In commodity trading, where volatility reigns supreme, finding an edge can be daunting. In this episode of The Derivative, we switch up our normal Commodity Trading talk with our non-trend following guest, Jae-Min Hyun of NWOne. With extensive experience in commodities trading and a quant background that spans Wall Street firms, hedge funds, and a new talent incubator, Jae-Min brings a wealth of knowledge to the table.
Join us as we explore the intricate intricacies of building a systematic commodity trading strategy using fundamental inputs instead of just price, why inefficiencies equal edge, the importance of risk management, the role of machine learning, and the rest of the challenges of navigating this dynamic market. Jae-Min Hyun’s insights shed light on the complexities of this niche within the world of finance, offering valuable perspectives for both seasoned traders and those seeking to understand the nuances of commodity trading — SEND IT!
_________________
_________________
From the Episode:
Semi-Annual Rankings Whitepaper
Follow along with Jae-Min on LinkedIn and for more information visit NWOne’s website www.nwone-llc.com
Systematic Commodity Trading (without Trend) with Jae-Min Hyun of NWOne
Jeff Malec 00:07
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. Hello there, it’s October and to quote Mark Twain This is one of the peculiarly dangerous months to speculate in stocks. You know, there’s our July January September, April, November, May, March June, December, August and February. love that quote. And it has been a little iffy here in October so far with stocks are down about seven and a half percent from their July highs. So we’ll see what happens through the rest of the year. Okay, how did this episode where we get to dig into one of the best commodity trading programs the past two years talking with Jae-Min Hyun of NWOne who surprisingly doesn’t use a trend following model, but instead of purely systematic strategy based on fundamental inputs. How does that work? Let’s find out. Send it. This episode is brought to you by RCMs managed futures group to help investors identify and invest in programs like this one, head over to rcmalts.com/podcast to see the write-up on this episode, or the RCM YouTube channel in this episode’s description, to find the link to the program’s performance and can explore the rest of the RCM database. And now back to the show. All right, Jae-Min, how are you?
Jae Min Hyun 01:36
Not too much, how are you?
Jeff Malec 01:37
Are you good? Am I saying that correctly? I believe so. Right? And what’s the etymology of the name there?
Jae Min Hyun 01:45
Well, there’s a Korean name behind it, but I’m not gonna pull your
Jeff Malec 01:52
well, we want to know what is it?
Jae Min Hyun 01:55
It’s actually meant to be. It’s a Korean name. So this is the Chinese character. I think Jae means that he’s meant to me, leader.
Jeff Malec 02:08
Oh, perfect. It’s apropos.
Jae Min Hyun 02:11
There you go. Thank you.
Jeff Malec 02:12
Yeah. And where are you in New York? I believe, right.
Jae Min Hyun 02:16
Yes, we are based in New York City. Midtown, New York.
Jeff Malec 02:19
All right. You get any of that leftover? What was it Ophelia? The storm leftovers?
Jae Min Hyun 02:25
In the has been a very wet weekend actually. Right now. It’s it’s raining outside?
Jeff Malec 02:33
And what’s your take on New York? Is it back? Is it dying? Is it fine?
Jae Min Hyun 02:37
No, I think New York is pretty much bad. Stuff, the reservations at various different restaurants and various different events. So yes, absolutely. No, he’s back.
Jeff Malec 02:49
Love it. And we’re in New York in Chicago. I can’t tell if that’s a siren here in Chicago or in New York. It must be in New York. We got him here all the time, too. So let’s start. Tell us a little bit about your background. Had some options trading back in the day? So yeah, let it let us know where it came from?
Jae Min Hyun 03:10
Yes, well, um, me as a person. I’m Korean. spent considerable amount of time the UK. Sometimes you need Asia. I had a very brief stint in France. And I’ve been in the US for close to 12 years now. So that’s my personal background. Yeah. It was my professional background. I started my career. And when we’re standing in London, I was transferred to New York 12 years ago. After leaving would study, I went to work for a couple of different hedge funds, Ellington management was my first race. And then I worked at today’s launch about trading, before setting up my own company and who
Jeff Malec 03:57
and so what were you doing at Morgan Stanley? Or I’ll go back even before that, where where were you in Indonesia? How long were you doing that?
Jae Min Hyun 04:05
I mean, that was like when my family moved to Indonesia when it’s my dad’s business by school in the United Kingdom
Jeff Malec 04:19
where do you go to school Eaten
Jae Min Hyun 04:22
into some polls?
Jeff Malec 04:24
I just know it all from that’s my, that pops up in the crossword every now and then. Sure, it does. So Morgan Stanley, what What were you doing there? Got your first taste of commodities?
Jae Min Hyun 04:36
Yes. So I did start stop doing a bunch of different things, convertible bonds, origination, interest rate derivatives, structuring equity, exotic structuring, but I ended up on the commodities trading desk. So I spent almost majority of my time trading commodities. Morgan Stanley, between London and New York.
Jeff Malec 04:57
And was that where you kind of banished to the commodities desk or was that a good place to be? Was it a coveted spot or no?
Jae Min Hyun 05:06
Indeed, yes. Morgan Stanley, I would say back in 2000s Toxic monitoring houses on the street between those Daniel Goleman. As you know, Morgan Stanley, we are more involved in physical commodities as Goldman’s they’re much more the paper side with their indices. But yeah, it was it was a major player in commodities trading.
Jeff Malec 05:37
And so what did that look like you were helping facilitate moving oil tankers and rail cars falling gray and all that good stuff.
Jae Min Hyun 05:44
Yes, sir. Yeah, absolutely. So, Tommy, say, again, we’re dominant in physical trading.
Jeff Malec 05:54
And so I’ve always thought was that easy to be on the trade desk? If you have the physical to back it up? Right. If you’re wrong, you can take delivery or vice versa? You can? Did it make it a little bit easier? A third level? Yes. But obviously, they’d still counted against your p&l. Yeah. Yeah, trade paper.
Jae Min Hyun 06:11
But I sat across from physical traders. So yes, I did learn a lot of ins and outs of physical trading. And also market color was helpful. But again, strictly we obviously we all had our own books.
Jeff Malec 06:25
And then when you say trading paper, you mean futures?
Jae Min Hyun 06:30
Yes. Futures and Options.
Jeff Malec 06:33
Then those are fighting words here in Chicago can’t call our beloved futures paper. So then moved out of there and went to what was the name of the hedge fund, Ellington management, Ellington. All right, what are we doing there?
Jae Min Hyun 06:48
So I was hired as a quant Portfolio Manager at Ellington Ellington predominantly a fixed income derivative house. Just set up a quant global macro fund. So I was hired as a commodity specialist for that fund.
Jeff Malec 07:07
And all pure quant yes systematic and but in even backing up the Morgan Stanley, were you pure quant then are you were learning the ropes and figured quant was more where you wanted to be?
Jae Min Hyun 07:19
No, I’m always sadly, we will discretion traders. Right. So there were few siSwati traders on the prop side of things as well as on the interesting side of things. But we are in the principle desks. So we’re discussing traders. So we’re marking making for big institutional blows. So we all discretionary traders, Ellington management was where I first learned the ropes of doing systematic trading.
Jeff Malec 07:51
And what what was your first foray there where you kind of trend to a trend following model as most newcomers to quant do
Jae Min Hyun 08:00
now mean, you obviously study various different strategies that are popular. So I did study trend, mean reversion carry type strategies when I was Eddington. But I focus mostly on what I knew, which is automotive fundamentals, training, and commodities. So my Wellington was to systematize a lot of the outputs that I traded, and I knew at Morgan Stanley
Jeff Malec 08:37
then you finally said, Hey, I’m gonna go hang my own shingle, so to speak. Yep, start out on my own. How was that nerve wracking?
Jae Min Hyun 08:44
Oh, absolutely. Not easy journey, by all means. But it has been rewarding.
Jeff Malec 08:55
And how do you think about that even now, as you hire your own people, right? Like, does it make you to those funds, they fear, obviously, you’re gonna go leave and start your own thing, but they also must support it and they want to train you well enough and teach you enough that you could go do it on your own right? So I kind of these firms strike that balance of like, we want to teach you everything we know. But we’re also don’t want you necessarily to leave but we probably know some percent of all you are going to leave.
Jae Min Hyun 09:23
Yeah. So that was the initiative behind you know, Launchpad trading, right. I went after Ellington management. The objective was to identify next set of stock traders, given the infrastructure, the capital and support to make them into qualified money managers, right. Yes, there are various different institutions, places out there that nurture relatively young traders, right but those seem to be very few and far between these days.
Jeff Malec 10:03
Yeah, harder and harder. And went in when the conversation now probably started with how, how well can you code how well can you do computerized things versus not? And then I buried a little. So what was your education? And were you a quant? By train? In your schooling?
Jae Min Hyun 10:22
Yes. So I started physics, I Cambridge, bachelor’s and master’s. So I did have a natural inclination to model things, right processes, mathematical models. So that obviously helped when formulating systematic strategies in commodities.
Jeff Malec 10:43
And do you feel the physics background? Do you still read? Do you think it’s a solvable puzzle? Or do you think it’s always right? It may be some far String Theory kind of astrophysics. Of like, there’s still unknowns. And we’re just have to kind of do as best we can to get close to the answer.
Jae Min Hyun 11:00
Absolutely. Well, markets always evolving, right? Physics, the physical world is relativity established, right? Yeah, this is a function of getting the right models, ideas, to fit, to come up with mathematical models and whatnot to fit the world. Whereas markets themselves are constantly evolving. So I think it’s a different challenge, right? When trying to fit mathematical models to market behavior, human behavior, says, there’s definitely different challenges involved.
Jeff Malec 11:37
Right now, I think you see, a lot of people come out of out of use science loosely, right, or mathematics or science, and they come in and they think it’s 100% solvable. And if they could just measure this, or if they could just measure that or get the right factor. They’d solve the machine and be able to print money and all that stuff. But sounds like you’re saying not not quite that easy.
Jae Min Hyun 11:58
No, I don’t I don’t think it is, I mean, certain areas of I think trading are close linkages, I would say. So if you talk about assets that are trading, equity, long shorts that are trading, statistical arbitrage trading, I think you probably see more of those PhD types. But most definitely non commodities, because commodities require domain knowledge. Commodities, and they don’t teach you that in schools. Right?
Jeff Malec 12:32
Right. There’s a great book on books that predictors. It’s about, they were out in Santa Fe, I’m gonna forget the name of the firm. But they basically a bunch of PhDs and they left and tried to model all this. And there’s chapters about how they didn’t realize there was slippage, and they didn’t realize commission costs and exchange. Right. So it’s not just the lessons they had to learn to get their models close to what they thought they could do. Were were enormous. Back in the 70s. We’ll put a link to that in the show notes if I can remember. So anything else we need to know the background had, where’d you come up with the NW? One name?
Jae Min Hyun 13:13
Oh, yeah, I’m just named after postcode in London, which is where I used to live. So I was inspired by my former boss, Mike Marino’s Eddington within Darlington after way, girl, so it’s a kind of a, you know, homage to like,
Jeff Malec 13:28
I love it. I didn’t even know how postal codes work. So NW one that sounds that sounds fancy. It sounds right in the middle. First postcode?
Jae Min Hyun 13:37
That’s right, Northwest one. So it’s very close to well, it includes Regent’s Park, which is where, where I used to live opposite Regent’s Park, which I used to live before I came to New York.
Jeff Malec 13:50
And how long you been in New York?
Jae Min Hyun 13:53
New 12 years, 12 years.
Jeff Malec 13:56
And how long’s NW one been up and running? We’ve been up and running six and a half years, six and a half. So what are what do you think? What are some of the best parts about running your own firm versus working at a hedge fund and some of the worst parts?
Jae Min Hyun 14:11
Um, I would say working at a obviously, having your own company enables you to make decisions that are purely based on your own thinking, right? So too many people. So with that flexibility comes more responsibility, right. But again, you know, I’ve worked at different institutions, big institutions, small institutions, and I you know, I, when I started the business, I thought I had a pretty good idea of what a perfect institution should be. And I am striving to achieve that. Working at pods. In other places, as their own challenges, right, on the upside, you’re given a lot of resources. But the downside? Capital obviously, is relatively competitive at those shops. And they are very strict risk limits that sometimes move as a function of the market or its opposite. So those are the challenges which prompted me to leave that space and start my own shop.
Jeff Malec 15:32
But funnily enough, you didn’t, it wasn’t like you were, well, I don’t know what you were like at the shops, but it wasn’t like you needed to be more risky, right? You’re not very risky now. So
Jae Min Hyun 15:41
know when the objective function is maximizing risk adjusted returns, so you know, in the business of offenses.
Jeff Malec 15:55
So let’s dive into the strategy event. For those you want to go download our rankings white paper, and this has been up there amongst the best over the last couple of years. So let’s just start with kind of a 30,000 foot view, what you’re trying to accomplish, what the strategy does, and then we’ll dig in from there.
Jae Min Hyun 16:15
Yeah, so a program, the objective is to extract alphas or inefficiencies, from commodity markets that are based on individual commodity market fundamentals. Right. So we mostly focus on market fundamentals, as opposed to technical factors, or heuristics, that a lot of our competitors that specialize in systematic strategies do. So that’s our biggest differentiator compared to competitors.
Jeff Malec 16:51
And pure commodity to right. Yeah, we only trade commodities. So pure, systematic commodity. And then that was interesting to hear you say that. So do you view alphas and inefficiencies interchangeably? Are those the same thing to you? Yes. But that implies that you have to be able to capture the inefficiency, right or profit from it.
Jae Min Hyun 17:12
Absolutely, yes. And again, there are all physical inefficiencies that are present in the market, that you may not be able to capture using systematic frameworks, right. So that’s the way I see it.
Jeff Malec 17:27
So pure commodities, how many? What are we talking to Global list? The most liquid commodities? How many commodities are we taught us?
Jae Min Hyun 17:34
We trade up to 20 different commodities. We trade in terms of instruments, we trade our way features into commodity spreads, calendar spreads. So we’re relatively varied in terms of exposure across six different commodity sectors.
Jeff Malec 17:53
So are the usual suspects their oil grains?
Jae Min Hyun 17:57
Yep, soft meats, precious metals, base metals. For base metals? We do not trade LME.
Jeff Malec 18:06
I have not tried that. Let
Jae Min Hyun 18:07
me know I’ve not traded on any contracts since leaving Morgan Stanley.
Jeff Malec 18:12
Was that so that wasn’t a result of their nickel debacle? No, it’s just that where they reneged on all the trades?
Jae Min Hyun 18:21
No. Obviously, in retrospect, I think it was the right thing to do. But LME contracts are relatively expensive to trade as far as I’m concerned. Because you have three to three months and then roll to the nearest futures look alike. So you, in principle, you end up paying transaction costs twice. That was going to be frustrating for me. So for those reasons, I decided not to trade on any contracts and our program.
Jeff Malec 18:50
Got it. And so those parts again, so you have directional futures strategy. And what was the other inter commodity spreads?
Jae Min Hyun 19:02
Oh, yeah, it’s a commodity spreads, because that’s part of the directional strategy. So that
Jeff Malec 19:07
I can be long coffee and short cocoa something like that. Yes. In
Jae Min Hyun 19:11
principle, yes.
Jeff Malec 19:15
Okay. And then we spoke before the future strategy is shorter term.
Jae Min Hyun 19:23
Yeah, right future strategy in the short term in terms of holding period as a holding period around one week. And the calendar spread strategy that has a long longer holding period between one to eight weeks and our last intraday strategy that trades are a features that has a holding period of between half an hour to one hour.
Jeff Malec 19:49
So let’s go through those in order. So the so the future strategy directional holds up to a week. How it’s not trend following it’s not mean version, it’s some something in between something different. So how are you generating without giving away the secret sauce? How are you generating those signals? What’s What are you looking at there? Yeah,
Jae Min Hyun 20:10
so for the ROI feature strategy and the calendar spread strategy, and to certain degree the intraday strategy driven by right now nine different classes of models, right. So each class a model, the objective is to identify orthogonal sets of inefficiencies or offers in the market. And the idea is to build collectively exhaustive way of extracting out where the individual model classes are orthogonal to each other. So within those nine different classes, you know, we have different ways of identifying certain deficiencies. Could be anything between pure fundamental data set will be seasonal, the risk premium could be related to the black relationships between various variables and asset prices. Some of it could be based on market positioning, oh, by the speculators, humans, always one type of stock
Jeff Malec 21:19
is an easy way to think about that, as I know, the price of production for oil is $76. And we’re trading at $86 or something, right, that relationship between that fundamental piece of data and the market price.
Jae Min Hyun 21:36
Yeah, that is a component Yes. So, we do look at the SMD dynamics, for various different commodities when the data is available, we will then run various different statistical tests against those variables and changes in either clockwise calendar spreads or more the spreads. And for those of a specific set of say, fundamental data set, if it has a high forecasting power for calendar spreads, then it will be considered for inclusion in the final portfolio. But we obviously go through a rigorous in sample out of sample testing. And after that we have a period of paper trading, before anything can be even considered to include in the portfolio
Jeff Malec 22:28
is the whole idea. There’s you’re using those fundamental factors to kind of assign the value price for the commodity. And then if you’re below that you will buy if you’re above that you’ll sell.
Jae Min Hyun 22:42
Yes, no. So we in terms of our Mala codes, we try to, we focus on continuous outputs, right. So we never have single step function type model output. So all our models generate a continuous spectrum, minus one minus bearish plus one less bullish and everywhere in between, everywhere in between. And then, by having continuous distribution, it’s much easier to combine different signals sets, and to minimize any data mine, right. Because if you have a step function, you could always tweak your, you know, your threshold. narrative or pick the most recent historical data, but we avoid that by using continuous spectrum.
Jeff Malec 23:31
That’s interesting. So So across all those models, we’ll stick with oil, or we’ll get off oil, because we’re gonna talk about that next but cotton, you trade cotton? Yeah, that’s a challenging market. So we’ll we’ll we’ll talk cotton. So model one is saying 26% long or something like that? Yeah, model model to saying 16% Sure, model three saying and then you’ll get the net of all those and say, Okay, we want to be 32% long.
Jae Min Hyun 24:01
Absolutely. Yes. So we apply different weighting function across this for different model outputs. And then the resulting aggregate of those individual outputs will be applied to either cotton flat price calendar spreads, or we don’t need to commodity spreads and cotton.
Jeff Malec 24:23
Got Yeah. And but the idea there is not that so you’re not necessarily saying cotton’s underpriced versus its real value, we want it to go up. You’re just saying over the next week, we think it’s gonna go wrong. The aggregate of the models believe the price is going to rise. Based on those factors. We don’t care why we’re just thinking,
Jae Min Hyun 24:43
Well, we do care individual models will then will have their own reasons to say why should be going out right. For example, one of the models could be based on fundamentals or say okay, fine in terms of s&p balances laters USD balances we get the model is bullish calling spreads on the flip side and we say, you know, the currency dynamics between US dollars versus various different other foreign currencies made the tapes bearish, right.
Jeff Malec 25:13
I like that. So that’s the directional futures. And then it could be in theory, could it be 100%? Long? Right, that one on all those are there’s limits within within each sector and within each market?
Jae Min Hyun 25:25
Yeah, there are the, so, individual model level will produce model signaling from minus one to plus one, which will then get aggregated based on our proprietary algorithm, they will get then get passed through to our portfolio construction engine. So at the portfolio construction engine, we face constraints, right? So we have constraints at various different levels. So what are the constraints we have set the sector level that say no single sector can take up more than 50% of the portfolio risk? within the sector? No single commodity can take on more than 30% of the risk that ensures our portfolio is diversified at any given time. Right? Because, you know, we are we trade across up to 2020 different commodities, we rely on diversification, right? So therefore, our portfolio has to be diversified in terms of exposure at any given time.
Jeff Malec 26:29
The what would you say some right trend followers that trade 120 markets would say that’s not enough diversification only 20 markets isn’t enough. Of course, there are a lot of those that they’re mentioning, we’re probably fixed income and currencies and bonds. But like, what’s your view on how much? How many markets is enough? Versus diminishing returns of adding more markets? How do you do that?
Jae Min Hyun 26:54
Yeah, um, so in trend followers, you know, what they’re trying to do is they’re trying to extract this unique set of alphas, right? That stem from either cognitive biases in the individual market participants or inefficiencies and information transmission mechanism, right? What they’re trying to do is they’re trying to extract that small piece of alpha, that are present in every single market. And the only way you can consistently do that is by deploying this strategy across many, many different markets, right? So they need diversification to extract a relatively small amount of alpha that’s present in each of the markets, right? As I approach is quite different, right? So we don’t we don’t look for top down heuristics based alphas. We studied individual markets, right? Say, Okay, what are the fundamental drivers of crude oil? Or corn, soya? Right? So we build understanding of brand division, commodity market, fit relevant models, rigorously test them, and then we build portfolio from bottom up. So our portfolio is very much bottom up. Therefore, by construction, you can say that we are our exposures are less correlated, right? Because and trend following. We’ve seen this many, many times, right? If the trend breaks in one single market, that tends to translate to the other segments of markets, right? Because people unwind certain set of strategies, has a knock on effect, or be treated as a market as his own his own thing? Do we get affected by some of these flows from Trend followers? Of course we do. But again, the way we construct our portfolios to be robust enough to withstand any of those flows,
Jeff Malec 28:59
is that is that an input even? Is that one of the data pieces in a way of those flows?
Jae Min Hyun 29:06
Yeah, I mean, we do monitor flows, we monitor flows of speculative community as well as aging community. And they do we have models that contribute signals from from those types of offers or inefficiencies.
Jeff Malec 29:31
So it’s the futures. Next was what calendar spreads or inter commodity spreads, or do you use those interchangeably? Calendar spreads
Jae Min Hyun 29:39
need to commodity spreads, falls into the bucket of our futures trading?
Jeff Malec 29:44
Okay, you’re basically just creating a new futures market. Yes, yeah. All right. What’s some of that? When will that get really crazy or there has to be some link it has to be soybean and being male or? Yeah, oil and gas. You wouldn’t do Something totally crazy.
Jae Min Hyun 30:02
No, we only look at pairs or spreads that are actively traded by, say,
Jeff Malec 30:09
hedgers. Okay. So it’s like a logical fundamental link as well as quantitatively. Alright. Yeah. Okay, so then the calendar spreads what’s going on with those?
Jae Min Hyun 30:22
Yeah, calendar spreads is obviously into delivery exposure across different commodities. And one nice thing about calendar spreads is that you’re hedged from currency, exposure point of view, right? So if you don’t write, say, cotton contract to long cotton, then you shoot us dollars, right. Whereas in calendar spreads, if you have going short, across different parts of the curve,
Jeff Malec 30:53
the whole thing will move up or down. Yeah,
Jae Min Hyun 30:57
mostly. So that’s nice characteristic of calendar spreads. And also, calendar spreads, obviously, I’ll link to storage dynamics or various different commodities. And storage obviously has is mostly driven by fundamentals. So trading calendar spreads is a is a pure way of expressing or extracting any outputs that are related to fundamentals.
Jeff Malec 31:32
Do you view that as trying to earn a yield sort of like almost like a carry strategy?
Jae Min Hyun 31:38
No, not necessarily. I mean, there are certain spreads that exhibit you know, either positive or negative betta, right? We want to express it. But our goal is not to extract two sets of premiums. If it’s there, fine. You know, we will extract it. But our goal is not exclusively extract those sets of themes,
Jeff Malec 32:05
it be more of a based on these fundamental factors that we’ve brought in as data, we think that it’s going to steep and it’s going to flatten,
Jae Min Hyun 32:14
right, yes.
Jeff Malec 32:18
Do you ever get curious and run that on bonds? Just to see what we’re where we’re gonna go? Where the yield curve is gonna go?
Jae Min Hyun 32:25
Ah, no, unfortunately, no, I tried to focus on the things that are that I know.
Jeff Malec 32:29
Exactly. Yeah. And so the futures was days to weeks, the calendar spreads weeks to month. That makes sense.
Jae Min Hyun 32:39
Yeah, each and a couple of months. And as you know, you know, calendar spreads exhibit lower dollar volatility. So therefore, you have to have a relatively long holding period to justify transaction costs, as well as the volatility of those contracts.
Jeff Malec 32:57
And but in the past, some people have been right, kind of a Widowmaker trade too, right? Because it has such low volatility until it doesn’t, until the spread totally, you know, who were the nat gas guys, famously, but amaranth. So how do you avoid that? Like, okay, I know, it’s a low volatility, trade, I’m going to hold it months, but it has the ability to get totally out of whack. Right, Hurricane Katrina, those, all the production was shutting down in New Orleans. So there’s outside events that can make those spreads go pretty crazy.
Jae Min Hyun 33:29
Yeah, I mean, that’s a definitely concern. So we mitigate that by number one, using relatively conservative estimates of volatility for sizing trades. And number two, with multi specialists, right, so we went through sets of spreads that can blow out or risk reward profiles. And we position our strategies or, or trades to be on the long convexity side of things, right. So those spreads, we talk about a lot of these like ask spreads, especially some gasoline spreads, some of the seasonal spec changes. They have a sec being long carry, right?
Jeff Malec 34:23
Yeah. Where it’s like a negative skew, right. I’m collecting, collecting, collecting until not.
Jae Min Hyun 34:29
So we tend to build our portfolio to avoid once we want to build a portfolio of complex expected returns right? positively skewed. So often, we take the other side of that so often, you know, some of our positions end up paying away small amount of decay. We’re expected weather to be big returns.
Jeff Malec 34:53
And so we have a lot of long volatility people on this pod. Do you view it as kind of a long volatility program? Is it going to be Do you believe it’ll do better in a way that March of 2020? Those kinds of periods when volatility is spiking?
Jae Min Hyun 35:07
That’s kind of difficult to say because, you know, we we don’t construct our portfolio or signals with those specific characteristics in mind, right? By construction, we tried to be agnostic. But we do better in those volatile periods of high elevated levels of macro volatility? Probably not, because our strategies or models are based on individual market fundamentals, right? So we want we need an environment where the individual market fundamentals can play out. If there’s an elevated amount of macro noise, in our signal to noise ratio, each model is expected to be lower. Right? You cannot expect our portfolio to perform, you know, when all hell’s breaking loose, right? But then you have long ball players out there, you have trend followers, or who’s who are supposed to take advantage of this market conditions, right. So we don’t we don’t aim to offer them to be like,
Jeff Malec 36:19
yeah, by the way, it’s interesting to hear you say you are cognizant of what we’re not putting on at least in the spread trades. Right, these huge negative skew negative convexity trades. But yeah, that’s it makes me think of back during COVID. Right. And they’re on the news saying, We’ve never seen the meat market like this. And there’s, we have to kill out, you know, call this herd and do all this stuff. So if that kind of talk is going on, it’s generally not fitting with your models, right? Because it’s outside the norm.
Jae Min Hyun 36:49
Yeah, that’s very good. It wasn’t particularly friendly for us. Huge demand destruction across all commodities, right. So those kind of idiosyncratic events have an overwhelming effect across all different markets, you can’t expect us to perform, because we haven’t designed or portfolio or strategies with those events in mind.
Jeff Malec 37:11
Right, but your risk control, you’re not going to blow up in those types of events, you’re just going to mainly be on the sideline or be pushed over conviction. And then the most interesting piece of these three to me at least is this intraday oil trade. So what’s going on there? Even though you said you can’t tell me too much about it? We’ll try and pull something out of it.
Jae Min Hyun 37:34
Yeah, so the intraday trading in the world, the idea is to model behavior of the physical traders who are active during certain periods in the day. Okay. Our objective is to trade around, right? So we anticipate the physical hedges are going to do and we trade around.
Jeff Malec 38:03
So fair to say flow based for lack of a better word or demand by side base, something like that.
Jae Min Hyun 38:10
Yeah, it’s flow that I derive from hedges. And as you know, hedgers, they tend to be less price sensitive, right?
Jeff Malec 38:23
Yeah. And who are we talking about? They’re like airlines and manufacturers, whatever, any bonafide oil hedger or even the majors who are trading in and out on both sides. Oh, yeah, everybody. Okay. So not just hedgers. But while they’re hedging on both sides, got it. I was thinking just a long time. Yeah. Great. All right. So then those three are always working in concert, is there a set budget, like a third of capital to each are they’re just each model is independent bottom up, you’re saying with the risk controls on top?
Jae Min Hyun 38:58
Yeah, so we do try to limit risk exposures across different strategies, right. So each strategy, the three strategies, dri teaches, calendar spreads and intraday strategies, they’re always competing for risk capital, right. So, with a one three view of risk allocation across different strategies, but again, similar risk management framework is in place. So, no single strategy can take up majority of the risk capital majority in the strategy level sense is 60%. So those single strategy is expect to take on more than 60% of risk capital portfolio level that ensures that you know our where in terms of our exposure and given time, or different groups of ours is diversified.
Jeff Malec 39:55
And so the 50% max sector and market inside each strategy so And then each strategy sleeve is 60%. Max.
Jae Min Hyun 40:02
Know the sector exposure that’s done on portfolio?
Jeff Malec 40:06
Level? Yes. Okay. Right, because then you could have 50 or 6030 3030. You could have 90 in one market if you didn’t, if you did it my way. Stick with your way. And so it’s done is quite unique and why just in oil for that intraday trade?
40:26
Um,
Jeff Malec 40:29
does it do do any other markets or No, for now, it’s just annoying.
Jae Min Hyun 40:32
I was just well, there’s interest what there is a nominal oil that’s run by fundamentals that induces in hedgers to behave in a certain way. That’s why it’s currently trades currently only trades, oil markets. But we’re always researching different markets for intraday strategies.
Jeff Malec 40:56
And let’s touch on that for a second. Just what your whole research process looks like. How long it took you to get to where you are now, what this might look like in two years will be the exact same are always progressing. What’s What’s the research side look like?
Jae Min Hyun 41:13
Yeah, so we’re always progressing. The nine different class models currently line and began with one class of models, right? We’ve added on many, many models, we’ve retired many models, the research process is always ongoing. But so it’s always incremental. Right, you can’t expect us to make any big changes. Because we need, you know, out of sample data to justify model inclusion or model retirement. So you’d expect the makeup of our models to expand overall, as we have more resources deploying, as they’re more datasets, market evolves, we’ll be adding on more and more models, more strategies. At the same time, we’ll continue to retire any strategies or models that are not working. As markets have structurally changed. So yeah, it’s an ongoing process.
Jeff Malec 42:20
And due back to that other question, the 20 markets, is there some limiting factors or point of diminishing returns? Right? Could you eventually have 500 models? 5000 models?
Jae Min Hyun 42:31
Yeah, absolutely. They’re all there is the element of diminishing returns, right? Because there is there’s so much inefficiencies out there, right. new efficiencies obviously, we try to extract using different models often could be overlapping to to a degree. So if that’s the case, then we’ll look to enter new markets. Right so we’d love to be able to trade Chinese commodities onshore and offshore goofy commodities, that’s in the cards also like to be able to trade FX contracts and also interest rate contracts. But that’s you know, much much further down the line
Jeff Malec 43:17
I know a guy who can help you get into China zooming back out a little bit we touched on a lot of this, but I just want to follow up with the why commodities question right why in touches on you want to get into interest rates you want to get into that stuff, but why why is it just commodities for now?
Jae Min Hyun 43:44
Well, number one, you know my background has been in commodities trading to mid 2000s. But another interesting aspect of commodity markets is that you have this one group of market participants right namely hedges who are willing to pay pay away hedging premium to the market, right. So you have a large group of participants that are paying away hedging premium to market now from speculates on the view, if you have relatively good models to model the price process, although the fundamentals you can pick up those edging premiums that have been paid away by by producers and consumers, right. That’s the unique aspect. So another participant in the market where a group of participants are willing to pay away X amount of value to the market right as opposed to having to fight with other like minded speculators are willing to be limited market inefficiency in the market, right. So that’s why I like in commodities mark is pretty unique, but point of view. Also, Marty markets relatively inefficient compared to other markets, right because Talk about commodities. You can talk about different grades of commodities. Number one, number two, were in geographical attributes associated with commodities, right? Is it soybeans growing the US? Is Brazil, right? So it’s not just like one single asset, talking about the spectrum. So that leads to more inefficiencies in the market. That can be extracted, right? If you’re an informed speculator. So this is why electricity, commodities,
Jeff Malec 45:42
you’re preaching to the choir, the how do you view? Right? The the traders that the big commodity houses who have billions of dollars, they have the same ideas, they have the same thoughts, and they have way more money. So how do you view that kind of competition? Of right, why don’t they do what you do?
Jae Min Hyun 46:05
Yeah, it’s a very good question, actually. So some of these, some of these big commodity trading houses, their primary business is, is in physical commodities right now, supporting these physical commodities from region to region be storing those commodities. And then obviously, hedging those exposures, send me my head. So they’re pretty much involved in trading physical commodities. And they use papers last futures market to hedge their positions, right. And we are exclusively focused on trading futures. So, in fact, a lot of the activity A lot of these trading activities that are done by physical commodity trading houses are useful for us. Right, we use information, the hedging activity of these physical traders to infer to augment the current unlimited market state for very simple commodities, right. So I don’t particularly view them as competitors, per se. Right. They are very, they played a very important role, obviously, the the global trade flows. And if anything, you know, their hedging activity tells us more about the current bond market state. And as to why, why are they not involved in trading futures? In the way that we do? I’m sure they do. But, you know, give you an example of, you know, a car company, for example. Right. So why is Mercedes not involved in running taxi business? So Uber, for example, right. Their primary focus is physical commodities. Right. And our primary focus is futures trading. Obviously, there are areas that overlap, but most part, they stick to what they know. And we stick to what we know, ultimately. Sure, you know, in 1015 years time, right? There will be some sort of vertical integration, right? Or the car industry, that’s what Elon Musk is trying to write. He’s trying to integrate have rover cars running around in commodity markets. Maybe the future that will come
Jeff Malec 48:40
that breaks my brain thinking about that, if you had, right, if you were inside one of these big places, and had all their information in their flow, and we’re trading against it. Right? Would it basically eliminate the inefficiency, because then there are other side of the business would be like, Oh, Jae min stoutness. We’re overpaying here basically by his p&l, so we’re going to, we’re going to lower that side and your p&l is going to come down. Right. So the it’s kind of hard to think about them existing in the same place because it would kind of cannibalize one of the other with each other. But you’d think like Citadel, or some of these big firms would say, hey, and they do right. They have commodity trading, they have oil trading. So I’m sure they’re looking at many of the much of the same similar stuff. But
Jae Min Hyun 49:25
yeah, there’s definitely you know,
Jeff Malec 49:28
but you’ve got your own flavor that does what does what it does. Just coming back to how do you view right? It’s you started as discretionary on that Morgan Stanley desk, and now totally systematic. Are there ever times where you’re like, Oh, I wish I don’t want to put on that position. That doesn’t make any sense or if I was discretionary, I do it separately. Like how’s that journey been coming from a discretionary trader to was purely quant systematic?
Jae Min Hyun 49:53
Yeah, it’s, um, it’s an ongoing journey. I think it’s a journey that though many people take maybe obvious reasons, but I’ve certainly been on it. It’s been it’s been interesting, right? Are there are times in which I want to do a race on my signals? Absolutely. Have I done it before early in my career? Yes, I have. The beauty of systematic strategies is that you’re looking for sets of signals of phenomenon that are repeatable, right? And a relatively persistent, right. Whereas if you’re a discretionary trader, you look for one of these, you look for one off relatively big opportunities that come by maybe two to three times a year, right? You’re limited to that, right? This is the metric, you’re able to extract relatively difficult to attain alphas, but because you’re systematic, you do it in a systematic manner, and you do it across many different strategies, you’re able to extract those sets of alphas. So I think it’s a much more sustainable way of extracting alphas from the market. That is not to say, you know, there are plenty of successful discretionary traders out
Jeff Malec 51:18
there, right. Less and less these days, but yeah,
Jae Min Hyun 51:21
I much prefer this approach. Compared to discretionary trading.
Jeff Malec 51:29
Right. And how do you view right the, like, the USDA report is notoriously error prone, lots of revisions. China’s been known to manipulate their orders and whatnot to get better prices for themselves. Right. So you feel like a discretionary trader could better see through those, that bad data, so to speak. So, yeah, how do you view that? Like, okay, I need to make sure this data is what I’m really seeing.
Jae Min Hyun 51:56
Yeah. So what we do here is we augment in a pure one minute data set with some other marketing arrived. Signals, right. So, if those two are contradictory, you know, ultimately, hybrid motor signals tend to be lower right? experimental model will say, x, and the market derived model will say Y, right, so that exposure will be less. So we have strategies and models that mitigate some of those types of nuances you just described. But you know, more often than not, these things are relatively short term in nature, right? So USDA, they may get it wrong. One or two crop cycles, overall, they do a pretty good job.
Jeff Malec 52:51
When, but for short term trader, that could be all the difference in the world, one or two cycles. But that just shows to me like it’s, it’s hard, right? We can sit here and talk about, oh, we just throw in this fundamental data and get these price picture and can train this stuff. But especially in commodities, that data is harder, right? It’s harder to come by or no? Like, do you use any old data sets, like satellite imagery or things like data? Or is it all rather standard stuff that comes out of Bloomberg?
Jae Min Hyun 53:22
Yeah, so we have studied some of the alternate datasets in the past. And there are many providers alternate datasets. But because often, when we go through model formulation, and model validation, so these alternative datasets not meet our standards. So we currently do not use any alternative datasets in production, or we keep track of many different data’s alternate data out there. So for the models that are in production, we predominantly use data coming out from government agencies, or well recognized trade bodies.
Jeff Malec 54:10
And I’ll come back one bit on commodity. So is, do you have a mandate to have commodity exposure? Like would someone switch out their 5% long commodities at a institutional portfolio for your strategy? Or they’re not not necessarily going to get the commodity they’ll get the commodity exposure, but not the correlation to upside and commodities?
Jae Min Hyun 54:33
Yeah, program. We don’t have a like a you don’t have a long bias, the commodity markets. So if you’re looking specifically for price related inflation protection, you cannot expect that from our program. Right. So our program is designed to extract inefficiencies or outliers that are driven by market fundamentals across different different essential commodities, right? So we are more absolute return focused, as opposed to offering data with respect to a given commodity market.
Jeff Malec 55:13
So throw it and throw it in your alt absolute return bucket. And then as you’ve been on both sides of this, I wanted to ask the question of kind of how do you view and coming back to your strategy design? The pros and cons of like, okay, I’ve developed 20 models, say, in house that I’m implementing, they’re non correlated, I’m getting different alphas versus I’m a pawn shop. And I’ve hired 20 different alpha sources to kind of write what are the what are the pros and cons of hiring it versus building it?
Jae Min Hyun 55:45
Yeah, I mean, I can’t speak much for hot shots, because, because I’ve never, we, I sort of worked at a pawn shop. But I can speak from on the funds perspective, right? So you can conceivably replicate a lot of the alphas that we look for by hiring different sector specialists, right. So you can hire someone who specializes in oil, not gas, grains, soft metals, meats. But then by doing that, you effectively, you know, it’s much more costly way of doing things, right. Because you have to pay. If you don’t perform the funds, you got to pay management fees, and incentive fees to individual manager, right. And after that, you don’t get in any effects. Yeah, group of traders may make money, another set of traders may lose money, right. But then you still have to pay fees, management fees, regardless, well, you have to pay incentive fees on the managers who make money, right. But you don’t get any call back from managers who lose money, right? So by trying to replicate by a diversified set of commodity options, the way that we do by Empire, discretionary traders, is a much, much more expensive way of doing things. Whereas for us, we deploy systematic strategies in place for discretionary traders, right. And we pass on automatic effects to the investor. Right? So that is a very cost efficient way of getting exposure to alphas that are related to individual commodity markets.
Jeff Malec 57:37
But how would you view right it’s like a bet on and I wasn’t necessarily saying a patch up trying to do exactly what you’re doing. But just in general, like, right, I’ve, because you’re kind of creating the same thing, all these different return drivers, right, and ensemble. And they’re doing the same thing, creating an ensemble, they’re not correlated, let’s generate the return. So it’s kind of just a question of the hive brain versus your brain, right? Like you’re, it’s all coming out of your brain, and your research and your team. But versus each of those in theory comes not only with their own model, but their own way of thinking about it. And although I would suspect after being in there for a while, they kind of get some groupthink and approach problems the same way. But so I don’t know if there’s a question in there, but it’s just interesting to me, right of the, the two paths of, hey, you can do this all with one brain? Or you can do this all with kind of a galaxy brain of these different people. Look, you know,
Jae Min Hyun 58:31
kind of these leads on to a nice segue, no, we all expanding, you know, it’s not just gonna be one brain race, I’m, we’re actively recruiting multiple continents to control this. We’re going to be value additive, right to the organization. So I don’t think it’s about our approach versus their approach. I think the approach is, it’s about dedicating resources to extract inefficiency from the market, right. And we are actively looking for more brain power. Protein,
Jeff Malec 59:10
which leads me into AI. So do you use AI to augment that already? And just being a physicist being a quant? What’s your views on the AI buzz recently?
Jae Min Hyun 59:22
Um, yeah, I mean, it’s been fantastic right. So the AI and machine learning investments into those areas have definitely been helpful for us. So we use the all the architecture we use in house is developed from the open source packages and tools available in the market. So has been very helpful to us. So in terms of, you know, using AI tools is less relevant for us, because an AI you typically need very, very large sets of data, right to train the models, right to formulate your new models, formulas that train and validate your models. Because we do automatic fundamentals driven trading. So we treat each individual market as is its own entity, right? There simply isn’t enough data available for a given market to train a relatively sophisticated AI model, right. So those are the reasons that prevent us from fully adopting AI. But, you know, we are using tools that are machine learning based, we actually using those tools for research, as well as portfolio generation.
Jeff Malec 1:00:47
But talk, explain that for a minute. Because corn goes back. Right, I think there’s 100 years of data on corn futures or whatnot. But you’re saying there’s not 100 years of USDA reports and, and some of the other fundamental data pieces?
Jae Min Hyun 1:01:01
Yeah. So this goes back to, you know, question of why on call the people doing what we do,
Jeff Malec 1:01:11
right, yeah.
Jae Min Hyun 1:01:15
Typical systematic strategies, such as monthly trading. As I mentioned earlier, we’ve seen it in Long, short equities or start up, right. Since that opsbase, you typically trade, you know, 2000 3000 stocks on the low side, and two 3000 stocks on the short side, right. So every day, you have up to 3000 data points. Right? Right. As long short, right. And if you were to back test those strategies across, many, many years, right, 10 years, there, you have enough data point for you to formulate, as well as validate any models that extract inefficiencies that are present in every single dose every single stocks, right. Whereas for us, we don’t have that luxury, right. So we only get to observe, if it’s end of day data, only one data point per commodity, right? So it’s very, the challenges in adopting some of this, off the shelf, our models and frameworks into our space due to lack of data. So the way we get around that is by augmenting fundamental knowledge, right? So we can really zoom in and say, okay, a lot of these feature sets, or potential data points are not really relevant for determining the price process for a given commodity, right. Because I know, my employees know, because we studied the market from the bottom up fundamentals point of view. So we are able to bridge the gap. And the gap that’s that’s missing, due to lack of data points. Yeah, that’s, that’s, you know, our philosophy. And that’s the way we have purchased by trading in our space.
Jeff Malec 1:03:17
Right. That’s a cool way to look at. Alright, I think we’re about to leave it there. You got any other last bits for us? Anything we missed?
Jae Min Hyun 1:03:26
No, I think we covered pretty much everything.
Jeff Malec 1:03:30
Everyone where they can find you? What’s the website?
Jae Min Hyun 1:03:34
website is www dot n, w o ne, hyphen llc.com. You can also reach us on LinkedIn. Like I said earlier, we are actively looking to hire content analysts, on traders on researchers. So please get in touch with us.
Jeff Malec 1:03:57
Nice. Yeah, we’ve got a lot of young student-level people listening to this. So give them a call. What is that market look like? Is it been tight and are competing with a lot of other firms for that talent?
Jae Min Hyun 1:04:10
Yeah, I mean, there is there is a little competition. The kind of people we look for are relatively different. To what what is positive for us? So yeah,
Jeff Malec 1:04:22
yeah. Or versus like going to work at Google or Amazon or something to write. Awesome. Well, thank you, Jay. Man, thanks for being here. And best of luck moving forward. We’ll come see you next time. We’re in New York.
Jae Min Hyun 1:04:27
Thank you, Jeff. Thank you.
Okay, that’s it for the podcast. Thanks to Jae-Min, thanks to RCM for sponsoring thanks to Jeff Burger for producing we’ll be back next week with don’t know yet but tune in to find out, peace.
This transcript was compiled automatically via Otter.AI and as such may include typos and errors the artificial intelligence did not pick up correctly.
