WEBVTT - BlackRock’s Jeff Shen on Systematic Edge and AI 0:00:14.640 --> 0:00:17.560 Welcome to Inside Active, a podcast about active managers that 0:00:17.600 --> 0:00:20.200 goes beyond sound bites at headlines and looks deeper into 0:00:20.200 --> 0:00:24.600 the processes, challenges, and philosophies and security selection. I'm David Cote, 0:00:24.600 --> 0:00:27.200 I lead mutual fund and active research at Bloomberg Intelligence. 0:00:27.560 --> 0:00:31.240 Today my coast is Christopher Kaine, us quantitative strategist at 0:00:31.240 --> 0:00:33.960 Bloomberg Intelligence. Chris, thanks for joining me today. 0:00:34.320 --> 0:00:34.919 Thank you so much. 0:00:35.000 --> 0:00:35.280 David. 0:00:36.120 --> 0:00:39.040 So you recently wrote a note about how low vall 0:00:39.159 --> 0:00:43.360 strategies held up during last year's Taraft driven volatility, but 0:00:43.360 --> 0:00:46.479 they're not behaving the same way now, even with geopolitical 0:00:46.600 --> 0:00:48.720 risk and the headlines, so it's changed. 0:00:49.840 --> 0:00:52.360 Yeah, that's a good question, David. I'll be honest, I'm 0:00:52.680 --> 0:00:55.320 rather surprised about the factor movements so far this year. 0:00:55.720 --> 0:00:57.680 To be clear, you know, coming into the year, we 0:00:57.760 --> 0:01:02.400 had value as our top our top factor. Basically, value 0:01:02.440 --> 0:01:04.520 is very cheap when you look at cheap stocks versus 0:01:04.520 --> 0:01:08.120 expensive stocks, and had some positive trends in its favor, 0:01:08.200 --> 0:01:10.960 so we were positive on value. That said, I didn't 0:01:11.000 --> 0:01:12.640 know a war in the Middle East was going to 0:01:12.680 --> 0:01:16.000 break out. I'm I'm pretty surprised how well value has 0:01:16.040 --> 0:01:19.360 held up, and I'm also surprised how the defensive strategies 0:01:19.440 --> 0:01:22.600 just have not worked, things like low volatility, things like quality. 0:01:22.959 --> 0:01:25.600 They're not getting killed, but they're not getting that safe 0:01:25.640 --> 0:01:28.080 haven bid that you would think you would get when 0:01:28.080 --> 0:01:30.640 you have a geopolitical event like this. So I think 0:01:30.720 --> 0:01:32.960 you could you could kind of interpret it as the 0:01:32.959 --> 0:01:35.959 market shrugging office war or looking past it, but we 0:01:36.080 --> 0:01:38.880 have not seen the safe haven flows internally that you 0:01:38.880 --> 0:01:39.520 would expect. 0:01:40.120 --> 0:01:40.600 Interesting. 0:01:41.000 --> 0:01:43.840 Well, speaking of factors, I'd like to welcome Jeff Sheen 0:01:43.959 --> 0:01:47.199 to the podcast. Jeff is co CIO of Systematic Active 0:01:47.200 --> 0:01:49.760 Equity at black Rock, and I'm sure you can help 0:01:49.800 --> 0:01:53.800 us understand the benefits of factors and systematic equity and 0:01:53.840 --> 0:01:56.200 active management. So I want to thank you for joining us. 0:01:56.240 --> 0:01:59.160 Jeff, great to be here, Thanks very much for having me. 0:02:00.160 --> 0:02:03.800 So. Systematic active equity kind of sits in that middle 0:02:03.840 --> 0:02:08.960 ground between traditional discretionary investing and pure quantitative How do 0:02:09.040 --> 0:02:11.120 you define its core edge today? And you know what 0:02:11.160 --> 0:02:15.320 do you think people still misunderstanding about it? Yeah? 0:02:15.320 --> 0:02:18.040 I mean I think you know, systematic active equity type 0:02:18.040 --> 0:02:20.880 of strategy had been around for a long time. We 0:02:20.960 --> 0:02:25.760 actually just celebrated our forty year anniversary last year, and 0:02:25.800 --> 0:02:30.240 I'll see that's it's actually how it differentiates. It's certainly 0:02:30.360 --> 0:02:33.560 I think historically probably is when we think about discretionary 0:02:33.639 --> 0:02:36.960 it's very deep, could. 0:02:36.800 --> 0:02:39.520 Be in a particular name, could be in a particular. 0:02:39.160 --> 0:02:42.519 Country or sector, and systematic qualitative strategy is a little 0:02:42.520 --> 0:02:46.480 bit broad, you know, five miles wide, but a. 0:02:46.400 --> 0:02:47.240 Few feet deep. 0:02:47.919 --> 0:02:52.120 And I think, you know, to Christopher's earlier point around factories, 0:02:52.200 --> 0:02:54.280 that's sort of typically with people think about it. I 0:02:54.280 --> 0:02:57.760 actually think that the evolution today is such that I 0:02:57.800 --> 0:03:04.200 think there's an opportunity, especially given additional data, additional algorithms 0:03:04.240 --> 0:03:09.200 and technology, you can be both broad and deep at 0:03:09.200 --> 0:03:12.720 the same time. So it's a bit of an exciting 0:03:12.800 --> 0:03:17.360 moment for the evolution for systematic investment on a forwardlooking basis. 0:03:18.200 --> 0:03:20.680 And I think the other thing that I think maybe 0:03:20.720 --> 0:03:25.840 in terms of misconception, is also that historically there was 0:03:25.880 --> 0:03:29.280 certainly a bit of a feeling of systematic strategies could 0:03:29.280 --> 0:03:32.400 be a bit of a black box. People don't exactly 0:03:32.520 --> 0:03:35.600 understand their all these labeling, but people don't exactly what's 0:03:35.600 --> 0:03:38.120 going on. And to the point that you know, a 0:03:38.200 --> 0:03:42.880 minimum volatility strategy may not be as defensive as one thinks. 0:03:43.280 --> 0:03:46.680 I actually think that that when there's an inflation story, 0:03:46.760 --> 0:03:50.520 there's also an AI story that may, you know, maybe 0:03:50.600 --> 0:03:54.400 underneath a hood of that particular conundrum. But I think 0:03:54.440 --> 0:03:57.040 I think the exciting thing on a forward looking basis, 0:03:57.080 --> 0:04:00.320 given some of a new technology, is that I think 0:04:00.720 --> 0:04:04.920 you can also understand these strategies in a much more 0:04:04.920 --> 0:04:09.960 transparent and humanly understandable way. So I think that's probably 0:04:10.200 --> 0:04:13.000 one of the misconceptions people say, Oh, I understand the stock, 0:04:13.440 --> 0:04:15.440 but I'm not social about the factors. I'm certainly not 0:04:15.520 --> 0:04:19.640 social about quantitative methods. With the aid of AI and 0:04:19.720 --> 0:04:23.520 large language model, some of these transparencies may also be 0:04:23.640 --> 0:04:25.880 evolving a forarlooking basis as well. 0:04:26.800 --> 0:04:28.960 So when you kind of step back from the models 0:04:29.000 --> 0:04:32.760 and the signals, what are the persistent sources of alpha 0:04:32.839 --> 0:04:34.760 your team is really trying to capture, and you know, 0:04:34.800 --> 0:04:36.600 do you think those have kind of evolved over time? 0:04:38.160 --> 0:04:41.040 I think, yeah, it's certainly evolved, And I think evolution 0:04:41.279 --> 0:04:46.240 is actually it's a feature of a successful strategy. Today's 0:04:46.279 --> 0:04:50.120 alpha is really going to be tomorrow's beta, maybe tomorrow's 0:04:50.160 --> 0:04:53.080 smart beta. So I think it's ultimately I think the 0:04:53.200 --> 0:04:56.000 edge here is about how can you innovate, how can 0:04:56.040 --> 0:05:00.440 you evolve? So that's, if you will, that the the 0:05:00.520 --> 0:05:04.480 secret sauce. It's trying to think about what can drive 0:05:05.600 --> 0:05:09.599 security selection, cross sectional returns, or macro returns. The drivers 0:05:09.600 --> 0:05:12.800 are as fundamentally different, you know. I think about back 0:05:12.800 --> 0:05:15.680 in nineteen eighty five when we first started trading to 0:05:15.800 --> 0:05:18.960 be able to get a price to book, price to 0:05:19.040 --> 0:05:22.200 earning and give it and yield information for five hundred 0:05:22.200 --> 0:05:25.360 stocks in S and P five hundred. That was big 0:05:25.440 --> 0:05:28.920 data machine learning back in nineteen eighty five. But fast 0:05:28.920 --> 0:05:32.320 forward today, I think on this point of evolution, it's 0:05:32.360 --> 0:05:35.440 certainly about the amount of data you can access, the 0:05:35.440 --> 0:05:39.160 amount of learners that you have in your system, and 0:05:39.320 --> 0:05:41.440 the type of execution you're going to have. So I 0:05:41.440 --> 0:05:46.000 also that maybe the core source of alpha is certainly 0:05:46.040 --> 0:05:50.720 anchored around innovation and an innovation today. I think it's 0:05:50.839 --> 0:05:53.120 also one of the most exciting times, if not the 0:05:53.200 --> 0:05:55.640 most exciting times over the last forty fifty years, is 0:05:55.640 --> 0:06:00.640 that the many innovations you can go for data learners execution. 0:06:01.839 --> 0:06:03.520 So I'm sure we'll dig into it, but I think 0:06:03.600 --> 0:06:07.360 it's it certainly has evolved quite a bit over the years. 0:06:07.839 --> 0:06:09.719 Jeff, I would love to give you get your opinion. 0:06:09.800 --> 0:06:12.920 You know, I cover you know, equity factors at BI, 0:06:13.320 --> 0:06:16.680 and I typically focus on the well known factors, you know, 0:06:16.720 --> 0:06:20.440 the value of mental quality, low volatility. One of the 0:06:20.520 --> 0:06:25.080 questions I constantly get from customers is can factor exposures 0:06:25.120 --> 0:06:28.200 be timed? And it's a bit of a uh, you know, 0:06:28.600 --> 0:06:30.920 controversial thing. Some people think they can, some people that 0:06:31.080 --> 0:06:33.400 they can't. I don't think it's not this similar to markets, 0:06:33.720 --> 0:06:35.600 O can you time the market? So I guess I'll 0:06:35.640 --> 0:06:38.039 ask you, like, do you think factor exposures could be 0:06:38.040 --> 0:06:41.440 timed in a somewhat reliable way? And what, if any 0:06:41.520 --> 0:06:43.640 does factor timing play in your process. 0:06:44.000 --> 0:06:45.800 That's a very very good point. 0:06:46.120 --> 0:06:49.360 The headline here is actually, I'll say that it's possible 0:06:50.560 --> 0:06:54.680 to time the factors. But at the same time, it's possible, 0:06:54.720 --> 0:06:56.240 but you've got to do it very carefully. 0:06:57.040 --> 0:07:00.160 That's probably the headline. I think. 0:07:00.560 --> 0:07:03.400 You know, when I say carefully, is that I think 0:07:03.680 --> 0:07:07.640 you know specifically, I mean, we are very familiar with 0:07:07.680 --> 0:07:10.400 the factory investment. You know, back when I talk about 0:07:10.440 --> 0:07:13.000 nineteen eighty five, it was a bunch of value factors 0:07:13.160 --> 0:07:16.680 and then you know, momentum becomes a big research topic 0:07:16.720 --> 0:07:18.680 for us. We've done quite a bit of work on that. 0:07:18.800 --> 0:07:22.480 And then earning quality, as you were mentioning, you know, 0:07:22.560 --> 0:07:27.120 certainly started around two thousand n Roun and some will come. 0:07:27.760 --> 0:07:29.600 Some of the blow ups there certainly caused people to 0:07:29.600 --> 0:07:33.640 think about think about this quality metric in a very 0:07:33.680 --> 0:07:38.000 careful way. And I would say that maybe right after 0:07:38.040 --> 0:07:41.640 the financial crisis, that's probably there is a bit of 0:07:41.640 --> 0:07:44.240 a watershed moment in terms of think about factor and 0:07:44.320 --> 0:07:48.560 factor timing financial crisis, a lot of these factoris actually 0:07:48.560 --> 0:07:52.760 had severe draw downs, and they all actually draw down together. 0:07:53.640 --> 0:07:55.760 So I would say that around that time there was 0:07:55.800 --> 0:07:59.040 a you know, twenty ten twenty eleven, there was a 0:07:59.080 --> 0:08:02.760 big debate in the industry that you know, are the 0:08:02.800 --> 0:08:04.600 old factor is going to come back and you just 0:08:04.640 --> 0:08:06.680 need to time them a little bit more, or their 0:08:06.720 --> 0:08:10.920 additional new new things that you should be looking at. 0:08:11.200 --> 0:08:13.560 We sort of took the second route, where we think 0:08:13.600 --> 0:08:17.760 that there were a lot of new innovations alternative data, 0:08:17.800 --> 0:08:21.560 machine learning AI that you can use for systematic investments. 0:08:21.560 --> 0:08:24.480 So we went for that sort of a route in 0:08:24.720 --> 0:08:28.120 a significant fashion. But when we look back at the 0:08:28.440 --> 0:08:31.480 you know the factor lend if you will. I think 0:08:31.520 --> 0:08:35.640 these are becoming very well known. Uh you know, drivers 0:08:35.679 --> 0:08:39.840 of across sectional return. They're pretty persistent drivers of return, 0:08:39.920 --> 0:08:42.760 even though the time variation to a point is becoming 0:08:42.840 --> 0:08:47.120 much more pronounced. And I think this attempt to time 0:08:47.240 --> 0:08:49.960 them is certainly I think it's you want to do 0:08:50.000 --> 0:08:53.000 it carefully in the sense that you want to identify 0:08:53.040 --> 0:08:56.920 the factor in a very careful fashion. Depending the definition, 0:08:57.040 --> 0:08:59.920 it can vary quite a bit. You want to InCor 0:09:00.080 --> 0:09:02.440 put it not only bottom up information, but you also 0:09:02.440 --> 0:09:05.600 want to look at top down information how the regime 0:09:05.720 --> 0:09:09.800 can change the factory rotation. And then last but not 0:09:09.840 --> 0:09:11.679 the least is also you've got to think about how 0:09:11.679 --> 0:09:15.400 to manage the risk of this kind of timing because 0:09:15.440 --> 0:09:20.199 by definition, it's a low breath type of strategy. It's 0:09:20.200 --> 0:09:23.160 a low information type of rotation. So you want to 0:09:23.200 --> 0:09:25.319 make sure that you manage the risk accordingly. And if 0:09:25.320 --> 0:09:28.400 you do this carefully, I think it's certain. We've certainly 0:09:28.440 --> 0:09:32.600 done you know this type of strategy in different form, 0:09:32.720 --> 0:09:35.920 and we think there's actually some promise, but it's not 0:09:36.080 --> 0:09:38.320 nearly as easy as people think it is. 0:09:39.000 --> 0:09:41.640 Yes, I want to echo that. You know, it's like 0:09:42.400 --> 0:09:45.280 Cliff Astas says sin a little. You know, it's like 0:09:45.400 --> 0:09:47.360 if you're going to do some sometime and maybe just 0:09:47.360 --> 0:09:51.640 just sin a little. I love that. So, you know, again, 0:09:51.720 --> 0:09:54.720 traditional equity factors like your value of momentums, et cetera. 0:09:55.040 --> 0:09:57.840 I would call them medium term horizon factors. You know, 0:09:57.920 --> 0:10:01.520 typically the rebalance every month coms of fundamental factors like 0:10:01.640 --> 0:10:04.440 quality or value. Could be a little bit lower frequency, 0:10:04.520 --> 0:10:06.960 like every quarter. You know, I'm not asking for secret 0:10:06.960 --> 0:10:09.959 sauce here, but like do you have do you incorporate 0:10:10.040 --> 0:10:14.320 other factors that rebounce maybe more frequently or even less 0:10:14.320 --> 0:10:17.440 frequently to that kind of more traditional cadence. 0:10:18.400 --> 0:10:20.640 Yeah, I mean that's a fascinating question. May I'll say 0:10:20.640 --> 0:10:23.400 that if I zoom it all slightly, I think it is. 0:10:23.840 --> 0:10:26.440 This kind of a medium frequency that you talked about 0:10:27.160 --> 0:10:29.960 is actually, in my mind, a very exciting place to 0:10:30.040 --> 0:10:33.920 be in the sense that a lot of the high 0:10:33.960 --> 0:10:39.720 frequency traders are essentially trading milliseconds intra day. So these 0:10:39.760 --> 0:10:45.280 are very much a technical oriented focus and that's a 0:10:45.400 --> 0:10:46.520 very exciting place. 0:10:46.320 --> 0:10:46.920 For a long time. 0:10:46.960 --> 0:10:49.720 And there are probably twenty thirty dominant players in there, 0:10:49.920 --> 0:10:52.080 and they've done very well, and there's a big competition 0:10:52.200 --> 0:10:55.920 going on. I think the exciting thing on for looking 0:10:55.960 --> 0:10:57.760 base is exactly to your point on this kind of 0:10:57.800 --> 0:11:02.880 a medium three quincy and medium investment horizon, I actually 0:11:02.920 --> 0:11:05.800 think that there's a this is uh, you know, if 0:11:05.800 --> 0:11:07.920 you will, I like to say that this is you 0:11:08.040 --> 0:11:12.320 need technology, computer science, engineering, but you also need the 0:11:12.320 --> 0:11:16.360 liberal arts. You also need economics and finance and understanding 0:11:16.360 --> 0:11:19.960 of policy and the geopolitics to really make that uh, 0:11:20.480 --> 0:11:24.160 to deliver consistent set of returns. So I think the 0:11:24.280 --> 0:11:28.920 nature of the game is actually quite different, and you know, 0:11:29.040 --> 0:11:32.400 in the high frequency or shorter horizon, you actually don't 0:11:32.440 --> 0:11:34.760 really need too much of a sensibility per se or 0:11:34.800 --> 0:11:39.000 economic intuition versus in the intermediate horizon to the you know, 0:11:39.000 --> 0:11:44.280 whether it's a value or momentum or minimum volatility, there 0:11:44.320 --> 0:11:48.280 is actually some economic story associated with it. So I 0:11:48.280 --> 0:11:52.440 think in that sense, I think the the big price 0:11:52.520 --> 0:11:54.880 I think in my mind on a forward looking basis 0:11:55.600 --> 0:11:59.559 is that, if you will, there are certain firms kind 0:11:59.559 --> 0:12:02.439 of sawt of the little bit of the high frequency 0:12:02.960 --> 0:12:06.800 trading game, and it's still very competitive and keep on evolving, 0:12:07.640 --> 0:12:12.479 But there's actually a set of very successful large incumbents. 0:12:12.520 --> 0:12:15.760 Now in this kind of medium horizon, I think the 0:12:15.800 --> 0:12:20.480 space is reasonably open, uh, discretionary, systematic, and you know 0:12:20.480 --> 0:12:24.160 maybe old fashion systematic, modern fashion, uh, you know, modern 0:12:24.240 --> 0:12:27.439 version of systematic. So that's a bit of a wide 0:12:27.440 --> 0:12:29.840 space that I think we'll see a bit more evolution 0:12:30.640 --> 0:12:31.720 forwardlooking basis. 0:12:33.200 --> 0:12:35.760 So one thing I think that's really interesting in systematic 0:12:35.800 --> 0:12:37.600 investing is, you know, I think a lot of folks 0:12:37.720 --> 0:12:40.440 see it as purely technical, but you know, there's still 0:12:40.480 --> 0:12:44.040 a research culture behind it. How do you foster idea 0:12:44.120 --> 0:12:47.920 generation and you know, intellectual diversity within you know, your 0:12:47.960 --> 0:12:48.840 systematic team. 0:12:50.120 --> 0:12:53.239 Yeah, that's I mean, David, that's probably one of the 0:12:53.280 --> 0:12:56.280 most important thing. Uh. You know, even though we're talking 0:12:56.320 --> 0:13:00.600 about numbers and uh, technology and systems, I think ultimately 0:13:00.679 --> 0:13:03.320 this is a people business. We want to make sure 0:13:03.320 --> 0:13:06.439 that we hire the best in the world. At the 0:13:06.480 --> 0:13:09.960 same time, I think from a cultural perspective, I'll say 0:13:09.960 --> 0:13:13.960 that two things that I think has actually made a 0:13:14.080 --> 0:13:17.240 big difference for us, just to sort of looking inside 0:13:17.280 --> 0:13:21.360 out a little bit. It's I think one part is 0:13:21.400 --> 0:13:25.719 actually this concept of collaboration. Now it's a little bit 0:13:25.880 --> 0:13:29.040 overused words to a certain extent. But I think you 0:13:29.080 --> 0:13:31.920 know what we did was, actually we provide a tremendous 0:13:31.920 --> 0:13:35.040 amount of transparency for folks who join us. 0:13:35.200 --> 0:13:36.880 So, you know, forty years. 0:13:36.559 --> 0:13:41.720 Of IP and people on day one would get a 0:13:41.720 --> 0:13:45.160 tremendous amount of transparency to how we do things, what 0:13:45.200 --> 0:13:48.199 we do, and so that they kind of realize that 0:13:48.880 --> 0:13:51.320 there are a lot of work that's already been done, 0:13:51.440 --> 0:13:54.960 lots of ideas have already been tried. It's very intimidating 0:13:55.760 --> 0:13:58.520 because when I first started, I kind of realized that, Okay, 0:13:58.800 --> 0:14:01.040 most of the question that I thought with smart questions 0:14:01.240 --> 0:14:04.240 have already been answered five years before I joined, so 0:14:04.320 --> 0:14:06.600 you really need to step up. So but that level 0:14:06.600 --> 0:14:12.000 of transparency really fostered the level of collaboration, making sure 0:14:12.040 --> 0:14:16.480 that there's actually a lot of disciplines involved in solving 0:14:16.559 --> 0:14:19.280 hard problems, and that there's actually a lot of people 0:14:19.320 --> 0:14:22.280 trying to help each other. So it's actually one big 0:14:22.800 --> 0:14:25.520 platform really leverage that. So so that that part, I 0:14:25.520 --> 0:14:29.960 think from a cultural perspective, really help help that. I 0:14:30.000 --> 0:14:33.440 think the second part that I think is also probably 0:14:33.480 --> 0:14:36.440 equally important, this is I'm going to speak a little 0:14:36.480 --> 0:14:39.600 bit from the San Francisco Silicon Value perspective is that 0:14:40.040 --> 0:14:43.560 this idea that the world can be a better place 0:14:44.160 --> 0:14:47.640 and whatever is being done, uh it's all good, but 0:14:47.800 --> 0:14:51.200 there will be something better and more and different. Uh. 0:14:51.280 --> 0:14:54.720 So if you will this kind of outlier type of thinking, 0:14:55.360 --> 0:14:58.080 I think it's actually very central to our culture. It's 0:14:58.080 --> 0:15:00.360 not just to say, oh, everybody's actually done this kind 0:15:00.400 --> 0:15:02.920 of things before and we're just going to execute a 0:15:02.920 --> 0:15:03.560 little bit better. 0:15:03.920 --> 0:15:05.520 No, I mean we want to actually. 0:15:05.160 --> 0:15:08.080 Think from the basic principle and to see if we 0:15:08.080 --> 0:15:12.080 can do things fundamentally different on a forwardlooking basis. So 0:15:12.120 --> 0:15:15.120 I think these are the two things devid to a point. 0:15:15.120 --> 0:15:19.000 I think from the people talent and the cultural perspective, 0:15:19.360 --> 0:15:21.680 I think if you can get smart people to collaborate 0:15:22.120 --> 0:15:25.520 and then really allow them but also push them to 0:15:25.640 --> 0:15:29.520 really go for a different way of thinking, a different 0:15:29.520 --> 0:15:33.200 way of investing, I think that's actually been very much 0:15:33.280 --> 0:15:34.680 central to our DNA. 0:15:36.200 --> 0:15:39.480 I think it's an interesting comment about the transparency because 0:15:39.480 --> 0:15:41.360 that really gets people up the speed. And like you said, 0:15:41.400 --> 0:15:44.040 I feel like us quants answer the same question in 0:15:44.080 --> 0:15:47.400 five you know many different ways. You're so kind of 0:15:47.400 --> 0:15:51.000 building on these questions, so you know, there's these well 0:15:51.040 --> 0:15:54.240 known factors that I call open secrets and finance that 0:15:54.320 --> 0:15:56.680 people know the value moments and quality low of all 0:15:57.520 --> 0:16:01.120 you know, generally like what percent didch what you say 0:16:01.160 --> 0:16:04.680 of your research is involved in those factors, trying to 0:16:04.720 --> 0:16:07.520 refine them, trying to get the best combination, trying to 0:16:07.560 --> 0:16:11.560 even time them, versus trying to find new things that 0:16:11.600 --> 0:16:13.760 are bespoke that people might not know about or are 0:16:13.840 --> 0:16:15.800 not well known to add value. 0:16:16.840 --> 0:16:20.160 Yeah, I mean I think today I'll say very little 0:16:20.520 --> 0:16:25.560 to try to refine the open secret type of factors. 0:16:25.960 --> 0:16:28.720 I think this is actually also we've been on this 0:16:28.800 --> 0:16:32.560 journey for sixteen seventeen years, so it's really around twenty 0:16:32.720 --> 0:16:36.440 nine twenty ten. You know, our global head of research, 0:16:36.520 --> 0:16:39.920 Run Kong, back in twenty ten basically say you know what, 0:16:40.360 --> 0:16:43.880 we are not going to do value version number eleven. 0:16:44.280 --> 0:16:48.880 We're not going to do quality version seventeen. And you know, 0:16:49.160 --> 0:16:51.120 while there are other people who may want to do that, 0:16:51.320 --> 0:16:54.240 we're going to go on a completely different route, which 0:16:54.280 --> 0:16:58.160 is to look for new signals, new ideas that nobody 0:16:58.200 --> 0:17:01.560 has actually really thought about before. You know, simple things 0:17:01.640 --> 0:17:03.920 like you know, if you look at the job posting 0:17:04.040 --> 0:17:07.680 data in the United States, there are you know, millions 0:17:07.680 --> 0:17:10.840 of these job posting data. You know, that's a you 0:17:10.880 --> 0:17:15.320 know on companies website, on aggregator's database. And if you 0:17:15.359 --> 0:17:18.200 have that data set, you can essentially get a sense 0:17:18.240 --> 0:17:21.399 of Okay, is this company hiring and what type of 0:17:21.440 --> 0:17:24.800 skill set are they hiring? Modern skill sets versus old 0:17:24.880 --> 0:17:27.760 fashioned skill set. We have this signal idea which is 0:17:28.320 --> 0:17:32.280 jokingly called lung Python short Excel. You know, people who 0:17:32.280 --> 0:17:35.280 are hiring more Python oriented skill sets are more ready 0:17:35.680 --> 0:17:38.840 for the machine learning AI world and versus actually quite 0:17:38.960 --> 0:17:42.080 like Excel. But we're bet against the company that's hiring 0:17:42.119 --> 0:17:47.200 heavily into Excel oriented skill set. So you can actually 0:17:47.760 --> 0:17:51.119 have a new data set and come up with new factors, 0:17:51.119 --> 0:17:53.920 if you will, new signals and asking a bunch of 0:17:54.000 --> 0:17:57.639 questions you never thought about asking. And so we've been 0:17:57.680 --> 0:18:00.119 on that journey for sixteen seventeen years and fast for 0:18:00.280 --> 0:18:04.080 today it is you know, one thy five hundred signals 0:18:04.119 --> 0:18:08.760 or factors if you will, far far beyond the traditional, 0:18:08.840 --> 0:18:14.040 open secret type of traditional factors. And that really allow 0:18:14.200 --> 0:18:19.560 us to you know, really tap into the data engineering piece, 0:18:20.240 --> 0:18:23.800 the different learner piece, the some of the new new 0:18:23.840 --> 0:18:26.480 things in terms of you know, natural language processing, and 0:18:26.560 --> 0:18:29.960 fast forward today large language model. So I think once 0:18:30.000 --> 0:18:32.919 you open it up, you do realize the world is 0:18:33.040 --> 0:18:35.360 very big and there are many many places to go. 0:18:35.760 --> 0:18:38.560 And I think given the development of AI today, I'll 0:18:38.560 --> 0:18:42.440 see that that number is going to go skyrocket even 0:18:42.480 --> 0:18:45.000 more on our forelooking basis, just given the technology is 0:18:45.000 --> 0:18:46.000 getting a lot better. 0:18:47.119 --> 0:18:49.840 So, I mean, it's obviously fascinating what's happened with AI 0:18:50.040 --> 0:18:52.639 and data. And if we just think about data, what 0:18:52.800 --> 0:18:54.840 do you think the real edges today? Is it having 0:18:55.000 --> 0:18:58.640 unique data or you know, being better at extracting signals 0:18:58.680 --> 0:19:01.359 from widely available data. 0:19:01.520 --> 0:19:03.760 Yeah, I mean, David, that's a very good question. I 0:19:03.760 --> 0:19:06.720 mean I would say that we probably don't necessarily see 0:19:06.800 --> 0:19:10.440 too much of a trade off. I think on one 0:19:10.520 --> 0:19:13.879 side of it is we are going for more and 0:19:13.920 --> 0:19:16.600 more data. You know, the number is like we have 0:19:16.840 --> 0:19:20.240 thirty thousand data sets on the platform, and that number 0:19:20.280 --> 0:19:23.480 I think is going to explode even more, especially in 0:19:23.600 --> 0:19:29.159 the world of synthetic data on top of you know, 0:19:29.200 --> 0:19:32.920 if you will sensor based or quote unquote real data. 0:19:33.080 --> 0:19:36.960 So I think that data explosion is going to continue. 0:19:37.280 --> 0:19:40.240 On the other side of it, I think given the 0:19:40.280 --> 0:19:42.280 more and more data that's available. I think it's also 0:19:42.280 --> 0:19:47.400 extraordinarily important to think about what are the interesting questions 0:19:47.520 --> 0:19:51.200 we're asking to this particular data set. To your point, 0:19:51.520 --> 0:19:54.840 extract more information from a particular data set. And I 0:19:54.880 --> 0:19:57.920 think that super exciting area because I think the quality 0:19:57.920 --> 0:20:00.000 of the questions in this kind of a big data 0:20:00.080 --> 0:20:05.359 machine learning era is very very important. And this ability 0:20:05.440 --> 0:20:09.840 to ask interesting questions I think now can also empowered 0:20:10.440 --> 0:20:13.520 by generative AI, so you can actually, you know, we 0:20:13.560 --> 0:20:18.200 have an interesting research which is actually essentially co working 0:20:18.600 --> 0:20:22.920 with large large language model try to extract more interesting 0:20:23.000 --> 0:20:26.000 questions that we have not thought about asking for a 0:20:26.000 --> 0:20:29.160 particular data set. So it is actually data to answer 0:20:29.160 --> 0:20:34.000 your question is actually broader but also deeper, more data sets. 0:20:34.240 --> 0:20:37.080 But it's also trying to make sure that we ask 0:20:37.160 --> 0:20:40.400 deeper questions to each individual data sets. And I think 0:20:40.480 --> 0:20:42.800 this is we can talk more about it. But I 0:20:42.800 --> 0:20:47.360 think this concept of scale in alpha generation, I think 0:20:47.359 --> 0:20:51.840 it's becoming increasingly critical. You need scale in data, you 0:20:51.920 --> 0:20:56.320 need scale in inside generation, you need scale in compute, 0:20:56.640 --> 0:21:00.439 you need scale in implementation. So how do you leverage 0:21:00.480 --> 0:21:03.680 the scale for innovation for alpha I think it's really 0:21:03.720 --> 0:21:06.119 going to be a bit of a defining theme in 0:21:06.160 --> 0:21:10.520 the years to come, you know. 0:21:10.520 --> 0:21:13.320 Jeff, you know obviously whatever you're willing to share here, 0:21:13.359 --> 0:21:15.560 but you know, kind of going back to the factor 0:21:15.600 --> 0:21:18.600 timing question, you know, and then let's just talk about 0:21:18.640 --> 0:21:21.480 traditional factors. And you said some of the extent of 0:21:21.560 --> 0:21:25.119 is like bottom up processes, or's top down processes. So 0:21:25.200 --> 0:21:27.159 let's focus on top down. So, like, if you can 0:21:27.200 --> 0:21:29.560 give us maybe just an idea of like the general 0:21:29.720 --> 0:21:33.919 things that maybe you could use the time factors, like 0:21:34.080 --> 0:21:38.840 is it the momentum of the factor, the evaluation differences, 0:21:39.000 --> 0:21:42.520 is some kind of economic regime, anything you could share 0:21:42.520 --> 0:21:43.600 that would be very valuable. 0:21:44.040 --> 0:21:47.040 Yeah, I mean I give you maybe one specific example 0:21:47.080 --> 0:21:49.640 that I find it to be to be interesting, right, 0:21:49.720 --> 0:21:52.520 I mean, I think you mentioned regime, and you know, 0:21:52.560 --> 0:21:56.640 this is always an interesting concept in the sense that intuitively, 0:21:56.720 --> 0:21:59.440 with the twenty twenty hindsight, you know, where one can 0:21:59.480 --> 0:22:03.760 write us story about a particular environment happens. Certain you know, 0:22:04.000 --> 0:22:06.119 strategy is supposed to work, and then they work, and 0:22:06.119 --> 0:22:09.760 it's very comforting to see. But ACCENTI, how do you 0:22:09.840 --> 0:22:13.040 define a regime and how do you link the regime 0:22:13.200 --> 0:22:16.920 to the factor performance in a rotational sense? So the 0:22:17.240 --> 0:22:20.280 interesting work that we've done I think I'm happy to 0:22:20.320 --> 0:22:23.679 share is actually try to define the regime in a 0:22:23.800 --> 0:22:28.480 way that that's much more machine learning driven. Is to say, 0:22:28.600 --> 0:22:31.080 let's look at everything going on in the world, you know, 0:22:31.119 --> 0:22:35.720 different asset prices, different economic variables, and you can actually 0:22:35.840 --> 0:22:40.520 use a machine learning algorithm to essentially define what type 0:22:40.560 --> 0:22:43.679 of environment that is given all the you know, if 0:22:43.720 --> 0:22:45.840 you will time serious data that you observe in a 0:22:45.880 --> 0:22:49.200