WEBVTT - The Math That Explains How Multi-Strategy Hedge Funds Make Money 0:00:03.120 --> 0:00:18.480 Bloomberg Audio Studios, Podcasts, Radio News. 0:00:20.079 --> 0:00:23.360 Hello and welcome to another episode of the All Thoughts Podcast. 0:00:23.480 --> 0:00:24.880 I'm Tracy Alloway. 0:00:24.560 --> 0:00:25.680 And I'm Joe Wysenthal. 0:00:26.000 --> 0:00:28.320 Joe, we're back on the multi strat beat. 0:00:28.920 --> 0:00:31.200 I love this beat. I think it's really interesting. There's 0:00:31.200 --> 0:00:34.240 a lot we've learned, but there's a lot we haven't learned. 0:00:34.720 --> 0:00:36.800 I love this beat. If you said we're going to 0:00:36.920 --> 0:00:38.920 just do ten episodes about this, I'd be like, yeah, 0:00:38.920 --> 0:00:39.479 that's fine. 0:00:39.800 --> 0:00:42.680 Well, yeah, I look forward to part six hundred and 0:00:42.720 --> 0:00:46.080 seventy eight in our ongoing attempt to understand multi strat 0:00:46.120 --> 0:00:48.839 hedge funds. But you know, we've been sort of learning 0:00:49.080 --> 0:00:51.720 as we go along, and there are a bunch of 0:00:51.800 --> 0:00:55.200 questions that I still have. One of them is there 0:00:55.240 --> 0:00:59.760 seem to be a lot of different opinions and variation 0:01:00.640 --> 0:01:03.960 pod shops right on how exactly they can be designed. 0:01:04.920 --> 0:01:08.160 Right, So there's different sort of structures that I understand. 0:01:08.319 --> 0:01:12.240 There's different compensation structures. There's different degrees to which the 0:01:12.400 --> 0:01:16.760 different pods so to speak, coordinate with each other. There's 0:01:16.800 --> 0:01:20.320 different degrees to which they like centralize ideas and research. 0:01:20.600 --> 0:01:23.959 So like I get that, there's still some big questions 0:01:24.000 --> 0:01:25.560 in my mind, and I'll just say one of the 0:01:25.560 --> 0:01:30.039 big ones right off the bat, which is, if you 0:01:30.120 --> 0:01:33.040 have a bunch of teams doing a bunch of different 0:01:33.480 --> 0:01:36.880 strategies and trading a bunch of things, why are the 0:01:36.920 --> 0:01:40.120 returns good instead of average? Because in my intuition, if 0:01:40.120 --> 0:01:43.000 you have a bunch of teams like, okay, you're diversifying 0:01:43.560 --> 0:01:45.479 alpha across a bunch of things, but great, But then 0:01:45.480 --> 0:01:48.040 you have a bunch, my gut intuition will be like, 0:01:48.120 --> 0:01:50.960 you don't get great returns, you get average returns, right, 0:01:51.280 --> 0:01:54.080 And yet many of them put up really impressive returns 0:01:54.160 --> 0:01:55.880 year after year after year, And I don't think I 0:01:56.000 --> 0:01:57.600 totally have a grasp of life. 0:01:57.760 --> 0:02:00.280 Well, yes, and this is a question that I have, 0:02:00.400 --> 0:02:02.880 which is eventually the pod shop. Some of them are 0:02:02.880 --> 0:02:05.240 getting very very big, right, and so if you have 0:02:05.440 --> 0:02:09.640 one thousand pods working under your roof, that's a bit extreme. 0:02:09.840 --> 0:02:12.440 But at some point aren't you just sort of replicating 0:02:12.560 --> 0:02:16.079 the market and that alpha opportunity as you just described 0:02:16.160 --> 0:02:19.040 kind of goes away. Well, on that note, I am 0:02:19.040 --> 0:02:22.680 happy to say we have the perfect guests to discuss 0:02:22.840 --> 0:02:26.280 all of this. So these sort of variations behind multistrap 0:02:26.360 --> 0:02:29.840 funds and also the math that actually powers it. We're 0:02:29.880 --> 0:02:32.240 going to be speaking with Dan Morillo. He is the 0:02:32.240 --> 0:02:36.720 co founder of Freestone Grove Partners and also ex Citadel, 0:02:36.840 --> 0:02:39.440 so again, the perfect person to be speaking to. 0:02:39.840 --> 0:02:40.080 Dan. 0:02:40.320 --> 0:02:41.160 Welcome to the show. 0:02:41.600 --> 0:02:42.799 Thank you, thank you for having me. 0:02:43.240 --> 0:02:45.560 I guess my first question is why are we talking 0:02:45.560 --> 0:02:45.760 to you? 0:02:45.919 --> 0:02:46.919 Yeah, why are we talking? 0:02:47.840 --> 0:02:49.799 Well, you're probably in a better position to answer than me, 0:02:49.840 --> 0:02:51.600 but I guess I'll tell you my background and hopefully 0:02:51.600 --> 0:02:53.640 that helps a little bit. So I've been about twenty 0:02:53.639 --> 0:02:57.240 five years now dating myself in the by side, on 0:02:57.280 --> 0:02:59.640 the hedge fund by side in particular, and I grew 0:02:59.720 --> 0:03:01.560 up on the quantitative side of the world. I thought 0:03:01.600 --> 0:03:03.240 I was going to be a professor, and then I 0:03:03.360 --> 0:03:06.600 realized that life is more exciting on the industry side 0:03:06.600 --> 0:03:09.959 of things. And I've done a wide range of roles 0:03:10.000 --> 0:03:12.680 in the quote quant side of the world, so everything 0:03:12.720 --> 0:03:14.799 from you know, at some point I was the lead 0:03:14.880 --> 0:03:17.880 of the global long short business at Parkley's Global Investors 0:03:17.880 --> 0:03:21.000 before Black Crok required them. At Blackrok, I stuck around 0:03:21.040 --> 0:03:22.840 for a bit. I at some point ran the research 0:03:22.880 --> 0:03:24.760 group for I Shares. I also was one of the 0:03:24.800 --> 0:03:27.919 founders of the model Solutions business there as you said, 0:03:27.919 --> 0:03:28.359 I was etc. 0:03:28.880 --> 0:03:29.600 Where I had. 0:03:29.560 --> 0:03:32.160 Responsibility for the Equity Quantitariy Research Group that did a 0:03:32.160 --> 0:03:33.720 lot of the stuff that you guys have talked about, 0:03:33.880 --> 0:03:36.040 risk model stuff and the hedging stuff and all of 0:03:36.040 --> 0:03:38.440 these sort of things. I also had responsibility for the 0:03:38.480 --> 0:03:41.000 Center Book where a lot of that central stuff that 0:03:41.040 --> 0:03:44.520 you also have talked about happens, and then most recently 0:03:44.720 --> 0:03:48.240 have founded co founded that also does the pod long 0:03:48.240 --> 0:03:51.040 short thing. So I'd like to think I have some expertise, 0:03:51.040 --> 0:03:52.320 but I guess you'd tell me after you ask me 0:03:52.360 --> 0:03:53.080 all these questions. 0:03:53.640 --> 0:03:56.280 I have a really rudimentary question, what does the word 0:03:56.360 --> 0:03:57.800 quant mean in finance? 0:03:58.720 --> 0:04:01.160 Actually? So this is a good point, right, I think 0:04:01.360 --> 0:04:03.280 it can mean lots of things. From my point of view. 0:04:03.320 --> 0:04:05.560 The thing that has always been attracted to me about 0:04:05.600 --> 0:04:07.160 the quant side of things is the idea that you 0:04:07.200 --> 0:04:10.520 can be disciplined in how you make decisions. 0:04:10.600 --> 0:04:10.840 Right. 0:04:11.280 --> 0:04:13.800 You can be quantitative in the purely mathematical sense, like 0:04:13.840 --> 0:04:16.720 you brand some code and there's lots of numbers, while 0:04:16.800 --> 0:04:19.440 still not actually applying that much judgment. You can also 0:04:19.520 --> 0:04:21.919 actually be quite disciplined and systematic without using a lot 0:04:21.960 --> 0:04:23.760 of quant tools. Right. I think the right way of 0:04:23.800 --> 0:04:26.880 doing quant is where you also mix these two together, right, 0:04:26.880 --> 0:04:29.960 when you have the ability to bring in the judgment 0:04:30.040 --> 0:04:32.520 that comes from understanding what the humans in the market 0:04:32.520 --> 0:04:34.479 are doing, but to do certa in a way that 0:04:34.680 --> 0:04:38.120 is repeatable and disciplined, and that tends to require quantitative 0:04:38.120 --> 0:04:41.960 modeling tools, whether that's risk models, focusing, evaluation, attribution, all 0:04:42.000 --> 0:04:43.840 of these sorts of things, right, And in fact, that's 0:04:43.839 --> 0:04:45.760 the sort of thing that attracted me. That is sort 0:04:45.800 --> 0:04:48.359 of a I guess a common threat through all of 0:04:48.360 --> 0:04:50.960 these jobs that I mentioned that I've had is the 0:04:51.040 --> 0:04:54.080 idea that you can do this this sort of systematic 0:04:54.120 --> 0:04:57.719 modeling work not just with the numbers themselves, but also 0:04:57.800 --> 0:05:00.440 with the humans that participate in the market. Are also 0:05:00.480 --> 0:05:03.520 subject to analysis, right, whether you think about sentiment measurement 0:05:03.720 --> 0:05:05.680 or the sort of questions you guys have asked in 0:05:05.720 --> 0:05:08.400 this podcast, right, what is the right way to organize 0:05:08.440 --> 0:05:10.000 a team? You know, how many teams should you have, 0:05:10.160 --> 0:05:12.960 how should you pay them? What fees should you charge 0:05:13.000 --> 0:05:15.560 with those? These are all subject to analysis, right. So 0:05:15.600 --> 0:05:17.600 I like the idea that you can do the quant 0:05:17.640 --> 0:05:19.760 thing on human behavior, right. 0:05:20.040 --> 0:05:22.520 Oh, this is exactly what I wanted to ask you about. Actually. 0:05:22.560 --> 0:05:25.200 So if you go to Freestone's website, you can see 0:05:25.240 --> 0:05:28.080 that there are two partners on the front page, and 0:05:28.160 --> 0:05:31.440 you are the quantitative one, and you do have a 0:05:31.520 --> 0:05:36.520 large number of quant researchers. What's the value added by 0:05:36.560 --> 0:05:39.640 those quants to a fundamental equities fund? 0:05:40.640 --> 0:05:42.640 Yeah? I think the way you want to think about 0:05:42.680 --> 0:05:46.280 it is that the insight that is associated with understanding 0:05:46.880 --> 0:05:49.159 the mechanics of a firm, which is the fundamental in 0:05:49.160 --> 0:05:50.800 this case for equities. You know, the job of the 0:05:50.839 --> 0:05:56.120 PM analyst is to understand what drives revenue, earnings, margins. 0:05:55.760 --> 0:05:56.120 Et cetera. 0:05:56.160 --> 0:05:58.880 And in portucur what is likely to be surprising about 0:05:58.880 --> 0:06:01.400 those next time they know earnings or over the next 0:06:01.400 --> 0:06:03.680 couple of quarters. Right, the way you make money is 0:06:03.920 --> 0:06:05.560 you have a view that is different from that of 0:06:05.640 --> 0:06:07.960 the market and people come to agree with you. Right, 0:06:07.960 --> 0:06:08.960 that's sort of success. 0:06:09.080 --> 0:06:09.320 Right. 0:06:10.080 --> 0:06:13.840 And in that effort, whether it's modeling the firms, whether 0:06:13.880 --> 0:06:16.719 it's understanding what about that surprise is really surprising about 0:06:16.760 --> 0:06:18.960 the firm versus something that's happening in the broader market. 0:06:19.400 --> 0:06:21.760 The data that comes into all of this, right, alternative 0:06:21.800 --> 0:06:24.000 data stuff. All of that requires a huge amount of 0:06:24.000 --> 0:06:27.240 investment on the technology side, the analytics side, the forecasting side. Right. 0:06:27.880 --> 0:06:29.359 It's no longer the case that you can be a 0:06:29.360 --> 0:06:31.680 smart guy reading tank q's in ten case, as might 0:06:31.720 --> 0:06:33.680 have been the case twenty five years ago, and just 0:06:33.800 --> 0:06:35.440 kind of see what the surprise is going to be. 0:06:35.720 --> 0:06:39.320 It requires a significant investment in being the most sophisticated 0:06:39.320 --> 0:06:41.600 person at doing that job. And that's not a thing 0:06:41.680 --> 0:06:44.400 you can do without all of that investment on the 0:06:44.440 --> 0:06:45.279 quantitative tooling. 0:06:45.360 --> 0:06:45.560 Right. 0:06:45.920 --> 0:06:48.760 There's also all the behavioral stuff. Right. Humans have the 0:06:48.800 --> 0:06:51.760 ability to really get into the detail of what the 0:06:51.800 --> 0:06:54.000 firm is doing. Right. Many of the people who are 0:06:54.080 --> 0:06:55.559 very good at there's are people who have been covering 0:06:55.560 --> 0:06:58.039 the same firm literally for a decade. Right. They know 0:06:58.080 --> 0:07:00.560 their CFO or the CEO, the product. You know, they've 0:07:00.600 --> 0:07:03.440 visited the factories, and so they have this ability to 0:07:03.480 --> 0:07:07.080 preak up on really subtle patterns. But they're also human, right, 0:07:07.560 --> 0:07:10.840 and humans come with biases. Right. You project your own 0:07:10.920 --> 0:07:13.040 patterns of sort of your view of the world into 0:07:13.080 --> 0:07:15.080 what's happening on the ground, and so it is also 0:07:15.080 --> 0:07:17.920 helpful to think about how do you become as disciplined 0:07:17.960 --> 0:07:20.320 as possible in that process, right, So you think about 0:07:20.400 --> 0:07:24.440 risk models, attribution questions, how can you tell luck versus skill? Right? 0:07:24.840 --> 0:07:26.880 Most humans, if you do well, you tend to think 0:07:26.880 --> 0:07:28.560 it's all about you. And if you do poorly, well, 0:07:28.560 --> 0:07:31.160 there wasn't my fault, my fault, right, And so these 0:07:31.200 --> 0:07:33.920 processes of how do you make sure that humans are 0:07:33.960 --> 0:07:37.680 as discipline as possible again requires huge investment in that 0:07:37.880 --> 0:07:40.720 quantitative analytical capability. So that's kind of what you bring 0:07:40.720 --> 0:07:41.240 to the table. 0:07:41.320 --> 0:07:46.600 Right, So we'll get into how you go about measuring 0:07:46.600 --> 0:07:49.400 the skill of your portfolio managers and breaking all these 0:07:49.400 --> 0:07:51.400 things down, and we'll talk about that a lot. When 0:07:51.440 --> 0:07:55.600 you founded Freestone Growth, you and your co founder Todd Barker, 0:07:56.240 --> 0:07:58.800 you must think there's an opportunity there, right, You must 0:07:58.840 --> 0:08:02.320 think there's like some opportunity out there to make money, 0:08:02.360 --> 0:08:04.720 to have a fund that's different than something that already 0:08:04.760 --> 0:08:07.520 exists on the market, that you bring something to the 0:08:07.560 --> 0:08:10.720 table that you could structure a company in some way 0:08:10.840 --> 0:08:15.400 that's advantageous. What is the sort of theory or thesis 0:08:15.440 --> 0:08:19.080 behind Freestone Growth such that you wanted to build something new? 0:08:19.680 --> 0:08:22.080 Yeah, so you're correct, We do think we can compete 0:08:22.120 --> 0:08:24.200 at the highest level of the industry, right, Otherwise we 0:08:24.200 --> 0:08:26.960 would have started this time. The way in which we 0:08:27.040 --> 0:08:30.120 think we can do this isn't some new magic thing, right, like, oh, 0:08:30.240 --> 0:08:31.600 only we can do X, Y and Z. 0:08:31.800 --> 0:08:32.000 Right. 0:08:32.600 --> 0:08:35.320 A lot of how we and this is what we 0:08:35.360 --> 0:08:39.640 tell our clients is that in having spent all this 0:08:39.720 --> 0:08:42.599 time looking at what works and what doesn't in the 0:08:42.720 --> 0:08:45.600 space I call it the multi strategy or MULTIPM space, 0:08:46.120 --> 0:08:47.920 we have a view that you can sort of be 0:08:47.960 --> 0:08:51.559 optimal around key business decisions, right, the number of analysts 0:08:51.559 --> 0:08:53.720 and pms you have in your platform, the way you 0:08:53.840 --> 0:08:56.560 organize them, the way you think about the incentives or 0:08:56.559 --> 0:08:59.920 how they're compensated, the right mix of counnitata versus fundamental 0:09:00.400 --> 0:09:02.320 in a way that sort of is the best of 0:09:02.360 --> 0:09:04.880 what we've seen around. Right. So it's not so much oh, 0:09:04.920 --> 0:09:07.319 there's this one thing that is massively different about us, 0:09:07.360 --> 0:09:10.480 and instead lots of little things that we think you 0:09:10.520 --> 0:09:12.240 can optimize in a way that many of the other 0:09:12.240 --> 0:09:16.200 platforms for various reasons, haven't gotten to, particularly with the 0:09:16.440 --> 0:09:18.360 advent of a lot of new ones. Right, where you 0:09:18.480 --> 0:09:21.319 end up with business design that we happen to think 0:09:21.480 --> 0:09:23.720 is not nearly as optimal as it could be. Right, 0:09:23.760 --> 0:09:25.680 So it's sort of optimized the business as sort of 0:09:25.679 --> 0:09:28.080 the pitch and then run each piece the best you can. 0:09:28.400 --> 0:09:29.120 Does that make sense? 0:09:29.200 --> 0:09:29.920 Yeah? 0:09:29.960 --> 0:09:33.120 Well, on this note, so there's something that kept coming 0:09:33.200 --> 0:09:36.559 up when we were preparing for this podcast. But people 0:09:36.640 --> 0:09:40.880 keep talking about Dan's math. Can you put your professorial 0:09:40.960 --> 0:09:44.240 hat on and explain to us what exactly is Dan's 0:09:44.280 --> 0:09:47.280 math and how does it come into play when it 0:09:47.320 --> 0:09:50.720 comes to designing and optimizing the size of your firm. 0:09:50.960 --> 0:09:51.160 Yeah? 0:09:51.200 --> 0:09:53.240 So, first, in my defense, I did not come up 0:09:53.240 --> 0:09:56.360 with that. I believe it was actually somebody from BOOMEERG 0:09:56.360 --> 0:09:58.080 that came up with that after some interview that they 0:09:58.120 --> 0:10:00.719 did with us early on. But yeah, question. Look, the 0:10:00.760 --> 0:10:03.120 point is that many of the things that you think about, 0:10:03.160 --> 0:10:06.080 which range from how many people should you having a platform, 0:10:06.200 --> 0:10:08.360 what sort of risk models should you run, what risks 0:10:08.360 --> 0:10:10.800 should you take, how should you do capital location, these 0:10:10.840 --> 0:10:14.000 are things that are subject to systematic analysis, right, and 0:10:14.080 --> 0:10:16.800 so this idea of quote de math is that many 0:10:16.840 --> 0:10:19.679 of these decisions you don't have to wave your hands around, right, 0:10:19.720 --> 0:10:23.440 there's sort of reasonably clear answers about them, right, there's 0:10:23.440 --> 0:10:25.040 a couple of ones that and we can chase down 0:10:25.080 --> 0:10:27.520 whichever ones as you like. But one of the ones 0:10:27.520 --> 0:10:30.440 that in my mind is the most important is there's 0:10:30.480 --> 0:10:33.080 been this sort of press in the industry with this 0:10:33.160 --> 0:10:35.480 idea that more is always better, Right. You want to 0:10:35.520 --> 0:10:38.880 have more porfan managers, more as more assets like that 0:10:38.880 --> 0:10:41.040 that scales a sort of underlying strength. 0:10:41.240 --> 0:10:41.440 Right. 0:10:42.320 --> 0:10:44.800 It actually goes back to your question around how come 0:10:44.880 --> 0:10:47.760 do you get good results out of lots of people? Right? 0:10:48.320 --> 0:10:50.400 And the answer is to listener, is actually not wrong. 0:10:50.520 --> 0:10:50.640 Right. 0:10:50.679 --> 0:10:52.720 There comes a point where adding more people actually doesn't 0:10:52.720 --> 0:10:56.280 make any difference. Right, And so if you just allow 0:10:56.360 --> 0:10:58.640 me two minutes to set up a little example. Right. So, 0:10:59.040 --> 0:11:01.559 the way this business works is you're hiring individual risk 0:11:01.600 --> 0:11:05.160 takers let's call them analysts. Right, So there's some potential 0:11:05.240 --> 0:11:08.080 pool of people you can hire, and assuming you have 0:11:08.120 --> 0:11:10.880 good hiring practices, you expect to hire people who have 0:11:10.920 --> 0:11:13.800 some mean performance. Think of that as a sharp ratio. 0:11:13.920 --> 0:11:16.280 Let's say that sharp ratio is point seventy five, right, 0:11:16.679 --> 0:11:19.120 So have shot ratio point seventy five means that if 0:11:19.120 --> 0:11:21.800 you take a risk of a dollar of risk, you 0:11:21.840 --> 0:11:25.120 expect to januarate seventy five cents of per that amount 0:11:25.120 --> 0:11:27.600 of risk that you deployed, right. And so you want 0:11:27.600 --> 0:11:29.960 to think of performance in sharp ratio space, right, because 0:11:30.080 --> 0:11:33.680 in different spaces people have different risk. Right. There's you know, 0:11:33.760 --> 0:11:36.680 biotech names are riskier than say, bank names, and so 0:11:36.720 --> 0:11:38.400 you want to adjust for that. So typically you want 0:11:38.400 --> 0:11:40.960 to think in sharp ratio space. So you hire folks 0:11:41.240 --> 0:11:44.839 you expect to have some mean distribution, some mean outcome. Right. 0:11:44.880 --> 0:11:46.400 So I hire a person. I don't know what their 0:11:46.440 --> 0:11:48.280 sharp ratio is going to be. I hope it's good. 0:11:48.640 --> 0:11:50.559 And on avers I get people who are let's say 0:11:50.559 --> 0:11:52.640 point seventy five right. Some people are going to be 0:11:52.640 --> 0:11:54.080 better than that. Some people are going to be worse 0:11:54.120 --> 0:11:55.920 than that. Maybe I end up needing to hire them, right, 0:11:55.920 --> 0:11:58.360 But I get some distribution of them, right, and then 0:11:58.440 --> 0:12:00.240 you give them capital and they run a couple over 0:12:00.240 --> 0:12:02.319 the time. Right. And so the magic of the versification 0:12:02.520 --> 0:12:05.920 is that you get a higher sharp ratio as you 0:12:06.000 --> 0:12:10.440 add people. Right. If the correlation was exactly zero, then 0:12:10.760 --> 0:12:12.959 the more people you add, essentially, the more your sharp 0:12:13.040 --> 0:12:16.360 ratio increases. It increases where there's square root of end. Essentially, 0:12:17.120 --> 0:12:20.080 if there's correlation. However, there's like a maximum limit of 0:12:20.120 --> 0:12:22.800 how much your aggregator sharp ratio can be. Right. So 0:12:23.280 --> 0:12:25.520 let's take a simple example. Let's say these point seventy 0:12:25.559 --> 0:12:28.120 five people that you have on average, Let's say they're 0:12:28.120 --> 0:12:30.360 correlated by ten percent, which most people will tell you 0:12:30.440 --> 0:12:32.120 that sounds kind of low, not a lot of correlation. 0:12:32.760 --> 0:12:35.640 Then there's a maximum limit of what your starbration can 0:12:35.679 --> 0:12:38.160 be about two point three even if you have an 0:12:38.200 --> 0:12:40.640 infinite number of people. So you're intuition that if you 0:12:40.679 --> 0:12:42.560 add lots and lots of people, you add some gate 0:12:42.760 --> 0:12:45.200 to some quote average return is correct, it's just what 0:12:45.280 --> 0:12:47.920 is the scale of that average return, right, And so 0:12:48.320 --> 0:12:50.439 if you add lots and lots of people, you get 0:12:50.480 --> 0:12:52.800 to that sort of maximum level. And the thing that 0:12:52.880 --> 0:12:56.800 really matters is the correlation, right, So it is incredibly 0:12:56.840 --> 0:13:00.000 hard to get zero correlation like that just doesn't really happen. 0:13:00.520 --> 0:13:02.520 So, just to be clear what we're talking about when 0:13:02.600 --> 0:13:09.000 you say correlation, you hire one PM and they trade semiconductors. 0:13:09.240 --> 0:13:12.440 You hire another PM and they trade interest rates, or 0:13:12.480 --> 0:13:15.240 maybe they trade banks or something like that. Yeah, but 0:13:16.120 --> 0:13:19.160 because things in the market are generally correlated, you could 0:13:19.160 --> 0:13:22.199 have these different people all around the world, and implicitly, 0:13:22.320 --> 0:13:24.480 even though it looks like they have their own focus 0:13:24.480 --> 0:13:27.880 on the market, they might all implicitly be making money 0:13:27.920 --> 0:13:29.840 based on their read of the FED or something like that, 0:13:30.160 --> 0:13:33.880 and thus their returns are correlated. And therefore, even if 0:13:33.880 --> 0:13:37.160 they're really all really good at their jobs, that caps 0:13:37.200 --> 0:13:40.360 the amount of firm wide sharp by virtue of the 0:13:40.360 --> 0:13:42.239 fact that they're not really adding diversification. 0:13:42.320 --> 0:13:44.679 That is exactly correct. So, and it's as simple as 0:13:44.720 --> 0:13:47.640 if you were to observe somebody's return literally every day, right, 0:13:48.080 --> 0:13:50.640 and we observe the other persons return every day. You 0:13:50.679 --> 0:13:53.400 can just computer correlation, put it in Excel computer correlation. 0:13:54.000 --> 0:13:56.640 And if that number is low, you get more juice 0:13:56.640 --> 0:13:58.679 out of adding more people. If that almost is hi, 0:13:58.800 --> 0:14:02.240 you get less juice. To point, it matters enormously. So 0:14:02.280 --> 0:14:05.680 in that example, that maximum is about two point four. 0:14:05.760 --> 0:14:08.560 If your mean person is point seventy five, like with 0:14:08.640 --> 0:14:11.440 an infinite number right at correlation of ten percent, Let's 0:14:11.440 --> 0:14:14.400 say your correlation is actually twenty percent, right, you know, 0:14:14.440 --> 0:14:16.400 it's obviously more, but it's still low in the grand 0:14:16.440 --> 0:14:18.960 scheme of things, then that maximum number is only one 0:14:18.960 --> 0:14:22.000 point six, right, So a little bit of correlation has 0:14:22.040 --> 0:14:24.480 an enormous impact on how much you can deliver in 0:14:24.520 --> 0:14:28.760 the end. Right, And more importantly, you get pretty close 0:14:28.760 --> 0:14:30.560 to that maximum without a lot of people. 0:14:30.720 --> 0:14:30.880 Right. 0:14:30.960 --> 0:14:33.400 So, in the example of point seventy five, in a 0:14:33.400 --> 0:14:36.880 correlation of ten percent, if I have forty five risk takers, 0:14:36.960 --> 0:14:39.080 think of them as analysts. Let's say I put them 0:14:39.120 --> 0:14:41.520 in teams of three. Right, PM team made out of 0:14:41.600 --> 0:14:45.040 three risk takers. You know, there's not that many teams, right, 0:14:45.160 --> 0:14:47.720 fifteen teams. That gives me about ninety five percent of 0:14:47.800 --> 0:14:50.920 that ultimate maximum. Right, So I don't need to have 0:14:51.040 --> 0:14:53.840 one hundred teams to get to my maximum. In fact, 0:14:53.880 --> 0:14:57.000 there comes a point where it is actually more important. 0:14:57.040 --> 0:14:58.880 Let's say you have a million dollars actually to spend on. 0:14:58.960 --> 0:15:02.880 Something could be I hire another person, but something could 0:15:02.920 --> 0:15:06.120 also be, Hey, I might produce a better piece of 0:15:06.160 --> 0:15:08.640 software to help me manage that correlation. To teach people 0:15:08.680 --> 0:15:11.800 to think about whatever their return is really independent of, 0:15:12.000 --> 0:15:14.240 you know, for example, interest rates. As you highlighted them, 0:15:14.720 --> 0:15:17.040 that actually might be a significantly better investment than adding 0:15:17.040 --> 0:15:19.080 a team, because if I reduce my correlation by a 0:15:19.120 --> 0:15:21.640 little bit, that actually gives me more juice than just 0:15:21.720 --> 0:15:25.080 adding people. Right, And to look back to that original question, 0:15:25.720 --> 0:15:27.200 what do we think might be different in terms of 0:15:27.240 --> 0:15:30.360 how you set up your business. Is again that a 0:15:30.400 --> 0:15:32.960 lot of people have gone from scale for scale, even 0:15:32.960 --> 0:15:35.320 though you don't have to, at least not for performance reasons. Right. 0:15:35.360 --> 0:15:36.920 There comes a point where you just kind of have 0:15:37.000 --> 0:15:39.800 the right scale and you're better invest better off investing 0:15:39.840 --> 0:15:40.480 in other things. 0:15:40.560 --> 0:15:40.760 Right. 0:15:41.080 --> 0:15:42.920 The reason people have gone for scale is because they 0:15:42.960 --> 0:15:44.760 want to run more money. It's not because that gives 0:15:44.800 --> 0:15:47.680 you more performance, right, at least a past a certain amount. Right. 0:15:48.200 --> 0:15:50.320 And in fact, if you think about scale, scale comes 0:15:50.320 --> 0:15:52.400 with lots of other issues. It comes with complexity. You 0:15:52.480 --> 0:15:54.960 maybe end up with more management layers, You have to 0:15:55.360 --> 0:15:58.440 worry a lot more about you know, offices and coordination 0:15:58.680 --> 0:16:01.560 and you know management, etcetera. You might actually end up 0:16:01.960 --> 0:16:05.200 reducing your performance. That that complexity costs money. Right, And 0:16:05.280 --> 0:16:06.840 so one of the key things that we say to 0:16:06.840 --> 0:16:09.120 our clients, just as an example, is we look to 0:16:09.240 --> 0:16:12.200 cap our size so that we can run the right 0:16:12.280 --> 0:16:14.800 number of people at the minimum complexity if possible, while 0:16:14.840 --> 0:16:16.880 still delivering pretty much sad level of performance. 0:16:17.040 --> 0:16:28.280 Right. 0:16:33.360 --> 0:16:38.120 Why do hedge funds promise uncorrelated returns at all? Because 0:16:38.160 --> 0:16:40.480 it feels to me, as you just said, it's very 0:16:40.480 --> 0:16:43.680 hard to get correlation down to zero. But the pitch 0:16:43.760 --> 0:16:47.080 to investors is always, here are a bunch of uncorrelated 0:16:47.120 --> 0:16:49.560 returns that we can do over and over again. And 0:16:49.600 --> 0:16:53.600 then what you see repeatedly is that when there is 0:16:53.760 --> 0:16:56.560 a big event in the market, they all have drawdowns 0:16:56.720 --> 0:16:59.680 at the same time. So why do they keep pitching 0:16:59.720 --> 0:17:03.040 on correlated returns and why do you investors keep putting 0:17:03.040 --> 0:17:03.600 money in them? 0:17:03.720 --> 0:17:05.600 Okay, so there seems to be two questions then there, 0:17:05.640 --> 0:17:07.719 which is how come are they correlated even though they 0:17:07.720 --> 0:17:09.760 claim not to be a number? One? And two is 0:17:09.800 --> 0:17:11.520 that why is that even a thing in the first place? Right, 0:17:11.560 --> 0:17:14.560 So let me start with the second one. The reality 0:17:14.600 --> 0:17:17.240 is most correlation is driven by some common effect. 0:17:17.520 --> 0:17:17.720 Right. 0:17:18.040 --> 0:17:20.960 You know you've had guests here talking about risk models 0:17:20.960 --> 0:17:23.439 where you think about sort of common factors, right. And 0:17:23.680 --> 0:17:26.720 a key reason why if you're an allocator, say you're 0:17:26.720 --> 0:17:30.000 a pension fund, you know, in university endowment, is that 0:17:30.440 --> 0:17:32.680 you get paid for taking risk. Right. A lot of 0:17:32.720 --> 0:17:34.639 the allocation is into things that are risky, and you 0:17:35.080 --> 0:17:36.640 expect to get paid for taking that risk. 0:17:36.720 --> 0:17:36.800 Right. 0:17:36.840 --> 0:17:38.520 That's sort of in a sense, that's the function of 0:17:38.560 --> 0:17:40.560 a big endowment or a big punch of fund. Right. 0:17:41.400 --> 0:17:43.479 The thing is, most of the risks that pay you 0:17:43.840 --> 0:17:46.399 those returns, whether that's you know, market as a whole, 0:17:46.480 --> 0:17:49.040 whether it's you know, individual factors like momentum that you 0:17:49.040 --> 0:17:51.600 can buy separately, you know interest rate risk, you know 0:17:51.640 --> 0:17:54.520 inflation risk. All of these things you can allocate to 0:17:54.600 --> 0:17:57.680 those for like essentially like a tenth of a cent 0:17:57.760 --> 0:18:00.760 on the dollar, right. And so if you're going to 0:18:00.840 --> 0:18:03.720 make an allocation to something else, you don't want that 0:18:03.760 --> 0:18:06.680 allocation to be the same thing you already have at 0:18:06.800 --> 0:18:07.840 essentially no fees. 0:18:07.960 --> 0:18:08.160 Right. 0:18:08.560 --> 0:18:10.520 So let's say you have a hedgehund who charges you, 0:18:10.520 --> 0:18:12.960