, you need to have long bit exposure. You cannot do shorts. It's not a market

where you would benefit. So that being said- Preston Pyshko Larkin Really? Even with the past two years and some of these drastic spikes lower? Too hard to grab? Igor Mikhailovich Kuznetsov Difficult to short. First of all, you can only short safely is Bitcoin and Ethereum. The rest are very hard. You'd be paying 80% borrow rate or something?

Yes. So going back to fund of fund or some systematic strategies, we also have yield, so more some sort of arbitrage. There are different ways, either staking, also like arbitrage, say cash and carry arbitrage, which is more or less the same. And also some beta, like pure beta exposure. And so last bit on that, you mentioned the staking and the yield farming. Is that even further a field than traditional finance

right like the coins are one thing but that seems to be what are your hesitations there what have you seen that scares you there what excites you there I think

so what excites you of course they offer rates higher than say traditional finance. Although I think there is definitely a protocol risk, a wallet risk. So those... I mean, you're basically it's an unsecured loan, right?

Yes. So that's the arguments against is like, hey, you could do an unsecured loan on people's cars and stuff and get high rates also. So it seems to me some of the arguments against are you're getting unsecured loans at rates lower than they should be. Yes.

But we'll let people argue that one. All right, Artur, any last thoughts? You're on Twitter there and your blog post too, right? Tell everyone where they can find you. Yes, so people can find me on Twitter. I try to be active. Nowadays, I think there is a lot of useful information. I wrote an email. What's your Twitter handle?

We'll put it in the links. Artursepp. All right. A-R-T-U-R-S-E-P-P. Yes. Quick. I went down to the office yesterday, which is near the Board of Trade there,

and I parked in what's called the Trader's Self Park. It's from 40 years ago, right, where people would trade when they went to the Board of Trade and did the hand-to-hand combat trading. So the elevator buttons are all like gold, wheat, T-bills, Euro dollars, right? All the floors are named after futures trades trades so i just i took a picture of that and tweeted it out and it's like going on 3 000 likes i'm like i'll put out some well thought out thread or something intelligent and it gets two likes and one one retweet take a picture

of an elevator buttons and it gets 3 000 likes so Twitter is a weird place. Not that anyone needed to know that. And then what's the blog? Yes, also my blog. I think we spoke a little bit at the beginning. For me, quant literature is my passion. So I really separate. That is my profession that makes something that creates a utility but I also is more like a

passion where I spent more time say understanding how things work. So using either maybe more technical stuff or maybe more something beyond. And of course, I try to keep the blog. There's a lot of, say, my older academic stuff, some almost like memoirs on quant and systematic trading, but also new stuff is coming, especially on crypto. Right now we are working on very interesting work. It's more academic, of course. Academic, not academic. It's like a modeling flavor.

But crypto, we didn't mention it. What's the URL? We'll send people there. We'll put it in the show notes, but tell the listeners. My website is also arthurseb.com easy you got you got the best ones it's nice having a unique name right all right arthur thanks so much i'll let you go enjoy your evening there my last question when you moved your head i saw a picture is of your kids and ski goggles. yes exactly. where do you ski? with my wife.

well we ski I mean it's lucky to be in Switzerland and we regularly go skiing. actually our son now is six years and he's perfect skier here yeah i really i i never expected that i would be able to ski with my son so like say when he's six right and and that's my main goal in life i keep in shape just so my son's 13 now so he's yeah almost faster me but right i want to be in shape for when he's 18 and going down super hard stuff and i got to keep up with him well that's my motivation all right well now we got to get a trip out to switzerland to ski together we'll bring uh bastion along get it get a little quant ski trip going that would be great all right artur thanks so much thank you yeah That would be great. All right, Artur. Thanks so much. Thank you, Jeff.

We'll talk to you soon. Yeah. All right, ciao. You've been listening to The Derivative. Links from this episode will be in the episode description of this channel. Follow us on Twitter at rcmalt and visit our website to read our blog or subscribe to our newsletter at rcmalt.com.

If you liked our show, introduce a friend and show them how to subscribe. And be sure to leave comments. We'd love to hear from you. This podcast is provided for informational purposes only and should not be relied upon as legal, business, investment, or tax advice. All opinions expressed by podcast participants are solely their own opinions and do not necessarily reflect the opinions of RCM Alternatives, their affiliates, or companies featured. Due to industry regulations, participants on this podcast are instructed not to make specific trade recommendations nor reference past or potential profits, and listeners are reminded that managed futures, commodity trading,

and other alternative investments are complex and carry a risk of substantial losses. As such, they are not suitable for all investors.