1 00:00:05,160 --> 00:00:19,119 Speaker 1: Bloomberg Audio Studios, Podcasts, Radio News. 2 00:00:20,760 --> 00:00:24,280 Speaker 2: Hello and welcome to another episode of The Odd Laws podcast. 3 00:00:24,360 --> 00:00:28,760 Speaker 2: I'm Jill Wisenthal, normally joined by my co host Tracy Alloway, 4 00:00:28,760 --> 00:00:31,640 Speaker 2: but she's on vacation today, so it's just me in 5 00:00:31,680 --> 00:00:34,400 Speaker 2: this intro. But in today's episode you will hear a 6 00:00:34,479 --> 00:00:40,080 Speaker 2: conversation taped live at Bloomberg's Reimagining Information form. On June twelfth, 7 00:00:40,280 --> 00:00:43,600 Speaker 2: we spoke with Gappi Pallioligo, global head of quantitative Research 8 00:00:43,640 --> 00:00:46,879 Speaker 2: at Ballysny Asset Management. He has a new book out, 9 00:00:46,920 --> 00:00:50,440 Speaker 2: it's called The Elements of Quantitative Investing. Neither of us 10 00:00:50,520 --> 00:00:52,599 Speaker 2: have read it because it would go way over our 11 00:00:52,640 --> 00:00:55,000 Speaker 2: heads because we're not quant so we don't know how 12 00:00:55,000 --> 00:00:57,880 Speaker 2: to read that stuff. But Gappy is great at explaining 13 00:00:57,920 --> 00:01:00,480 Speaker 2: all of this stuff in clear English. So we had 14 00:01:00,480 --> 00:01:03,040 Speaker 2: a great conversation and we hope you enjoy listening to it. 15 00:01:03,160 --> 00:01:04,720 Speaker 3: So just to begin, I'm going to start with a 16 00:01:04,760 --> 00:01:08,680 Speaker 3: really really dumb question, possibly, but isn't all investing quant 17 00:01:08,760 --> 00:01:13,160 Speaker 3: investing nowadays? I mean every investor has access to some 18 00:01:13,319 --> 00:01:15,240 Speaker 3: form of quantitative or using numbers. 19 00:01:15,319 --> 00:01:20,880 Speaker 4: Yeah, I guess yes. End of answer. Yeah, I think so, 20 00:01:21,000 --> 00:01:23,480 Speaker 4: I think so I mean pretty much everybody uses some 21 00:01:23,600 --> 00:01:26,759 Speaker 4: kind of quantitative overlay, right, but two different degrees. So 22 00:01:26,840 --> 00:01:29,440 Speaker 4: I have a friend who worked for one of the 23 00:01:29,600 --> 00:01:35,039 Speaker 4: Tiger cubs, and they they refused to use sharp They 24 00:01:35,160 --> 00:01:39,200 Speaker 4: refused to use logs in a spreadsheet because they said 25 00:01:39,200 --> 00:01:41,760 Speaker 4: that they were dangerous. Probably they took the log of 26 00:01:41,800 --> 00:01:45,000 Speaker 4: a negative number. And so yeah, no, two different degrees. 27 00:01:45,040 --> 00:01:48,920 Speaker 4: But yes, there is some quantitative culture seeping through. 28 00:01:49,200 --> 00:01:53,000 Speaker 3: Okay, so what defines quantitative investing? How would you differentiate 29 00:01:53,040 --> 00:01:56,280 Speaker 3: that from I don't know, value investing, discretionary investing. 30 00:01:56,920 --> 00:02:00,840 Speaker 4: Okay, I think that there are several possible answers. So 31 00:02:01,440 --> 00:02:05,760 Speaker 4: I'm going to go with the one answer that I 32 00:02:05,840 --> 00:02:10,560 Speaker 4: read in my life as a quant I think it's 33 00:02:10,600 --> 00:02:13,200 Speaker 4: a wily book. It's a very good book, by the way, 34 00:02:13,560 --> 00:02:21,040 Speaker 4: And I think Cliff Asnes defined quantitative investing as basically 35 00:02:21,120 --> 00:02:26,600 Speaker 4: investing in a large cross section of assets, having a 36 00:02:26,680 --> 00:02:31,160 Speaker 4: relatively low edge low expected return in all of them. 37 00:02:31,639 --> 00:02:36,360 Speaker 4: And so that's its definition, but it's not quite I 38 00:02:36,360 --> 00:02:39,480 Speaker 4: think complete enough at this point, because you can also 39 00:02:39,520 --> 00:02:44,560 Speaker 4: be a quantitative investor trading a relatively narrow cross section 40 00:02:44,680 --> 00:02:51,240 Speaker 4: of assets but with high high frequency. Right, So What 41 00:02:51,360 --> 00:02:55,040 Speaker 4: matters really is the number of bets in a sense 42 00:02:55,080 --> 00:02:57,600 Speaker 4: that you are going to take right. So I think 43 00:02:57,639 --> 00:03:00,200 Speaker 4: that probably is if you have a large number of 44 00:03:00,240 --> 00:03:04,200 Speaker 4: independent bets or quasi independent bets, this means that you 45 00:03:04,280 --> 00:03:07,440 Speaker 4: need to be able to scale your method to a 46 00:03:07,520 --> 00:03:11,200 Speaker 4: large number of independent bets, and this means that you 47 00:03:11,400 --> 00:03:13,399 Speaker 4: are in some way a quantitative investor. 48 00:03:13,800 --> 00:03:16,440 Speaker 2: Speaking of roles and jobs, what do you Global head 49 00:03:16,520 --> 00:03:19,440 Speaker 2: of quantitative research at Pelisney. What's your job? You've been 50 00:03:19,440 --> 00:03:22,440 Speaker 2: there about six months. What does the job intel at 51 00:03:22,480 --> 00:03:25,120 Speaker 2: a at a fund, at a firm like PALSNT. 52 00:03:25,360 --> 00:03:31,120 Speaker 4: Okay, global head of quantitative research. Okay, So basically I 53 00:03:31,160 --> 00:03:36,080 Speaker 4: am the head of quantitative research for equities, and maybe 54 00:03:36,080 --> 00:03:38,240 Speaker 4: one day in the future I will do you know, 55 00:03:38,320 --> 00:03:43,720 Speaker 4: some commodities or fixed income, but I'm perfectly happy to 56 00:03:43,920 --> 00:03:48,480 Speaker 4: serve equities, you know, both discretionary and systematic. What we 57 00:03:48,600 --> 00:03:51,400 Speaker 4: do is I mean my group mostly, I mean I 58 00:03:51,440 --> 00:03:53,840 Speaker 4: am in meetings, so I don't do any work. So 59 00:03:54,040 --> 00:03:59,440 Speaker 4: we in a sense provide centralized quantitative services for the firm. 60 00:03:59,600 --> 00:04:04,480 Speaker 4: So the first backbone thing that we do is you 61 00:04:04,600 --> 00:04:10,200 Speaker 4: develop factor models wherever you can, right, so for equities 62 00:04:10,440 --> 00:04:14,440 Speaker 4: at different horizons. Ideally you would like to develop them 63 00:04:14,480 --> 00:04:17,280 Speaker 4: for other asset classes. But you know, factor models are 64 00:04:17,680 --> 00:04:22,240 Speaker 4: the backbone of a lot of quantity investing nowadays. And 65 00:04:22,360 --> 00:04:26,159 Speaker 4: then hedging at the firm level and at the individual 66 00:04:26,160 --> 00:04:30,039 Speaker 4: PM levels, which is apparently very simple, but actually it's 67 00:04:30,120 --> 00:04:32,680 Speaker 4: very deep as a problem. And then we do portfolio 68 00:04:32,720 --> 00:04:36,520 Speaker 4: advisory services, which is basically you go two pms. You 69 00:04:36,600 --> 00:04:41,120 Speaker 4: help them construct better portfolios, You help them understand their performance, 70 00:04:41,120 --> 00:04:44,800 Speaker 4: which is extremely important, manage their risk, manage their drawdown 71 00:04:45,360 --> 00:04:49,440 Speaker 4: on occasion, be their therapists. But this is what we do. 72 00:04:49,720 --> 00:04:51,919 Speaker 3: I know you're in meetings all day, but you know, 73 00:04:52,000 --> 00:04:54,640 Speaker 3: if you were someone on your team, how would you 74 00:04:54,680 --> 00:04:57,960 Speaker 3: be coming up with actual ideas for factors. I hear 75 00:04:58,000 --> 00:05:01,360 Speaker 3: people who sometimes come up with ideas for all thoughts episodes. 76 00:05:01,560 --> 00:05:05,440 Speaker 3: Some of them have even turned out reasonably okay, But 77 00:05:05,520 --> 00:05:07,760 Speaker 3: how does idea generation work? You sit down, You're like, 78 00:05:07,760 --> 00:05:09,640 Speaker 3: I need to come up with a new factor today. 79 00:05:10,279 --> 00:05:11,720 Speaker 3: What are you doing? What are you looking at? 80 00:05:12,200 --> 00:05:16,440 Speaker 4: Okay? I want to specify a little bit more what's 81 00:05:16,480 --> 00:05:21,679 Speaker 4: a factor because otherwise gets a little bit too vague. 82 00:05:21,880 --> 00:05:24,920 Speaker 4: So like, there are factors and factors, So there are 83 00:05:24,960 --> 00:05:28,440 Speaker 4: some factors that are real factors and what are those? 84 00:05:28,560 --> 00:05:34,200 Speaker 4: Those are essentially attributes of some kind that you can 85 00:05:34,440 --> 00:05:42,279 Speaker 4: assign to your investable universe. And there are sources of 86 00:05:42,760 --> 00:05:49,320 Speaker 4: returns that affect the individual securities through this characteristic, and 87 00:05:49,920 --> 00:05:53,560 Speaker 4: they are pervasive. So every asset is in some form 88 00:05:54,000 --> 00:05:58,599 Speaker 4: affected by the systematic source of return number one, So 89 00:05:58,839 --> 00:06:02,160 Speaker 4: they've got to be pervasive. The second thing is they 90 00:06:02,200 --> 00:06:06,119 Speaker 4: got to be persistent, right, So it's not the case 91 00:06:06,200 --> 00:06:09,719 Speaker 4: that I have a lot of factor returns for two 92 00:06:09,760 --> 00:06:13,040 Speaker 4: months and then nothing for ten ten months, right, So 93 00:06:13,080 --> 00:06:17,160 Speaker 4: that's not really a factor. And then possibly the third 94 00:06:17,240 --> 00:06:20,599 Speaker 4: characteristic is that they have to be interesting, so they 95 00:06:20,680 --> 00:06:24,960 Speaker 4: have to be in some way vaguely interpretable. So when 96 00:06:25,000 --> 00:06:29,760 Speaker 4: you you know you match these requirements, it's a factor. Now, 97 00:06:30,880 --> 00:06:35,760 Speaker 4: now imagine that you have the Trump factor. Let's see 98 00:06:35,960 --> 00:06:39,919 Speaker 4: if Trump wins, a few stocks will definitely benefit, a 99 00:06:39,920 --> 00:06:43,599 Speaker 4: few stocks will definitely not benefit from the election of 100 00:06:43,720 --> 00:06:50,320 Speaker 4: Trump versus Kamala Harris. Another source could be well tariffs, right, 101 00:06:51,200 --> 00:06:54,920 Speaker 4: Another source could be AI. Okay, AI definitely right. Doesn't 102 00:06:54,920 --> 00:06:58,919 Speaker 4: fit the characteristic of being pervasive because there is a 103 00:06:58,960 --> 00:07:03,320 Speaker 4: relatively small universe that's affected by the AI theme is 104 00:07:03,640 --> 00:07:07,640 Speaker 4: likely not to go not going to be persistent. So 105 00:07:07,720 --> 00:07:10,400 Speaker 4: it wasn't here like a few years ago, and it 106 00:07:10,560 --> 00:07:13,680 Speaker 4: will probably not be here in five years because everything 107 00:07:13,720 --> 00:07:17,640 Speaker 4: will be to some extent. AI it's interesting, but that's 108 00:07:17,640 --> 00:07:19,880 Speaker 4: a theme, it's not a factor. That's what I would 109 00:07:19,920 --> 00:07:24,920 Speaker 4: call a theme. And there are also some mathematical characteristic 110 00:07:25,120 --> 00:07:29,320 Speaker 4: of a factor versus a theme, which so basically you 111 00:07:29,360 --> 00:07:33,960 Speaker 4: can create a portfolio that tracks a factor, and this 112 00:07:34,080 --> 00:07:39,920 Speaker 4: portfolio will have a relatively small idiosyncratic risk, so it 113 00:07:39,960 --> 00:07:43,920 Speaker 4: will be truly a reproduction of the systematic source of 114 00:07:44,000 --> 00:07:48,240 Speaker 4: return that you were observing through the assets. So imagine 115 00:07:48,280 --> 00:07:52,840 Speaker 4: that this systematic source exists, but you do not observe 116 00:07:52,880 --> 00:07:55,840 Speaker 4: it directly. It's latent, it's out there, but you can 117 00:07:55,880 --> 00:08:00,400 Speaker 4: actually reconstruct it with a portfolio. A theme is let's 118 00:08:00,440 --> 00:08:04,880 Speaker 4: say ten assets, you cannot really reconstruct it the same 119 00:08:04,920 --> 00:08:09,000 Speaker 4: way because ten assets are just too few to diversify 120 00:08:09,040 --> 00:08:13,680 Speaker 4: away the idiosyncratic source of returns of the individual assets. 121 00:08:13,920 --> 00:08:17,440 Speaker 2: So when you're like thinking about factor identification, how much 122 00:08:18,480 --> 00:08:21,440 Speaker 2: of the money that you make the actual returns come 123 00:08:21,520 --> 00:08:25,760 Speaker 2: from essentially factor identification or being able to identify a 124 00:08:25,840 --> 00:08:31,320 Speaker 2: factor before other measure identify a factor that exists before 125 00:08:31,520 --> 00:08:33,480 Speaker 2: other competitors out there in the market. 126 00:08:33,800 --> 00:08:36,840 Speaker 4: Okay, that's a great question because I I think I 127 00:08:36,880 --> 00:08:40,440 Speaker 4: know the answer. Okay, great, But the reality is this, 128 00:08:40,760 --> 00:08:44,400 Speaker 4: I think you know somebody else's factor is my alpha, 129 00:08:44,440 --> 00:08:47,040 Speaker 4: and vice versa. Right, So say more, there are well 130 00:08:47,080 --> 00:08:54,400 Speaker 4: known factors, let's say, some variety of value and momentum 131 00:08:54,600 --> 00:08:58,560 Speaker 4: or reversion, and you can bet on those and you 132 00:08:58,640 --> 00:09:03,360 Speaker 4: diversify away everything else, and what you get is, basically 133 00:09:03,360 --> 00:09:05,840 Speaker 4: you get some returns that are priced priced in the 134 00:09:05,880 --> 00:09:09,400 Speaker 4: sense that, as you know, you pay basically some risk 135 00:09:09,600 --> 00:09:12,720 Speaker 4: for that. Right, So this is priced return and that's great. 136 00:09:13,679 --> 00:09:17,000 Speaker 4: But once upon a time like these were non not 137 00:09:17,400 --> 00:09:20,840 Speaker 4: public knowledge. If you were lucky enough to be a 138 00:09:20,840 --> 00:09:23,959 Speaker 4: hedge fund in the eighties, and I've met a few, 139 00:09:24,679 --> 00:09:27,480 Speaker 4: you know, and you were maybe also investing in Europe, 140 00:09:28,520 --> 00:09:32,720 Speaker 4: these factors were really working very well, and they were alpha. 141 00:09:32,760 --> 00:09:36,200 Speaker 4: They were not called factors. You know. The first I 142 00:09:36,200 --> 00:09:40,880 Speaker 4: think published paper is probably eighty nine for momentum. Right now, 143 00:09:41,280 --> 00:09:45,000 Speaker 4: there is alpha, and alpha is basically ideally would be 144 00:09:45,040 --> 00:09:48,400 Speaker 4: a return that has no asocidate risk to it. It 145 00:09:48,559 --> 00:09:53,679 Speaker 4: hardly ever exists. So what you really have are factors 146 00:09:54,720 --> 00:09:58,560 Speaker 4: that exist at some frequency or in some universe, or 147 00:09:58,640 --> 00:10:02,640 Speaker 4: with some characteristic that nobody else has found yet, and 148 00:10:02,679 --> 00:10:04,160 Speaker 4: so they can be exploited. 149 00:10:04,240 --> 00:10:07,320 Speaker 3: More, how do you make sure that when you're isolating 150 00:10:07,320 --> 00:10:11,840 Speaker 3: a particular factor, you're not accidentally taking into accounts some 151 00:10:11,920 --> 00:10:14,880 Speaker 3: other dynamics. So, you know, maybe you want to invest 152 00:10:14,920 --> 00:10:17,800 Speaker 3: in a bunch of companies with like pricing power during 153 00:10:17,840 --> 00:10:21,640 Speaker 3: the tariffs, but actually your cohort of companies ends up 154 00:10:21,840 --> 00:10:24,400 Speaker 3: just looking like a bunch of big tech companies or 155 00:10:24,440 --> 00:10:25,160 Speaker 3: something like that. 156 00:10:26,800 --> 00:10:30,600 Speaker 4: The short answer without explanation, is that you can. But 157 00:10:30,880 --> 00:10:34,000 Speaker 4: the long answer is a little bit more involved. If 158 00:10:34,040 --> 00:10:39,480 Speaker 4: you have true characteristics, like I don't know, a tariff 159 00:10:39,760 --> 00:10:45,240 Speaker 4: and a tech classification that are one hundred percent correlated, 160 00:10:45,480 --> 00:10:48,360 Speaker 4: well then you really have only one. You don't need both, right, 161 00:10:48,520 --> 00:10:52,920 Speaker 4: So okay, But if I have in my let's say, 162 00:10:53,040 --> 00:10:56,880 Speaker 4: arsenal of factors, if I have multiple factors they're somewhat 163 00:10:57,040 --> 00:11:01,760 Speaker 4: overlapping but not completely overlapping, then you can build a 164 00:11:01,840 --> 00:11:06,320 Speaker 4: portfolio that separates the impact of one from the other. 165 00:11:06,480 --> 00:11:08,000 Speaker 3: So you try to isolate you can you. 166 00:11:07,880 --> 00:11:10,600 Speaker 4: Can isolate them, you can kind of purify them. Now 167 00:11:10,720 --> 00:11:15,200 Speaker 4: there is also the scenario where there are factors that 168 00:11:15,320 --> 00:11:19,240 Speaker 4: are not in the model and they should be and 169 00:11:19,240 --> 00:11:22,880 Speaker 4: and basically those they complicate the picture a little bit. 170 00:11:23,679 --> 00:11:27,559 Speaker 4: But otherwise, if you have a reasonable model, you are 171 00:11:27,800 --> 00:11:30,280 Speaker 4: you're going to be able to separate them to understand 172 00:11:31,320 --> 00:11:36,160 Speaker 4: what's the relationship. You can create a portfolio that exploits 173 00:11:36,200 --> 00:11:38,560 Speaker 4: the first one, and then create a second portfolio that 174 00:11:38,679 --> 00:11:47,719 Speaker 4: is uncorredly to the first one that exploits the second one. 175 00:11:59,200 --> 00:12:01,319 Speaker 2: Just zooming out for a second again. And this sort 176 00:12:01,320 --> 00:12:04,000 Speaker 2: of relates to Tracy's first question, but also, I guess 177 00:12:04,000 --> 00:12:06,840 Speaker 2: relates to my first question. If you have a fund 178 00:12:07,360 --> 00:12:10,360 Speaker 2: and it has various pms and analysts in there, is 179 00:12:10,400 --> 00:12:14,599 Speaker 2: there a difference between quant at your level, which is 180 00:12:14,640 --> 00:12:18,400 Speaker 2: at the fund level, versus say a POD or a 181 00:12:18,400 --> 00:12:22,280 Speaker 2: PM whose specialty is quant trading. And there are different 182 00:12:22,320 --> 00:12:25,520 Speaker 2: definitions or different senses in which that term can apply. 183 00:12:26,040 --> 00:12:30,200 Speaker 4: Yeah, the fact is that you know, quant is is 184 00:12:30,240 --> 00:12:35,920 Speaker 4: a very very generic label nowadays. Yeah, so there are many, 185 00:12:36,400 --> 00:12:39,800 Speaker 4: many quants and they do all sorts of very interesting jobs. 186 00:12:40,400 --> 00:12:44,040 Speaker 4: Some of them are are just differentiated because they live 187 00:12:44,120 --> 00:12:50,120 Speaker 4: in different constructs. So nowadays, in a platform, especially in 188 00:12:50,120 --> 00:12:54,800 Speaker 4: a quantitative one, it's not impossible to see pods and 189 00:12:54,880 --> 00:12:59,840 Speaker 4: center groups. Okay, so that's one distinction. So what's the 190 00:13:00,400 --> 00:13:03,840 Speaker 4: what's the difference. In a pod, you typically have a 191 00:13:03,960 --> 00:13:07,120 Speaker 4: siloed group. I'm probably not stating the obvious, but you 192 00:13:07,120 --> 00:13:10,440 Speaker 4: know you have a silot group. They don't communicate with 193 00:13:10,520 --> 00:13:14,040 Speaker 4: other pods. You want at the firm level to have 194 00:13:14,120 --> 00:13:17,920 Speaker 4: independent sources of alphas, and their payout typically is a 195 00:13:17,960 --> 00:13:22,200 Speaker 4: percentage of their P and L after costs. Okay, and 196 00:13:22,200 --> 00:13:25,480 Speaker 4: then you're a quant in a pod. In a center group, 197 00:13:25,600 --> 00:13:30,560 Speaker 4: typically you are part of a larger group and the 198 00:13:30,600 --> 00:13:37,880 Speaker 4: group will hopefully have large capacity. So these have you know, 199 00:13:38,280 --> 00:13:43,320 Speaker 4: a larger program, like a larger research program. Their compensation 200 00:13:43,559 --> 00:13:47,800 Speaker 4: tends to be more discretionary. And that's a center group. 201 00:13:48,880 --> 00:13:51,120 Speaker 4: Then you have all sorts of other quants. So you 202 00:13:51,200 --> 00:13:55,200 Speaker 4: have people like me who serve the firm at the 203 00:13:55,240 --> 00:14:00,400 Speaker 4: center level. I also serve the leadership of the firm. 204 00:14:00,840 --> 00:14:04,600 Speaker 4: And then you have people doing who are doing, for example, 205 00:14:04,920 --> 00:14:09,960 Speaker 4: execution research. She's extremely extremely complex and interesting, right, So 206 00:14:10,920 --> 00:14:13,840 Speaker 4: it's not black and white like you can do execution 207 00:14:13,960 --> 00:14:18,400 Speaker 4: research and be responsible for some P and L. It's 208 00:14:18,600 --> 00:14:21,840 Speaker 4: very very very rich nowadays and very specialized. 209 00:14:22,600 --> 00:14:24,840 Speaker 3: I was actually going to ask about execution because when 210 00:14:24,840 --> 00:14:26,880 Speaker 3: we're talking about quant investing, I think a lot of 211 00:14:26,960 --> 00:14:30,560 Speaker 3: questions are around factors and idea generation. But you have 212 00:14:30,640 --> 00:14:35,360 Speaker 3: all the I would assume boring stuff like liquidity trading 213 00:14:35,440 --> 00:14:38,320 Speaker 3: costs that you also have to think about how do 214 00:14:38,320 --> 00:14:41,239 Speaker 3: you actually incorporate those into your strategies. 215 00:14:43,240 --> 00:14:44,960 Speaker 4: So you can do it in a variety of ways. 216 00:14:45,040 --> 00:14:48,800 Speaker 4: It depends first of all, what position the firm occupies 217 00:14:48,840 --> 00:14:51,920 Speaker 4: in the ecosystem. So if you are a high frequency 218 00:14:52,600 --> 00:14:58,520 Speaker 4: trading company, most likely you are using your own capital 219 00:14:58,600 --> 00:15:02,520 Speaker 4: because you are capacity constraint, so you know you don't 220 00:15:02,560 --> 00:15:08,440 Speaker 4: need a lot of capital. So those firms exploit market 221 00:15:08,480 --> 00:15:13,840 Speaker 4: microstructure level information. Okay, so in a sense, a high 222 00:15:13,880 --> 00:15:20,240 Speaker 4: frequency trading firm does not have a market impact model 223 00:15:20,240 --> 00:15:23,560 Speaker 4: in the traditional sense. They don't see parent orders, right, 224 00:15:23,680 --> 00:15:28,520 Speaker 4: they execute at the microscopic level. If you are a 225 00:15:28,600 --> 00:15:32,800 Speaker 4: hedge fund, typically you trade a lot, you have your 226 00:15:32,840 --> 00:15:37,520 Speaker 4: own data set of orders. These data sets differ a lot, 227 00:15:37,680 --> 00:15:40,560 Speaker 4: so you could have a market impact model for a 228 00:15:40,640 --> 00:15:45,520 Speaker 4: quantitative trading group or a strategy, and you could have 229 00:15:45,560 --> 00:15:48,240 Speaker 4: a different market impact model for hedging and a different 230 00:15:48,520 --> 00:15:51,520 Speaker 4: market impact model for fundamental investing. And then what you 231 00:15:51,640 --> 00:15:56,520 Speaker 4: get is basically a term function that you place in 232 00:15:56,560 --> 00:16:02,200 Speaker 4: your optimization problem that hopefully help to size the portfolio 233 00:16:02,480 --> 00:16:06,080 Speaker 4: or trade the portfolio optimally. And this is extremely important. 234 00:16:06,560 --> 00:16:10,760 Speaker 4: Uh you know, market impact is it is a very 235 00:16:10,880 --> 00:16:16,160 Speaker 4: very sizeable uh fraction of the lost P and L 236 00:16:17,200 --> 00:16:19,840 Speaker 4: of of a firm. 237 00:16:20,080 --> 00:16:24,440 Speaker 2: What as of today, what value is there in your 238 00:16:24,520 --> 00:16:30,600 Speaker 2: world of specifically generative AI, l O, MS, et cetera. 239 00:16:30,720 --> 00:16:34,000 Speaker 2: What how do you how do you currently or not 240 00:16:34,160 --> 00:16:39,560 Speaker 2: currently get actual value out of them? 241 00:16:39,720 --> 00:16:44,320 Speaker 4: Okay, so on this I have really relatively little to say. 242 00:16:44,320 --> 00:16:45,480 Speaker 4: That's that's original. 243 00:16:45,560 --> 00:16:48,520 Speaker 3: But tell us everything your employer is doing with AI. 244 00:16:48,920 --> 00:16:54,640 Speaker 4: Yes, that'll send you the resum, thank you, But I think, okay, 245 00:16:55,080 --> 00:16:57,960 Speaker 4: just let's recap the basics. Right, So the basics are, 246 00:16:58,440 --> 00:17:01,360 Speaker 4: at least for the time being, everybody is trying to 247 00:17:01,360 --> 00:17:05,600 Speaker 4: be more productive with AI. Right, So you want to 248 00:17:05,640 --> 00:17:09,199 Speaker 4: have all your documents you want to have now you 249 00:17:09,240 --> 00:17:14,000 Speaker 4: know what? Perplexity has a finance module. I think one 250 00:17:14,080 --> 00:17:19,359 Speaker 4: day soon maybe Bloomberg will not have the keywords any longer. 251 00:17:19,400 --> 00:17:23,760 Speaker 4: You just give you know, Bloomberg a task and it 252 00:17:23,880 --> 00:17:26,840 Speaker 4: will grab all the pieces of information and hand it 253 00:17:26,880 --> 00:17:29,560 Speaker 4: over to you and maybe you can schedule it. All 254 00:17:29,600 --> 00:17:34,760 Speaker 4: of this is relatively table stakes. I mean, the the 255 00:17:34,880 --> 00:17:38,200 Speaker 4: agentic aspect is not yet, but it will become pretty soon. 256 00:17:38,320 --> 00:17:40,919 Speaker 4: I think it's going to be very hard to compute 257 00:17:40,960 --> 00:17:46,240 Speaker 4: with the likes of maybe Bloomberg, but for sure, let's 258 00:17:46,280 --> 00:17:52,400 Speaker 4: say you know the big hyperscalers. So that's one. At 259 00:17:52,440 --> 00:17:57,240 Speaker 4: the investment level, it's it's much more complicated. So in 260 00:17:58,240 --> 00:18:03,399 Speaker 4: strategies where there is a natural richness in data, you 261 00:18:03,440 --> 00:18:08,160 Speaker 4: can definitely use if not deep learning, but you or AI, 262 00:18:08,320 --> 00:18:13,160 Speaker 4: but you can definitely use very advanced machine learning algorithms 263 00:18:13,480 --> 00:18:16,560 Speaker 4: and you do not have a data snoopin problem, you 264 00:18:16,560 --> 00:18:19,280 Speaker 4: do not have a back testing problem, and so you 265 00:18:19,359 --> 00:18:21,600 Speaker 4: are in a data rich environment and you can do that. 266 00:18:22,280 --> 00:18:25,879 Speaker 4: And it's not a secret that, for example, XTX has 267 00:18:26,160 --> 00:18:31,320 Speaker 4: a very large on premu you know number of in 268 00:18:31,400 --> 00:18:33,800 Speaker 4: Nvidia cards I don't remember h one hundreds or something 269 00:18:33,840 --> 00:18:37,800 Speaker 4: like that. So that's one thing, right. The question is 270 00:18:37,840 --> 00:18:42,520 Speaker 4: really what's going to happen to the slower investment styles. 271 00:18:42,840 --> 00:18:49,600 Speaker 4: And my view is that hopefully large firms like mine 272 00:18:49,600 --> 00:18:52,680 Speaker 4: will have an advantage. But will see right why because 273 00:18:52,680 --> 00:18:55,120 Speaker 4: we do have we do have the scale. We have 274 00:18:55,760 --> 00:18:58,119 Speaker 4: a large number of pms, We have a lot of 275 00:18:58,240 --> 00:19:00,560 Speaker 4: historical data, we have a lot of propriety every data 276 00:19:01,080 --> 00:19:04,400 Speaker 4: that nobody else has. So maybe that that will work out. 277 00:19:04,880 --> 00:19:07,919 Speaker 4: But how to make it happen, I don't know because 278 00:19:08,040 --> 00:19:10,879 Speaker 4: things are changing so fast. And also I'm, you know, 279 00:19:11,000 --> 00:19:13,879 Speaker 4: relatively a tourist in the areas I'm trying to learn 280 00:19:13,920 --> 00:19:15,040 Speaker 4: a little bit more about. 281 00:19:30,520 --> 00:19:33,040 Speaker 3: You mentioned proprietary data, and this comes up a lot 282 00:19:33,040 --> 00:19:35,960 Speaker 3: where people talk about, well, the competitive advantage nowadays really 283 00:19:36,080 --> 00:19:39,000 Speaker 3: is that data set? I mean, is is that true? 284 00:19:39,040 --> 00:19:41,280 Speaker 3: If I get something really cool and unique, I can 285 00:19:41,359 --> 00:19:44,280 Speaker 3: automatically become I don't know, a billionaire trader, if I 286 00:19:44,280 --> 00:19:46,199 Speaker 3: can figure out how to execute on it. Is that 287 00:19:46,240 --> 00:19:46,639 Speaker 3: all there is? 288 00:19:46,840 --> 00:19:49,200 Speaker 4: Maybe yes, I have very weak beliefs on this. I 289 00:19:49,200 --> 00:19:52,360 Speaker 4: don't know. Maybe yes, we'll find out. 290 00:19:52,640 --> 00:19:55,640 Speaker 3: Well, so where are people getting interesting data sets from? 291 00:19:56,560 --> 00:19:59,199 Speaker 4: I mean, you get interesting data from observing human beings 292 00:19:59,680 --> 00:20:03,600 Speaker 4: actually investing, and you don't get to see a great 293 00:20:03,760 --> 00:20:08,120 Speaker 4: PM investing, but I do. That's that's the benefit. 294 00:20:08,960 --> 00:20:12,800 Speaker 2: So from your central position, you just get to see 295 00:20:12,840 --> 00:20:14,840 Speaker 2: a lot of activity and you get to see novel 296 00:20:14,960 --> 00:20:17,840 Speaker 2: data that other people don't get to see simply by 297 00:20:17,880 --> 00:20:20,560 Speaker 2: being in the center of all of these different trades 298 00:20:20,600 --> 00:20:22,879 Speaker 2: and everything, and that gives you a sort of higher 299 00:20:22,880 --> 00:20:25,600 Speaker 2: abstraction layer or whatever it is that someone else in 300 00:20:25,640 --> 00:20:26,439 Speaker 2: the market doesn't have. 301 00:20:26,680 --> 00:20:31,680 Speaker 4: Yeah, and it's possible that not in the distant future, 302 00:20:32,320 --> 00:20:36,840 Speaker 4: good pms will become good because they can improve on 303 00:20:37,000 --> 00:20:41,919 Speaker 4: themselves by basically playing or training or having a baseline 304 00:20:42,080 --> 00:20:46,040 Speaker 4: of an agent that reproduces their behavior. So you know 305 00:20:46,080 --> 00:20:48,760 Speaker 4: there is an alter gapy, well ano a PM, but 306 00:20:48,800 --> 00:20:52,400 Speaker 4: an alter whatever who says what would you do right? 307 00:20:52,480 --> 00:20:54,680 Speaker 4: And you get a baseline behavior and then you can 308 00:20:54,680 --> 00:20:56,920 Speaker 4: think about it and you could say, well, I would 309 00:20:56,960 --> 00:21:00,600 Speaker 4: do something different, and then that becomes an example in 310 00:21:00,640 --> 00:21:05,320 Speaker 4: a reinforcement learning process where the AI keeps learning from 311 00:21:05,359 --> 00:21:08,760 Speaker 4: you and you keep improving because the baseline is changing. 312 00:21:09,520 --> 00:21:12,320 Speaker 3: So before we came out here, I asked perplexity to 313 00:21:12,359 --> 00:21:14,919 Speaker 3: come up with a new factor, and it came up 314 00:21:14,960 --> 00:21:18,920 Speaker 3: with something called the policy agility factor, which is supposed 315 00:21:18,960 --> 00:21:23,520 Speaker 3: to be that countries that display policy flexibility have better 316 00:21:23,640 --> 00:21:27,040 Speaker 3: outperformance over the longer term. Countries that are able to 317 00:21:27,320 --> 00:21:31,680 Speaker 3: more quickly adapt to changing situations are outperformers over the 318 00:21:31,720 --> 00:21:34,800 Speaker 3: long run. Can you grade that factor. I didn't do 319 00:21:34,840 --> 00:21:37,800 Speaker 3: a back tust. But like if someone brought you an 320 00:21:37,800 --> 00:21:42,560 Speaker 3: idea like that, not me, perplexity, I don't want you 321 00:21:42,600 --> 00:21:45,440 Speaker 3: to insult me over the next five minutes. What would 322 00:21:45,520 --> 00:21:47,920 Speaker 3: you say to them? What are the problems with this? 323 00:21:48,720 --> 00:21:53,639 Speaker 4: I mean no major problems, there are questions. So the 324 00:21:53,680 --> 00:21:56,000 Speaker 4: first thing that you want to make sure is that 325 00:21:56,800 --> 00:22:01,480 Speaker 4: if AI whatever it means, brings to you definition, right, 326 00:22:02,040 --> 00:22:06,520 Speaker 4: that definition should be at a point in time and 327 00:22:06,600 --> 00:22:09,719 Speaker 4: should not be trained on all the on all the 328 00:22:09,720 --> 00:22:11,960 Speaker 4: past data, right. So number one, you want to do 329 00:22:12,000 --> 00:22:17,800 Speaker 4: that because if you back test that feature and in 330 00:22:17,840 --> 00:22:22,119 Speaker 4: a way perplexity has already tested it, it's not a 331 00:22:22,160 --> 00:22:24,760 Speaker 4: fair play. You know, the performance will the back test 332 00:22:24,800 --> 00:22:29,800 Speaker 4: will look great. So, unfortunately, we live in a world 333 00:22:30,000 --> 00:22:35,600 Speaker 4: where some factors will never be back testable. So you 334 00:22:35,640 --> 00:22:39,160 Speaker 4: don't know whether they work or they don't work, right, 335 00:22:39,280 --> 00:22:42,720 Speaker 4: You just know that you cannot test them in advance, 336 00:22:42,800 --> 00:22:45,880 Speaker 4: like a policy agility. This seems to be a very 337 00:22:45,920 --> 00:22:49,119 Speaker 4: low turnover factor, right, and it seems to be probably 338 00:22:49,200 --> 00:22:49,960 Speaker 4: a very low. 339 00:22:49,800 --> 00:22:52,320 Speaker 3: Sharp factor in a low universe and. 340 00:22:52,320 --> 00:22:56,080 Speaker 4: A small universe. So how do you how do you know? Well, 341 00:22:56,119 --> 00:22:58,080 Speaker 4: probably you want to make sure that it makes sense, 342 00:22:58,119 --> 00:23:02,520 Speaker 4: and maybe you can start putting a small volativity allocation 343 00:23:02,640 --> 00:23:04,560 Speaker 4: to it and. 344 00:23:04,520 --> 00:23:05,800 Speaker 3: Then you would build it up as you. 345 00:23:05,720 --> 00:23:08,240 Speaker 4: Watched it out for yes, out of sample. 346 00:23:08,520 --> 00:23:10,159 Speaker 3: Okay, so speaking of back to us, I have one 347 00:23:10,160 --> 00:23:14,000 Speaker 3: more question. But it seems like quant investing. Part of 348 00:23:14,000 --> 00:23:16,640 Speaker 3: the issue with this is you are looking back at 349 00:23:16,680 --> 00:23:18,960 Speaker 3: historical data. That's all you have. You don't have data 350 00:23:18,960 --> 00:23:22,119 Speaker 3: about the future. Unfortunately, it strikes me as hard to 351 00:23:22,160 --> 00:23:24,760 Speaker 3: deal with regime changes. So when you have a big 352 00:23:24,880 --> 00:23:29,760 Speaker 3: break in how something works in finance or markets or 353 00:23:29,800 --> 00:23:32,960 Speaker 3: the global economy, how does quant investing actually take into 354 00:23:32,960 --> 00:23:36,159 Speaker 3: account those sorts of risks, Like say, you know, a 355 00:23:36,160 --> 00:23:38,600 Speaker 3: lot of investing is based on the idea that bonds 356 00:23:38,600 --> 00:23:40,680 Speaker 3: and stocks are going to move inversely to each other, 357 00:23:40,720 --> 00:23:44,040 Speaker 3: and then in twenty twenty two they started moving together. 358 00:23:46,760 --> 00:23:51,040 Speaker 4: I think that most people with a quantitative background in 359 00:23:51,080 --> 00:23:55,880 Speaker 4: finance will tell you that regime change is very difficult 360 00:23:55,920 --> 00:23:59,119 Speaker 4: to detect and to act on in an effective manner. 361 00:23:59,200 --> 00:24:02,400 Speaker 4: So I think that's been my experience at least, right 362 00:24:02,440 --> 00:24:07,280 Speaker 4: so in every possible application I've tried, and you know, 363 00:24:07,359 --> 00:24:09,560 Speaker 4: it never, it really never works for me. Maybe it 364 00:24:09,600 --> 00:24:12,480 Speaker 4: works for somebody else. What I think it's a bit 365 00:24:12,560 --> 00:24:17,280 Speaker 4: easier to do is to detect regime change in a 366 00:24:17,359 --> 00:24:22,560 Speaker 4: human being. So instead of trying to use you know, 367 00:24:22,640 --> 00:24:25,399 Speaker 4: there are many many algorithms for regime change. You know, 368 00:24:25,440 --> 00:24:31,880 Speaker 4: there are MARKT based q sum, completely non parametric. Instead 369 00:24:31,880 --> 00:24:36,040 Speaker 4: of trying to act on regime changes in the environment, 370 00:24:36,359 --> 00:24:41,240 Speaker 4: try to detect changes in the behavior of a portfolio 371 00:24:41,240 --> 00:24:45,160 Speaker 4: manager and act on that because that works, I think, 372 00:24:45,840 --> 00:24:50,800 Speaker 4: and usually you know, jives with experience with so that 373 00:24:50,800 --> 00:24:52,359 Speaker 4: that is something that can be exploited. 374 00:24:52,600 --> 00:24:54,320 Speaker 2: I want to go back to an answer you gave 375 00:24:54,400 --> 00:24:56,920 Speaker 2: early on, which is sort of like the old school 376 00:24:57,119 --> 00:25:00,000 Speaker 2: factor investing and like the original versions and maybe they 377 00:25:00,160 --> 00:25:03,040 Speaker 2: sort of an international factor or a liquidity factor, or 378 00:25:03,040 --> 00:25:06,200 Speaker 2: the small cap factor, the value factor. And it feels 379 00:25:06,240 --> 00:25:09,640 Speaker 2: like a lot of these things haven't worked in ages, 380 00:25:10,119 --> 00:25:12,280 Speaker 2: and there's this debate that seems like, Okay, is this 381 00:25:13,320 --> 00:25:17,320 Speaker 2: the long cycle and eventually it's gonna come back, or 382 00:25:17,840 --> 00:25:20,720 Speaker 2: is it that everybody not only knows about these factors 383 00:25:20,720 --> 00:25:24,280 Speaker 2: that have discussed them to death, they're also extremely commodified 384 00:25:24,400 --> 00:25:26,400 Speaker 2: in the sense that you could just buy an ETF 385 00:25:26,400 --> 00:25:28,280 Speaker 2: of them, right, You could just buy a small cap ETF. 386 00:25:28,320 --> 00:25:31,800 Speaker 2: It's trivial to execute. You could just buy a momentum ETF. 387 00:25:31,840 --> 00:25:34,680 Speaker 2: It's trivial to execute a value ETF, et cetera. Like 388 00:25:34,960 --> 00:25:38,440 Speaker 2: my intuition would be, since everyone knows about them and 389 00:25:38,480 --> 00:25:42,480 Speaker 2: they're completely commodified technologically, they're just gone. But there is 390 00:25:42,520 --> 00:25:44,439 Speaker 2: still debates. Some people think it's totally a matter of 391 00:25:44,440 --> 00:25:46,880 Speaker 2: time before these come back in vogue, and that it's 392 00:25:46,880 --> 00:25:49,440 Speaker 2: the long cycle, et cetera. I'm curious how you think 393 00:25:49,440 --> 00:25:51,800 Speaker 2: about some of the original factors that people discussed in 394 00:25:51,880 --> 00:25:53,040 Speaker 2: their prospects going forward. 395 00:25:53,760 --> 00:25:59,040 Speaker 4: Well, so some factors were identified, but then somehow they 396 00:25:59,080 --> 00:26:04,879 Speaker 4: got demoted so famously. Size, right, so conditional on having 397 00:26:05,040 --> 00:26:10,640 Speaker 4: other characteristics of a stock. Size doesn't really explain much 398 00:26:10,640 --> 00:26:15,240 Speaker 4: of your returns, and so it's a combination of other factors. Okay, 399 00:26:15,320 --> 00:26:21,040 Speaker 4: well that's one case. Then there are cases where it 400 00:26:21,160 --> 00:26:25,560 Speaker 4: seems that some factors have been exploited. You know, their 401 00:26:25,600 --> 00:26:30,920 Speaker 4: capacity has been exhausted, and so you can't make an 402 00:26:30,920 --> 00:26:35,119 Speaker 4: attractive return of them. There are some factors that still 403 00:26:35,160 --> 00:26:38,639 Speaker 4: have a low sharp, but they still have a positive sharp, 404 00:26:38,720 --> 00:26:43,840 Speaker 4: and so you know every positive sharp deserves, however small 405 00:26:44,480 --> 00:26:45,200 Speaker 4: and allocation. 406 00:26:45,760 --> 00:26:46,680 Speaker 2: What's an example of. 407 00:26:46,600 --> 00:26:50,600 Speaker 4: That medium term momentum. Right, medium ton momentum is treadable 408 00:26:50,600 --> 00:26:53,960 Speaker 4: and it's relatively high capacity. Then you have the whole 409 00:26:54,040 --> 00:26:57,440 Speaker 4: term structure of momentum, so you know there is a 410 00:26:57,480 --> 00:27:03,320 Speaker 4: shorter horizon reversal and whatnot. Short interest worked great until 411 00:27:03,640 --> 00:27:07,960 Speaker 4: it didn't really work so consistently any longer. And then 412 00:27:08,000 --> 00:27:12,560 Speaker 4: they also assume different characteristics, right, so you start having 413 00:27:12,920 --> 00:27:14,159 Speaker 4: more crashes and the like. 414 00:27:14,840 --> 00:27:15,840 Speaker 2: Is there a mean factor? 415 00:27:17,720 --> 00:27:20,920 Speaker 4: I don't think so. But is there change any theme 416 00:27:21,040 --> 00:27:25,239 Speaker 4: or something like that it's a theme or a theme? Yeah, 417 00:27:25,359 --> 00:27:27,480 Speaker 4: I don't know that ESG is a factor either. I 418 00:27:27,480 --> 00:27:28,720 Speaker 4: don't think so. 419 00:27:28,240 --> 00:27:29,640 Speaker 3: Oh why do you say that? 420 00:27:30,280 --> 00:27:32,320 Speaker 4: Because I don't think it's really that persistent. 421 00:27:33,440 --> 00:27:35,120 Speaker 3: I mean, it doesn't affect human behavior. 422 00:27:36,480 --> 00:27:40,840 Speaker 4: I think that just there is also this feature, right 423 00:27:40,840 --> 00:27:44,680 Speaker 4: the moment that you say that a factor exists, it's reflexive, right, 424 00:27:44,920 --> 00:27:48,879 Speaker 4: there is reflexibity in this, right, But I don't know 425 00:27:48,960 --> 00:27:53,280 Speaker 4: that it really explains much of the returns in recent times. 426 00:27:53,280 --> 00:27:56,000 Speaker 3: So I'm going to ask one more question because I 427 00:27:56,040 --> 00:27:58,040 Speaker 3: started with a dumb one, and so I will finish 428 00:27:58,080 --> 00:28:01,240 Speaker 3: with another dumb one. Is there good and bad alpha 429 00:28:01,800 --> 00:28:06,440 Speaker 3: or is bad alpha just beta? No? 430 00:28:06,680 --> 00:28:10,879 Speaker 4: Every alpha signal is you know, God's little child. There 431 00:28:10,960 --> 00:28:11,840 Speaker 4: is no bad alpha. 432 00:28:12,880 --> 00:28:15,880 Speaker 3: All right, Gappy, thank you so much for coming back 433 00:28:15,920 --> 00:28:28,480 Speaker 3: on odd Lots. 434 00:28:30,760 --> 00:28:32,960 Speaker 2: We're gonna leave it there. That was our conversation with 435 00:28:33,040 --> 00:28:36,080 Speaker 2: Gappi Pallioligo. I'm Jill Wisenthal. You can follow me at 436 00:28:36,080 --> 00:28:40,400 Speaker 2: the Stalwart, Follow Tracy at Tracy Alloway. Follow our guest Gappy, 437 00:28:40,480 --> 00:28:45,440 Speaker 2: He's at Underscore Underscore Polioligo. Follow our producers Kerman Rodriguez 438 00:28:45,480 --> 00:28:48,440 Speaker 2: at Kerman armand dash O Bennett at Dashbuck and kel 439 00:28:48,480 --> 00:28:51,160 Speaker 2: Brooks at Keil Brooks. From our odd Lots content, go 440 00:28:51,240 --> 00:28:53,360 Speaker 2: to Bloomberg dot com slash odd Lots. We have a 441 00:28:53,440 --> 00:28:56,600 Speaker 2: daily newsletter and all of our episodes, and if you 442 00:28:56,720 --> 00:28:59,360 Speaker 2: enjoy the show, please leave us a positive review on 443 00:28:59,440 --> 00:29:03,720 Speaker 2: your favorite podcast platform. Remember, Bloomberg subscribers get to listen 444 00:29:03,720 --> 00:29:06,600 Speaker 2: to Odd Laws and free on Apple podcasts. Just go 445 00:29:06,680 --> 00:29:09,760 Speaker 2: to the Apple podcast app and follow the instructions there. 446 00:29:10,000 --> 00:29:10,720 Speaker 2: Thanks for listening.