WEBVTT - Matthew Granade Discusses Quantitative Investment (Podcast) 0:00:02.240 --> 0:00:06.800 This is Master's in Business with Barry Ridholts on Bloomberg Radio. 0:00:09.920 --> 0:00:12.840 This week on the podcast, I have an extra special guest. 0:00:13.119 --> 0:00:17.520 His name is Matthew Grenade and he is a senior God. 0:00:17.560 --> 0:00:21.840 How do I describe his role? His title really doesn't 0:00:21.840 --> 0:00:26.800 do it justice. His official title is Chief Market Intelligence 0:00:26.840 --> 0:00:31.360 Officer at Point seventy two, follow the progression that has 0:00:31.400 --> 0:00:34.320 taken place. Stevie Cohen was running Sack Capital for a 0:00:34.360 --> 0:00:38.680 long time. That was eventually converted into a family office 0:00:38.720 --> 0:00:42.960 which was Point seventy two that reopened to outside investors 0:00:43.040 --> 0:00:49.240 last year in UM and Grenade has been working there 0:00:49.720 --> 0:00:52.000 for a good couple of years. Previously he was at 0:00:52.080 --> 0:00:55.920 Bridgewater with Ray Dalio. You'll hear all about that during 0:00:55.920 --> 0:01:01.480 our conversation, but more importantly, you'll hear about the intersection 0:01:01.560 --> 0:01:05.440 between man and machine, between the way models can be 0:01:05.560 --> 0:01:10.200 used to not only manage assets, but improve the entire process, 0:01:11.000 --> 0:01:15.840 along with a variety of big data and other approaches 0:01:16.640 --> 0:01:20.040 UH that are really quite fascinating if you are at 0:01:20.040 --> 0:01:26.479 all interested in quantitative investing, machine learning, hedge funds, UH, 0:01:26.520 --> 0:01:30.200 the state of investing today and what anybody who is 0:01:30.240 --> 0:01:33.880 pursuing alpha must do to stay current, then you're gonna 0:01:33.920 --> 0:01:38.319 find this to be an absolutely fascinating conversation. So, with 0:01:38.360 --> 0:01:41.440 no further ado, here is my conversation with Point seventy 0:01:41.480 --> 0:01:49.200 two's Matthew Grenade. My extra special guest this week is 0:01:49.280 --> 0:01:54.120 Matthew Grenade. He is the chief market intelligence officer at 0:01:54.160 --> 0:01:58.240 Point seventy two. That is Stevie Cohen's new hedge fund, 0:01:58.640 --> 0:02:04.360 which employs about out people and manages about thirteen billion dollars. 0:02:04.800 --> 0:02:08.079 Point seventy two asset management was converted into a hedge 0:02:08.080 --> 0:02:13.639 fund in and last year it reopened to external investors. 0:02:14.280 --> 0:02:19.200 Matthew comes to us by way of Bridgewater Associates Domino 0:02:19.360 --> 0:02:23.360 Data Lab, and he got his both undergraduate and graduate 0:02:23.600 --> 0:02:27.920 NBA at Harvard Business School, where at undergraduate he was 0:02:28.000 --> 0:02:32.840 the president of the Harvard Crimson. Matthew Grenade, Welcome to Bloomberg. 0:02:33.000 --> 0:02:35.040 Thank you for having me. So let's start. Let's start 0:02:35.080 --> 0:02:39.600 with the most unusual thing on your um resume. You're 0:02:39.680 --> 0:02:44.120 president of the Harvard Crimson, not exactly a hotbed of 0:02:44.200 --> 0:02:49.000 future hedge funds officers. How did that come about? What 0:02:49.040 --> 0:02:52.480 was that experience? Like, well, a couple of things, I 0:02:52.520 --> 0:02:55.120 mean in terms of that experience. Running in newspaper is 0:02:55.160 --> 0:02:57.600 one of the most amazing things in the world, and 0:02:57.639 --> 0:02:59.840 I got to run a small one at Harvard. But I, 0:03:00.080 --> 0:03:02.320 you know, I think it's just an incredible job because 0:03:02.760 --> 0:03:05.600 you're in the middle of so much information, You're helping 0:03:05.639 --> 0:03:10.079 shape the debate, you're investigating things, so many interesting people. Um, 0:03:10.160 --> 0:03:12.320 Harvard's an awesome place to do that at. And so 0:03:12.600 --> 0:03:14.480 there are there are a few jobs that I've loved 0:03:14.480 --> 0:03:16.600 as much as I loved that one. It was incredible 0:03:17.320 --> 0:03:19.839 how you get that job. There's sort of a couple 0:03:19.840 --> 0:03:22.000 of things. One, there's a bit of a path um so, 0:03:22.160 --> 0:03:24.920 it generally is a newsperson, a reporter. Um. So I 0:03:24.960 --> 0:03:26.800 was a reporter for my first couple of years, and 0:03:26.800 --> 0:03:29.639 then I was the head of the central what's called 0:03:29.639 --> 0:03:32.919 the central Administration beat that covers the president of the university. 0:03:33.200 --> 0:03:35.720 That's also kind of a traditional stepping stone. Then there's 0:03:35.720 --> 0:03:38.200 a process called the Turkey Shoot. Um. The Turkey Shoot 0:03:38.240 --> 0:03:40.400 runs for about a month leading up to Thanksgiving, where 0:03:40.440 --> 0:03:42.680 they picked the next president. There's all sorts of sort 0:03:42.680 --> 0:03:45.320 of arcane rules, but probably the most interesting is that 0:03:45.680 --> 0:03:48.480 every outgoing member of the paper gets to vote, and 0:03:48.480 --> 0:03:51.240 if more than three disagree, you're blackballed, and so you 0:03:51.320 --> 0:03:54.760 resually hold in uh sort of and you know, in 0:03:54.800 --> 0:03:57.280 sort of in veto mode for as long as that goes. 0:03:57.320 --> 0:04:02.680 And so the deliberations generally run about teen hours straight exactly, 0:04:02.720 --> 0:04:04.560 and then and then there's a big party and whatever 0:04:04.640 --> 0:04:06.880 sort of once the sort of unlocking happens. But I 0:04:06.960 --> 0:04:09.200 had I think six or seven opponents for the job, 0:04:09.280 --> 0:04:11.920 and uh, you know, you have to you know, a 0:04:11.960 --> 0:04:14.760 little politics, a little politics, a little message, a little 0:04:14.760 --> 0:04:16.919 of this, um, and that's that's how it works. But 0:04:17.080 --> 0:04:20.160 it was an amazing opportunity. So that's an unusual background 0:04:20.279 --> 0:04:23.680 as a journalist and someone who's publishing the paper to 0:04:24.040 --> 0:04:28.080 really being a data scientist for a financial services shop. 0:04:28.440 --> 0:04:32.760 How did that career path unwind? Well, they're probably more 0:04:32.839 --> 0:04:34.479 similar than you think, because I mean a lot of 0:04:34.520 --> 0:04:39.279 it comes down to information, collecting information, using information, UM. 0:04:39.360 --> 0:04:40.840 And so you know, I've always been someone who likes 0:04:40.880 --> 0:04:42.640 to know what's going on, um, you know, what's going 0:04:42.640 --> 0:04:44.080 on in the world. I like to sort of be 0:04:44.120 --> 0:04:46.679 ahead of other people and knowing things, and so that's 0:04:46.720 --> 0:04:48.839 the that's the similarity. But the you know, but the 0:04:48.880 --> 0:04:52.719 career arc was um, I went from from college to Mackenzie, 0:04:53.120 --> 0:04:55.640 was a business analyist there, uh, and then went to 0:04:55.720 --> 0:04:58.920 business school like you mentioned, and then ended up at Bridgewater. 0:04:59.200 --> 0:05:02.520 Um and which is also a fascinating place. Is a 0:05:02.560 --> 0:05:05.560 fascinating place. So I was there for six years, um 0:05:05.640 --> 0:05:08.520 and Uh, it's a phenomenal place to work. I'm a 0:05:08.560 --> 0:05:11.800 big fan of Ray Dalio. I find his philosophy just 0:05:11.960 --> 0:05:15.560 totally intriguing. I think Bridgewater kind of gets a bad rap. 0:05:15.640 --> 0:05:18.520 People have called it a cult and have criticized the 0:05:18.600 --> 0:05:22.440 radical transparency. You survived there for six years. Can't be 0:05:22.480 --> 0:05:24.600 all bad, right, how to be pretty good? No, it's 0:05:24.600 --> 0:05:26.680 not all bad at all. In fact, I think it's 0:05:26.880 --> 0:05:30.440 you know one uh, you know as investors go, um, 0:05:30.480 --> 0:05:32.440 you know they're they're as good as it gets um, 0:05:32.560 --> 0:05:35.000 and you know, just phenomenal at it. And look, I 0:05:35.040 --> 0:05:40.520 think the differences of the culture there get overstated, um, 0:05:40.680 --> 0:05:44.080 meaning the radical transparency and meaning like how different it 0:05:44.120 --> 0:05:45.520 is from you know. Look, I mean like you know, 0:05:45.600 --> 0:05:50.159 I would say everywhere I've ever worked McKenzie point seventy two, Domino, Bridgewater, Um, 0:05:50.200 --> 0:05:52.600 you know, they've all been ambitious people who are trying 0:05:52.640 --> 0:05:54.320 to get to the right answer. Who wanted to do 0:05:54.360 --> 0:05:57.400 great things. Um. And you know, like at core, like 0:05:57.480 --> 0:05:59.800 that's that's a lot of what Bridgewater is about. And 0:06:00.160 --> 0:06:02.480 you know, Ray and the team, they're are very thoughtful 0:06:02.520 --> 0:06:06.200 about ways to um, you know, just sort of apply 0:06:06.560 --> 0:06:08.520 certain ideas. Um. You know, like you want to. You 0:06:08.520 --> 0:06:10.600 always want to make sure you're getting the best opinions right, 0:06:10.600 --> 0:06:12.919 And so they're very explicit about you know, who should 0:06:12.920 --> 0:06:15.039 you listen to about things? But you know, I see, 0:06:15.160 --> 0:06:17.039 I see Steve asked that question all the time, you know, 0:06:17.080 --> 0:06:18.480 like why am I listening to you? You know, I 0:06:18.480 --> 0:06:21.040 should be listening to this person instead. Um. And so 0:06:21.400 --> 0:06:23.400 I think Bridgewater is great at sort of scaling it. 0:06:23.480 --> 0:06:26.080 But but um, but I think that the ideas are 0:06:26.800 --> 0:06:29.000 are not not quite as radical as the media would 0:06:29.000 --> 0:06:32.039 want you to believe. And then the transparency, Um, it's 0:06:32.080 --> 0:06:34.440 just great. I mean I love the idea. Yeah, I 0:06:34.440 --> 0:06:35.520 mean I was. I always just saying like it's a 0:06:35.560 --> 0:06:36.920 very clean place to live. And the reason it's a 0:06:37.000 --> 0:06:39.000 very clean place to live at Bridgewater is you just 0:06:39.080 --> 0:06:42.000 don't say things behind people's back. You just say things 0:06:42.040 --> 0:06:46.479 to their face. Um, and you're just He writes about 0:06:46.480 --> 0:06:49.200 that in his first book in a chapter where he 0:06:49.279 --> 0:06:55.000 describes raise people problem. I mean, most founders and chairman 0:06:55.600 --> 0:06:59.160 don't spend the chapter describing the wrong people person. That's 0:06:59.240 --> 0:07:01.800 fairly trying its parent. Yeah, I mean I think that's 0:07:01.839 --> 0:07:05.400 fairly transparent. And and that's just how you're expected to operate, 0:07:05.440 --> 0:07:07.839 you know. I mean, if you're gonna say something about Ray, 0:07:08.000 --> 0:07:11.080 you say it to him. And um and I have 0:07:11.120 --> 0:07:13.840 many stories of of of saying things to Ray that 0:07:13.880 --> 0:07:18.240 I think people would find not horrifying. There's they were 0:07:18.280 --> 0:07:20.200 me being honest and him and I trying to sort 0:07:20.200 --> 0:07:22.280 of work out differences. But you know, the only rule 0:07:22.360 --> 0:07:24.360 was just don't say it behind his back, and and 0:07:24.400 --> 0:07:27.400 that's you know, it's it's interesting that that's considered so radical, 0:07:27.520 --> 0:07:29.440 you know what I mean, It's not it's not that radical. 0:07:29.960 --> 0:07:32.280 So now let's let's take this phote. You'll end up 0:07:32.320 --> 0:07:36.600 at at point seventy two. Your title is Chief Market 0:07:36.640 --> 0:07:40.840 Intelligence Officer. I've never even seen c M I O 0:07:41.040 --> 0:07:44.640 as a abbreviation. What does a c M I O do? 0:07:45.360 --> 0:07:47.920 That title was the title I had when I got there. Um, 0:07:48.000 --> 0:07:51.520 and I was really focused at that point on proprietary research. 0:07:51.680 --> 0:07:53.440 And so what we mean by that is how do 0:07:53.520 --> 0:07:57.800 we take UM data sets or surveys or web scraping 0:07:57.880 --> 0:07:59.520 or sort of all the different things you can do, 0:08:00.120 --> 0:08:03.840 UM and make that useful to our portfolio managers and analysts. UM. 0:08:03.920 --> 0:08:06.040 Since then, my job has evolved to include a couple 0:08:06.080 --> 0:08:09.040 other things. So I also oversee our central book at 0:08:09.040 --> 0:08:11.680 this point, UM, which is our sort of a systematic 0:08:11.680 --> 0:08:14.360 best ideas book we have and also receive venture capital. 0:08:14.480 --> 0:08:17.640 And we just haven't really changed the title. Quite fascinating. 0:08:18.040 --> 0:08:21.000 Let's talk a little bit about big data and machine 0:08:21.080 --> 0:08:25.760 learning and artificial intelligence. Help me make a little sense 0:08:25.800 --> 0:08:30.600 about those buzzwords which have come into vogue for a while. 0:08:31.280 --> 0:08:34.360 But but your shop has been using these things for 0:08:34.360 --> 0:08:39.520 for quite a while. UM. What's the state of the industry, uh, 0:08:39.559 --> 0:08:44.400 in terms of machine learning and big data and artificial intelligence? Well, 0:08:44.400 --> 0:08:46.840 I think the you know, the thing to sort of 0:08:46.880 --> 0:08:50.040 contextualize all those terms, UM. And you know, I agree 0:08:50.040 --> 0:08:52.320 with you, they're they're very buzzy. UM. But but the 0:08:52.320 --> 0:08:53.760 way I like to think about it as being model 0:08:53.800 --> 0:08:55.640 what I call model driven UM. And so you can 0:08:55.640 --> 0:08:58.640 talk about model driven businesses or model driven processes, and 0:08:58.720 --> 0:09:01.240 really the idea of a model is it takes in data. 0:09:01.480 --> 0:09:04.240 It could be big data, it could be not big data. UM. 0:09:04.280 --> 0:09:07.160 It runs a certain set of logic on that UM 0:09:07.280 --> 0:09:10.360 and then it produces a prediction of some variety UM. 0:09:10.520 --> 0:09:13.440 And you know, basically it tries to close the loop 0:09:13.480 --> 0:09:16.960 around that data so that you know, you're constantly improving 0:09:17.120 --> 0:09:19.800 the logic or the algorithms. And so Netflix is a 0:09:19.800 --> 0:09:22.680 model driven business intensents a model driven business UM. And 0:09:22.720 --> 0:09:25.120 obviously finance and and you know the hedge funds we're 0:09:25.120 --> 0:09:27.679 talking about there, you know, they're they're they're very model driven. 0:09:27.960 --> 0:09:29.719 What I would you know, what I would say is that, 0:09:29.760 --> 0:09:32.320 you know, the state of the industry, uh, in that 0:09:32.440 --> 0:09:35.720 regard is that UM, you know, these techniques are highly 0:09:35.800 --> 0:09:38.840 highly relevant to kind of almost everything we're doing, you know, 0:09:38.840 --> 0:09:41.960 whether it be extracting signal from data sets or you know, 0:09:41.960 --> 0:09:44.360 all the way up to making trading decisions. Uh. And 0:09:44.400 --> 0:09:46.280 so you know, we're investing, you know, like a lot 0:09:46.320 --> 0:09:48.480 of hedge funds were investing a lot in you know, 0:09:48.640 --> 0:09:52.280 people with the data science capabilities and with the machine 0:09:52.360 --> 0:09:57.000 learning capabilities as well. So ron course Ferry you famously said, 0:09:57.080 --> 0:09:59.640 torture the data long enough and it will confess to 0:09:59.679 --> 0:10:02.400 whatever you want. How do you avoid running into that 0:10:02.520 --> 0:10:06.280 problem of when you're building models and putting a ton 0:10:06.320 --> 0:10:09.880 of different quantitative information into it, how do you avoid 0:10:09.960 --> 0:10:12.960 that bad outcome of Hey, if we back test this 0:10:13.120 --> 0:10:15.400 enough and we make these tweaks, we could get this 0:10:15.480 --> 0:10:17.959 to say whatever we want. Yes, I think there's I 0:10:18.000 --> 0:10:19.760 think there's a couple of different ways you do that. 0:10:19.800 --> 0:10:22.720 I mean one is UM. You know, you want to 0:10:22.760 --> 0:10:26.240 have a fundamental intuition of some variety around what you're doing, 0:10:26.400 --> 0:10:27.920 you know, I mean, you're not just sort of running 0:10:27.920 --> 0:10:30.280 everything through a machine and some some people do, but 0:10:30.280 --> 0:10:32.200 but not not. That's not how I like to do it. 0:10:32.320 --> 0:10:34.120 You're not just sort of running everything through and sort 0:10:34.120 --> 0:10:36.400 of seeing, you know, seeing what fits, because to your point, 0:10:36.480 --> 0:10:38.480 something will fit UM, and it may be a real 0:10:38.520 --> 0:10:40.319 thing or it maybe you know, a very short lived 0:10:40.360 --> 0:10:42.200 thing UM. And then you know, you have to have 0:10:42.240 --> 0:10:45.080 a lot of discipline in terms of looking at your UM. 0:10:45.280 --> 0:10:48.200 You know it's called out a sample, uh sorry, basically 0:10:48.200 --> 0:10:50.560 in sample, out of sample and live UM. And what 0:10:50.600 --> 0:10:53.080 that basically means is where are you allowing yourself to 0:10:53.080 --> 0:10:55.520 to fit the parameters where you're sort of just looking 0:10:55.559 --> 0:10:57.720 at the results but still in a in a backwards 0:10:57.720 --> 0:10:59.960 looking way, and when are you sort of really trying 0:11:00.080 --> 0:11:02.200 it out? And you know, we have very strict rules 0:11:02.240 --> 0:11:05.720 about how we segment those different things before we start, 0:11:05.840 --> 0:11:08.880 you know, using you know, putting money against a certain strategy, 0:11:09.120 --> 0:11:11.360 so and out of sample, just to put a little 0:11:11.360 --> 0:11:15.160 flesh on that. If you're testing on a large cap us, hey, 0:11:15.240 --> 0:11:17.240 let's see how the status in the past. Let's see 0:11:17.240 --> 0:11:19.960 how it does overseas, not just the area you're looking 0:11:20.000 --> 0:11:22.520 forward to see if it's really something to the model. 0:11:22.559 --> 0:11:24.600 Is that a fair descriptor yeah. So let's say you 0:11:24.640 --> 0:11:27.680 were using um, you know, credit card data to trade Chipotle, 0:11:27.880 --> 0:11:30.120 you know, or something like that. Um. You know, what 0:11:30.160 --> 0:11:31.440 you would do is you would sort of, you know, 0:11:31.440 --> 0:11:33.280 you build some rules, um, and you would sort of 0:11:33.280 --> 0:11:36.240 fit those rules to some sub some set of data 0:11:36.280 --> 0:11:38.360 some time period, you know, three or four years. Then 0:11:38.520 --> 0:11:40.240 you would stop fitting the rules and you would sort 0:11:40.240 --> 0:11:41.600 of look at the next three or four years and 0:11:41.600 --> 0:11:43.920 sort of see it, does that those two match. Do 0:11:43.960 --> 0:11:46.679 they look the same or is the behavior very different? 0:11:46.920 --> 0:11:48.600 And then you would and then you would basically start 0:11:48.679 --> 0:11:51.120 running the model live from today and then see again 0:11:51.160 --> 0:11:53.040 if those match the other two periods and so you're 0:11:53.160 --> 0:11:55.720 looking sort of for a consistency across that and if 0:11:55.720 --> 0:11:57.439 you're not seeing that, then that's a good sign that 0:11:57.480 --> 0:12:00.280 you're overfitting it. It's also you know, because going back 0:12:00.320 --> 0:12:01.839 to my original point, you know, you want to think 0:12:01.880 --> 0:12:04.200 about whether or not there's a real intuition there. You know, 0:12:04.200 --> 0:12:07.200 I mean, should credit card and chipotle a make sense together? Right? 0:12:07.360 --> 0:12:08.840 It probably does because a lot of people use a 0:12:08.840 --> 0:12:10.679 credit card at chipole. But you know, if you were 0:12:10.800 --> 0:12:13.600 using uh, you know, credit card to trade ge you know, 0:12:13.640 --> 0:12:15.560 you might you might start scratching your head about what 0:12:15.600 --> 0:12:18.000 you're doing. Right, might just be a random correlation as 0:12:18.040 --> 0:12:21.800 opposed to a real causal relationship. So so let's talk 0:12:21.840 --> 0:12:26.679 about some of these unusual UM data sources. I know, 0:12:27.640 --> 0:12:31.520 alternative satellite data is all the rage these days. People 0:12:31.559 --> 0:12:33.920 are looking at parking lots, how filled they are. They're 0:12:33.960 --> 0:12:38.080 looking at how deep transport ships are sitting, uh in 0:12:38.160 --> 0:12:40.720 the water, how far below the waterline they might actually be. 0:12:41.440 --> 0:12:47.520 How esoteric can we get with these alternative types of data? Well, 0:12:47.520 --> 0:12:50.160 I think you can. I think you can get quite esoteric. 0:12:50.200 --> 0:12:52.720 I mean I think satellite, um you know, satellite has 0:12:52.720 --> 0:12:54.480 been around for a while and to your point. I mean, 0:12:54.520 --> 0:12:58.000 it's it's very widely used. Um. You know, you know 0:12:58.040 --> 0:13:01.000 what we think much more about now is um, you know, 0:13:01.000 --> 0:13:04.080 sort of much more specific data sets. UM. You know 0:13:04.160 --> 0:13:06.840 kind of that that give you, you you know, a read 0:13:06.880 --> 0:13:10.360 into a limited number of tickers, often via some sort 0:13:10.360 --> 0:13:14.120 of payment system or something like that. UM. And uh, 0:13:14.160 --> 0:13:16.600 you know I think that we're I think we're just 0:13:16.800 --> 0:13:19.160 you know, we're probably in the third inning of something 0:13:19.240 --> 0:13:21.160 or something like that in the in the in the 0:13:21.240 --> 0:13:26.600 data movement in investing. That's that's quite fascinating. So let's 0:13:26.640 --> 0:13:29.240 talk a little bit about complexity. You know, we could 0:13:29.240 --> 0:13:31.720 go back a hundred years and just look at Graham 0:13:31.720 --> 0:13:36.040 and Dodd simple p ratio and more expensive stocks over 0:13:36.080 --> 0:13:40.160 time perform less well and have lower expected turns than 0:13:40.679 --> 0:13:44.440 less expensive stocks. Are we running the risk of making 0:13:44.480 --> 0:13:49.120 things too complex? At at what point does complexity get 0:13:49.200 --> 0:13:53.880 outweighed by its own internal complications? Well, I think, um, 0:13:53.920 --> 0:13:55.560 you know, I think this goes back to the point 0:13:55.600 --> 0:13:59.400 I was making about, you know, about an intuition. Um. 0:13:59.480 --> 0:14:02.120 And you know, at the end of the day a 0:14:02.200 --> 0:14:05.320 point of two. You know, we are we are fundamental investors, 0:14:05.360 --> 0:14:07.839 you know, we believe that Uh, that you know that 0:14:07.920 --> 0:14:11.320 companies ultimately, you know, trade on how they're doing as 0:14:11.360 --> 0:14:13.160 a business and the kind of cash flows they're going 0:14:13.200 --> 0:14:16.760 to produce UM and you know, everything we do, I mean, 0:14:16.800 --> 0:14:20.040 we will use very sophisticated data science to predict a 0:14:20.080 --> 0:14:23.000 revenue stream or something like that, but we're at core 0:14:23.160 --> 0:14:25.480 trying to do something fairly simple. You know, we're trying 0:14:25.480 --> 0:14:27.720 to understand what the revenues are, what the costs are, 0:14:28.080 --> 0:14:31.320 you know, what the growth profile of the earnings are UM, 0:14:31.440 --> 0:14:33.720 and you know, we never sort of lose that grounding 0:14:34.160 --> 0:14:35.920 UM and so you know, look, there's a lot of 0:14:35.920 --> 0:14:38.160 ways to make money in the markets UM, and I'm 0:14:38.200 --> 0:14:40.400 only I'm not an expert in a lot of them. 0:14:40.440 --> 0:14:43.480 I'm only familiar with some of them. But but for us, 0:14:43.560 --> 0:14:47.680 I think that grounding back to pretty simple principles U 0:14:48.160 --> 0:14:50.320 is very important and not something that we lose track of. 0:14:50.920 --> 0:14:53.520 It's interesting that you I think of you guys as 0:14:53.560 --> 0:14:57.920 a quant shop, but you keep referring to intuition. What's 0:14:57.960 --> 0:15:02.520 the intersection like between man machine? Is it really UM 0:15:03.400 --> 0:15:07.760 technology aiding human decision making or is it mostly hey, 0:15:07.840 --> 0:15:10.560 let's go and make the decisions and we'll just see 0:15:10.560 --> 0:15:13.320 what happens, so it points in me too. We do UM, 0:15:13.360 --> 0:15:15.880 we do UM. We do a mix of of three things. 0:15:15.920 --> 0:15:19.600 We have a very large discretionary business that's global long 0:15:19.680 --> 0:15:23.120 short equity you know, people driven its portfolio managers and 0:15:23.160 --> 0:15:26.800 analysts UM looking at some subset of the of stock 0:15:26.920 --> 0:15:31.200 universe UM, meeting with management teams, looking at data sets UH, 0:15:31.240 --> 0:15:35.280 and then making decisions in a in a fairly discretionary fashion. UM. 0:15:35.320 --> 0:15:39.120 We also have a systematic business that's running on algorithms UM. 0:15:39.280 --> 0:15:41.920 And then we have a people plus machine business, which 0:15:41.960 --> 0:15:44.520 is the one that I oversee, which is the you 0:15:44.520 --> 0:15:47.160 know what what what you call the central book earlier UM, 0:15:47.360 --> 0:15:50.480 where what we're doing there is we're looking at UM 0:15:50.520 --> 0:15:52.680 what the behavior of all the people is as one 0:15:52.680 --> 0:15:55.280 of the important inputs UM. But we're also looking at 0:15:55.280 --> 0:15:58.840 the data sets and we're running algorithms to essentially helped 0:15:58.880 --> 0:16:01.000 make decisions out of that. So one way of thinking 0:16:01.040 --> 0:16:04.080 about it is that historically Steve had a best ideas 0:16:04.080 --> 0:16:07.000 book that he he ran as a discretionary investor, and 0:16:07.080 --> 0:16:09.920 over time we've built that up into a systematic best 0:16:09.960 --> 0:16:13.280 ideas book UM. But but a lot of the input 0:16:13.280 --> 0:16:16.480 of that is from discretionary investors and so UM. So 0:16:16.600 --> 0:16:18.400 you know, one of the kind of key questions we're 0:16:18.440 --> 0:16:20.640 always asking is what are the people best at and 0:16:20.680 --> 0:16:23.120 what are the machines best at? And you know, our view, 0:16:23.520 --> 0:16:26.120 UM is that you know, in terms of of really 0:16:26.160 --> 0:16:30.480 being able to interpret fairly nuanced and complicated situations inside 0:16:30.520 --> 0:16:35.240 a specific company, that people are still um, really really good. UM. 0:16:35.280 --> 0:16:37.440 You know, there's other things that machines do very very well. 0:16:38.000 --> 0:16:39.440 But you know, if you're going to meet with the 0:16:39.440 --> 0:16:42.400 management team and interpret a large set of data that 0:16:42.400 --> 0:16:44.600 that has a lot of sort of nuanced and specifics 0:16:44.640 --> 0:16:46.880 to it, UM, the people still beat the machines at that. 0:16:47.080 --> 0:16:48.680 And so we have a you know, we have several 0:16:48.720 --> 0:16:52.240 hundred people that do that. Do you see that edge 0:16:52.320 --> 0:16:57.200 of humans over machines continuing indefinitely or at at some 0:16:57.280 --> 0:17:02.400 point in the future, will smart um computers and artificial 0:17:02.400 --> 0:17:06.680 intelligence be able to do that also well? And definitely 0:17:06.760 --> 0:17:09.679 is a very long time. So I'm gonna I'm not 0:17:09.720 --> 0:17:12.240 gonna I'm not gonna comment on indefinitely. What I will 0:17:12.280 --> 0:17:15.080 say is that our our thesis is a firm right 0:17:15.080 --> 0:17:19.000 now over the next call it you know, seven to 0:17:19.080 --> 0:17:23.000 ten years, is that UM, is that it is people 0:17:23.040 --> 0:17:26.879 plus machines UM, and that the people are very good 0:17:26.920 --> 0:17:30.040 at the nuanced situation, at the idea generation, at the 0:17:30.280 --> 0:17:34.000 interpreting the thin data at the synthesis UM. And that 0:17:34.080 --> 0:17:37.840 the machines are very good at conducting, UH, correcting for 0:17:38.119 --> 0:17:42.119 behavioral bias at portfolio construction, at trade execution. And you know, 0:17:42.119 --> 0:17:43.600 what we're trying to do is figure out how you 0:17:43.640 --> 0:17:46.040 marry those two up in a really smart way UM. 0:17:46.119 --> 0:17:48.080 And that that is essentially the you know, the next 0:17:48.119 --> 0:17:51.560 wave of hedge fund but UM. But you know, like 0:17:51.600 --> 0:17:53.680 where where we are ten or fifteen years in terms 0:17:53.720 --> 0:17:56.920 of what people can do versus machines, I don't think 0:17:56.920 --> 0:18:00.600 I can comment on that quite quite fascinating. Let's talk 0:18:00.600 --> 0:18:04.840 about the venture capital work you guys do. UM. What 0:18:05.040 --> 0:18:09.960 makes you different from traditional vcs? Well, I think a 0:18:10.040 --> 0:18:13.760 couple of things make us UM different than traditional vcs, 0:18:13.800 --> 0:18:17.920 But probably the most important is we we are extremely 0:18:18.200 --> 0:18:22.399 expertise focused in how we are designed, so UM, we 0:18:22.440 --> 0:18:25.399 have no generalists. UM. We have certain practice areas. Right now, 0:18:25.400 --> 0:18:28.440 we have three different three or four different practice areas UM, 0:18:28.480 --> 0:18:31.479 all of which are led by people who have worked 0:18:31.480 --> 0:18:33.720 in that space and invested in that space for quite 0:18:33.760 --> 0:18:36.239 some time, and kind of one of the standards I 0:18:36.400 --> 0:18:41.320 use is, you know, when when when portfolio companies are 0:18:41.320 --> 0:18:44.560 meeting with the investors on our team, do they believe 0:18:44.600 --> 0:18:47.240 that the person they're sitting across from is the one 0:18:47.240 --> 0:18:49.480 of the world's leading experts on the area that they're 0:18:49.520 --> 0:18:52.160 working in. UM. So that that's one difference. I think 0:18:52.160 --> 0:18:54.399 that the other different side point too is we're extremely 0:18:54.400 --> 0:18:57.960 outbound in how we operate. So one of our challenges was, 0:18:58.320 --> 0:18:59.679 you know, we don't have a we don't have a 0:18:59.720 --> 0:19:02.120 brand end NVC you know the way a sequoia does 0:19:02.240 --> 0:19:04.480 or something like that. And so, you know, one of 0:19:04.520 --> 0:19:06.520 the biggest concerns you gotta have in venture investing is 0:19:06.560 --> 0:19:09.480 adverse selection UH. And you probably don't want to be 0:19:09.520 --> 0:19:12.359 taking what's coming through the door. UM. So you know, 0:19:12.400 --> 0:19:15.760 what we focus on is um themes that we think 0:19:15.840 --> 0:19:18.880 are gonna be big money makers, where we think real 0:19:19.160 --> 0:19:22.879 change is happening, where technology is is um uh is 0:19:23.280 --> 0:19:26.160 driving really important impact UH. And then we go try 0:19:26.200 --> 0:19:27.840 to find the companies that we want to invest in 0:19:27.960 --> 0:19:32.119 and knock on their door proactively, look extremely proactive. Almost 0:19:32.160 --> 0:19:35.800 almost all of it is an outbound motion like ninety 0:19:36.320 --> 0:19:39.359 eight percent of it UM and then UM, and so 0:19:39.400 --> 0:19:41.600 that it would be the two big differences. I'd also 0:19:41.680 --> 0:19:45.000 say that UM, you know, you know, probably as firms go, 0:19:45.440 --> 0:19:48.560 our diligence is more intense than a lot of venture firms. 0:19:48.560 --> 0:19:50.919 I think that comes from Steve Um. You know Steve 0:19:51.200 --> 0:19:54.520 uh Um. Steve's one of Steve's sayings is do the 0:19:54.560 --> 0:19:57.679 work um. And you know, when we go into an 0:19:57.680 --> 0:20:00.560 investment committee to talk about something, uh, there's kind of 0:20:00.560 --> 0:20:02.840 only one answer, which is I did the work um. 0:20:03.440 --> 0:20:05.960 Otherwise the meaning's gonna end very soon. And so we 0:20:05.960 --> 0:20:07.600 we hold a pretty high bar in terms of the 0:20:07.600 --> 0:20:09.480 amount of research we're gonna do when we're looking into 0:20:09.480 --> 0:20:10.879 a company. So there would be the three things I 0:20:10.960 --> 0:20:13.800 point to. So, once you decide to make an investment 0:20:13.920 --> 0:20:17.960 in a startup or an existing company, how actively involved 0:20:18.320 --> 0:20:22.040 um with the corporate management are you? Are you guys 0:20:22.080 --> 0:20:25.919 they're giving them advice assistance? Or is it more of 0:20:25.960 --> 0:20:29.439 an arms length here's some money, now, now go do 0:20:29.560 --> 0:20:34.280 something great. It varies, but I would say we're fairly active. 0:20:34.320 --> 0:20:36.920 And the reason we end up being active is goes 0:20:36.920 --> 0:20:40.160 back to this expertise thing that I was describing, which 0:20:40.200 --> 0:20:42.760 is that, Um, you know, because the team is made 0:20:42.840 --> 0:20:45.399 up of people who are very deep experts, it tends 0:20:45.400 --> 0:20:48.360 to be that the entrepreneurs want them on the boards 0:20:48.880 --> 0:20:51.240