WEBVTT - Giuseppe Paleologo on Quant Investing at Multi-Strat Hedge Funds 0:00:05.160 --> 0:00:19.119 Bloomberg Audio Studios, Podcasts, Radio News. 0:00:20.760 --> 0:00:24.280 Hello and welcome to another episode of The Odd Laws podcast. 0:00:24.360 --> 0:00:28.760 I'm Jill Wisenthal, normally joined by my co host Tracy Alloway, 0:00:28.760 --> 0:00:31.640 but she's on vacation today, so it's just me in 0:00:31.680 --> 0:00:34.400 this intro. But in today's episode you will hear a 0:00:34.479 --> 0:00:40.080 conversation taped live at Bloomberg's Reimagining Information form. On June twelfth, 0:00:40.280 --> 0:00:43.600 we spoke with Gappi Pallioligo, global head of quantitative Research 0:00:43.640 --> 0:00:46.879 at Ballysny Asset Management. He has a new book out, 0:00:46.920 --> 0:00:50.440 it's called The Elements of Quantitative Investing. Neither of us 0:00:50.520 --> 0:00:52.599 have read it because it would go way over our 0:00:52.640 --> 0:00:55.000 heads because we're not quant so we don't know how 0:00:55.000 --> 0:00:57.880 to read that stuff. But Gappy is great at explaining 0:00:57.920 --> 0:01:00.480 all of this stuff in clear English. So we had 0:01:00.480 --> 0:01:03.040 a great conversation and we hope you enjoy listening to it. 0:01:03.160 --> 0:01:04.720 So just to begin, I'm going to start with a 0:01:04.760 --> 0:01:08.680 really really dumb question, possibly, but isn't all investing quant 0:01:08.760 --> 0:01:13.160 investing nowadays? I mean every investor has access to some 0:01:13.319 --> 0:01:15.240 form of quantitative or using numbers. 0:01:15.319 --> 0:01:20.880 Yeah, I guess yes. End of answer. Yeah, I think so, 0:01:21.000 --> 0:01:23.480 I think so I mean pretty much everybody uses some 0:01:23.600 --> 0:01:26.759 kind of quantitative overlay, right, but two different degrees. So 0:01:26.840 --> 0:01:29.440 I have a friend who worked for one of the 0:01:29.600 --> 0:01:35.039 Tiger cubs, and they they refused to use sharp They 0:01:35.160 --> 0:01:39.200 refused to use logs in a spreadsheet because they said 0:01:39.200 --> 0:01:41.760 that they were dangerous. Probably they took the log of 0:01:41.800 --> 0:01:45.000 a negative number. And so yeah, no, two different degrees. 0:01:45.040 --> 0:01:48.920 But yes, there is some quantitative culture seeping through. 0:01:49.200 --> 0:01:53.000 Okay, so what defines quantitative investing? How would you differentiate 0:01:53.040 --> 0:01:56.280 that from I don't know, value investing, discretionary investing. 0:01:56.920 --> 0:02:00.840 Okay, I think that there are several possible answers. So 0:02:01.440 --> 0:02:05.760 I'm going to go with the one answer that I 0:02:05.840 --> 0:02:10.560 read in my life as a quant I think it's 0:02:10.600 --> 0:02:13.200 a wily book. It's a very good book, by the way, 0:02:13.560 --> 0:02:21.040 And I think Cliff Asnes defined quantitative investing as basically 0:02:21.120 --> 0:02:26.600 investing in a large cross section of assets, having a 0:02:26.680 --> 0:02:31.160 relatively low edge low expected return in all of them. 0:02:31.639 --> 0:02:36.360 And so that's its definition, but it's not quite I 0:02:36.360 --> 0:02:39.480 think complete enough at this point, because you can also 0:02:39.520 --> 0:02:44.560 be a quantitative investor trading a relatively narrow cross section 0:02:44.680 --> 0:02:51.240 of assets but with high high frequency. Right, So What 0:02:51.360 --> 0:02:55.040 matters really is the number of bets in a sense 0:02:55.080 --> 0:02:57.600 that you are going to take right. So I think 0:02:57.639 --> 0:03:00.200 that probably is if you have a large number of 0:03:00.240 --> 0:03:04.200 independent bets or quasi independent bets, this means that you 0:03:04.280 --> 0:03:07.440 need to be able to scale your method to a 0:03:07.520 --> 0:03:11.200 large number of independent bets, and this means that you 0:03:11.400 --> 0:03:13.399 are in some way a quantitative investor. 0:03:13.800 --> 0:03:16.440 Speaking of roles and jobs, what do you Global head 0:03:16.520 --> 0:03:19.440 of quantitative research at Pelisney. What's your job? You've been 0:03:19.440 --> 0:03:22.440 there about six months. What does the job intel at 0:03:22.480 --> 0:03:25.120 a at a fund, at a firm like PALSNT. 0:03:25.360 --> 0:03:31.120 Okay, global head of quantitative research. Okay, So basically I 0:03:31.160 --> 0:03:36.080 am the head of quantitative research for equities, and maybe 0:03:36.080 --> 0:03:38.240 one day in the future I will do you know, 0:03:38.320 --> 0:03:43.720 some commodities or fixed income, but I'm perfectly happy to 0:03:43.920 --> 0:03:48.480 serve equities, you know, both discretionary and systematic. What we 0:03:48.600 --> 0:03:51.400 do is I mean my group mostly, I mean I 0:03:51.440 --> 0:03:53.840 am in meetings, so I don't do any work. So 0:03:54.040 --> 0:03:59.440 we in a sense provide centralized quantitative services for the firm. 0:03:59.600 --> 0:04:04.480 So the first backbone thing that we do is you 0:04:04.600 --> 0:04:10.200 develop factor models wherever you can, right, so for equities 0:04:10.440 --> 0:04:14.440 at different horizons. Ideally you would like to develop them 0:04:14.480 --> 0:04:17.280 for other asset classes. But you know, factor models are 0:04:17.680 --> 0:04:22.240 the backbone of a lot of quantity investing nowadays. And 0:04:22.360 --> 0:04:26.159 then hedging at the firm level and at the individual 0:04:26.160 --> 0:04:30.039 PM levels, which is apparently very simple, but actually it's 0:04:30.120 --> 0:04:32.680 very deep as a problem. And then we do portfolio 0:04:32.720 --> 0:04:36.520 advisory services, which is basically you go two pms. You 0:04:36.600 --> 0:04:41.120 help them construct better portfolios, You help them understand their performance, 0:04:41.120 --> 0:04:44.800 which is extremely important, manage their risk, manage their drawdown 0:04:45.360 --> 0:04:49.440 on occasion, be their therapists. But this is what we do. 0:04:49.720 --> 0:04:51.919 I know you're in meetings all day, but you know, 0:04:52.000 --> 0:04:54.640 if you were someone on your team, how would you 0:04:54.680 --> 0:04:57.960 be coming up with actual ideas for factors. I hear 0:04:58.000 --> 0:05:01.360 people who sometimes come up with ideas for all thoughts episodes. 0:05:01.560 --> 0:05:05.440 Some of them have even turned out reasonably okay, But 0:05:05.520 --> 0:05:07.760 how does idea generation work? You sit down, You're like, 0:05:07.760 --> 0:05:09.640 I need to come up with a new factor today. 0:05:10.279 --> 0:05:11.720 What are you doing? What are you looking at? 0:05:12.200 --> 0:05:16.440 Okay? I want to specify a little bit more what's 0:05:16.480 --> 0:05:21.679 a factor because otherwise gets a little bit too vague. 0:05:21.880 --> 0:05:24.920 So like, there are factors and factors, So there are 0:05:24.960 --> 0:05:28.440 some factors that are real factors and what are those? 0:05:28.560 --> 0:05:34.200 Those are essentially attributes of some kind that you can 0:05:34.440 --> 0:05:42.279 assign to your investable universe. And there are sources of 0:05:42.760 --> 0:05:49.320 returns that affect the individual securities through this characteristic, and 0:05:49.920 --> 0:05:53.560 they are pervasive. So every asset is in some form 0:05:54.000 --> 0:05:58.599 affected by the systematic source of return number one, So 0:05:58.839 --> 0:06:02.160 they've got to be pervasive. The second thing is they 0:06:02.200 --> 0:06:06.119 got to be persistent, right, So it's not the case 0:06:06.200 --> 0:06:09.719 that I have a lot of factor returns for two 0:06:09.760 --> 0:06:13.040 months and then nothing for ten ten months, right, So 0:06:13.080 --> 0:06:17.160 that's not really a factor. And then possibly the third 0:06:17.240 --> 0:06:20.599 characteristic is that they have to be interesting, so they 0:06:20.680 --> 0:06:24.960 have to be in some way vaguely interpretable. So when 0:06:25.000 --> 0:06:29.760 you you know you match these requirements, it's a factor. Now, 0:06:30.880 --> 0:06:35.760 now imagine that you have the Trump factor. Let's see 0:06:35.960 --> 0:06:39.919 if Trump wins, a few stocks will definitely benefit, a 0:06:39.920 --> 0:06:43.599 few stocks will definitely not benefit from the election of 0:06:43.720 --> 0:06:50.320 Trump versus Kamala Harris. Another source could be well tariffs, right, 0:06:51.200 --> 0:06:54.920 Another source could be AI. Okay, AI definitely right. Doesn't 0:06:54.920 --> 0:06:58.919 fit the characteristic of being pervasive because there is a 0:06:58.960 --> 0:07:03.320 relatively small universe that's affected by the AI theme is 0:07:03.640 --> 0:07:07.640 likely not to go not going to be persistent. So 0:07:07.720 --> 0:07:10.400 it wasn't here like a few years ago, and it 0:07:10.560 --> 0:07:13.680 will probably not be here in five years because everything 0:07:13.720 --> 0:07:17.640 will be to some extent. AI it's interesting, but that's 0:07:17.640 --> 0:07:19.880 a theme, it's not a factor. That's what I would 0:07:19.920 --> 0:07:24.920 call a theme. And there are also some mathematical characteristic 0:07:25.120 --> 0:07:29.320 of a factor versus a theme, which so basically you 0:07:29.360 --> 0:07:33.960 can create a portfolio that tracks a factor, and this 0:07:34.080 --> 0:07:39.920 portfolio will have a relatively small idiosyncratic risk, so it 0:07:39.960 --> 0:07:43.920 will be truly a reproduction of the systematic source of 0:07:44.000 --> 0:07:48.240 return that you were observing through the assets. So imagine 0:07:48.280 --> 0:07:52.840 that this systematic source exists, but you do not observe 0:07:52.880 --> 0:07:55.840 it directly. It's latent, it's out there, but you can 0:07:55.880 --> 0:08:00.400 actually reconstruct it with a portfolio. A theme is let's 0:08:00.440 --> 0:08:04.880 say ten assets, you cannot really reconstruct it the same 0:08:04.920 --> 0:08:09.000 way because ten assets are just too few to diversify 0:08:09.040 --> 0:08:13.680 away the idiosyncratic source of returns of the individual assets. 0:08:13.920 --> 0:08:17.440 So when you're like thinking about factor identification, how much 0:08:18.480 --> 0:08:21.440 of the money that you make the actual returns come 0:08:21.520 --> 0:08:25.760 from essentially factor identification or being able to identify a 0:08:25.840 --> 0:08:31.320 factor before other measure identify a factor that exists before 0:08:31.520 --> 0:08:33.480 other competitors out there in the market. 0:08:33.800 --> 0:08:36.840 Okay, that's a great question because I I think I 0:08:36.880 --> 0:08:40.440 know the answer. Okay, great, But the reality is this, 0:08:40.760 --> 0:08:44.400 I think you know somebody else's factor is my alpha, 0:08:44.440 --> 0:08:47.040 and vice versa. Right, So say more, there are well 0:08:47.080 --> 0:08:54.400 known factors, let's say, some variety of value and momentum 0:08:54.600 --> 0:08:58.560 or reversion, and you can bet on those and you 0:08:58.640 --> 0:09:03.360 diversify away everything else, and what you get is, basically 0:09:03.360 --> 0:09:05.840 you get some returns that are priced priced in the 0:09:05.880 --> 0:09:09.400 sense that, as you know, you pay basically some risk 0:09:09.600 --> 0:09:12.720 for that. Right, So this is priced return and that's great. 0:09:13.679 --> 0:09:17.000 But once upon a time like these were non not 0:09:17.400 --> 0:09:20.840 public knowledge. If you were lucky enough to be a 0:09:20.840 --> 0:09:23.959 hedge fund in the eighties, and I've met a few, 0:09:24.679 --> 0:09:27.480 you know, and you were maybe also investing in Europe, 0:09:28.520 --> 0:09:32.720 these factors were really working very well, and they were alpha. 0:09:32.760 --> 0:09:36.200 They were not called factors. You know. The first I 0:09:36.200 --> 0:09:40.880 think published paper is probably eighty nine for momentum. Right now, 0:09:41.280 --> 0:09:45.000 there is alpha, and alpha is basically ideally would be 0:09:45.040 --> 0:09:48.400 a return that has no asocidate risk to it. It 0:09:48.559 --> 0:09:53.679 hardly ever exists. So what you really have are factors 0:09:54.720 --> 0:09:58.560 that exist at some frequency or in some universe, or 0:09:58.640 --> 0:10:02.640 with some characteristic that nobody else has found yet, and 0:10:02.679 --> 0:10:04.160 so they can be exploited. 0:10:04.240 --> 0:10:07.320 More, how do you make sure that when you're isolating 0:10:07.320 --> 0:10:11.840 a particular factor, you're not accidentally taking into accounts some 0:10:11.920 --> 0:10:14.880 other dynamics. So, you know, maybe you want to invest 0:10:14.920 --> 0:10:17.800 in a bunch of companies with like pricing power during 0:10:17.840 --> 0:10:21.640 the tariffs, but actually your cohort of companies ends up 0:10:21.840 --> 0:10:24.400 just looking like a bunch of big tech companies or 0:10:24.440 --> 0:10:25.160 something like that. 0:10:26.800 --> 0:10:30.600 The short answer without explanation, is that you can. But 0:10:30.880 --> 0:10:34.000 the long answer is a little bit more involved. If 0:10:34.040 --> 0:10:39.480 you have true characteristics, like I don't know, a tariff 0:10:39.760 --> 0:10:45.240 and a tech classification that are one hundred percent correlated, 0:10:45.480 --> 0:10:48.360 well then you really have only one. You don't need both, right, 0:10:48.520 --> 0:10:52.920 So okay, But if I have in my let's say, 0:10:53.040 --> 0:10:56.880 arsenal of factors, if I have multiple factors they're somewhat 0:10:57.040 --> 0:11:01.760 overlapping but not completely overlapping, then you can build a 0:11:01.840 --> 0:11:06.320 portfolio that separates the impact of one from the other. 0:11:06.480 --> 0:11:08.000 So you try to isolate you can you. 0:11:07.880 --> 0:11:10.600 Can isolate them, you can kind of purify them. Now 0:11:10.720 --> 0:11:15.200 there is also the scenario where there are factors that 0:11:15.320 --> 0:11:19.240 are not in the model and they should be and 0:11:19.240 --> 0:11:22.880 and basically those they complicate the picture a little bit. 0:11:23.679 --> 0:11:27.559 But otherwise, if you have a reasonable model, you are 0:11:27.800 --> 0:11:30.280 you're going to be able to separate them to understand 0:11:31.320 --> 0:11:36.160 what's the relationship. You can create a portfolio that exploits 0:11:36.200 --> 0:11:38.560 the first one, and then create a second portfolio that 0:11:38.679 --> 0:11:47.719 is uncorredly to the first one that exploits the second one. 0:11:59.200 --> 0:12:01.319 Just zooming out for a second again. And this sort 0:12:01.320 --> 0:12:04.000 of relates to Tracy's first question, but also, I guess 0:12:04.000 --> 0:12:06.840 relates to my first question. If you have a fund 0:12:07.360 --> 0:12:10.360 and it has various pms and analysts in there, is 0:12:10.400 --> 0:12:14.599 there a difference between quant at your level, which is 0:12:14.640 --> 0:12:18.400 at the fund level, versus say a POD or a 0:12:18.400 --> 0:12:22.280 PM whose specialty is quant trading. And there are different 0:12:22.320 --> 0:12:25.520 definitions or different senses in which that term can apply. 0:12:26.040 --> 0:12:30.200 Yeah, the fact is that you know, quant is is 0:12:30.240 --> 0:12:35.920 a very very generic label nowadays. Yeah, so there are many, 0:12:36.400 --> 0:12:39.800 many quants and they do all sorts of very interesting jobs. 0:12:40.400 --> 0:12:44.040 Some of them are are just differentiated because they live 0:12:44.120 --> 0:12:50.120 in different constructs. So nowadays, in a platform, especially in 0:12:50.120 --> 0:12:54.800 a quantitative one, it's not impossible to see pods and 0:12:54.880 --> 0:12:59.840 center groups. Okay, so that's one distinction. So what's the 0:13:00.400 --> 0:13:03.840 what's the difference. In a pod, you typically have a 0:13:03.960 --> 0:13:07.120 siloed group. I'm probably not stating the obvious, but you 0:13:07.120 --> 0:13:10.440 know you have a silot group. They don't communicate with 0:13:10.520 --> 0:13:14.040 other pods. You want at the firm level to have 0:13:14.120 --> 0:13:17.920 independent sources of alphas, and their payout typically is a 0:13:17.960 --> 0:13:22.200 percentage of their P and L after costs. Okay, and 0:13:22.200 --> 0:13:25.480 then you're a quant in a pod. In a center group, 0:13:25.600 --> 0:13:30.560 typically you are part of a larger group and the 0:13:30.600 --> 0:13:37.880 group will hopefully have large capacity. So these have you know, 0:13:38.280 --> 0:13:43.320 a larger program, like a larger research program. Their compensation 0:13:43.559 --> 0:13:47.800 tends to be more discretionary. And that's a center group. 0:13:48.880 --> 0:13:51.120 Then you have all sorts of other quants. So you 0:13:51.200 --> 0:13:55.200 have people like me who serve the firm at the 0:13:55.240 --> 0:14:00.400 center level. I also serve the leadership of the firm. 0:14:00.840 --> 0:14:04.600 And then you have people doing who are doing, for example, 0:14:04.920 --> 0:14:09.960 execution research. She's extremely extremely complex and interesting, right, So 0:14:10.920 --> 0:14:13.840 it's not black and white like you can do execution 0:14:13.960 --> 0:14:18.400 research and be responsible for some P and L. It's 0:14:18.600 --> 0:14:21.840 very very very rich nowadays and very specialized. 0:14:22.600 --> 0:14:24.840 I was actually going to ask about execution because when 0:14:24.840 --> 0:14:26.880 we're talking about quant investing, I think a lot of 0:14:26.960 --> 0:14:30.560 questions are around factors and idea generation. But you have 0:14:30.640 --> 0:14:35.360 all the I would assume boring stuff like liquidity trading 0:14:35.440 --> 0:14:38.320 costs that you also have to think about how do 0:14:38.320 --> 0:14:41.239 you actually incorporate those into your strategies. 0:14:43.240 --> 0:14:44.960 So you can do it in a variety of ways. 0:14:45.040 --> 0:14:48.800 It depends first of all, what position the firm occupies 0:14:48.840 --> 0:14:51.920 in the ecosystem. So if you are a high frequency 0:14:52.600 --> 0:14:58.520 trading company, most likely you are using your own capital 0:14:58.600 --> 0:15:02.520 because you are capacity constraint, so you know you don't 0:15:02.560 --> 0:15:08.440 need a lot of capital. So those firms exploit market 0:15:08.480 --> 0:15:13.840 microstructure level information. Okay, so in a sense, a high 0:15:13.880 --> 0:15:20.240 frequency trading firm does not have a market impact model 0:15:20.240 --> 0:15:23.560 in the traditional sense. They don't see parent orders, right, 0:15:23.680 --> 0:15:28.520 they execute at the microscopic level. If you are a 0:15:28.600 --> 0:15:32.800 hedge fund, typically you trade a lot, you have your 0:15:32.840 --> 0:15:37.520 own data set of orders. These data sets differ a lot, 0:15:37.680 --> 0:15:40.560 so you could have a market impact model for a 0:15:40.640 --> 0:15:45.520 quantitative trading group or a strategy, and you could have 0:15:45.560 --> 0:15:48.240 a different market impact model for hedging and a different 0:15:48.520 --> 0:15:51.520 market impact model for fundamental investing. And then what you 0:15:51.640 --> 0:15:56.520 get is basically a term function that you place in 0:15:56.560 --> 0:16:02.200 your optimization problem that hopefully help to size the portfolio 0:16:02.480 --> 0:16:06.080 or trade the portfolio optimally. And this is extremely important. 0:16:06.560 --> 0:16:10.760 Uh you know, market impact is it is a very 0:16:10.880 --> 0:16:16.160 very sizeable uh fraction of the lost P and L 0:16:17.200 --> 0:16:19.840 of of a firm. 0:16:20.080 --> 0:16:24.440 What as of today, what value is there in your 0:16:24.520 --> 0:16:30.600 world of specifically generative AI, l O, MS, et cetera. 0:16:30.720 --> 0:16:34.000 What how do you how do you currently or not 0:16:34.160 --> 0:16:39.560 currently get actual value out of them? 0:16:39.720 --> 0:16:44.320 Okay, so on this I have really relatively little to say. 0:16:44.320 --> 0:16:45.480 That's that's original. 0:16:45.560 --> 0:16:48.520 But tell us everything your employer is doing with AI. 0:16:48.920 --> 0:16:54.640 Yes, that'll send you the resum, thank you, But I think, okay, 0:16:55.080 --> 0:16:57.960 just let's recap the basics. Right, So the basics are, 0:16:58.440 --> 0:17:01.360 at least for the time being, everybody is trying to 0:17:01.360 --> 0:17:05.600 be more productive with AI. Right, So you want to 0:17:05.640 --> 0:17:09.199 have all your documents you want to have now you 0:17:09.240 --> 0:17:14.000 know what? Perplexity has a finance module. I think one 0:17:14.080 --> 0:17:19.359 day soon maybe Bloomberg will not have the keywords any longer. 0:17:19.400 --> 0:17:23.760 You just give you know, Bloomberg a task and it 0:17:23.880 --> 0:17:26.840 will grab all the pieces of information and hand it 0:17:26.880 --> 0:17:29.560 over to you and maybe you can schedule it. All 0:17:29.600 --> 0:17:34.760 of this is relatively table stakes. I mean, the the 0:17:34.880 --> 0:17:38.200 agentic aspect is not yet, but it will become pretty soon. 0:17:38.320 --> 0:17:40.919 I think it's going to be very hard to compute 0:17:40.960 --> 0:17:46.240 with the likes of maybe Bloomberg, but for sure, let's 0:17:46.280 --> 0:17:52.400 say you know the big hyperscalers. So that's one. At 0:17:52.440 --> 0:17:57.240 the investment level, it's it's much more complicated. So in 0:17:58.240 --> 0:18:03.399 strategies where there is a natural richness in data, you 0:18:03.440 --> 0:18:08.160 can definitely use if not deep learning, but you or AI, 0:18:08.320 --> 0:18:13.160 but you can definitely use very advanced machine learning algorithms 0:18:13.480 --> 0:18:16.560 and you do not have a data snoopin problem, you 0:18:16.560 --> 0:18:19.280 do not have a back testing problem, and so you 0:18:19.359 --> 0:18:21.600 are in a data rich environment and you can do that. 0:18:22.280 --> 0:18:25.879 And it's not a secret that, for example, XTX has 0:18:26.160 --> 0:18:31.320 a very large on premu you know number of in 0:18:31.400 --> 0:18:33.800 Nvidia cards I don't remember h one hundreds or something 0:18:33.840 --> 0:18:37.800 like that. So that's one thing, right. The question is 0:18:37.840 --> 0:18:42.520 really what's going to happen to the slower investment styles. 0:18:42.840 --> 0:18:49.600 And my view is that hopefully large firms like mine 0:18:49.600 --> 0:18:52.680 will have an advantage. But will see right why because 0:18:52.680 --> 0:18:55.120 we do have we do have the scale. We have 0:18:55.760 --> 0:18:58.119 a large number of pms, We have a lot of 0:18:58.240 --> 0:19:00.560 historical data, we have a lot of propriety every data 0:19:01.080 --> 0:19:04.400 that nobody else has. So maybe that that will work out. 0:19:04.880 --> 0:19:07.919 But how to make it happen, I don't know because 0:19:08.040 --> 0:19:10.879 things are changing so fast. And also I'm, you know, 0:19:11.000 --> 0:19:13.879 relatively a tourist in the areas I'm trying to learn 0:19:13.920 --> 0:19:15.040 a little bit more about. 0:19:30.520 --> 0:19:33.040 You mentioned proprietary data, and this comes up a lot 0:19:33.040 --> 0:19:35.960 where people talk about, well, the competitive advantage nowadays really 0:19:36.080 --> 0:19:39.000 is that data set? I mean, is is that true? 0:19:39.040 --> 0:19:41.280 If I get something really cool and unique, I can 0:19:41.359 --> 0:19:44.280 automatically become I don't know, a billionaire trader, if I 0:19:44.280 --> 0:19:46.199 can figure out how to execute on it. Is that 0:19:46.240 --> 0:19:46.639 all there is? 0:19:46.840 --> 0:19:49.200 Maybe yes, I have very weak beliefs on this. I 0:19:49.200 --> 0:19:52.360 don't know. Maybe yes, we'll find out. 0:19:52.640 --> 0:19:55.640 Well, so where are people getting interesting data sets from? 0:19:56.560 --> 0:19:59.199 I mean, you get interesting data from observing human beings 0:19:59.680 --> 0:20:03.600 actually investing, and you don't get to see a great 0:20:03.760 --> 0:20:08.120 PM investing, but I do. That's that's the benefit. 0:20:08.960 --> 0:20:12.800 So from your central position, you just get to see 0:20:12.840 --> 0:20:14.840 a lot of activity and you get to see novel 0:20:14.960 --> 0:20:17.840 data that other people don't get to see simply by 0:20:17.880 --> 0:20:20.560 being in the center of all of these different trades 0:20:20.600 --> 0:20:22.879 and everything, and that gives you a sort of higher 0:20:22.880 --> 0:20:25.600 abstraction layer or whatever it is that someone else in 0:20:25.640 --> 0:20:26.439 the market doesn't have. 0:20:26.680 --> 0:20:31.680 Yeah, and it's possible that not in the distant future, 0:20:32.320 --> 0:20:36.840 good pms will become good because they can improve on 0:20:37.000 --> 0:20:41.919 themselves by basically playing or training or having a baseline 0:20:42.080 --> 0:20:46.040 of an agent that reproduces their behavior. So you know 0:20:46.080 --> 0:20:48.760 there is an alter gapy, well ano a PM, but 0:20:48.800 --> 0:20:52.400 an alter whatever who says what would you do right? 0:20:52.480 --> 0:20:54.680 And you get a baseline behavior and then you can 0:20:54.680 --> 0:20:56.920 think about it and you could say, well, I would 0:20:56.960 --> 0:21:00.600 do something different, and then that becomes an example in 0:21:00.640 --> 0:21:05.320 a reinforcement learning process where the AI keeps learning from 0:21:05.359 --> 0:21:08.760 you and you keep improving because the baseline is changing. 0:21:09.520 --> 0:21:12.320 So before we came out here, I asked perplexity to 0:21:12.359 --> 0:21:14.919 come up with a new factor, and it came up 0:21:14.960 --> 0:21:18.920 with something called the policy agility factor, which is supposed 0:21:18.960 --> 0:21:23.520 to be that countries that display policy flexibility have better 0:21:23.640 --> 0:21:27.040 outperformance over the longer term. Countries that are able to 0:21:27.320 --> 0:21:31.680 more quickly adapt to changing situations are outperformers over the 0:21:31.720 --> 0:21:34.800 long run. Can you grade that factor. I didn't do 0:21:34.840 --> 0:21:37.800 a back tust. But like if someone brought you an 0:21:37.800 --> 0:21:42.560 idea like that, not me, perplexity, I don't want you 0:21:42.600 --> 0:21:45.440 to insult me over the next five minutes. What would 0:21:45.520 --> 0:21:47.920 you say to them? What are the problems with this? 0:21:48.720 --> 0:21:53.639 I mean no major problems, there are questions. So the 0:21:53.680 --> 0:21:56.000 first thing that you want to make sure is that 0:21:56.800 --> 0:22:01.480 if AI whatever it means, brings to you definition, right, 0:22:02.040 --> 0:22:06.520 that definition should be at a point in time and 0:22:06.600 --> 0:22:09.719 should not be trained on all the on all the 0:22:09.720 --> 0:22:11.960 past data, right. So number one, you want to do 0:22:12.000 --> 0:22:17.800 that because if you back test that feature and in 0:22:17.840 --> 0:22:22.119 a way perplexity has already tested it, it's not a 0:22:22.160 --> 0:22:24.760 fair play. You know, the performance will the back test 0:22:24.800 --> 0:22:29.800 will look great. So, unfortunately, we live in a world 0:22:30.000 --> 0:22:35.600 where some factors will never be back testable. So you 0:22:35.640 --> 0:22:39.160 don't know whether they work or they don't work, right, 0:22:39.280 --> 0:22:42.720 You just know that you cannot test them in advance, 0:22:42.800 --> 0:22:45.880 like a policy agility. This seems to be a very 0:22:45.920 --> 0:22:49.119 low turnover factor, right, and it seems to be probably 0:22:49.200 --> 0:22:49.960 a very low. 0:22:49.800 --> 0:22:52.320 Sharp factor in a low universe and. 0:22:52.320 --> 0:22:56.080 A small universe. So how do you how do you know? Well, 0:22:56.119 --> 0:22:58.080 probably you want to make sure that it makes sense, 0:22:58.119 --> 0:23:02.520 and maybe you can start putting a small volativity allocation 0:23:02.640 --> 0:23:04.560 to it and. 0:23:04.520 --> 0:23:05.800 Then you would build it up as you. 0:23:05.720 --> 0:23:08.240 Watched it out for yes, out of sample. 0:23:08.520 --> 0:23:10.159 Okay, so speaking of back to us, I have one 0:23:10.160 --> 0:23:14.000 more question. But it seems like quant investing. Part of 0:23:14.000 --> 0:23:16.640 the issue with this is you are looking back at 0:23:16.680 --> 0:23:18.960 historical data. That's all you have. You don't have data 0:23:18.960 --> 0:23:22.119 about the future. Unfortunately, it strikes me as hard to 0:23:22.160 --> 0:23:24.760 deal with regime changes. So when you have a big 0:23:24.880 --> 0:23:29.760 break in how something works in finance or markets or 0:23:29.800 --> 0:23:32.960 the global economy, how does quant investing actually take into 0:23:32.960 --> 0:23:36.159 account those sorts of risks, Like say, you know, a 0:23:36.160 --> 0:23:38.600 lot of investing is based on the idea that bonds 0:23:38.600 --> 0:23:40.680 and stocks are going to move inversely to each other, 0:23:40.720 --> 0:23:44.040 and then in twenty twenty two they started moving together. 0:23:46.760 --> 0:23:51.040 I think that most people with a quantitative background in 0:23:51.080 --> 0:23:55.880 finance will tell you that regime change is very difficult 0:23:55.920 --> 0:23:59.119 to detect and to act on in an effective manner. 0:23:59.200 --> 0:24:02.400 So I think that's been my experience at least, right 0:24:02.440 --> 0:24:07.280 so in every possible application I've tried, and you know, 0:24:07.359 --> 0:24:09.560 it never, it really never works for me. Maybe it 0:24:09.600 --> 0:24:12.480 works for somebody else. What I think it's a bit 0:24:12.560 --> 0:24:17.280 easier to do is to detect regime change in a 0:24:17.359 --> 0:24:22.560 human being. So instead of trying to use you know, 0:24:22.640 --> 0:24:25.399 there are many many algorithms for regime change. You know, 0:24:25.440 --> 0:24:31.880 there are MARKT based q sum, completely non parametric. Instead 0:24:31.880 --> 0:24:36.040 of trying to act on regime changes in the environment, 0:24:36.359 --> 0:24:41.240 try to detect changes in the behavior of a portfolio 0:24:41.240 --> 0:24:45.160 manager and act on that because that works, I think, 0:24:45.840 --> 0:24:50.800 and usually you know, jives with experience with so that 0:24:50.800 --> 0:24:52.359 that is something that can be exploited. 0:24:52.600 --> 0:24:54.320 I want to go back to an answer you gave 0:24:54.400 --> 0:24:56.920 early on, which is sort of like the old school 0:24:57.119 --> 0:25:00.000 factor investing and like the original versions and maybe they 0:25:00.160 --> 0:25:03.040 sort of an international factor or a liquidity factor, or 0:25:03.040 --> 0:25:06.200 the small cap factor, the value factor. And it feels 0:25:06.240 --> 0:25:09.640 like a lot of these things haven't worked in ages, 0:25:10.119 --> 0:25:12.280 and there's this debate that seems like, Okay, is this 0:25:13.320 --> 0:25:17.320 the long cycle and eventually it's gonna come back, or 0:25:17.840 --> 0:25:20.720 is it that everybody not only knows about these factors 0:25:20.720 --> 0:25:24.280 that have discussed them to death, they're also extremely commodified 0:25:24.400 --> 0:25:26.400 in the sense that you could just buy an ETF 0:25:26.400 --> 0:25:28.280 of them, right, You could just buy a small cap ETF. 0:25:28.320 --> 0:25:31.800 It's trivial to execute. You could just buy a momentum ETF. 0:25:31.840 --> 0:25:34.680 It's trivial to execute a value ETF, et cetera. Like 0:25:34.960 --> 0:25:38.440 my intuition would be, since everyone knows about them and 0:25:38.480 --> 0:25:42.480 they're completely commodified technologically, they're just gone. But there is 0:25:42.520 --> 0:25:44.439 still debates. Some people think it's totally a matter of 0:25:44.440 --> 0:25:46.880 time before these come back in vogue, and that it's 0:25:46.880 --> 0:25:49.440 the long cycle, et cetera. I'm curious how you think 0:25:49.440 --> 0:25:51.800 about some of the original factors that people discussed in 0:25:51.880 --> 0:25:53.040 their prospects going forward. 0:25:53.760 --> 0:25:59.040 Well, so some factors were identified, but then somehow they 0:25:59.080 --> 0:26:04.879 got demoted so famously. Size, right, so conditional on having 0:26:05.040 --> 0:26:10.640 other characteristics of a stock. Size doesn't really explain much 0:26:10.640 --> 0:26:15.240 of your returns, and so it's a combination of other factors. Okay, 0:26:15.320 --> 0:26:21.040 well that's one case. Then there are cases where it 0:26:21.160 --> 0:26:25.560 seems that some factors have been exploited. You know, their 0:26:25.600 --> 0:26:30.920