00:44:43 · Brett Caughran
[00:44:43] S1: I'll give you my hypothesis on this. Software coding happens so fast. There's no friction between better technology, better outcome. But ultimately, talk about capital flows. Capital flows are ultimately dictated by limited partner flows, foundations, endowments, writing the big checks. Those are incredibly slow moving organizations for the right reason. It's taken sort of the reallocation away from single manager to multi manager funds has been a fifteen year journey. And, you know, we have a prior in this of value investing where value investing basically got gutted by factor investing, right? Like if something's cheap now, it's cheap for a reason because the good stuff is sort of captured in a multi factor value quality model, right? And so good companies don't get cheap the same way they were because the sort of $2,000,000,000,000 factor investing industry stops that up. Like Buffett's investment style, as AQR is sort of written eloquently, is like a multi factor model now. Yeah. Residualizing that Buffett had no alpha, which is sort of a funny exercise. But it took factor investing, having this market footprint of multiple trillions of asset to really do that. And it will take, if we really want to systematize multi manager alpha, it will take actually chips, a big chip stack to start do that. Like having the concept of a working system is not enough. You need to have that deployed with a large chip stack. I don't think that's happening yet. I know that's not happening, right? Like maybe you have Bridgewater, a few of these sort of running capital behind it, but I'm talking more about the true fundamental investor in the field, research insights, shooting against guidance, meeting with management, channel checks. Like that's what I'm talking about. When I talk about systematizing fundamental investing, No one has done that yet because the technology is not there yet. Think it's probably, it may be there yet. Should discuss that. But the context strategy isn't there yet. That's why when you sort of talk about mosaic theory scale and systematizing alternative data research, I'm like, Oh, that's interesting. Like a that's a real unlock and actually pressing the go button on this.
00:47:02 · Alistair Smallwood
[00:47:02] S0: Yeah, I think it's a really interesting point. I mean, I'll pick up the first point around that factor piece. I think that don't have a quant background, so I'm sure that this can could be debated somewhat. But the factors are based on human biases almost always, where momentum is based on human bias, value is often based on human bias. And the really interesting thing going forward is that as humans participate less as a percentage in equity markets, to what extent are those biases going to be back testable and therefore investable? And the really interesting thing is that if every if model updates come every three months, then there will be no way to backtest existing factors, and therefore factors will evolve significantly over time. And what that then means for risk models and all of those things that sit within risk management teams, whether you're long only or market manager or single manager, will be really fascinating to see how that evolves. And then the second piece, to what extent do I think the technology is there? I fundamentally think that the technology capability is there and has been there for some time. The massive missing piece is I think we have a real skills gap problem within fundamental equity research, where in order to build these skills and sorry, these systems effectively, have to have deep domain expertise in equity research, and frankly, you have to be good at it. You also have to have deep AI expertise in order to marry those two things together and build a bit of a self fulfilling loop system. I think that our conversations with multi managers, single managers, long onlys, whether those be global or country specific, that is the biggest shortfall currently. You have internal technology teams that are very AI capable, and then you have internal fund management teams who are very market capable. But ultimately connecting those, where you're not spending the first twelve months building the plumbing that sits behind any system that you're going to build, is a massive problem. And I think it's a real barrier to the frontier of AI equity research. And I think there is a fundamental advantage to smaller funds that have managed to hire people that are capable at this.
00:49:18 · Alistair Smallwood
I think the way we fall short is I think there is this broader assumption that a very well crafted system can do the whole job end to end, which includes risk management, includes trading, includes genuinely acting on the decisions that it's got. I still think that we're a long way from that. And the reason being is because for as long as equity markets have human participants, I think that they will still require human judgment to sit on top to understand what's going on and provide that layer of, frankly, sensibility rather than having a system run end to end. I'm sure that there will be funds that come out that build these systems and invest autonomously, but I don't think that that is an existing fundamental team that switches itself to be fully autonomous. I think we're still quite a long way from the same level of care and diligence and frankly fiduciary duty to underlying investors in order for people to do that. Given the technology is moving at such a rate, and ultimately the way that people do the job and market structure will change, I think building a system that is repeatable and reliable is going to be an ongoing constant job. It won't be a single setup and here we go. We're now going to generate alpha into infinity.
00:50:34 · Brett Caughran
[00:50:34] S1: You're so right about the tyranny of the backtest, like walk into a quant fund today and pitch them a systematic fundamental investment strategy I have. And the first question and the most sort of operative question will be like, how do we back test this? And there's like just no way to do an out of sample back test based on the sort of parameter weights. And so it's sort of like a trust me, right, sort of thing. And like, one can have full conviction that this will sort of generate alpha and be a one or two sharp strategy. So I think the scaling will come incredibly slowly for that reason. I think the other thing, even in fundamental investing, I was at firms that had long track records. When you go through a drawdown, you're like, It's okay, guys, we've seen this before. Look at this history, long short spread. We always bounce back. Let's double check our models. And the sort of there's this faith and conviction. During the 2010s, many single managers in the Tiger Cub world built systematic businesses. Basically what happened is when things were going great, great. But then when they go off the rails, there's not that level of conviction because it's a bit of a black box in terms of judgment of portfolio construction. You then get the human element of pushback LPs, sort of you end up putting a bullet in the head of the problem strategy and it's like, hey, Remenco is fine. And there's just this sort of human friction around. And I don't think we get away from that in this sort of systematic, right? Many firms now are sort of building small sleeves of systematic capital that's starting to sort of get live in sample testing and hearing some intriguing things. But guess what? There is variance. Those strides will go off the rails. And if they're large enough with large enough negative variance, which is almost inevitably distributionally, and you go to your LP who you're trying to maintain capital with, like it's a very easy... It's very hard to sort of like, Hey, we have faith in this thing that caused a thousand basis points of alpha detraction this year. Please don't redeem your money, right? Much more natural impulse is kill that strategy and stick with what we've been good at for twenty years.
00:52:49 · Alistair Smallwood
[00:52:49] S0: Exactly. I suppose the sad nature of black swan events is that they're black swan for a reason, that we haven't seen them before, and therefore you're gonna hear exactly the same questions around your strategy and process every single time that they happen, whether it's a systematic black swan or maybe a black swan specifically within five stocks that you held in that portfolio at a given time. And the questioning and frankly, the accountability question is still the same. And the accountability of a fully outsourced system is still going to be quite difficult. And I wonder whether this is always something I've thought about in this the whole question around the viability of consultancy businesses going forward and AI is going to do a huge amount of that consultancy. But I've always had this view of every time I've had a public company say that we just had McKinsey come in and do X and Y study. There's no way that that management team couldn't have done it themselves. But ultimately, they wanted to put the accountability of firing half the workforce on somebody else, or a huge project plan. Ultimately, if you then go to the board with an external view, there is some accountability and who are we passing this blame off to if it doesn't work. With AI systems, there isn't really that. And we've actually experienced it somewhat in the way that we build internally. Is that if I ask a question to another engineer about why is this broken? If they then reply saying Codex thinks this, there is a very strange human reaction to that, to he's not taking accountability for a system that he built, albeit using coding agents. And although that's exactly how we all now deal with any question that gets asked of us, particularly in my particular work environment, but there's an accountability issue that exists there. You have to own the decision that was made, albeit it might not have been made by you. And therefore, this very careful, pragmatic thinking and testing of anything that you're doing with AI is still super, super important because the accountability still has to exist within you, the human.
00:54:39 · Brett Caughran
[00:54:39] S1: Yeah, and sensibility is such a great way to frame it, I mean, it's sort of taking out the hypothetical, again, not investment advice on Nike. Like if people scale these systems and one of their signals is dividend cut and that sort of leads to crowding in sort of Nike, companies are always trying to beat the shorts and sort of like do creative things. Like maybe Nike sells their Chinese business for a sum and that allows them to sort of stick, pay the dividend until the business turns around. Which always can be an outcome that drives a short squeeze. But maybe that's an epic short squeeze because that signal takes all these systematic fundamental guys out at the same time. Nike's up 40% in a day and all of these strategies just have a huge P and L hole in them. Then they have to degrowth. So all the other crowded systematic fundamental shorts squeeze at the same time. And then there's to your point of like quant crash or flash crash or sort of more episodic crowding. Those are the sort of things that are like, it seems simple on the surface, just press the button and I'll make money when Nike cuts a dividend. Markets aren't that simple. Markets are adaptive. The bar shifts, crowding matters, positioning matters. And so the human judgment to come in and be like, Hey, maybe they can do this. And having the system to a degree where like, I know this is consensus because this is what the sort of first order, cloud based outcome is. Let me think about something more creatively or let me not play in this scenario.