Chris Meredith is co-Chief Investment Officer and Director of Research at O’Shaughnessy Asset Management. In this episode, we focus on the latter title and talk all about what it means to develop a strong research program.
Our conversation centers around what Chris considers to be the three key pillars: data, tools, and people.
Chris provides insight into how data sets have changed since the beginning of his career, starting with highly structured price and fundamental data to so-called “pointy,” highly specific data sets and now completely unstructured blobs of information. He offers his thoughts into how this growing information set represents both an opportunity for researchers as well as a risk, requiring careful forethought into how it is going to be attacked.
Our discussion of tools covers both the digital and the physical. We talk about the influence of open-source software, the growing role of machine learning, and the operational benefits of treating each researcher’s laptop like a stand-alone research sandbox.
It is easy to tell that while Chris has a passion for the data and tools, he truly believes that they are for naught without the right people and he shares some of his ideas on how to maximize the potential of his team. Chris also sheds light on the OSAM research partners program, which grants 3rd party researchers access to the OSAM data platform. This new initiative is a highly unusual approach for a traditionally secretive industry, but early papers coming from their collaborations suggest it may bear significant fruit.
Please enjoy my conversation with Chris Meredith.
Corey Hoffstein 00:00
Okay, you’re ready, ready? All right 321 Let’s do it. Hello and welcome, everyone. I’m Corey Hoffstein. And this is flirting with models, the podcast that pulls back the curtain to discover the human factor behind the quantitative strategy.
Corey Hoffstein Is the co founder and chief investment officer of new found research due to industry regulations. He will not discuss any of new found research funds on this podcast. All opinions expressed by podcast participants are solely their own opinion and do not reflect the opinion of newfound research. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of newfound research may maintain positions and securities discussed in this podcast for more information is it think newfound.com.
Corey Hoffstein 00:54
Chris Meredith is CO Chief Investment Officer and director of research at O’Shaughnessy Asset Management. In this episode, we focus on the ladder title and talk all about what it means to develop a strong research program. Our conversation centers around what Chris considers to be the three key pillars, data, tools and people. Chris provides insight into how data has changed since the beginning of his career. Starting with highly structured price and fundamental data to so called plenty highly specific data sets, and now completely unstructured blobs of information. He offers his thoughts into how this growing information set represents both an opportunity for researchers as well as a potential risk requiring careful forethought into how it’s going to be attacked. Our discussion of tools covers both the digital and the physical. We talk about the influence of open source software, the growing role of machine learning, and the operational benefits of treating each researchers laptop like a standalone sandbox. It is easy to tell that while Chris has a passion for the data and tools, he truly believes that there for not without the right people, and he shares some of his ideas on how to maximize the potential of his team. Chris also sheds light on the Oh Sam Research Partners Program, which grants third party researchers access to the Oh Sam data platform. This new initiative is a highly unusual approach for a traditionally secretive industry. But early papers coming from their collaboration suggested may bear significant fruit. Please enjoy my conversation with Chris Meredith. Chris, thank you very much for joining me today. Thanks for Chris maybe we can start off for the listeners who maybe aren’t as familiar with you but know O’Shaughnessy most likely maybe can give a bit of background. How’d you get your start in the field? And what was your journey that led you to O’Shaughnessy today? Sure.
Chris Meredith 02:56
So Cory for the introduction. I’m Chris Barrett, the CO Chief Investment Officer here at O’Shaughnessy Asset Management. I have been with Jim since 2004. One way or at Bear Stearns together. The journey for me was one where undergrad I came out in 1995, I had an inkling about the internet and the technology change that was about to happen. And so I thought I’d get into that I was not smart enough, though, to know which side to get into. So I did it on the corporate route working for GE and Oracle. And while I was working there, I’m from upstate New York, mom’s a teacher, my dad’s an engineer, I had no idea of finance as a profession, liberal arts college, Colgate University, so no real outreach to Wall Street. But I was stocking money away and Oracle stock and the 9090 90 opened one envelope that said you had like $40,000. And then I opened one a year later and said I had $140,000. And I was like, Wow, that’s good. And so I paid off my school loans. I thought stock thing is good. I gotta keep an eye on this. And then the next envelope showed up and it was like, back down to like $35,000. And I said, it’s not good. I don’t know what to do with that. So I started trying to do research. And this is me just trying to understand and dig in. And the more I read, the more I realized that I couldn’t get any traction on. Nobody had any idea. Like there was no cohesion on any of the things that were out there. And so I was like, alright, I’ll try figuring this out for myself. And so what I did was with having a technology background, and just my mind works scientifically, I started downloading Yahoo scraping all the analyst estimates data, building my own models at home, my wife thought I was crazy. I’d like disappear into my office and just started pulling data down and keeping it including like news articles and the rest and I had a friend, my wife’s best friend’s husband worked in a hedge fund and I was telling him what I was doing. He was like, you know, that’s a career. It’s what I do. I work in a hedge fund. He’s like, if you really love it, you should probably go do that. And I thought about it. And so I started looking at business schools and Cornell had a program called the Cayuga Fund, which is one that is a great platform. It was set up by Charles Lee and Swami Baskin Swaminathan who have pivoted from academics to industry. So Charles Lee while I was there at Cornell moved and went became the head of Research at BGI. And Swami left and became the head of research Good LSV so pretty good for them. They set up a platform, they gave me access to all the academic data, Chris copies, data, etc. And I started building my own models using legit data and they gave me a platform to work off of and part of that was also then Cornell had ties to Bear Stearns, I was able to go and interview and I met Jim and it was kind of fortuitous. I had three people be like, you gotta meet Jim O. Shaughnessy, including one classmate who used to work for Jim. I met him I started working on the summer internship at the end of the summer, he was like, You’re great, you give me a job. And from there research analyst, system Portfolio Manager all the way up to today.
Corey Hoffstein 05:32
And as they say, the rest is history. So today’s conversation is going to be a little bit different than I think a lot of our traditional conversations where we get deep in the weeds of different factors and investment strategies, we’re actually going to focus more on your role at O’Shaughnessy as director of research. So you hold both the Director of Research and CO Chief Investment Officer titles. How do you think about distinguishing between the two? What does it mean to be director of research versus what does it mean to be Chief Investment Officer.
Chris Meredith 06:01
So for Chief Investment Officer, the way I think of it is I hold myself responsible for all investment related activities at the firm research is essentially, it drives the principles for how you’re going to invest. So you do all the research, you say, Okay, here’s the things that we’re gonna use to select stocks, here’s how we’re going to organize those themes into a portfolio, how concentrated is it. And then there’s a part about taking that called investment model. So the end of the day, what you wind up with is a model, saying this is the best way we think the best expression of how you can go about investing, we then hand it over to portfolio management that takes that model takes every client’s holdings and translates it into trades. So there’s a part about then executing on that. And that’s what portfolio management and trading come through. And then there’s a technology layer that supports that. So the way that I think about separating is that research is kind of the upfront work that sits around establishes the principles. Portfolio Management and trading are basically executing and they’re supported by technology. And for me, the way that I’ve we’ve organized the team is I have incredibly strong leads in the heads of portfolio management, trading and technology. So I get to focus on research, which is really kind of a passion of mine anyways, which is trying to figure out how the stock market works.
Corey Hoffstein 07:10
And I know you spend a lot of time not just in the weeds research itself, but also thinking about the research platform as a whole, enabling your team to do research in a more, not just thoughtful, but expeditious manner, making sure they have the tools at their fingertips. For you, what does a full stack research platform look like today? And how has that picture evolved, say, going back to the earliest days of you just scraping Yahoo data.
Chris Meredith 07:37
First of all, it’s better than scraping Yahoo data. I’ll start with that those were terrible models. But it was, the idea is I separated into three parts of a research platform, there’s obviously the data, that you’re going to aggregate as much information as possible, build a mosaic of publicly available information around a company. There’s the tools like the computers, the code stacks, of how you go about analyzing that data. So data plus analysis and putting together and then there’s the people, the team itself, who’s going about utilizing the tools and the data to reach conclusions and how it goes through. So the part of those three, and how they’ve evolved over time is everything is heading towards more sophistication. So for data, the breadth and depth of data available now is much, much higher than what we’ve seen in the past. So when you started back in 2004, even back at Cornell, you’re dealing essentially with just copies for the financial statements, you’re dealing with pricing for risk. And then you’re dealing with ibis, possibly for analyst estimates along the way. And those are data that’s going to go back to the 1960s on a quarterly basis for the financials in the 1920s. For the pricing, what we’ve seen is just that there’s just because of what’s happened with computing, the growth of Moore’s Law every 18 months doubling what happened with storage costs going down. It’s just this explosion of data. So it started coming out with originally some structured datasets, things like auction metrics, and you’re looking at derivatives data, you’re looking at trace starting up in 2000, where you’re getting the corporate bond data that’s coming through all these things like ownership data now, where your people are taking the 13 F filings and structuring those so you can know which mutual funds are owning which company on a certain quarterly basis, and building those, they’re building out professionals data to know who the executives are, what’s the executive compensation, what’s their tenure, what’s the board, what’s their makeup, and so then there’s all this column broadening of it, but there’s also like these pointy datasets that are coming out as well. So things like the ESG data, which is like you have Sustainalytics, or the MSCI data, which is coming out and telling you what is the value based or the risk based portion of certain of the UN PRI is a 17 structure they put out there for things that you should look at for sustainability and ESG awareness, you’re seeing other ones that come through where they’re talking about supply chain data, it was so interesting one was s&p comes through there’s a lot of vendor management on this for data. So there’s one where s&p comes through and they say they’ve got a new data set now that saying, they’re looking at every shipping manifest coming into the country, and they’re saying what’s going through in the past Out of what’s loaded into that cargo bin, where’s it headed? What’s the address it’s coming from and where it’s going to. So there’s these interesting pointing datasets, like I said, ones that have very specific application, and you have to weigh in judge what’s the cost benefit of those and how you’re going to integrate them over time. So, on top of that, on the data side, there’s also unstructured data now. So and this is where it gets really interesting in which is this part about how we are incorporating in more and more data that used to not be available for analysis. So the parts that I like to talk through are just things that were originally in books. So recent exercise we did was we scraped all the old Moody’s manuals, and we built financials for sales and earnings back to 1925, just to start seeing if there’s anything interesting that comes from a Great Depression and analyzing value. So that’s when were you physically you scrape the data, you put it into raw, just line by line, kick everything, using machine learning algorithms to classify which ones sales, which one’s net income, trying to identify those use algorithms to look for data outliers to see if somebody fat fingered a number in there. And then here, you’re going through manually piecing that data together at the end. So that’s where you’re building out a dataset that’s structured and unstructured. Another one, though, would be reading quite plainly natural language processing. So an interesting exercise we did was, we took all the 10 K’s that were out there. And we ran it through this piece of technology called Dr. Beck and Dr. Beck essentially vectorized his workspace, it uses the probability loadings of the words and how close they are next to each other, and then turns it into basically a principal component analysis for a document and saying, Here’s for all the 10 ks that are out there, and all the management discussion analysis sections, it turns it into a loading of numbers saying this one is you lose the interpretability, you don’t know what the numbers are applied to. But you can group them and say these are similar on this one, these are similar on another factor that’s out there, and then use that to analyze and what’s really interesting, by the way, I believe this is the technology that’s behind Google translation. So they did it on wiki, and they did it for English. And they did it for French. And it turned out the loadings were the same, so they could just map those along the way. But what’s interesting for me is that when I started this, you asked about change back in 2005, there was this barrier between what a fundamental analyst could do versus what a quantitative analyst could do. And a lot of that was just brute force information consumption, the idea of listening to the calls, reading the analyst reports, that’s breaking down. Now, that’s one where we’re able to incorporate that into our process. And it’s gonna be really interesting to see where this shakes out over the next call it five to 10 years on what you’re able to incorporate, because it’s not just being able to read that you’re reading them to any section that said, we’re able to read 3000 of them in a minute, which is what you can’t do on a fundamental side. So it’s that question of where’s the edge coming from on the fundamental versus quantitative processes.
Corey Hoffstein 12:44
Talk to me a little bit more about this pointy data set, sort of the depth aspect, you mentioned things like ESG, shipping manifests, it strikes me at least, and there are a large number of vendors. Now some of these sort of fall into the alternative space. But it strikes me that you have a balance there between this idea of raw data and somewhat pre processed data, as you look to acquire those datasets. What’s your preference, you have a preference as to whether you’re just getting everything raw on your team as the ability to build it up and structured as you want? Or do you want a vendor that’s cleaning and going through and giving it to you hopefully in a more valuable format.
Chris Meredith 13:22
So there’s value add propositions from vendors, like on that shipping, one that they have the manifest data is publicly available, what they feel their value add is, is they’ve done the mapping of the physical name and address to accompany. So there’s a benefit to that. And so this is where from my seat and when I talked about vendor management, there’s always this balance of, we’re a boutique, we have a budget, we’ve got to work with how you spend the money. And this idea of okay, they can bring this to us, but we can also scrape the manifest. So what’s your value add on top of this, and that’s where Oh, that mapping exercise, that’s fairly difficult. Same thing with something like a news analytics source where there’s providers of out there of just scraping the news or scraping raw news. It’s not a trivial exercise, but it’s not that hard. There’s some HTML processing and the rest. But that idea of mapping what an article is to a entity is something so there’s a company talking about General Electric and Honeywell, which is main person of that article. So you’ve got to do to access a building that metal data layer, and then there’s an analytics part of it of saying, Okay, what’s the sentiment of this, etc. So it’s the way I think look through this is there’s a data processing aspect, which by the way, is the least fun part of our job. It’s like I talked about the manually mapping of that SNP data going through and just checking in being like, okay, that number looks weird that that is the not fun. Don’t talk to your kids over the dinner table on this part of it. But there’s the analytics part of it, which you don’t want to outsource. So if they wrap everything into and say, by the way, here’s the sentiment of that news article. That’s where we say no, I would like to build my own sense. And then I would like to understand all the pieces for it and again to this platform and how it’s changed, which is more granularity on our side, more ability to build and do our own analytics on building these data points. So it used to be one where you would work within a FactSet framework, and you would just take their earnings yield number. And now we build our own earnings yield number from the pieces and just understanding all the parts from it. It’s the same thing for these other datasets for just have it where, again, you want more and more flexibility on how you’re putting these things together.
Corey Hoffstein 15:17
So you’ve mentioned some of these pointed datasets, you’ve mentioned a couple of tools, which I would consider to be more on the forefront of data exploration things more on the machine learning side of the equation. But you also mentioned the concept of budget, you don’t have an infinite amount of time, you don’t have an infinite amount of money. How do you think about the problem of trying to integrate these new ideas and these new frontiers of both data and quantitative techniques, when you don’t necessarily know which will bear fruit for you,
Chris Meredith 15:46
we have a huge graveyard of ideas. It’s one where we have tested a number of things in research, it’s a low percentage shot, you’re going in with ideas, it’s rare to find something that is truly great. There’s a lot of the research we do that’s incremental. So it’s about trying to just rework prompts you already have, when you’re going about trying a new idea, the most important thing we start with is just is there an idea behind it? Is there something that intuitively makes sense? Is it something that should have the ability to forecast earnings or to be able to predict what’s going on inside of a company. And that’s when we’re again, starting with the idea is the most important part. Some of this is where you would come up with ideas in academic literature and doing surveys and reviewing, just trying to figure out what other people are doing, that can give you some idea of whether there’s some meat on the bones for some of these ideas. My concern is academic papers, I’m not sure if they’re sharing as much as they used to. But you still do the surveys of literature to try to figure out what’s out there. But there’s also this part of a formula that we put in place in the research team, which is trying to understand what the value add potentially is the way that we’ve done it is an every research template that somebody comes up and and has a proposal, there’s a modified information ratio, which is essentially what’s your expected return out of this? Or is it a expected decrease in transaction costs? Scale? By the risk? Are you going to be increasing returns? Are you gonna be lowering transaction costs? Or are you going to be or overall lowering the risk of the portfolio versus the benchmark? And then multiply that by the assets of what you think that you’re going to impact for our existing client base? Or if it’s a new product? Where do you think this fits into our overall business? So there’s that modify call it view of just trying to quantify what the potential impact of one of these are. And then there’s obviously the cost aspect, which is if somebody’s coming to us and saying, I’ll license you that news data source, by the way, it’s a quarter of million dollars a year. And we expect that to be on the other side, where it’s going to potentially keep us from buying. And it’s going to have like 20 basis points of impact on our trading model. We’re like, I’m not sure. That’s the right trade to make. So there’s always this balance of trying to evaluate, get up front, I will say this, what’s really interesting for us as we have been exploring relationships with partners, it seems to be that some of the groups that are out there are marrying themselves to those pointy datasets. And so there’s a lot of people in the industry who have spent years and years exploring one idea one group that we have as a research partner, which we’ll talk more about research partners that have Shaughnessy, they’re a group in Europe called quanto P and they have built out they’ve scraped the European Patent and Trademark Office database. And they have been spending five years going through and exploring how patents have relationships with stock returns. So they have built up deep expertise in that one area. Whereas we are generalists and try to apply broadly to portfolios. They’re trying to build an innovation portfolio, but they had less access to a deep financial historical database. So they brought expertise to the table on one thing, we brought our products for teasing our research platform along the way. And then we started exploring the ideas together. And the idea was, can we jointly publish research out of this, what I like about this is that it saves us the part about obviously, they’ve got expertise in this, they build it out. But I had this on the board as an idea to explore by ourselves. In order to do that, we would have to do all the work of scraping the data, integrating it with our platform, etc. And they’ve covered that off. And now we’re finding a partnership and being able to explore the ideas together.
Corey Hoffstein 18:53
There’s a lot to unpack there. But I think I just want to jump right to this third party research endeavor that you guys have undertaken, which I think is really, really unique. So maybe first, for the listeners who haven’t heard of it before, you can maybe describe what you guys have done in opening up your research platform, and getting these outside researchers access and partnership and trying to tap into some of that thought leadership. But maybe you can also talk about some of the lessons that you guys have learned along the way in practice, both on the investment side if you’re open to it, as well as just what it takes to integrate with third parties and doing research.
Chris Meredith 19:30
First of all, it’s Patrick’s initiative he had when he was promoted to CEO at the beginning of last year. One of his concepts was how to expand the points of leverage within the firm. We’ve spent 10 years and $10 million. I think it’s the number that we quote internally on building our research platform. And then we have a team of people inside of the company who’s basically utilizing it, but that’s limited on our salary and what we can have. And so the idea is how do we expand the leverage points of that? And the research partners was one that came around because Patrick has been through his podcasts interacting with a lot of People in industry and people out of industry quite plainly, there are people who are not working professionally as an asset manager or doing the research but are curious, like minded people who want to their problem is they don’t have any budget. And they’re working off of just the public information that’s out there. So we thought the scalability of this is you approach somebody, the first person was Jesse Livermore, who’s an anonymous blogger. And what happened was, he had been publishing some incredible work out there. And he actually, apparently has received sight unseen, to an anonymous person job offers from various companies, very well respected companies that are out there. And he said, No, he basically has a job that he’s doing right now. And in a lifestyle. He’s like, I’m doing this for fun. And so everybody wanted him to work for their company. And he didn’t want to give that up. And so Patrick came and said, How does this work for you? And he said, I’d love to have it where I just have access to the data. I work on my own timeframe, I work on my own schedule, and I publish when I want to, but we work together on that. It’s a duck, like that’s it there. So we started this up, it was a first experiment with Jesse and and it was phenomenal right out of the gate, there was a part about getting somebody up on your platform. So that was just from old management project management days was we wrote up a document it said, here’s the data, here’s the layout of it, here’s the data structure. By the way, here’s a bunch of queries, that’ll get you started on building returns, building factor profiles, deciles, and get you started. And once we gave them that kind of tutorial, it was off to the races, then it came the integration with our culture. So this is the part also that I think it’s probably the most interesting part about the partners, which is that our team, a lot of that’s the stewardship of the culture, which is how you have your sharing ideas, how you’re exploring ideas, what was really interesting was integrating him into that. So he’s remote. So slack phone calls, great ways to share ideas and work through that we’re working off of a cloud research platform where we were bouncing code back and forth. The adaptation is he was working nights and weekends. So I had to start working nights and weekends. My wife had some questions around that one originally. But that worked out fine. In the long run, it was just about initially getting him up and running. It was great. I mean, people talk about thought, diversity. This truly brought thought diversity to our group. What you want to avoid in any research team is some sort of institutionalization where everybody’s thinking like, you need to have the part where you’re bringing in new ideas. This was truly an injection of why do you do it this way? I don’t understand, is this the best way to really do it, you’re like, well, that’s how we’ve always done it. And then you start questioning things. But the original work that we did, he and I started exploring this idea of what’s happening with value investing. And this is where the paper factors from scratch came around. I was trying to decompose what was happening with value. And I had worked through multiple academic frameworks, fall to 2000, to try to sit there and go through that agenda. 2009 paper, I was like, nothing was working when I couldn’t replicate the results even, which was frustrating for me. But then it was just trying to figure out and he came back with this really elegant paradigm. He was like, let’s just do it and decompose and say, it’s either earnings that are growing or the multiples that are changing inside of a company, what’s driving long term returns. And as soon as we pivoted to that, we were off to the races was part of I coded up a model that was he had originally done a sequel, I did C sharp, and it was more tying out to the shear positions and everything else, because I was able to do that in the weekend. And it’s like all things that took one weekend to build then took a month to clean. And then we got through it. And we started seeing the results from it, it was fascinating, which is really this idea of almost my opinion, a unifying principle of how factors work, which is, it’s all about this idea of earnings growth or rewriting. And so value is one where they’re trading at a 30% discount over the next three years, but their earnings are only gonna go down 15%. So there’s some Distress but the market is overreacting to it, and pricing. And then you watch as the earnings basically normalized, and then the multiples expand back out. And then momentum is one where you see that you’re paying a 10% premium over the next year, but you’re getting 15% excess returns, which is where you get that 5% of the premium every year. And so it’s it’s really interesting concept. And Patrick got in there. And we were all basically bouncing the ideas. And we realized there’s this whole part about rebalancing and understanding how to account for the mechanics of that. But when it was done, we felt like we just expanded our knowledge of how our factors work. And when that happened only because we were able to bring in a third party like a Jesse Livermore, who asked questions and had his own ideas for how to propose a framework. And it came around was great. That was like a huge win for us. And that’s where we felt like this is going to work. And so since then we have expanded the platform, adding up to seven partners now, all of which have various ways that they interact with us at various levels. So some would be like that quad IP group where we’re exploring patents. Another one that’s out there is Kevin said Google, he’s an ex Google employee who’s when I talked on the phone, he said My main goal is to stay semi retired. So he wants to just have it where his income is all investment income. So he’s like, I kind of want to make sure that I’m doing that the best I can. Other than that, he manages I think 10 fantasy dynasty, fantasy football leagues. And then just this on the side, but he was really interesting for us because he’s got a PhD in math and computer science from MIT. And so as we’ve been exploring, like you said, these advanced machine learning algorithms Here we are able to bring in truly an expert, I believe he said his claim to fame, hopefully I’m not getting this wrong is that he on the Google search bar where you type it in and it autocompletes for you. He said, I worked on that, basically. So somebody who can do these machine learning algorithms come into our firm, whereas, and teach us he came in for the day did like a four hour tutorial on boosted trees. So we have a good understanding of how those work we’ve been reading. And we’ve been unpacking those by ourselves, but to be able to bring in that skill as a partner for essentially, nominal cost. And we put the partners on a very small stipend to make their consultants along the way. But this part about Patrick had the right way, where what we had done was there are people out there with unconventional ways they want to interact with you, instead of a 40 hour work week job. And we met them instead of having them try to meet us. And so this idea of finding people who are incredibly talented, who have this curiosity that drive them, and us being adaptive in our call it work environment, in order to satisfy that allowed us to expand our team. And now we’re utilizing getting much higher productivity off that investment of a research platform that
Corey Hoffstein 26:06
we’ve made. So one of the things about having such a strong research platform is it leads to obviously a lot of different research ideas, a lot of research projects. But as anyone who knows who’s lived in the space long enough, the vast majority of those projects end up going nowhere, this idea of a research graveyard that you mentioned, which I love that concept. And I know I’ve got a very expansive research graveyard, but mine just ends up in sort of this labyrinth of folders since directory structure of failed projects. And I was listening to Patrick on TED site, his podcast the other day, and he mentioned something really interesting, which was when you guys have a research project, even if there is a failure, even if you don’t believe the research goes anywhere, you don’t actually just drop it, you institutionalize the knowledge in your process. And you make sure that the process is set up to continue tracking that research going forward to make sure that that time series, whatever is being evaluated continues so that someone can return to it and see how things have worked out of sample. Can you speak to that a little bit because I thought it was a really fascinating extra step that I don’t think a lot of firms go through, but I could see how incredibly valuable it would be to put in that extra legwork.
Chris Meredith 27:18
Yeah, it’s one where again, the investment is made on the I think through technologies back from my old days, and it which is, it’s an asset that you build, and then depreciate. So the idea is that you keep track of these things you put together. So like the ownership data, the first thing that we were trying to look at to see is that if owner operators have the ability to generate excess returns for their company, so somebody that owns 20% of the company, or more. And our research, it’s a sound idea. And it was one where we had a client who was asking us to take a look at this. So we were exploring this, and we turned the data over. And it was one of those frustrating ones where you felt like there was like, maybe 80 basis points that was coming out of alpha inside of this. And then you flip it another way and it was like, a little bit flatter. And the rest. I was like okay, at some point, you’re like this doesn’t feel like it’s going to generate something significant for us. Still some interesting things that came out of it. Things like private equity, board membership, and how that looks within companies that are still some ideas that we’re exploring along the way, but we put a pin in it. But that said, we still have it where we are tracking that owner operator metric. It’s in our graveyard. And it’s one where we’re saying let’s see if there’s anything that potentially comes out because there’s also say that’s through our initial analysis, when you’re going through and looking at a linear basis of saying, Does this one factor explain anything? Now, there’s also going to be in my opinion, when you’re going to sit there and incorporate more nonlinear methods and looking at factor interactions, there’s potential ways that that factor may not by itself, do something. But it may be a conditional inside of something else. So there’s a part about keeping track of it, letting you still examine those factors keeping as part of the graveyard, as I said, and even though you don’t put it inside, you still calculate it and have it every time you do that we’ll build on your factor database.
Corey Hoffstein 28:52
I want to return to the tooling side of things that sort of three aspects you laid out sort of the data, the tools, the people, well, maybe but the data and the tools as you continue to expand the data in the platform, whether that’s incorporating new datasets, or whether that’s datasets that you guys have internally created and curated and are now tracking, as well as incorporate new tools, new ideas, like machine learning, how do you think about managing sort of the actual operational burden of maintaining platform flexibility, this idea of you have a platform that worked four years ago, you start incorporating all these tools, and all of a sudden, it’s this unwieldy duct tape together machine. How do you think about keeping it well oiled so that the team can keep producing at high efficiency? First of all,
Chris Meredith 29:38
it’s a great point, which is that there’s a way that things get clunky over time. But we have it where I think this is at Oh, Sam four times that we rebuild from scratch the platform where you learn from what isn’t you haven’t where you incorporate all the best ideas and what you’ve learned from the data over time. And we went through that last one and back in like 2017, where we refactored all the code and basically put it back together. Part of this is that we have a called Gold standard research area that we do, and then we spin it out. So everybody has one on their laptop, we bought the heaviest laptops you could possibly imagine, for people on the research team that are like 12 core 64 Gig RAM laptops that apparently are 20 volt. We didn’t know this the first time, so you can’t plug him in on a plane. So that would Danny, one of our research analysts learned that on his way to Thailand or over to the Pacific, I forget where he’s going. But he had a 18 hour flight in his laptop died in 40 minutes, he was like, I can’t do anything on this anymore. So but we have it where everybody kind of has their and they’re literally called sandboxes. And the idea is you can mess up your sandbox, all you want. And what will happen is when you need it, you get a refresher on that gold standard and put it through. So part of having it where you have handy research team that is able to work quickly and rapidly. And this is a part of we believe in a entrepreneurial environment where you’ve got to work and move quickly. And this is when we’re having them with their own environment, their own tools. That’s one aspect of letting them run with those and build something out and proving it, there’s a part about then translating that into production. So that’s a lot of also that my job that I feel like is taking the work that comes from the research end of the desk, and helping that get over to the tech end of the desk where they’re running that on a production on a daily basis. So a lot of that is once they get it done. Okay, how do we integrate it with this, what’s the code base we’re going to use Version Control is tapping into all of our source control, that’s where best practices on it suddenly come in play. So there’s kind of this work in your sandbox, get the results, be as messy as you want. So we’re gonna wipe that out and start over again. And by the way, that sandbox ID is why the research partners was so clean, we will take those sandboxes and just put them out for people and say, Here’s a copy of this and go ahead and work with our derived data. You’re not working with the raw, you’re working with all the pieces of how we cleaned it and put it together. And so everybody has their own environment that they’re doing on that, then also you’re not working, stepping on each other, we learned that probably five years ago, when you have one big server and everybody’s working off of it, particularly the latest CPU intensive algorithms, you can wind up having it where you have to share time on those and people get tired of me waiting for me. We shifted our everybody now has their own environment over that’s kind of the long winded way of saying how we go about managing these environments to try to get as you say, keep it as clean as possible. But allowing people the flexibility to work and create when you’re pulling apart a new dataset or you’re trying something new, there’s gonna be all these little widgets that you’re building on the site, etc. So then we just take those and figure out how to translate it back into production code.
Corey Hoffstein 32:28
If we rewind the clock, call it 20 years ago, a lot of the software tooling that was available was highly proprietary, closed source and very expensive. And now we live in a world where there’s just really an abundance of free open source software of very high quality, how has the open source movement really affected and impacted the research team at O’Shaughnessy, and the way you think about managing the platform that you do
Chris Meredith 33:01
when it talks about the acceleration? I think open source is one of the key aspects of that acceleration. So like you were talking about, there are multiple of these items that we use out there that having 1000 people working on a codebase in the cloud, that are making it. So the tools, think about the barriers to entry to try some of these things has gone dramatically down. So when I showed up first working for Jim full time, and two, I did the internship gave me the offer night show for in 2005. And his very first thing he said was I have here all about neural networks build me one, tell me if these things have any practice. And I was like, okay, so I started reading all the literature, I went up to like the Columbia AI department like their computer science department. And I was asking them, I went into the code, and I started building when it took me three months to build a neural network backpropagation network that was like three layers deep, and I ran it, it was awful. It was like it basically, they went all in on value, and then it’s 1998. And then at the end of 1999, switched over to momentum and then fell apart like four times after that in 2003. I crushed again on the second half of the year growth rally and we so but that exercise of three months to build this algorithm. And to put it together to test it to find out that it didn’t work. A neural network now is maybe 10 lines of code, you take this open source package, you basically the algorithms all done and there’s this problem is sort of like the equivalent of me coding and with Yahoo, and scraping Yahoo, there’s this vetted by multiple PhDs on a computer science background building these things, and you have access to them. So that barrier to entry on these has gone down to the point where it’s, again, accelerate. Now you have to be careful with them. Because there’s a lot of dependencies that are stacking up from people working in open source. So there are certain things you need to know like how to freeze a Python environment, like to figure out all the versions for everything that’s coming through and how do you make sure that those are lined up so what you’re putting together like somebody could update a one of the underlying dependencies of psychics learn or something like that and that could affect downstream what happens so you gotta understand production eyes and code, the willingness to work with those, can it have this tremendous impact where you’re essentially leveraging the best in the industry, specifically, and bringing them into your organization through this open source? Now you have to evaluate those tools. This is the harder part, I think now, which is, it used to be the harder part was building the tools. Now the harder part is so many of these, which ones are you using? What’s the right one to do? And then figuring out which ones you’re gonna spend your time figuring out to adopt the skill and just learn how they work? And just basically, where are you spending people’s time inside of these because of all the infinite number of tools that are out there?
Corey Hoffstein 35:34
Yeah, I think too. You mentioned Python there. I think there’s like at least three deep neural network GPU based platforms out there that are all very high quality. You’ve got pytorch Keras. And I’m blanking on the last one, but to your point, TensorFlow, TensorFlow, yes, thank you. And to your point, they’re all incredibly well supported, some of them have are coming out of places like Google and Facebook, choice of one over another can have an impact. But ultimately, to your point, you are getting the benefit of this unbelievable access to 1000s of high quality developers touching these types of things, which hopefully would accelerate the research process and make things a little easier. But your point, it does lower the barriers to entry. And it seems like maybe that will hurt analytical edges going forward. How do you think about that, from a competitive perspective?
Chris Meredith 36:25
Well, what will be interesting is like said, lowering the barrier entry. And there’s this part about people thinking the latest one’s got to be the best one. And so there’s ones like the deep neural networks, and obviously what happened with the AlphaGo and the Alpha zero, wherever they got very excited about it. And so there’s part about taking those tools and unpacking them. And some people are highly convinced that must be better than the algorithm before. But there’s a part of understanding that is basically, if I know this, right, and again, there are people much smarter than me, I’m just trying to use these tools along the way. But the way I understand it is that it’s based on like, essentially, your visual network and the seven layer part of how you’re, it’s mainly for image processing along the way. So there’s a part about the neural networks with deep learning where it’s like, Okay, does that think about how a visual cortex worked with that work, I’m trying to figure out the interrelationship between value momentum and what’s going on inside of a company’s earnings, projection, it may not versus like boosted trees and cuts and things like that. So there’s this aspect of trying to evaluate the tools and figure out how they are. But I think what’s going to happen is lowering the barrier to entry is going to give more people access, and allow them to generate some of those false positives that you need to be aware of inside of any research process. The less robust methodology is where somebody comes back, they get a little bit of result, they think they claim victory. And so this is one where, again, having a sound research process, trying to be able to understand, does it work in all market environments? Those are how many are there times where it falls apart? Is it working in large cap and small cap the same way? Is it working domestically and internationally, these parts about making sure that it’s really going and having a strong process and a skeptical mind that you’re applying to these things, I think is a key aspect of doing research. So I think that the proliferation of tools, what’s going to happen with the access to them, it’s going to lead to more and more people having the ability to at least generate a test. But that doesn’t mean that it’s gonna come up with a good result.
Corey Hoffstein 38:11
We’ve danced around machine learning a little bit, I’d love to just go directly at it and ask your thoughts. I know this has been an area you guys have been exploring? How much are you encouraging your team nowadays to focus on machine learning driven research projects? If you’re willing to share? Have you noticed any areas of machine learning that you’re finding to be fruitful in the research other areas, maybe that you guys have moved away from just really overall? How are you thinking about incorporating it?
Chris Meredith 38:37
So it’s really interesting, I think you and I were on a panel, maybe a year and a half ago, and somebody asked the question about machine learning. And I was like, Yeah, we tried it. We looked at it, and it’s terrible. I think that was what I said on that panel, which goes to show that I should always hold myself caveat of I don’t know everything at any given point. I don’t know what really what I do know is that everything I had done up to that point had not worked. And our philosophy in house had been, we had use machine learning, we’ve unpacked what the state of the art was at the time back in like 20 1314. So like support vector machines, I think we’re back then and naive Bayes and some decision trees and we said, Okay, let’s try to predict returns. And that failed miserably. So we’re talking about the graveyard now where it’s like, Okay, let’s try to do that. Okay. Can we time factors at the least? And that was terrible. That was worse. Can we at least Wait, which one of sales price or earnings price within a value composite? Which one would be better can at least figure that out? And it was worse than equally weighting as soon as they got out of sample. And so we were like, everything we did with it was right for not being that good. And we basically walked away saying, Yeah, we don’t think this is going to add much value or what we have right now. Now, the call innovation of what we’re thinking about, like where we’re spending our time has been more about I think there’s a reason those failed. I think there’s a very good reason those failed, and that’s where you are applying a non linear algebra. rhythm, which is prone to overfitting, and are prone to figure out whatever the most, it’ll map exactly to whatever data you give it. But then you got to sit there and say, is that a relationship that has stability outside of that to the out of sample data, and we did it on returns. And what we realized is that returns are driven, not by one thing, but by hundreds of things. And so in each one of those hundreds of things are going to have their own relationship with between the data and the events. So there’s a part where what we have said, as our next leg of research is, when we proved it out, we took and said, Okay, predict returns, and show us how that works. And literally every algorithm is almost if you look at an AUC metric, which is an area under the curve, and basically 0.5 is guessing, I think the best example I heard was like rating in class, you get a 60, you get a D, and a 70, you get a, a C, etcetera. And like that we’re at like a 53 on predicting returns, it was awful, it was no good at all. And then but we started taking saying, What about predicting things like Ford earnings? All sudden, our AUC was like, point seven, five and Olson we, I was like, wow, we’re, we’re a C student, all of a sudden, we can predict something, which is good, that was a place to start. So that idea of being able to predict something we have taken instead, okay, let our research going forward, it’s going to be less returns focused and more on outcome focused, what’s going on with the stock? What are the events that could potentially have it? What events out there, first of all, one had associated returns with them? And what’s your ability to predict those events in isolation, and we are, let’s say making progress, and that’s probably the best way to put it.
Corey Hoffstein 41:29
So let’s talk about the future of research for you and the team. Oh, Sam, at this point is a fairly mature firm and at least outwardly, it strikes me that a lot of the principles, and the ideas at least seem to remain very consistent with a lot of the ideas that Jim put forth in his original work, and obviously, some of those who have evolved. But as you look into the future, what do you think research endeavors will look like? What are areas that you’re excited to continue exploring
Chris Meredith 41:56
that what I just talked about is one that’s obviously very exciting for us. And that the idea of improving within factors and within things. So Jim, we are very committed to the principles Jim laid out in the book of value and momentum and yield and quality that we have where it’s idea of being able to select stocks and all those. So what’s interesting is the research since the beginning of that initial piece with Jesse Livermore, of talking about factors from scratch, there’s a kind of a an arc to it, that is pretty straightforward. Where we did that factors from scratch showing that essentially, it’s all about earnings and Ford earnings, alpha within factors. The next piece we did showed how within a theme like value, essentially, what we’re doing with our other characteristics is we’re showing that the range of outcomes within value if you can have any ability to predict forward earnings. So that papers pretty pretty report when you look at something like quality themes and momentum. Screening on the negative side of those within value is simply about avoiding the value traps, the ones that are sitting there, and they they’re cheap, but they’re cheaper reason. And that’s because on a 412 month basis, their earnings going to be way below where they are right now. And so they’re actually pricing, maybe something like our case of a Blackberry, they’re not even priced appropriately, it’s just going to continue to go down. So it looks cheap. It’s got PFI, but there’s some reason for that there’s some distress on that, right. But this idea of having it where we’re able to go through and all of our themes. And so this is the logical procession of what we’ve got, which is, we’d like value, we like the idea of being paying as little as possible to get into an earnings stream of a company. No problem, we just got to make sure then that the earning stream is stable as much as we can or even growing. That’s where we use our other themes to sit there and verify what’s going inside of it. We’re doing that through linear expressions, momentum, quality, earnings, growth, et cetera. And then we’re saying, Okay, what happens if you take a nonlinear approach to that simply weighed them differently. So that’s exciting for us. And it’s a whole new toolkit, a whole different way of going through and saying, Can you predict earnings? So that’s a big area of research for us. Other areas are things like this specialty areas, like ESG impact investing, and we’re going through and we’re seeing, are there ways that we can take our platform and provide solutions to clients that are things that are maybe not returned to focus, but there’s another values based methodology for what they want to do for what they want to do with their capital? And this is one of saying, can we offer that on some sort of sustainability and some sort of values based investing? We are looking for new areas, we’re looking for things outside of value, momentum and yield, we got a couple of ideas in the fire for that. And that’s one where are there other ways that stocks can generate returns that you can do in a predictable fashion? So that’s kind of a, what’s called the furthest afield from what we’ve got right now. And that would be separate products and new things that are out there. I would say more about our platform. And this is that this idea of customization, as an asset manager, and one of the things that’s interesting, it’s become more apparent over time for us is that we have done all this work on all this research, and then we bundled it up into 12 to 15 products that were what we thought, our best expression for it. And they fit within the style boxes of allocators. And we said here’s how you access us, which is simply you take a market leaders value portfolio and you invest in that. One of the things that’s interesting about technologies it allows for a different way where customers can come to you To say, is there a better way to get this as I have a different way that I want to access you? Is that something that I can do? And that’s where we’re exploring that and trying to figure out there’s ways we can offer more personalized solutions. And that’s when we’re really this future, the firm. It’s interesting for the industry, I think what we’re seeing more and more is that there are allocators and people out there that want to engage with us. On a broader level, they want to have ecosystem engagement, right, where they talk to us about not just the portfolio we’re providing, but they want to access our thought leadership, they want to access our performance attribution, how we think about presenting returns, they want to access more with us, instead of a more of a transaction of just having more they invest in one product. And I think that’s the future of what happens with our firm is that more and more, we’re going to be engaging in a broader level with clients.
Corey Hoffstein 45:47
That idea of customization is really fascinating to me, not only is it a research problem on the investment side of the equation, it’s a business scalability problem, because it’s not just can you take this investment strategy, and slightly customize it, and what does that do to the strategy? Does it hurt the efficacy of the signals in any sort of way? But then it’s when you have 1000 variations of the same core concept? How do you end up communicating to your clients, their particular performance attribution, waterfall, graphs, commentaries, that sort of thing? That is a big business scale problem, as you send people into the field to go do their quarterly meetings with clients? How do you guys think about that?
Chris Meredith 46:29
So what’s interesting is we’ve talked about the acceleration of O’Shaughnessy on the research side, but we haven’t talked about as the acceleration on call it the business side and the operation side where one of the things that happened was, we’ve always been a separately managed account platform. We’ve got 3000 accounts. And we went through a process starting back in 2009, on automating how we go about delivering, so when you’re talking about the scalability, there’s an operational component to it. And we had a project internally called Project alpha, I didn’t name it, I actually didn’t like the name of it. But it was called Project alpha, about how to get an operational risk out of our system and have it so it’s scalable. And part of that was when we first spun out. And we formed O’Shaughnessy from Bear Stearns, we wanted a best in class platform. And part of that from our I talked about that chain of research to portfolio management to trading. What we realized was, the sooner we did things to our traders, the better off our market impact was every time we started implementing a transaction cost analysis system, and we started looking at the time of day when things got to them into Aaron, our head trader, and she was basically saying, I would like everything by 830. So we said, how do we get everything by 830? Now we’re getting everything to her by like 755. So this idea of operational excellence where you continue to invest, part of that as as a firm, the technology investment we’ve made have made a strategic decision to build not buy. And so everything that we’ve done on the operational flexibility is one that has been built up. So this is one where my ability to communicate with technology and our team’s ability to communicate with technology, and sort of say, here’s how we think it should work. And can you do that, and then our technology team’s ability to execute on that has been a huge asset for the firm on the scalability. The second thing you talked about and scalability for performance attribution, our firm has invested heavily on how we communicate to clients about attribution. It’s really interesting, the evolution of that as well. When we first started, the first one I can remember was JP Morgan acquire Bear Stearns and their due diligence people came by and they said, let’s look at your performance attribution. This was in 2008. And we showed them Brinson attribution, they’re like, Okay, but how did you get your sector weightings? We said, well, the factors to that. And they said, well, where’s your report on that? We said, We just talked to them. And they said, no, no, we need we need a factor attribution framework. So we went from Brinson to factor attribution frameworks, we tried originally, all the off the shelf ones that were out there. And they all had some misspecification between their risk model and our factor loadings, which created confusion for clients. So most of the time we were on with clients, we were spending our time talking about their risk model instead of our investment process. So we just said, Let’s build our own. And since then, we have built three different very unique platforms for adding where we express factors, and how we talked about that part of portfolio. So we have built a scalable methodology for that as well, talking with clients and just the impact that they’re that you can layer in and very cleanly. See, here’s the choices you make in your portfolio. Here’s the alignment of those choices with the attribution process. So the scalability, we have an operations, the scalability we have inside of our performance leaves us with a platform that allows us to do a lot more. And that’s where we feel like we have the ability to generate more solutions for clients.
Corey Hoffstein 49:36
So we’ve skipped over it a bit. But I want to go back to the third pillar you mentioned which is the people then you mentioned the importance of the people and this was a couple questions ago but you very briefly mentioned this idea of this modified information ratio that you use for research proposals, this idea of looking at a research proposal and talking about whether how much AUM it affects The expected return transaction costs lowering risk. How do you think about running an efficient research team, both in making sure that they’re focused on the right projects, but also making sure that they feel inspired to continue doing the work and have the freedom to be creative?
Chris Meredith 50:19
The idea here is, we have not just our internal research team, we have the research partners, we even have interns who are a great asset for the firm and how they come through, and how do you sit there and make sure that those are running as efficiently as possible. A couple of thoughts on that. The first one is that if I’ve learned anything, I have learned that people work three times as hard on their own ideas versus ideas that are put on them. So the very first part is letting people explore their own ideas, you want to make sure that they fit into the goals of the firm, though, and this is where it gets really interesting, which is aligning the interests and incentives of the individual researcher with the incentives of the firm. And that’s where that modified information ratio comes about, which is what are we trying to do, which is we’re trying to make our process better for as many clients as possible. So in the context of things understanding, okay, there’s gonna be some incremental improvement on this one project, is it gonna be a quick hit project is it gonna be one that’s going to was going to suck up the whole team for three months, knowing that value cost benefit of what you were doing on in weighing that beforehand, is upfront, that helps you make it so it’s an efficient team, but you need to have it where people are able to go through and have it where they feel motivated for having it where they’re exploring their own ideas. There’s a project management aspect of it, what resources you need. This was the background on project management from it, which is the triangle of any project, which is, what are your goals of the projects, what are the resources you’re putting against it versus what’s the time you’re going to take to get it done, you basically can set two of those things. And then the third one is kind of like the outcome profit, right? So you have your goals that you want to do, and you have the resource you can put against it, and you need to be flexible on time. If your time is a fixed and your people are fixed, and you have to be flexible on your goals. So all of those are things of just knowing how to balance those and work work them out. And I would say the last aspect of running an efficient team is around the culture. More and more, I’ve had the opportunity to speak with people at other shops, there are some cultures out there where people feel like they need to protect their ideas. They need to have it where their ideas that they’ve generated in the past, or the reason that they’re getting compensated and, and staying at the firm. And so there’s a part about like this natural hesitation to share and have an openness to it. And our firm, we’ve gotten around that it’s truly one where people, the idea is you’re compensated not for the ideas you have generated, you’re compensated for whatever ideas you’re going to have next. And part of that is we know that some of them don’t work. So you could have a year where you are doing great research, and there’s nothing that comes out of it, that impacts our process. But that’s okay, that’s a normal part of the process. It’s like, I’m a baseball aficionado, I love baseball. And then you can have a Derek Jeter go like three 457 or something he’s still Derek Jeter, he’s going to be fine. So that’s where you can have these times a period where you’re gonna just have it, you’re in a rut, but you’re gonna get through it. So this idea of having it where you have people that are just highly motivated, because they like the ideas, I think the ideas are critical for any research team. The part about having a project management incentive structure around it, making sure the incentives are aligned. And then having a good strong culture where everybody’s sharing ideas and helping each other out, we have it all the time where I talked about that project management where one person is trying to hit a time goal. And so somebody else can pull off a project and help them along the way just to make sure they can get done. It’s a great culture, it’s just one that I continue to try to help shepherd and make better. So you
Corey Hoffstein 53:23
build this strong foundation on these three pillars, Data Tools, people, in theory, that sort of strong research platform should hopefully lead you to new evidence and new ideas that might at some point imply that you should change your process in some way. Thinking to the idea that O’Shaughnessy is, again, a mature firm with a pretty sizable asset base, how do you think about introducing and communicating these ideas of improvements to your clients who maybe had bought in to something you had done in the past
Chris Meredith 53:58
the way but his clients don’t like change, but they want you to evolve? So the part is they have bought you? Or is it gonna, quote unquote, bought you for what they believe is what you hold out to do in your investment process. And we have a high degree of sensitivity to that every time we do a research project, one of the very first things is, what’s the tracking error to the previous version? Are we ripping this off? does it behave completely differently? Is it a you want to see as perfect correlation moving up 20 basis point, and it’s the exact same strategy is going to have the same areas where it struggles in the same areas where it outperforms and shines, but you’re just doing a better job of it. So this idea of having it were having this strong fidelity of strategy is probably the right way to put it and understand that people that hired you, for that mandate, want to continue to see that. So this is part of the mindset of where you’re going to continue to move and where the research goes. I mean, setting it in the context of what I just said in machine learning. There’s a very logical reason for why we’re going down that path, which is we buy stocks based on value, momentum or yield within that We select on certain factors that severity and look for the winners within value, momentum and yield. And that’s those other quality themes, valuation, things that are saying, if you remove these, you’re gonna see you’re gonna work away from companies are gonna see a decline in earnings, we want to explore beyond linear methods of those themes. And just to see if we can do a better job of predicting future earnings. That’s a logical evolution of how you go over time. If I was to go to a client say, yeah, everything we did, we just throw it out. Now we just give you this black box machine learning algorithm, they would throw their hands up and say, I can’t do that I’m out. So there’s a part about understanding the logic of the procession of what you’re doing, how it’s within the context of all the philosophy that you’ve had over time. And then being able to communicate that? Well. It’s one of saying, this is a service industry. At the end of the day, we’ve got to be able to provide the service to clients and say, here’s how we’re going about investing for you, and show why it’s a natural progression from what we’ve done in the past.
Corey Hoffstein 55:52
What can court researchers learn from discretionary researchers
Chris Meredith 55:56
a ton. There’s another great thing Patrick did, which he established some learning lunches where, again, he’s met some of the best people in the industry, he brings them in for us to talk to this is great, because there’s people that have expertise. Like he brought in one person, very well known analysts who are on financial services and understanding banks. And immediately we got to the point of, yeah, by the way, you want to understand there’s banks with like mortgage loans and people with car loans, and those are very different banks. And so understanding like what’s on the debt on the actual balance sheet for the loan book, can wind up with a very different part of that. That was really interesting. We brought in an m&a expert, we brought in another series of quants from a different large multi strat hedge fund platform. All these are that’s I guess, learning from quants are not quite the same thing. But the idea is all of these people that we bring in what they do is they give business context for whereas in quants, you can not necessarily have that because you haven’t been in the space for so long. So there’s definitely a lot that we can learn. And we continue to learn,
Corey Hoffstein 56:53
speaking of continuing to learn, as you open up your platform to more parties, folks like Jesse Livermore, how do you think about maintaining research integrity, when you might just keep looking at the same data over and over again? How do you avoid sort of the curve fitting overfitting potential of just hammering the same data to try to find something new?
Chris Meredith 57:16
Yeah, that’s the torture the data long enough, it’ll tell you exactly what you want, you got to be careful that I think part of this is turning over the same idea. There’s robustness checks, though, that are a big part of this. Like I said before, does it work in all cap ranges? Does it work in different geographies has worked in different time ranges? Is there anything one specific item of the data that’s making it, where it’s getting to live certain, there’s a whole bunch of things that you need to do and the particular also just changing a few of the parameters inside of it and seeing how much it moves. That was the one like that ownership example that I gave was, there was one that was just a little bit, you know, they look pretty good when you did it. But you could change the parameters around it all just a tiny bit of a collapsed. But it was one I’m very mindful of there was a great chart, I saw that it was on the x axis, the number of back tests and on the y axis, the average T stat that comes out of those back tests. And it’s what you’d expect, right? It was upward to the right line with like bands. And by the time we got to 200, back tests, on average, are no matter what your 3% is, right? I don’t know where they source that chart from. But I was like, it was just a visual part of exactly that problem, which is, the more you turn this over, the more likely you are to see something. And it’s part of also why more and more, we set a higher bar for what comes out of research, you move the bar up a bit, you apply haircuts to what’s coming out of it, you put all the robustness tests you can against it. And then most importantly, there’s sound logic for why things work at the very, very start of it, and making sure that it lines up your original premise for what you thought would happen.
Corey Hoffstein 58:40
So you mentioned you guys did a full platform rebuild in 2017. So this question, maybe you’re closer than you were in the past. But let’s say you could just snap your fingers and rebuild the platform from scratch, no cost, you could make it your dream platform instantaneously. What would you change and why
Chris Meredith 58:59
part here would be obviously all the data in the world, put it all all in there and perfectly normalized. So everything relates with one another. And there’s, and it’s clean. And there’s no issue, I saw a great chart, one of the feature engineering books on machine learning said that 70% of the work is just putting the data together. They also did a chart on what’s the least favorite part of the job, and it’s putting the data together. So it was like that was 70% of the displeasure for people in the job, right. So you want it all there, you want the infinite computing resources where you don’t have to wait for anything and just instantly stuff gets done and access to all the tools that are available out there. And the truth is, like, here’s the thing is we’ve got a lot of data, we’re not short on data right now. And we’ve got really good computers and tools through open source and through the rest of it. I’m not worried about those, like those are gonna continue to get better and continue to evolve as they have in the past. The team is the most important part of this. The team is really the critical aspect for how you generate the insights. The platform for me is there’s a part where having it where we are able to have really bright people who have the ability to not only come up with good ideas, but then have have the technology skills to go and explore those ideas, and then have it where they are also, again, being invented, where they’ve got innovation, and they’re coming through and coming up with new ideas to work through those, and then have it where they’re able to fit and work alongside one another within our culture, which is one of, again, more of a team based environment, having it those are the key aspects, I will say, adding more people to that, that we can find that have the like mindset of all of those aspects, opening up to more research partners that have that same ability to work with us having this pipeline again, of intern talent, where I will say the people coming out of universities right now, are much, much more equipped to do this. And we were went quick, when at least I’ll talk to myself when I was when I came out. It’s a great thing. This is part of the reason I go back to Cornell and I teach. And I’ve also set up a research platform up there to try to work with students and understand
Corey Hoffstein 1:00:53
I actually want to talk about that. Will you expand on that a bit, because I think that’s a great initiative. So
Chris Meredith 1:00:57
I talked about the unifund. I went through that program and how it allowed me to pivot from technology to asset management. I go back and I teach that now. So the Cornell came back to me and one of the professors had done it for eight years, he was looking to expand his portfolio within academics, etc. And he said, Can you come back and help teach this. So I go back, and it’s a quant basically, model that they do fundamental analysis on top of, so there’s a quant research platform out there that we’ve built, which is really interesting. So Cornell, through their academics that are PhD research has access to a lot of the data we have here. In fact, there’s some data that we don’t have here they’ve got access to. And that’s ones where there’s we have, obviously data they don’t have. So there’s this little bit of expansion on that platform for the data. And the tools are all very straightforward. If you know how to put it all together, it’s mostly open source, etc. And then it comes to the team or it’s able to find students and undergraduate group, one of the students there called spark stone that he put together, I became an advisor on and they have like, this cadre of like eight to 10 undergrads who do nothing but work through that platform, and they have their own version of that platform they’ve taken off on their own, but they basically do research. And I tap into that and try to help them out every once awhile. But that’s also led to us having hiring from them and interns that come from them. So it’s as part about pipeline of talent, understanding what’s coming through in academics, it’s also been a huge part for us on celebrating our ability to adopt the machine learning tools and bring them into our research toolkit. It’s just having access, where they’re getting classes and all the parts with it. So the way I look at that is it’s great. It also, by the way helps me because I you know, I learned a ton to where it’s, I don’t want to be institutionalized. No Sam either. So it’s about having where I get to learn, if you want to learn something, there’s nothing better than going standing from a classroom and teaching. And I will tell you that you have to know it. And so it’s been just a great experience. And it’s one where it continues to add to what I know and my ability to explore the stock market
Corey Hoffstein 1:02:39
as a Cornell alum, first of all, thank you. Second of all, I really wish I had known about this when I was at Cornell, because I would have loved it.
Chris Meredith 1:02:45
It’s one way that continues to expand. And this was the thing when I went through Cornell, I basically, there was no quant program, I had to piece it together, I had to talk my way into the PhD room where they had all the word the word research, data service and the rest of it. But that’s when we’re I’m trying to make that more accessible, like this access Park habit where you can have the ability to explore the data and bring the tools and brains to as many people as possible because there’s some really smart people that corner.
Corey Hoffstein 1:03:09
Alright, Chris, last question of the podcast and same question. I’m asking everyone this season, total departure from our conversation, but I think you’ll like the question anyway. And the question is this. Imagine a scenario where you have to sell all of your investable assets. So we’re going to liquidate everything. And you can only invest in one thing for the rest of your life. Now that one thing could be a portfolio, it could be an investment strategy. It could be an asset class or an individual security, whatever you want. But once you buy it, you can never sell it. What would it be? And why? What’s the horizon? I
Chris Meredith 1:03:44
got here? So this is for the rest of my life for the rest of your life. Oh, boy. Yeah, we go equity markets. And I would do factor investing. I know it’s the boring answer, because it’s what I do. It’s my own knitting. I really believe in this. It’s one where and let me say this, I would bet on ohsms portfolios, my own cooking. I have most of my investments in this anyways, everything except I can’t kids have 529 plans that I’m not impressed. But I believe first of all, one, I think equity markets in the long term do return higher. So you went up and you have to have the ability to stay with the strategy. You have that where you can go through and whether we’re going to be pullbacks and periods of distress inside of this. But I believe firmly that you can select stocks within equity markets and I think that we have the ability to generalize. So it’s kind of a boring answer. It’s my own cooking, but that’s what I would say,
Corey Hoffstein 1:04:26
Chris, this has been a blast. Thank you. Thank you for