WEBVTT - Michael Ervolini on Separating Skill From Luck 0:00:15.120 --> 0:00:18.119 Welcome to Inside Active, a podcast about active managers that 0:00:18.160 --> 0:00:20.799 goes beyond sound bites and headlines and looks deeper into 0:00:20.840 --> 0:00:25.040 their processes, challenges, and philosophies and security selection. I'm David 0:00:25.079 --> 0:00:28.200 cone I, lead mutual fund and active Research at Bloomberg Intelligence. 0:00:28.640 --> 0:00:31.920 Selecting active managers has been one of the biggest challenges 0:00:31.960 --> 0:00:36.440 in investing. While performance, risk, attribution, and other portfolio metrics 0:00:36.680 --> 0:00:39.800 can tell us how fun performed, they don't necessarily tell 0:00:39.880 --> 0:00:42.600 us whether that performance was the result of repeatable skill 0:00:43.000 --> 0:00:47.680 or simply favorable market conditions. For institutional investors, consultants, and 0:00:47.720 --> 0:00:52.400 financial advisors, separating genuine investment skill from luck remains one 0:00:52.400 --> 0:00:56.160 of the industry's most difficult and most important questions. Recent 0:00:56.240 --> 0:01:00.080 advances in portfolio analytics are challenging some long held assumption 0:01:00.480 --> 0:01:04.240 about how active managers should be evaluated. Rather than focusing 0:01:04.280 --> 0:01:07.800 primarily on investment outcomes, a growing body of research suggests 0:01:08.000 --> 0:01:11.440 investors may gain deeper insights by examining the decision makers, 0:01:12.080 --> 0:01:14.880 what they buy, what they sell, how they seize positions, 0:01:14.920 --> 0:01:18.600 and how consistently those decisions add value over time. So 0:01:18.680 --> 0:01:21.280 today I wanted to explore how investors should think about 0:01:21.280 --> 0:01:25.160 measuring manager's skill, why traditional performance metrics may not tell 0:01:25.200 --> 0:01:28.200 the full story, and whether a more decision based approach 0:01:28.280 --> 0:01:31.720 could improve active manager selection. So join me to discuss 0:01:31.760 --> 0:01:35.640 that is Michael Ervolini, author of Skill Versus Luck taking 0:01:35.680 --> 0:01:38.440 the guests out of Equity fund selection. Mike, thanks for 0:01:38.520 --> 0:01:39.199 joining me today. 0:01:39.680 --> 0:01:42.240 It's a pleasure, David, and I appreciate the opportunity to 0:01:42.319 --> 0:01:44.560 speak with you and your audience great. 0:01:45.000 --> 0:01:47.240 So what problem were you trying to solve when you 0:01:47.280 --> 0:01:48.920 wrote this book Skill Versus Luck? 0:01:50.040 --> 0:01:56.560 Fundamentally, it was helping the industry overcome conventional thinking. It's 0:01:56.600 --> 0:02:00.480 a challenge in every industry trying to move forward rather 0:02:00.520 --> 0:02:04.360 than holding fast to the current activities and beliefs we have. 0:02:05.080 --> 0:02:09.000 And as you mentioned in your intro, there's quite a 0:02:09.120 --> 0:02:12.600 few activities going on right now where people are using 0:02:12.760 --> 0:02:17.280 decision based analytics to give better insights into manager skill. 0:02:19.520 --> 0:02:23.320 That area hasn't received as much traction as it needs, 0:02:23.760 --> 0:02:26.680 and so in writing the book, I was hoping to 0:02:26.720 --> 0:02:30.120 stimulate an industry wide conversation about why we need to 0:02:30.160 --> 0:02:35.600 go beyond outcome based analytics and look at decision based 0:02:35.600 --> 0:02:40.639 analytics in combination so that people can get a deeper 0:02:40.680 --> 0:02:44.799 and richer understanding of what skills are driving the fund results, 0:02:45.160 --> 0:02:49.040 what skills are persistent, and all of that can allow 0:02:49.480 --> 0:02:53.960 investors to be more confident in selecting managers that have 0:02:54.080 --> 0:02:58.560 a greater likelihood of outperforming in the future. And really 0:02:58.800 --> 0:03:02.519 that's what cation decisions are all about, is who you 0:03:02.560 --> 0:03:05.240 want to give your money to in order to get 0:03:05.360 --> 0:03:06.800 better returns going forward. 0:03:08.080 --> 0:03:10.200 So you're the title of the book kind of suggests 0:03:10.200 --> 0:03:14.320 we consistently are constantly confused luck with skill. Why do 0:03:14.360 --> 0:03:17.519 you think intelligent people make that mistake so often? 0:03:19.080 --> 0:03:22.400 Well, two points on that one is I really believe 0:03:22.440 --> 0:03:26.559 that skill and luck are just opposite ends of a continuum, 0:03:27.280 --> 0:03:32.120 and so when a fund manager has more skill, they're 0:03:32.240 --> 0:03:36.080 less dependent upon luck for their results. So that's part 0:03:36.120 --> 0:03:39.520 one is skill and luck are involved in every decision. 0:03:39.880 --> 0:03:42.720 You just want managers that are more skilled. But a 0:03:42.760 --> 0:03:46.240 more specific answer to your question is, when you're making 0:03:46.320 --> 0:03:51.280 decisions using conventional analytics, they bring you to the threshold 0:03:52.120 --> 0:03:56.840 of understanding skill, but then you need to make a jump. 0:03:57.040 --> 0:04:00.839 You need to overinterpret the results or see in them 0:04:00.880 --> 0:04:02.920 something that you want to see. In other words, you 0:04:03.000 --> 0:04:07.320 have to guess a little bit. And I think because 0:04:07.360 --> 0:04:14.000 of that inadequacy of conventional analytics at addressing skill it 0:04:14.040 --> 0:04:16.640 makes it hard to figure out which managers are skilled 0:04:16.680 --> 0:04:21.000 and which aren't, and that has been plaguing the industry 0:04:21.040 --> 0:04:23.120 for a while. It continues to do so. But I 0:04:23.200 --> 0:04:28.039 have worked with institutional investors in particular who've been using 0:04:28.080 --> 0:04:33.560 decision based analytics, and their ability to sort out superior 0:04:33.640 --> 0:04:37.000 managers has improved as a result of this. So we 0:04:37.839 --> 0:04:40.640 can get past this. But most of the confusion has 0:04:40.640 --> 0:04:44.039 to do with working with inferior analytics as opposed to 0:04:44.640 --> 0:04:45.880 people not trying their best. 0:04:46.720 --> 0:04:48.640 So do you think we're kind of wired to judge 0:04:48.640 --> 0:04:50.400 outcomes rather than decisions. 0:04:50.760 --> 0:04:53.440 Yeah. One of the confusions that we have is how 0:04:53.480 --> 0:04:58.240 we make decisions. And what is not commonly understood is 0:04:58.240 --> 0:05:01.080 that more than ninety five percent the decisions we make 0:05:01.120 --> 0:05:06.840 every day are made completely automatically by our unconscious and 0:05:06.920 --> 0:05:12.279 so a lot of what's happening, whether it's a buying 0:05:12.279 --> 0:05:16.080 a stock, picking a manager, or ordering something off the menu, 0:05:16.839 --> 0:05:21.400 these automatic decisions generally work. But they only work really 0:05:21.440 --> 0:05:25.240 well where there's been great feedback about how good the 0:05:25.279 --> 0:05:28.080 decision is and how good the outcome has been. And 0:05:28.120 --> 0:05:32.839 so in that automatic decision environment. When you're using outcome 0:05:32.880 --> 0:05:36.640 based analytics, you see in them what you want, and 0:05:36.680 --> 0:05:40.920 so I think when asset owners or allocators are shown 0:05:42.040 --> 0:05:47.560 decision based results against outcome based results, they can take 0:05:47.600 --> 0:05:51.000 the best of both. But lacking any kind of decision 0:05:51.040 --> 0:05:55.880 based analytics and using only outcome based analytics, our natural 0:05:55.920 --> 0:05:59.719 instinct is to see in those results something that seems 0:05:59.720 --> 0:06:02.760 reason and often what we see is completely wrong. 0:06:03.560 --> 0:06:07.800 Okay, and you know you redefine skill as the combined 0:06:07.839 --> 0:06:12.000 effect of expert judgment and investment processes. Why did you 0:06:12.000 --> 0:06:14.120 feel the industry needed a new definition? 0:06:15.120 --> 0:06:18.440 Well, I think partly is because we had been looking 0:06:18.440 --> 0:06:20.520 at skill the wrong way. We've been looking at it 0:06:20.520 --> 0:06:23.360 as outcomes, and when you look at that, it's hard 0:06:23.400 --> 0:06:26.200 to figure out what's in there. So if we're going 0:06:26.279 --> 0:06:29.720 to look at decisions, then we want to focus on 0:06:29.760 --> 0:06:32.880 what is it the manager brings to the fund and 0:06:32.960 --> 0:06:37.440 the manager brings to qualities, their expert judgment and their 0:06:37.440 --> 0:06:40.240 investment process, and that's what they should be judged on. 0:06:40.720 --> 0:06:45.040 And fortunately, decision based analytics can allow you to rigorously 0:06:46.120 --> 0:06:51.799 evaluate and quantify both of those elements judgment and process, 0:06:52.360 --> 0:06:56.560 and when you understand them well you can then get 0:06:56.600 --> 0:07:02.120 a sense of how consistently good the men is making buying, selling, 0:07:02.200 --> 0:07:07.400 sizing decisions, and using that, I think what we've seen 0:07:07.480 --> 0:07:10.960 in the industry is that people then have more confidence 0:07:11.280 --> 0:07:14.840 that the manager they're working with, well not perfect, has 0:07:14.880 --> 0:07:17.920 a better than average chance about performing going forward. 0:07:19.080 --> 0:07:21.640 So then how do you actually separate you know, a 0:07:21.640 --> 0:07:24.040 good decision from a lucky outcome, Like, how do you 0:07:24.080 --> 0:07:24.840 go in and do that? 0:07:25.440 --> 0:07:29.400 Well, there's always going to be some luck. Okay, that's 0:07:29.520 --> 0:07:31.160 just we'll call that the market. 0:07:31.320 --> 0:07:32.680 You know, if you were. 0:07:32.600 --> 0:07:36.880 Lucky enough to buy the Magnificent seven some years ago 0:07:38.400 --> 0:07:42.200 and overweight them extensively, part of that was a good decision, 0:07:42.240 --> 0:07:44.840 and part of that is they've just done remarkably well. 0:07:44.920 --> 0:07:46.920 So there's a little bit of luck, or if your 0:07:46.960 --> 0:07:52.160 style is in favor and so forth. But given that 0:07:53.000 --> 0:07:57.040 we can look at individual decisions, let's say a buying decision. 0:07:57.600 --> 0:07:59.600 What we can do is over a period of time, 0:07:59.640 --> 0:08:02.880 we use ten years just for easy number. We look 0:08:02.920 --> 0:08:09.560 at every position that was purchased over that ten year period, 0:08:10.200 --> 0:08:16.000 and then we evaluate, independently from how the manager sized 0:08:16.000 --> 0:08:20.160 a position or when they sold it, on average, how 0:08:20.200 --> 0:08:25.600 did those positions do in aggregate that they generate excess 0:08:25.640 --> 0:08:29.360 return or not on their own. And so we can 0:08:29.400 --> 0:08:31.640 do that for sizing, we can do that for adding 0:08:31.720 --> 0:08:33.320 on the way up, adding on the way down. We 0:08:33.360 --> 0:08:38.080 can look at decisions broadly or very very narrowly, and 0:08:38.120 --> 0:08:42.040 the idea is to isolate individual types of decisions and 0:08:42.160 --> 0:08:46.679 evaluate if they, on average are helping to generate more 0:08:46.800 --> 0:08:50.240 excess return or in fact are taking away excess return, 0:08:50.320 --> 0:08:52.160 which is obviously a negative skill. 0:08:52.800 --> 0:08:55.560 Okay, and you know, one of the most interesting concepts 0:08:55.559 --> 0:08:58.640 I think in the book is the counterfactual portfolio. Can 0:08:58.640 --> 0:09:01.280 you kind of explain that idea and plan English. 0:09:01.360 --> 0:09:08.040 Sure, a counterfactual portfolio is simply a copy of the 0:09:08.280 --> 0:09:17.040 actual portfolio, but then one decision type is adjusted to 0:09:17.160 --> 0:09:22.160 understand its impact on the portfolio. So a very straightforward 0:09:22.160 --> 0:09:28.199 example is this, there's a fund where the manager periodically 0:09:28.760 --> 0:09:33.360 purchases stocks that are outside the benchmark of the fund, 0:09:34.280 --> 0:09:37.760 and presumably these out of benchmark purchases are intended to 0:09:37.840 --> 0:09:42.160 generate excess return or alpha. Well, one way to evaluate 0:09:42.240 --> 0:09:45.840 that is to create a copy of the fund and 0:09:45.880 --> 0:09:48.920 then go through it every day, and whenever the manager 0:09:49.080 --> 0:09:54.719 buys a position that's outside the benchmark. Just simply eliminate 0:09:54.800 --> 0:09:59.319 that from the copy or the counterfactual. And so when 0:09:59.360 --> 0:10:05.240 you're done doing that exercise, the counterfactual portfolio on every 0:10:05.320 --> 0:10:10.439 single day owns exactly what's in the actual portfolio, except 0:10:10.480 --> 0:10:15.360 it doesn't own any out of benchmark positions. And now 0:10:15.480 --> 0:10:20.240 with that counterfactual portfolio, you generate its returns and its 0:10:20.720 --> 0:10:25.880 excess return. You can do information ratios, whatever metrics you want, 0:10:26.080 --> 0:10:28.680 and compare those to the actual. Now, if we're looking 0:10:28.760 --> 0:10:33.600 at relative return, if the actual portfolio has a higher 0:10:33.679 --> 0:10:39.320 relative return than the counterfactual, that tells us owning those 0:10:39.360 --> 0:10:45.360 out of benchmark positions really helped it generated more excess return. Conversely, 0:10:46.120 --> 0:10:50.839 if the actual portfolio underperforms the counterfactual, it tells us 0:10:51.040 --> 0:10:54.760 that owning those out of benchmark positions actually hurt the results. 0:10:55.240 --> 0:10:58.120 So that's the idea of a counterfactual. You take one 0:10:58.160 --> 0:11:01.400 decision type at a time, you either eliminate it or 0:11:01.440 --> 0:11:05.199 alter it, and then compare it back to the actual 0:11:05.559 --> 0:11:09.199 and that way you're able to see whether specific decisions 0:11:09.240 --> 0:11:15.040 are additive or detracting from overall success in the fund. 0:11:15.840 --> 0:11:17.439 Now it makes sense, it makes a lot of sense. 0:11:17.720 --> 0:11:21.400 It's really interesting. You know. I also couldn't help but 0:11:21.440 --> 0:11:24.240 think that, you know, what you're looking at applies even 0:11:24.280 --> 0:11:27.320 beyond investing. Do you think your framework could be used 0:11:27.320 --> 0:11:31.319 to evaluate, you know, even CEOs or even sports coaches. 0:11:32.559 --> 0:11:37.040 I'm not sure I've thought about that. It's possible. It 0:11:37.160 --> 0:11:43.920 definitely was meant to be used across any publicly treated 0:11:44.640 --> 0:11:49.600 asset type, that's for sure, so option swaps, fixed income. 0:11:50.760 --> 0:11:54.640 But beyond that, we've never tackled anything like that, So 0:11:55.080 --> 0:11:57.440 I'm not sure it's possible, But I'm not sure. 0:11:58.600 --> 0:12:02.560 It's definitely interesting concept. And you know, obviously the biggest, 0:12:02.600 --> 0:12:04.320 you know, one of the biggest criticisms you've had is 0:12:04.360 --> 0:12:07.800 allocators spend way too much time evaluating performance, not enough 0:12:07.880 --> 0:12:11.160 time evaluating decisions. If you were sitting on an investment 0:12:11.160 --> 0:12:13.760 community today, what would you stop measuring and what would 0:12:13.760 --> 0:12:14.559 you start measuring? 0:12:16.480 --> 0:12:19.360 Well, I'm not sure I would stop measuring anything, but 0:12:19.520 --> 0:12:22.920 I would add a whole lot of decision based analytics. 0:12:23.240 --> 0:12:26.320 I would want to know how effective the manager is 0:12:26.360 --> 0:12:29.360 at buying, selling, and sizing, and then I would want 0:12:29.400 --> 0:12:32.439 to look at them from multiple perspectives. I would want 0:12:32.480 --> 0:12:36.000 to look, let's say, ten years again, how good was 0:12:36.200 --> 0:12:39.520 buying over ten years, how good was it year by 0:12:39.600 --> 0:12:42.599 year to see how consistent it was, How good is 0:12:42.640 --> 0:12:45.640 the buying across sectors, to make sure that we're not 0:12:45.720 --> 0:12:48.079 just leaning on a few sectors, but the manager's good 0:12:48.120 --> 0:12:50.679 across the board. And I might want to look at 0:12:50.679 --> 0:12:54.800 it by factor. If there's a certain element in the 0:12:54.880 --> 0:13:00.800 strategy that the managers expressing in their coversations or their 0:13:00.800 --> 0:13:04.679 marketing material about we look for stocks with these characteristics, 0:13:05.000 --> 0:13:08.440 I would want to use those factors to evaluate the 0:13:08.480 --> 0:13:11.360 buying as well. As an example, if the manager says 0:13:11.360 --> 0:13:15.720 they'd like to buy high quality companies, well, we could 0:13:15.840 --> 0:13:19.680 use any number of factors to evaluate the quality of 0:13:19.679 --> 0:13:22.000 a company balance sheet, debt, or whatever we want to 0:13:22.000 --> 0:13:26.920 do and see if in fact, higher quality balance sheet 0:13:27.040 --> 0:13:31.600 companies show up as being the stronger buys in the portfolio. 0:13:32.080 --> 0:13:34.440 So I'd want to do a lot of work like that, 0:13:34.559 --> 0:13:37.400 And when you get to that place, you know that 0:13:37.440 --> 0:13:40.120 the manager is doing what they said they're going to do, 0:13:41.080 --> 0:13:44.640 They're doing it very well every year and over time, 0:13:44.679 --> 0:13:49.079 it's really adding value to the fund. That, in combination 0:13:49.200 --> 0:13:53.040 with conventional analytics, allows you to know that over the 0:13:53.080 --> 0:13:55.520 past few years the results have been great too. The 0:13:56.520 --> 0:14:00.439 amount of risk taking was not disproportional to the return, 0:14:00.559 --> 0:14:03.400 and they're well within their peer group, and all that 0:14:03.480 --> 0:14:07.000 kind of stuff. All these conventional analytics are important, they 0:14:07.040 --> 0:14:10.439 simply don't provide enough information about skill. 0:14:11.280 --> 0:14:15.040 Okay, you know you obviously spent decades looking at how 0:14:15.280 --> 0:14:17.960 professional investors make decisions. What do you think the biggest 0:14:18.000 --> 0:14:21.600 misconception people have about how decisions are actually made? 0:14:22.920 --> 0:14:27.960 Well, I mentioned one, which is this unconscious bit thinking. 0:14:28.520 --> 0:14:34.760 Recognizing that most of your decisions are unconscious really convinces 0:14:34.840 --> 0:14:38.760 me and a lot of other people that it's imperative 0:14:38.760 --> 0:14:41.800 that you have a strong process because that can arrest 0:14:42.280 --> 0:14:45.280 some of the automatic decision making and make sure that 0:14:45.320 --> 0:14:48.600 you stay in the area where you make your best decisions. 0:14:49.160 --> 0:14:54.480 Another misconception is emotions. We hear frequently, at least I 0:14:54.520 --> 0:14:59.320 have in my career an investor or money manager or 0:14:59.360 --> 0:15:02.120 other profession I will say, well, we take all the 0:15:02.160 --> 0:15:04.640 emotions out of our decision making. We do everything here 0:15:04.640 --> 0:15:08.640 by the numbers. Well, that sounds great, but it's actually impossible. 0:15:09.480 --> 0:15:14.760 What neuroscience tells us is it's literally impossible to make 0:15:14.840 --> 0:15:19.640 a rational choice without emotions. You can add up a 0:15:19.680 --> 0:15:23.800 column of numbers without emotions because there's no judgment involved. 0:15:23.880 --> 0:15:26.200 You're just adding numbers. But when you're trying to make 0:15:26.240 --> 0:15:29.960 a choice, what discount rates to use? What fund am 0:15:30.000 --> 0:15:33.640 I ultimately going to pick for my allocation? What research 0:15:33.680 --> 0:15:35.400 am I going to use to help me get to 0:15:35.440 --> 0:15:41.120 that place? Those choices, you get to difficult spots where 0:15:41.160 --> 0:15:45.040 it's not easy to differentiate one option from another, and 0:15:45.080 --> 0:15:48.920 a little emotional spark is what helps in the tie breakers. 0:15:49.600 --> 0:15:52.520 And so understanding that your emotions play a role in 0:15:52.560 --> 0:15:55.760 your decision making is another reason you want to have 0:15:55.880 --> 0:15:59.040 a good process because it can slow you down and 0:15:59.160 --> 0:16:05.600 allow you to arrest. Automatic decision making allow you to 0:16:05.800 --> 0:16:10.600 harness emotions in a positive way. And when you embrace 0:16:10.680 --> 0:16:14.640 those two concepts and use them to help build your process, 0:16:15.200 --> 0:16:17.680 that result is you get much better decision making. 0:16:18.960 --> 0:16:21.360 Do you think there's you know, a danger we've become 0:16:21.560 --> 0:16:23.960 obsessed with measuring everything, you know, in other words, can 0:16:24.080 --> 0:16:26.040 skill itself become over engineered. 0:16:26.560 --> 0:16:29.160 I think we can overmeasure things. Yeah, we can lose 0:16:29.200 --> 0:16:34.520 sight of what we're really after. And if that happens, 0:16:35.160 --> 0:16:37.480 it's because we didn't have a good plan going in. 0:16:38.120 --> 0:16:40.240 You know, we have to understand why we're measuring skill 0:16:40.280 --> 0:16:42.000 and what we want to come out of it. But 0:16:42.080 --> 0:16:46.960 I will caution this frequently, the fear of overmeasurement can 0:16:47.000 --> 0:16:50.320 be used as an argument to hang onto convention, to 0:16:50.480 --> 0:16:53.280 not take a step forward to state where we are 0:16:53.280 --> 0:16:57.880 more comfortable in presenting these analytics. To some fund managers, 0:16:58.480 --> 0:17:01.560 sometimes the pushback is, well, you know, I don't want 0:17:01.560 --> 0:17:04.200 you coming in and messing up what I'm doing now. Well, 0:17:04.320 --> 0:17:06.879 so I don't want to see this stuff now. I 0:17:06.880 --> 0:17:10.640 think there's a little fear in there about learning what 0:17:10.680 --> 0:17:12.879 you do well and what you don't. So there's a 0:17:12.920 --> 0:17:16.639 balance point. Yes, we can't go overboard and measure too 0:17:16.720 --> 0:17:20.320 much or follow analytics too far, because no analytic is perfect. 0:17:21.000 --> 0:17:26.359 But reasonable people can take on new analytics and learn 0:17:26.400 --> 0:17:28.639 to make better decisions. And that's what it's all about. 0:17:28.680 --> 0:17:33.480 Being proportional, being reasonable and understanding that the purpose here 0:17:33.520 --> 0:17:36.359 is to make better decisions, not just to measure things 0:17:36.440 --> 0:17:39.560 or to follow analytics one hundred percent. 0:17:40.560 --> 0:17:44.320 You know your approach depends on analyzing thousands of decisions. 0:17:44.560 --> 0:17:47.280 Could someone argue that investing is still too uncertain to 0:17:47.320 --> 0:17:48.080 isolate skill. 0:17:49.160 --> 0:17:54.360 I don't think so now, because I myself have been 0:17:54.359 --> 0:17:58.560 involved in analyzing four or five six hundred funds, and 0:17:58.600 --> 0:18:01.600 there's no question that skill can be measured. And there 0:18:01.600 --> 0:18:05.560 are other companies out there that have collectively looked at 0:18:06.000 --> 0:18:11.000 perhaps eight or nine thousand funds, and so skill can 0:18:11.040 --> 0:18:14.320 be measured, can be isolated, can be quantified. It's done 0:18:14.320 --> 0:18:18.320 differently by different analytics shops, but there's no question that 0:18:18.359 --> 0:18:23.560 can be isolated. It simply can't be done using conventional analytics. 0:18:23.960 --> 0:18:27.359 Okay, you know Warren Buffett has said that you know, 0:18:27.440 --> 0:18:31.480 temperament matters more than IQ. Does your framework capture qualities 0:18:31.520 --> 0:18:35.200 like patients, discipline, conviction or is it only you know, 0:18:35.280 --> 0:18:36.520 kind of measurable decisions. 0:18:37.320 --> 0:18:42.320 No, In fact, it can help you determine some of 0:18:42.359 --> 0:18:49.200 those things. One example, let's say patients. Okay, we could 0:18:49.240 --> 0:18:54.040 look at when a manager is buys stocks and they 0:18:54.080 --> 0:18:59.160 take off relatively early after purchase. Does the manager hang 0:18:59.200 --> 0:19:03.640 on to them or does the manager sell them far 0:19:03.720 --> 0:19:06.600 too quickly and give up a lot of alpha. Now 0:19:06.600 --> 0:19:10.960 there's a patience issue. So by looking at new winners 0:19:11.520 --> 0:19:14.639 and evaluating whether they're being sold too quickly or not, 0:19:15.040 --> 0:19:19.120 that's a way of getting into patients conviction. We can 0:19:19.160 --> 0:19:21.879 look at when positions are purchased and maybe in the 0:19:21.920 --> 0:19:26.560 first few months they take a dip in price. Sometimes 0:19:26.720 --> 0:19:30.000 an early dip in price is an indication that the 0:19:30.040 --> 0:19:33.359 manager bought a bad stock. Other times it's just that 0:19:33.440 --> 0:19:36.080 the market has had a bad reaction to that company, 0:19:36.320 --> 0:19:39.640 but the fundamentals really haven't changed, and more often than not, 0:19:39.800 --> 0:19:42.960 those stocks are going to bounce back. And by evaluating 0:19:43.040 --> 0:19:47.280 that behavior, we can see whether the manager has sufficient 0:19:47.359 --> 0:19:51.400 conviction in stocks that wiggle a little bit right after 0:19:51.400 --> 0:19:54.919 they're purchased, but ultimately do really, really well. So we 0:19:55.000 --> 0:19:59.120 can look for behavioral kinds of aspects in the manager 0:20:00.440 --> 0:20:01.920 looking at the decisions they make. 0:20:02.600 --> 0:20:05.080 Okay, you know another thing that you know kind of 0:20:05.080 --> 0:20:07.560 really jumped out was that you know you've concluded that 0:20:07.600 --> 0:20:10.280 buying is usually the biggest drive of alpha. Why do 0:20:10.320 --> 0:20:13.960 you think buying is more important than selling. Well, if 0:20:13.960 --> 0:20:14.320 you're not. 0:20:14.280 --> 0:20:17.040 A good buyer, you're basically starting out in a hole, 0:20:18.240 --> 0:20:20.840 and it's hard to sell your way out of a 0:20:20.920 --> 0:20:24.640 hole and generate alpha. It can be done, but it's 0:20:24.720 --> 0:20:28.320 really hard. But if you start out buying good stocks, 0:20:29.119 --> 0:20:31.560 you can have a few mistakes on sizing and selling 0:20:31.880 --> 0:20:35.200 and still do pretty well, so it's just a question 0:20:35.240 --> 0:20:39.840 of starting off on the strongest platform possible. And that's 0:20:39.840 --> 0:20:45.600 with good buying. In general speaking, most successful managers are 0:20:45.720 --> 0:20:48.600 really really good buyers. There's a few that are not, 0:20:49.480 --> 0:20:52.359 but and I can explain why if that's of interest, 0:20:52.400 --> 0:20:55.320 But generally speaking, they're good buyers. 0:20:56.240 --> 0:20:57.840 So that kind of brings me to the next part 0:20:58.000 --> 0:21:00.240 is you know the you argue that selling is probably 0:21:00.320 --> 0:21:03.280 that the industry's weakest skill. You know, and when I 0:21:03.320 --> 0:21:05.639 speak to managers, I always I don't want to know 0:21:05.800 --> 0:21:08.639 just what their buying philosophy is, but you know what 0:21:08.800 --> 0:21:10.720 makes them sell? And so what do you think a 0:21:10.720 --> 0:21:12.720 lot of professionals struggle so much? Was selling? 0:21:13.119 --> 0:21:15.920 It's the least understood. If you were to do a 0:21:16.000 --> 0:21:22.400 quick search on equity investing strategy, everything you. 0:21:22.480 --> 0:21:24.720 Found would be about buying. 0:21:26.640 --> 0:21:31.040 When it so that's the whole industry in academy itself, 0:21:31.600 --> 0:21:36.440 when they think about the strategic decisions inside of equity 0:21:36.480 --> 0:21:39.800 investing or any other asset class, it's generally they're looking 0:21:39.800 --> 0:21:46.040 at buying. Selling is often a discipline. Well, if buying 0:21:46.119 --> 0:21:48.919 is strategic and selling is a discipline, it sort of 0:21:48.960 --> 0:21:52.439 tells you that selling is sort of the orphan of 0:21:52.560 --> 0:21:56.600 thinking in asset management, and so it's not well understood. 0:21:57.320 --> 0:22:00.399 The second thing is what is a good sell. Often 0:22:00.440 --> 0:22:03.560 people believe that a good sell is after I sell 0:22:03.600 --> 0:22:08.400 the stock the price goes down. Well, that's not necessarily 0:22:08.800 --> 0:22:13.200 a good description. A better one is A good sell 0:22:13.960 --> 0:22:19.480 is when you sell a stock and after you sell it, 0:22:19.480 --> 0:22:24.879 its return is less than the fund's average. So, in 0:22:24.880 --> 0:22:27.320 other words, you want to hang onto stocks that can 0:22:27.440 --> 0:22:30.879 lift up the fund average return, and you want to 0:22:30.920 --> 0:22:34.879 sell stocks that will take down the fund's average return, 0:22:37.320 --> 0:22:42.359 meaning that if the market is going down, holding onto 0:22:42.400 --> 0:22:46.560 a stock that only goes down five percent when a 0:22:46.560 --> 0:22:49.200 lot of other stocks are going down ten percent, well 0:22:49.640 --> 0:22:52.000 you shouldn't sell that stock even though it's priced going down. 0:22:52.520 --> 0:22:54.800 So we get confused a little bit about what a 0:22:54.840 --> 0:22:59.000 good sell is. And again it's selling a stock before 0:22:59.119 --> 0:23:02.840 it drags down the average. That's the fundamental test. And 0:23:02.880 --> 0:23:09.080 then the fact that academia and other analytics really don't 0:23:09.119 --> 0:23:13.960 look at selling particularly well. Conventional analytics are really nonexistent 0:23:14.240 --> 0:23:18.280 in their ability to help you understand selling. The decision 0:23:18.320 --> 0:23:21.560 based analytics do a real nice job at looking at 0:23:21.600 --> 0:23:25.080 selling selling when stocks are up selling when stocks are down. 0:23:26.240 --> 0:23:29.680 You can look at selling all different manners and really 0:23:29.760 --> 0:23:34.080 understand it. But most managers do not have good analytics 0:23:34.160 --> 0:23:37.520 to help them understand their selling, and that's why they 0:23:37.560 --> 0:23:40.720 try their best, but they really can't improve it because 0:23:40.720 --> 0:23:41.800 they don't have the right feedback. 0:23:42.320 --> 0:23:44.560 Okay, and you know, if you had to pick one 0:23:44.720 --> 0:23:48.480 investment habit, you know you think every active manager should 0:23:48.480 --> 0:23:50.600 stop doing tomorrow. What do you think that would be? 0:23:51.600 --> 0:23:57.040 Reaching for conviction too quickly? Conviction is an important element. 0:23:58.119 --> 0:24:00.880 When you talk to managers, they talk about it. When 0:24:00.920 --> 0:24:03.640 consultants took the managers, they ask them about their conviction. 0:24:04.080 --> 0:24:08.119 Conviction is a well bandeed word in the industry and 0:24:08.160 --> 0:24:09.119 it's important. 0:24:09.880 --> 0:24:10.520 But what the. 0:24:10.480 --> 0:24:15.440 Research shows is coming to it slowly and allowing it 0:24:15.480 --> 0:24:20.440 to develop, particularly using your analysts, using the investment committee, 0:24:21.359 --> 0:24:26.080 using your research team. Allowing conviction to develop slowly allows 0:24:26.119 --> 0:24:29.959 the manager to be more confident that they're really dealing 0:24:30.000 --> 0:24:33.879 with well airned conviction and not just bluster. And telling 0:24:33.920 --> 0:24:37.600 the difference when you're excited is really really hard. So 0:24:37.680 --> 0:24:40.800 that again brings me back to why process is so important. 0:24:41.520 --> 0:24:45.600 Getting to a discipline process where you can let your 0:24:46.000 --> 0:24:49.800 unconscious do its best job. Let your emotions help you 0:24:49.880 --> 0:24:52.800 as much as they can, but let others in your 0:24:52.880 --> 0:24:57.840 organization offer questions, criticism, or support. All of that allows 0:24:57.880 --> 0:25:03.040 the manager to develop real fiction and make their best decisions. 0:25:03.600 --> 0:25:07.480 Okay, you know we're obviously entering an error where AI 0:25:07.640 --> 0:25:11.639 can analyze enormous amounts of information. Do you think AI 0:25:12.080 --> 0:25:14.560 makes skill more valuable or less valuable? 0:25:16.440 --> 0:25:20.160 Right now? There's no question it makes skill more valuable. 0:25:20.800 --> 0:25:25.040 There's a recent research report from Lauren Cohen out of 0:25:25.080 --> 0:25:28.320 Harvard just came out in the past seven or ten 0:25:28.400 --> 0:25:32.399 days where they use some large language models to mimic 0:25:34.000 --> 0:25:38.480 a number of mutual funds over an extended period of time, 0:25:39.040 --> 0:25:42.040 and what they found out is that these largers language 0:25:42.040 --> 0:25:48.000 models were really really good at anticipating something between seventy 0:25:48.040 --> 0:25:53.600 five and eighty percent of the stocks that the fund 0:25:53.640 --> 0:25:58.400 actually purchased over time. What they also found is that 0:25:59.560 --> 0:26:04.879 the stock the large language models didn't buy were the 0:26:04.880 --> 0:26:09.360 ones that contained most of the alpha. So there's something 0:26:09.480 --> 0:26:15.560 about human judgment at the edge of the marketplace that 0:26:15.960 --> 0:26:20.159 is pretty impressive today. It allows people to really do 0:26:20.240 --> 0:26:24.800