1 00:00:01,440 --> 00:00:12,321 Welcome to Top Traders Unplugged. In markets success doesn’t 2 00:00:12,321 --> 00:00:24,411 come from predicting what happens next, it comes from being 3 00:00:24,411 --> 00:00:31,665 prepared for what you can’t predict. In each episode we go deep 4 00:00:04,620 --> 00:00:10,450 with some of the world’s most thoughtful minds in investing, economics, 5 00:00:10,450 --> 00:00:16,280 and beyond to understand how they think, how they prepare, and 6 00:00:16,280 --> 00:00:21,580 how they decide, and the experiences that shaped how they 7 00:00:21,580 --> 00:00:26,880 see the world. No noise, no short-cuts, just real conversations 8 00:00:26,880 --> 00:00:31,650 to help you think better and invest with confidence. 9 00:00:35,740 --> 00:00:45,112 Welcome and welcome back to this week's edition of the Systematic 10 00:00:45,112 --> 00:00:52,780 Investor series with Richard Brennan and I, Niels Kaastrup-Larsen, 11 00:00:52,780 --> 00:01:03,004 where each week we take the pulse of the global markets through 12 00:01:03,004 --> 00:01:08,116 the lens of a rules-based investor. This week we have a very 13 00:00:38,872 --> 00:00:44,614 special guest, namely the one and only Dave Dredge, who really 14 00:00:44,614 --> 00:00:49,834 needs no introduction, but who officially is the Chief Investment 15 00:00:49,834 --> 00:00:55,576 Officer at Convex Strategies in Singapore. Rich and Dave, it is 16 00:00:55,576 --> 00:01:01,318 really wonderful to have you both here this week. Hope you're 17 00:01:01,318 --> 00:01:08,104 doing well. What's going in your part of the world, so to speak? 18 00:00:35,740 --> 00:01:08,120 I'll let Dave go first since he is the guest. 19 00:01:10,680 --> 00:01:17,973 It's summer in Singapore, which is just like any other time 20 00:01:17,973 --> 00:01:27,255 of the year in Singapore. So, it's hot and humid and rains a lot. 21 00:01:27,255 --> 00:01:34,548 But business wise we're super busy. There's obviously a lot going 22 00:01:34,548 --> 00:01:42,504 on in the world and the need for effective explicit loss mitigation 23 00:01:42,504 --> 00:01:46,482 that effective explicit, negatively correlating, explicit, 24 00:01:46,482 --> 00:01:50,460 true diversifying has never been higher. And so, we're pretty 25 00:01:14,380 --> 00:01:24,555 busy, and markets are active across all kinds of asset classes, 26 00:01:24,555 --> 00:01:32,880 and I've got some presentations coming up. I'm down 27 00:01:32,880 --> 00:01:43,980 in Australia in a couple of weeks speaking at a conference down 28 00:01:43,980 --> 00:01:50,455 there. So, everything's going here as usual. 29 00:01:50,730 --> 00:01:53,513 That's fantastic. Good stuff. How about you, Rich? How are you 30 00:01:53,513 --> 00:01:53,766 doing? 31 00:01:54,330 --> 00:01:57,983 It's a bit colder for me, Niels. So, down in the southern hemisphere 32 00:01:57,983 --> 00:02:01,355 here. I'm not too far away from Dave in Singapore, but sufficiently 33 00:02:01,355 --> 00:02:05,008 south to know it's winter down here. So, I noticed that Dave and 34 00:02:05,008 --> 00:02:08,661 I have both got our merch on for Top Traders Unplugged. And the 35 00:02:08,661 --> 00:02:12,314 only one who's absent of that, Niels, is you. And I assume it's 36 00:02:12,314 --> 00:02:15,686 because it's a boiling summer heat up there and you can't afford 37 00:02:15,686 --> 00:02:17,934 to have anything over that top of yours. 38 00:02:19,100 --> 00:03:00,840 Well, I mean that is definitely one excuse. The other 39 00:03:00,840 --> 00:03:46,754 more truthful excuse is I completely forgot it. But there we 40 00:03:46,754 --> 00:04:41,016 are. I will have to improve my prep skills, that is for sure. It's 41 00:02:22,479 --> 00:03:03,027 great to have you here and it is really wonderful knowing what 42 00:03:03,027 --> 00:03:36,817 topics we're going to be talking about, which I appreciate 43 00:03:36,817 --> 00:04:20,744 both of you help put together. And I think really the audience is 44 00:04:20,744 --> 00:04:41,018 in for a treat this week. Now, it is coming up to month end in July, 45 00:02:35,990 --> 00:02:52,880 and my trend barometer yesterday finished at 45 which is 46 00:02:52,880 --> 00:03:13,148 pretty neutral, probably more or less in line with what we see 47 00:03:13,148 --> 00:03:35,105 in terms of the indices. And for those who listen on a regular 48 00:03:35,105 --> 00:03:53,684 basis, you'll notice that I skipped the ‘what's been on your 49 00:03:53,684 --> 00:04:13,952 radar’ because we’ve got so much stuff to cover with Dave and 50 00:04:13,952 --> 00:04:35,909 Rich that I thought we would just simply go onto that a little 51 00:04:35,909 --> 00:04:40,976 bit quicker today. Now we're recording on Thursday. We’ve got 52 00:02:28,935 --> 00:02:50,010 one and a half days left of the month of July, so we'll see how 53 00:02:50,010 --> 00:03:08,275 it goes. But so far the data looks like the following, and this 54 00:03:08,275 --> 00:03:23,730 is as of Tuesday evening. BTOP50, down 23 basis points, up 55 00:03:23,730 --> 00:03:41,995 8.32% for the year. SG CTA index down 79 basis points for the 56 00:03:41,995 --> 00:03:58,855 month, and up 8.50% for the year. SocGen Trend down 75 basis 57 00:03:58,855 --> 00:04:15,715 points for the month, up 8.3% for the year. And the Short-Term 58 00:04:15,715 --> 00:04:28,360 Traders Index (that's struggling a bit more), down 1.71% 59 00:04:28,360 --> 00:04:41,005 in July, and up 3.34% so far this year. In the traditional world, 60 00:02:25,408 --> 00:02:45,909 MSCI world is down 1.63%, and if we exclude US and Canada (which 61 00:02:45,909 --> 00:03:07,987 was suggested by Andrew - not the Andrew that comes on as a co-host 62 00:03:07,987 --> 00:03:26,911 but another Andrew) that is down 26 basis points only for the 63 00:03:26,911 --> 00:03:47,412 month, and up 9.56% so far this year. The S&P US Aggregate Bond 64 00:03:47,412 --> 00:04:07,913 Index down 91 basis points, so no relief on that, and it's flat 65 00:04:07,913 --> 00:04:29,991 for the year. And the S&P 500 Total Return down 2.39% so far in 66 00:04:29,991 --> 00:04:41,030 July, up 7.58% so far this year. Before we dive into all the 67 00:02:36,476 --> 00:03:08,332 topics, just wanted to ask whether there was anything from the 68 00:03:08,332 --> 00:03:37,292 month of July. I mean obviously there's been some activity 69 00:03:37,292 --> 00:04:14,940 I think, to say the least, over in Taiwan if I'm not mistaken. 70 00:04:14,940 --> 00:04:41,004 But anything else going on from your vantage point? 71 00:04:41,680 --> 00:04:46,750 Well, there's a lot going on in my world. So, you've got sort 72 00:04:46,750 --> 00:04:50,260 of multi-year, multi-decade highs in long-dated bond yields in 73 00:04:50,260 --> 00:04:55,330 Japan, in the US, in Britain. You've got a whole bunch of central 74 00:04:55,330 --> 00:04:59,620 bank policy meetings over the end of the month. Obviously, the 75 00:04:59,620 --> 00:05:03,130 Fed overnight last night, standing pat with three dissents 76 00:05:03,130 --> 00:05:07,030 in the vote. Very unusual there. And a huge disappointment 77 00:05:07,030 --> 00:05:12,880 in the bond market. You've had a ton of noise out here, as you mentioned, 78 00:05:12,880 --> 00:05:14,830 across some of the equity… 79 00:05:14,910 --> 00:05:18,100 Yeah, I said Taiwan, Dave, but I meant Korea, didn't I? 80 00:05:18,110 --> 00:05:24,457 Korea, but Taiwan also. Taiwan hasn't been quiet. Japan hasn't been 81 00:05:24,457 --> 00:05:30,804 quiet. And so in particular the big semi chip manufacturers in 82 00:05:30,804 --> 00:05:36,574 Japan, in Taiwan, in Korea. But Korea most particularly, where 83 00:05:36,574 --> 00:05:42,921 you've had this proliferation of levered ETFs that has fueled this 84 00:05:42,921 --> 00:05:48,114 boom/bust there that's been extraordinary. I was just looking 85 00:05:48,114 --> 00:05:55,615 at it is these two times levered ETFs in Hong Kong on the 86 00:05:55,615 --> 00:06:02,539 listed chip makers in Korea. So, one of those being SK Hynix. 87 00:05:18,110 --> 00:05:40,327 So just over the last three months, SK Hynix itself as of the 88 00:05:40,327 --> 00:06:02,544 close today is up 1%. The two times levered ETF is down 47%. 89 00:06:03,200 --> 00:06:03,680 Wow. 90 00:06:03,840 --> 00:06:08,268 And so, it's such a perfect example, and I'm sure we'll talk 91 00:06:08,268 --> 00:06:12,696 about it, of what in the business is known as volatility drag. 92 00:06:12,696 --> 00:06:17,124 And so, at some point in that three month period, the underlying 93 00:06:17,124 --> 00:06:22,659 stock was up as much as 125% and is now all the way back down 94 00:06:22,659 --> 00:06:27,456 to up 1%. But the volatility drag of the 2x leverage, which is 95 00:06:27,456 --> 00:06:32,991 a 2x the daily move, not 2x over the period, has been up as much 96 00:06:32,991 --> 00:06:38,157 as 300%. It is now down 47% from the start. And something I say 97 00:06:38,157 --> 00:06:42,585 all the time, and Rich will get this particularly, over a long 98 00:06:42,585 --> 00:06:45,537 enough time horizon, all levered ETFs approach zero. 99 00:06:47,650 --> 00:06:52,726 Not to hijack the conversation on that topic right now, but you 100 00:06:52,726 --> 00:06:56,956 kind of wonder why authorities keep giving people these nuclear 101 00:06:56,956 --> 00:07:00,763 instruments, right? I mean you really have to wonder. 102 00:07:01,580 --> 00:07:06,176 In fact, Korea's talking a lot today. There's an article today from 103 00:07:06,176 --> 00:07:10,772 one of the big normal ETF managers in Korea saying we shouldn't 104 00:07:10,772 --> 00:07:14,602 allow these things. These things are again, they're just volatility 105 00:07:14,602 --> 00:07:18,432 drag. It's just transferring wealth away from the end investor 106 00:07:18,432 --> 00:07:22,645 and it's inevitable that way. It's simply the mathematics of geometric 107 00:07:22,645 --> 00:07:27,241 returns and the impact of that variance on the draw down versus 108 00:07:27,241 --> 00:07:31,837 the draw up over time. It's inevitable that they go to zero. 109 00:07:32,230 --> 00:07:36,845 Yeah. Anything in your part of the world, Rich, that we want to 110 00:07:36,845 --> 00:07:38,620 highlight before we move on? 111 00:07:39,190 --> 00:07:43,348 It's just been a volatile time for us, Niels. It's a bit like, what 112 00:07:43,348 --> 00:07:46,615 is it, those magnificent men in their flying machines, goes up, 113 00:07:46,615 --> 00:07:50,773 and then it goes down, and it goes up, and then turns around. And 114 00:07:50,773 --> 00:07:54,040 so, that's what we're getting for this month, especially with the 115 00:07:54,040 --> 00:07:57,307 KOSPI, that was a significant decline. So, it's been a challenging 116 00:07:57,307 --> 00:07:59,089 couple of months for us, Niels. 117 00:08:00,830 --> 00:08:20,390 Yeah, absolutely. All right, let's move on. Let me try and set 118 00:08:20,390 --> 00:08:43,210 up the conversation the best I can and then we move into these questions, 119 00:08:43,210 --> 00:09:01,140 topics that we prepared. So, obviously, Rich, you've been a regular 120 00:09:01,140 --> 00:09:20,700 contributor to the podcast for many years, and Dave has been one 121 00:09:20,700 --> 00:09:40,260 of the most popular guests each time he has joined this show. 122 00:09:40,260 --> 00:09:58,190 But something has been kind of nagging me over the years. Now, you 123 00:08:04,678 --> 00:08:20,070 both come at markets from completely different directions. 124 00:08:20,070 --> 00:08:41,234 So, Dave, from the world of options, convexity and tail protection. 125 00:08:41,234 --> 00:09:00,474 And Rich, you come from trend following and quantitative research. 126 00:09:00,474 --> 00:09:12,018 Different backgrounds, different tools, different terminology 127 00:09:12,018 --> 00:09:35,106 entirely, actually. And yet when I think back to all those conversations 128 00:09:35,106 --> 00:09:58,194 we've had, the two of you sound like echoes of each other. 129 00:08:00,830 --> 00:08:18,980 So, I went looking for why, and I think I found it. Both of you 130 00:08:18,980 --> 00:08:31,080 see markets as complex adaptive systems rather than as statistical 131 00:08:31,080 --> 00:08:41,970 distributions to be sampled and estimated. That one commitment, 132 00:08:41,970 --> 00:08:55,280 and almost everything else you believe, seems to follow from that. 133 00:08:55,280 --> 00:09:11,010 So that is what I want to explore today; take that idea, examine 134 00:09:11,010 --> 00:09:24,320 it properly, and follow it wherever it goes. Because it takes 135 00:09:24,320 --> 00:09:36,420 both of you to the same uncomfortable conclusion that the 136 00:09:36,420 --> 00:09:48,520 wrong mathematics is being deployed in markets and that risk 137 00:09:48,520 --> 00:09:58,200 is not volatility, it's something deeper than that. So, I'm 138 00:08:03,728 --> 00:08:22,565 going to put questions to each of you, knowing full well that we 139 00:08:22,565 --> 00:08:39,953 will most likely end up going in a little bit different directions 140 00:08:39,953 --> 00:08:57,341 and it'll be more of a fluent conversation. But anyways, I hope 141 00:08:57,341 --> 00:09:19,076 you're up for it. And this may turn out to be kind of the lecture 142 00:09:19,076 --> 00:09:35,015 where the student, being me, has two professors to ask questions 143 00:09:35,015 --> 00:09:52,403 to and will feel a little bit underprepared as the lecture begins. 144 00:09:52,403 --> 00:09:58,199 But there we are. 145 00:09:59,360 --> 00:09:59,879 Wonderful. 146 00:09:59,879 --> 00:10:04,433 Anyways, let's start with kind of the shared lineage. So, Dave, 147 00:10:04,433 --> 00:10:09,815 what do you actually mean when you say that market is a complex 148 00:10:09,815 --> 00:10:14,369 adaptive system? And what does the conventional view assume it to 149 00:10:14,369 --> 00:10:14,783 be? 150 00:10:16,010 --> 00:10:20,703 I think I'll give Rich his flowers. I think he said it best, 151 00:10:20,703 --> 00:10:25,035 maybe ever, on one of your interviews with him when he said 152 00:10:25,035 --> 00:10:28,645 that the future isn't hidden, it hasn't been written yet. 153 00:10:29,290 --> 00:10:29,770 True. 154 00:10:29,770 --> 00:10:48,118 We don't know what it is because there's too many moving parts. 155 00:10:48,118 --> 00:11:14,111 And so, the way we think of it and we talk about it here is Ed Lorenz's 156 00:11:14,111 --> 00:11:29,401 Butterfly Wings, that it's infinite paths inside a finite space 157 00:11:29,401 --> 00:11:43,162 chasing an unattainable strange attractor. And our job is 158 00:11:43,162 --> 00:11:58,452 always to do everything we can to understand initial conditions 159 00:11:58,452 --> 00:12:16,800 and where we are; understand the now and then be resilient to 160 00:12:16,800 --> 00:12:22,916 an unknowable future path. And so, you guys have heard before, everybody's 161 00:10:39,002 --> 00:10:51,696 heard my race car analogy. It's not about forecasting the future 162 00:10:51,696 --> 00:11:04,390 circumstances of the race. It's about making your car as resilient 163 00:11:04,390 --> 00:11:20,546 as it can possibly be and then being able to react and respond to 164 00:11:20,546 --> 00:11:33,240 the unknowable path of the future. Because the future is so 165 00:11:33,240 --> 00:11:47,088 much more than simply this shape of the race course. It's what 166 00:11:47,088 --> 00:11:58,628 other drivers doing. It's what you're doing. It's weather, it's 167 00:11:58,628 --> 00:12:12,476 wetness, it's wear and tear. It's all these things that are impossible 168 00:12:12,476 --> 00:12:22,862 to predict that all interact and affect each other. And, as Rich 169 00:10:33,163 --> 00:10:45,604 will say so eloquently, the emergent outcome of all those different 170 00:10:45,604 --> 00:10:58,045 parts that create a whole that is unexplainable by the individual 171 00:10:58,045 --> 00:11:13,879 parts. And so, we see the world that way and then we see our 172 00:11:13,879 --> 00:11:27,451 task on behalf of the investors that we work with is to 173 00:11:27,451 --> 00:11:39,892 prepare them with a car that has braking capacity, that makes 174 00:11:39,892 --> 00:11:53,464 it the best car for achieving that resilience and the ability to 175 00:11:53,464 --> 00:12:05,905 accelerate and decelerate and react as things occur in the ever 176 00:12:05,905 --> 00:12:14,953 evolving initial conditions chasing that strange attractor through 177 00:12:14,953 --> 00:12:22,870 this infinite loop of a finite universe. 178 00:12:23,330 --> 00:12:27,300 Yeah, absolutely. Now, as Dave alluded to, Rich, you've arrived 179 00:12:27,300 --> 00:12:32,064 at the same place but in a completely kind of different way. 180 00:12:32,064 --> 00:12:34,446 What has convinced you about this? 181 00:12:35,090 --> 00:12:39,133 Well, first big fan of yours, Dave, have been a long-term big fan 182 00:12:39,133 --> 00:12:43,176 of yours and sort of, I'll quietly go to sleep at night reading 183 00:12:43,176 --> 00:12:47,530 one of your memos, and yeah, I just feel as though we're very much 184 00:12:47,530 --> 00:12:49,085 aligned in the same space. 185 00:12:50,050 --> 00:13:46,594 But my interest stems from a strong passion in science which co-evolved 186 00:13:46,594 --> 00:14:33,714 with my strong interest in finance. But recognizing that the 187 00:14:33,714 --> 00:15:25,546 two weren't aligned. I think you see that science has progressed 188 00:15:25,546 --> 00:16:17,378 from the world of linear outcomes, the world of Newton, into 189 00:16:17,378 --> 00:17:09,210 a world that's much more dynamic, a lot less stationary. It's 190 00:17:09,210 --> 00:18:05,754 a world of reflexivity, where matter tells space how to curve and 191 00:18:05,754 --> 00:18:34,026 space tells matter how to move. There's this combined reflexivity 192 00:13:12,982 --> 00:14:04,579 which Einstein found which overturned Newton. And that, therefore, 193 00:14:04,579 --> 00:15:01,909 changed concepts such as what is length? Because length became 194 00:15:01,909 --> 00:15:59,239 a very frame dependent posture. And I feel that finance 195 00:15:59,239 --> 00:17:08,035 is sort of maybe 100 years behind where the science is today 196 00:17:08,035 --> 00:18:16,831 in their understanding of what these systems are when we talk about 197 00:18:16,831 --> 00:18:34,030 Complex adaptive systems. When we use the word complex and when 198 00:13:33,042 --> 00:14:32,156 we use the word adaptive, I don't think people really understand 199 00:14:32,156 --> 00:15:25,896 what we're talking about there. We are truly talking about 200 00:15:25,896 --> 00:15:58,140 complexity, and non-stationarity, and evolution of 201 00:15:58,140 --> 00:16:46,506 systems. We're talking about processes here. We're not talking 202 00:16:46,506 --> 00:17:50,994 about a world of things, we're talking about a world of processes. 203 00:17:50,994 --> 00:18:33,986 And these processes are very contingent on constraints. And when 204 00:12:54,168 --> 00:13:22,994 you map all of the things that can affect an outcome, which is typically 205 00:13:22,994 --> 00:13:45,643 what we see in the statistical residue of outcomes, that's what 206 00:13:45,643 --> 00:14:10,351 I call a statistical residue of a process. The deep process itself 207 00:14:10,351 --> 00:14:33,000 shows that there are so many multi variables affecting the outcome 208 00:14:33,000 --> 00:14:55,649 that it is impossible to forecast an outcome with a degree 209 00:14:55,649 --> 00:15:22,416 of fidelity. And so, the way I see these systems, finance sort of 210 00:15:22,416 --> 00:15:47,124 has gone back to the world of Euclid linearity, the straight line. 211 00:15:47,124 --> 00:16:07,714 I think Mandelbrot first exposed that in 1960, showing that 212 00:16:07,714 --> 00:16:30,363 there was roughness at every scale in the financial markets, a 213 00:16:30,363 --> 00:16:55,071 bit like the coastlines of England. And you couldn't use a standard 214 00:16:55,071 --> 00:17:17,720 ruler to measure at each scale because each gave different results. 215 00:17:17,720 --> 00:17:40,369 You had to look at the dimensions or the scaling property 216 00:17:40,369 --> 00:18:00,959 that existed between those links to then understand how that 217 00:18:00,959 --> 00:18:23,608 system sort of could be understood in terms of its volatility 218 00:18:23,608 --> 00:18:33,903 and those sort of things. But finance has really resisted that 219 00:13:14,044 --> 00:14:06,031 move for such a long time. It's stuck in this world of Euclid 220 00:14:06,031 --> 00:14:54,019 where it's maths, it's a correct form of maths, but it's not 221 00:14:54,019 --> 00:15:42,007 a maths to apply to this domain we call the reality, which 222 00:15:42,007 --> 00:16:17,998 is the financial markets we observe today. Mandelbrot exposed 223 00:16:17,998 --> 00:17:01,987 that. And I think there's two key things. Firstly, there's roughness 224 00:17:01,987 --> 00:17:41,977 at every scale. There's a fractal geometry, there's these attractors, 225 00:17:41,977 --> 00:18:29,965 as Dave so eloquently put. We have a system that is computationally 226 00:18:29,965 --> 00:18:33,964 irreducible. So, what that means is, as Dave mentioned, the 227 00:13:28,741 --> 00:14:11,731 complexity and the decisions that especially in a reflexive system 228 00:14:11,731 --> 00:14:59,020 where you get the agents within the system reacting and adapting 229 00:14:59,020 --> 00:15:46,309 to the system and the system forcing that adaption and reaction, 230 00:15:46,309 --> 00:16:20,701 there's this reflexivity involved which makes our ability 231 00:16:20,701 --> 00:16:59,392 to deal with simple linear rulers non-functional. We'll find 232 00:16:59,392 --> 00:17:46,681 that the system that we are inquiring and investigating is inherently 233 00:17:46,681 --> 00:18:16,774 non-linear, it's unpredictable. The future is being 234 00:18:16,774 --> 00:18:33,970 written at every step. And to understand what happens in five days 235 00:13:30,514 --> 00:14:21,094 time, you need to understand the decisions and consequences of 236 00:14:21,094 --> 00:15:21,790 all of the agents who will be transacting over that period and 237 00:15:21,790 --> 00:16:12,370 their conglomerated positions and all of their effects, which is 238 00:16:12,370 --> 00:17:08,008 incalculable from the domain that sits in the now. And therefore, 239 00:17:08,008 --> 00:18:18,820 it's not like a future that is hidden in fog. It's a future that 240 00:18:18,820 --> 00:18:33,994 is inherently unpredictable. So, I view these systems, particularly 241 00:13:12,982 --> 00:13:51,202 these, what we call complex adaptive systems, which are particularly 242 00:13:51,202 --> 00:14:33,244 inherent in living systems where we have agents who react to 243 00:14:33,244 --> 00:15:15,286 that system and the system reacts to those agents. There's this 244 00:15:15,286 --> 00:15:53,506 dynamic reflexivity. When we have that occurring, which is typically 245 00:15:53,506 --> 00:16:35,548 in living systems, we find that they are complex adaptive systems. 246 00:16:35,548 --> 00:17:21,412 A bit different to, say, the weather, which is a physical system. 247 00:17:21,412 --> 00:18:03,454 The molecules in the weather, they're not reflexive in the same 248 00:18:03,454 --> 00:18:34,030 sense that an agent in our market is. I go back to that famous 249 00:13:16,174 --> 00:14:04,068 quote by Richard Feynman. He said, imagine how hard physics would 250 00:14:04,068 --> 00:14:56,316 be if electrons have feelings. Now, that is what the market is. 251 00:14:56,316 --> 00:15:48,564 And that's why it is a complex system and incredibly difficult to 252 00:15:48,564 --> 00:16:36,458 crack. And despite the Nobel laureates we've had, the very clever 253 00:16:36,458 --> 00:17:24,352 quants you will be really surprised at the fact that there 254 00:17:24,352 --> 00:18:07,892 is very little evidence that anyone has significantly beaten the 255 00:18:07,892 --> 00:18:34,030 market. The market is the master. rememberthatmovieMarginCall,withKevinSpaceyandJeremyIrons.JeremyIronswaseatinghissteakrecountingthelitanyofcollapsesthathaveoccurredoverthelast200yearsandhowtheseverycleverpeoplehavealwayssortofexperiencedthesecollapseswithoutbeingabletoresolvewhytheseoccurred.Andit'sbecausethey'reusingamathematicsthatisnotappropriateforthedomain-therealityweexistin.Sothat'swheremyinterestis.That'swhyIthinkDaveandIareveryalignedinourviews. 256 00:18:36,760 --> 00:18:38,069 I remember that movie Margin Call, with Kevin Spacey and Jeremy 257 00:18:38,069 --> 00:18:39,378 Irons. Jeremy Irons was eating his steak recounting the litany of 258 00:18:39,378 --> 00:18:40,806 collapses that have occurred over the last 200 years and how these 259 00:18:40,806 --> 00:18:41,996 very clever people have always sort of experienced these collapses 260 00:18:41,996 --> 00:18:43,305 without being able to resolve why these occurred. And it's because 261 00:18:43,305 --> 00:18:44,614 they're using a mathematics that is not appropriate for the domain 262 00:18:44,614 --> 00:18:46,161 - the reality we exist in. So that's where my interest is. That's 263 00:18:46,161 --> 00:18:47,589 why I think Dave and I are very aligned in our views. 264 00:18:48,680 --> 00:19:01,840 Yeah, well, excellent, great stuff. All right, so obviously we're 265 00:19:01,840 --> 00:19:17,632 in the world of investing that is directional. We like when price 266 00:19:17,632 --> 00:19:33,424 moves. So let me sort of focus on that a little bit. And we know 267 00:18:50,186 --> 00:18:56,712 that also, from the work of people like Mike Green, that a lot 268 00:18:56,712 --> 00:19:02,736 of money has been flowing into the markets recently that never looks 269 00:19:02,736 --> 00:19:09,262 really at the price, but it has an impact on price, I guess. 270 00:19:09,262 --> 00:19:14,784 So, I wonder maybe from your perspective, Dave, what does that 271 00:19:14,784 --> 00:19:21,812 do to price that you have this, and I call it new but it's 272 00:19:21,812 --> 00:19:28,338 not that new, but it's been there for a little while, this type 273 00:19:28,338 --> 00:19:32,856 of investor that looks at price differently, meaning price 274 00:19:32,856 --> 00:19:33,358 insensitive. 275 00:19:35,200 --> 00:19:41,992 Yeah, this is something we could go on about for the entire 276 00:19:41,992 --> 00:19:48,218 time period. As I think you know, I'm reasonably good friends 277 00:19:48,218 --> 00:19:55,576 with Mike Green and with Hari Krishnan. And huge fan of their work 278 00:19:55,576 --> 00:20:01,802 and their most recent paper on passive investing and equity flows, 279 00:20:01,802 --> 00:20:08,594 which they were kind enough to even give me an acknowledgment in 280 00:20:08,594 --> 00:20:14,820 for some feedback I had provided. I'm known for years and 281 00:20:14,820 --> 00:20:21,612 years of what some people fondly refer to as Dredgisms in the 282 00:20:21,612 --> 00:20:27,838 world. And one of my famous Dredgisms is, ‘positioning is the 283 00:20:27,838 --> 00:20:34,630 only thing that matters’. And I love all the time where people 284 00:20:34,630 --> 00:20:41,988 want to justify why a market moved, or this, or that reason. I 285 00:20:41,988 --> 00:20:48,214 say that positioning is the only thing that matters and sort 286 00:20:48,214 --> 00:20:51,044 of that passive flow dynamic. And again, something that I know 287 00:19:39,940 --> 00:19:49,420 you and Rich talk about all the time, talking about price insensitive 288 00:19:49,420 --> 00:19:54,950 investors versus price sensitive investors. Similarly, when 289 00:19:54,950 --> 00:20:03,640 we model markets, we break markets into, and again this will 290 00:20:03,640 --> 00:20:13,910 sound very familiar to you guys, we break them into what I call 291 00:20:13,910 --> 00:20:21,020 rational accounting man, which would be price insensitive people 292 00:20:21,020 --> 00:20:28,130 that are following accounting rules and regulation or instructions 293 00:20:28,130 --> 00:20:36,030 that are non-economic related, and what we would call boundedly 294 00:20:36,030 --> 00:20:43,930 rational agents. People who are making decisions based upon the 295 00:20:43,930 --> 00:20:49,460 risk they're taking, skin-in-the-game, their own capital, 296 00:20:49,460 --> 00:20:51,040 tacit knowledge. 297 00:20:54,670 --> 00:20:58,030 And what we would call boundedly rational agents. People 298 00:20:58,030 --> 00:21:01,910 who are making decisions based upon the risk they're taking skin 299 00:21:01,910 --> 00:21:05,710 in the game, their own capital, tacit knowledge, right? 300 00:21:06,990 --> 00:21:11,230 And we model things that way. And so you can think of that passive 301 00:21:11,470 --> 00:21:15,950 as a subset of bounded of sorry, rational accounting man. And 302 00:21:00,486 --> 00:22:16,094 we model things that way. And so, you can think of that passive 303 00:22:16,094 --> 00:23:14,254 as a subset of rational accounting man. And rational accounting 304 00:23:14,254 --> 00:24:12,414 man dominates most markets because all bank traders, all pension 305 00:24:12,414 --> 00:25:04,758 fund traders, all insurance traders, all central bank traders, 306 00:25:04,758 --> 00:25:51,286 all FX reserve managers are rational accounting men. When they 307 00:21:01,650 --> 00:21:40,040 buy something, they're not thinking about the value of it for 308 00:21:40,040 --> 00:22:07,960 their retirement, they're thinking about the regulatory requirements 309 00:22:07,960 --> 00:22:42,860 around accounting and risk and usually an annual compensation period. 310 00:22:42,860 --> 00:23:17,760 That is a very different thought process than somebody who's 311 00:23:17,760 --> 00:23:59,640 trying to make a 30 year investment decision to fund his retirement 312 00:23:59,640 --> 00:24:34,540 where losses are his and failures to achieve his investment 313 00:24:34,540 --> 00:25:09,440 goals impair his own lifestyle. Those are two very different 314 00:25:09,440 --> 00:25:51,320 decision making and very much what Rich talks about all the time. 315 00:20:54,670 --> 00:21:36,194 Now, one of the things I stress a lot to Mike and Hari on 316 00:21:36,194 --> 00:22:14,752 every time we talk about their work is yes, I agree, but you're 317 00:22:14,752 --> 00:22:50,344 missing the biggest elephant in the room, which is the bond market. 318 00:22:50,344 --> 00:23:22,970 They talk about the percent of passive taking over and eventually 319 00:23:22,970 --> 00:23:55,596 becoming a problem when it gets beyond a certain amount above 320 00:23:55,596 --> 00:24:28,222 and beyond what we would call boundedly rational, which you might 321 00:24:28,222 --> 00:25:00,848 call active agents. Well, the bond market went well through that 322 00:25:00,848 --> 00:25:45,338 at the peak of what I call the mother of all bubbles of bond markets, 323 00:25:45,338 --> 00:25:51,270 late 2020. And we saw the implications of that, and I know 324 00:21:28,380 --> 00:22:02,090 you guys benefited on behalf of your investors, the implications 325 00:22:02,090 --> 00:22:42,542 of at least the initial shock unwind of that imbalance in 2022 326 00:22:42,542 --> 00:23:26,365 when interest rates went up and you had one of the biggest opportunities 327 00:23:26,365 --> 00:24:13,559 in trend that you've had in a long time. And I'll argue we're still 328 00:24:13,559 --> 00:24:50,640 in a big problem of unwinding that complexity. Because those rational 329 00:24:50,640 --> 00:25:24,350 accounting men, even in losses, do not behave like price 330 00:25:24,350 --> 00:25:51,318 sensitive investors because they don't mark the market. The central 331 00:21:01,568 --> 00:21:39,507 banks who bought all the bonds are sitting on massive unrealized 332 00:21:39,507 --> 00:22:13,997 losses. The FX reserve managers are sitting on massive unrealized 333 00:22:13,997 --> 00:22:48,487 losses. The banks are sitting on massive unrealized losses. The 334 00:22:48,487 --> 00:23:26,426 pension funds and insurance companies, that are a big part of 335 00:23:26,426 --> 00:24:07,814 some of the other side of stuff we do, with duration extending 336 00:24:07,814 --> 00:24:35,406 callable note structures, with embedded Bermuda swaptions, that 337 00:24:35,406 --> 00:25:09,896 are masking duration in the form of structured complex derivative 338 00:25:09,896 --> 00:25:40,937 products, have really massive unrealized losses and nobody's cut 339 00:25:40,937 --> 00:25:51,284 the positions yet. Nobody's freed the capital that's tied up 340 00:21:45,315 --> 00:23:04,900 in those things or the lack of capital. Nobody's recognized the 341 00:23:04,900 --> 00:24:38,955 accounting losses. And so, to me, this is one of the biggest drivers 342 00:24:38,955 --> 00:25:51,305 of all things in the financial system is exactly this. It's exactly 343 00:21:05,078 --> 00:22:07,526 what you and Rich talk about all the time, that interaction and 344 00:22:07,526 --> 00:22:59,566 the growing dominance, at times, of the price insensitive investor, 345 00:22:59,566 --> 00:23:46,402 and, of course, the application of leverage in particular 346 00:23:46,402 --> 00:24:43,646 in the regulated space where volatility is the measure of risk 347 00:24:43,646 --> 00:25:35,686 and low volatility means low risk and adding leverage doesn't 348 00:25:35,686 --> 00:25:51,298 increase the risk. The risk is the volatility, no matter how much 349 00:21:38,437 --> 00:22:31,930 leverage you apply to the super senior tranche of a subprime 350 00:22:31,930 --> 00:23:30,286 CDO. Now obviously Rich and I know very well that the leverage 351 00:23:30,286 --> 00:24:33,505 is the risk. It's the reason why it's a left skewed return distribution. 352 00:24:33,505 --> 00:25:31,861 It's the reason why tails are far fatter than a Gaussian normal 353 00:25:31,861 --> 00:25:51,313 distribution would hint at. The lack of capital is the strange 354 00:21:10,763 --> 00:21:31,454 attractor, at times, pulling the market that direction because 355 00:21:31,454 --> 00:21:56,743 of the proliferation of loss management as that leverage has to 356 00:21:56,743 --> 00:22:24,331 get cleansed from the system. So, to me that's what moves prices. 357 00:22:24,331 --> 00:22:45,022 It's the accumulating imbalances as some sort of price 358 00:22:45,022 --> 00:23:05,713 insensitive flow builds, and builds on itself, reflexively building 359 00:23:05,713 --> 00:23:28,703 as the application of leverage through a flawed regulatory construct, 360 00:23:28,703 --> 00:23:47,095 not just regulatory, but flawed understanding, builds that 361 00:23:47,095 --> 00:24:16,982 and then reverses the other way with the fat tail on the other 362 00:24:16,982 --> 00:24:42,271 side as the mistake becomes recognized, the king has no clothes. 363 00:24:42,271 --> 00:25:07,560 The knowledge game recognizes that problem and it moves back the 364 00:25:07,560 --> 00:25:42,045 other way and in what will be a very fat tail. And that's the way 365 00:25:42,045 --> 00:25:51,241 we see the world. 366 00:25:51,730 --> 00:25:56,170 Yeah, absolutely. Now you did an interesting exercise, Rich, because 367 00:25:56,170 --> 00:26:01,054 you built, “a market where nobody really is responding to price”. 368 00:26:01,054 --> 00:26:02,830 What happened with that? 369 00:26:03,170 --> 00:26:45,510 Well, I suppose this model was influenced by Jean-Philippe Bouchaud's 370 00:26:45,510 --> 00:27:23,616 sort of examination of market microstructure, what moves price? 371 00:27:23,616 --> 00:28:10,190 And what I'd say is that, well, it's the transaction event 372 00:28:10,190 --> 00:28:56,764 that moves price. And all other theories, you might have fundamentals, 373 00:28:56,764 --> 00:29:34,870 you might have macro investing, you might have liquidity 374 00:29:34,870 --> 00:30:21,444 flows, all of these different theories, they're what I regard as 375 00:30:21,444 --> 00:30:59,550 downstream of the transaction. The thing, the mechanistic thing 376 00:30:59,550 --> 00:31:37,656 that is actually moving price is the transaction itself. So, the 377 00:26:12,726 --> 00:27:14,840 way I see these markets is a composite, it's an ecosystem of participants 378 00:27:14,840 --> 00:28:07,398 in those markets and each of those particular species in that 379 00:28:07,398 --> 00:28:55,178 ecosystem, they create a directed force based on the models 380 00:28:55,178 --> 00:29:38,180 that they adopt in that market. So, fundamental investors 381 00:29:38,180 --> 00:30:25,960 will have a particular force implied to evaluation. Trend followers 382 00:30:25,960 --> 00:31:23,296 will be implying a sort of an amplifying force as they respond 383 00:31:23,296 --> 00:31:37,630 to price moving. So, therefore, it sort of comes to this 384 00:26:58,914 --> 00:28:29,498 point of what I call or it's feedback in the systems that creates 385 00:28:29,498 --> 00:29:39,178 this fractal structure and it's actually what is creating this 386 00:29:39,178 --> 00:31:09,762 movement in price. So, when I look at the feedback, I see there's 387 00:31:09,762 --> 00:31:37,634 two forms of feedback. There's the positive feedback or what I call 388 00:26:36,618 --> 00:27:18,428 amplifying feedback, and there is negative feedback, which is more 389 00:27:18,428 --> 00:28:04,419 the convergent premise, the reduction or the return to the mean, 390 00:28:04,419 --> 00:28:50,410 or the mean reverting, the counter trend fade. So, these two 391 00:28:50,410 --> 00:29:28,039 forces, divergent and convergent can be expressed by all 392 00:29:28,039 --> 00:30:09,849 of these different ecosystems of different species. And these ecosystems, 393 00:30:09,849 --> 00:30:51,659 and they all have these different forces being applied in 394 00:30:51,659 --> 00:31:25,107 the markets. Some are amplifying, some are destructive 395 00:31:25,107 --> 00:31:37,650 or mean reverting. Now, when I put an agent-based model together, 396 00:26:47,034 --> 00:27:58,313 looking at what I call… let's look at the species of a market, 397 00:27:58,313 --> 00:29:04,109 I'm going to call one the random trader. They are the trader 398 00:29:04,109 --> 00:30:04,422 who has no model that has a preference, a directional preference. 399 00:30:04,422 --> 00:31:04,735 They can be amplifying or they could be, negative feedback. They're 400 00:31:04,735 --> 00:31:37,633 what I call the random trader. There's a convergent trader who's 401 00:26:53,850 --> 00:28:35,210 always fading the move. So, the markets are oscillating, they're 402 00:28:35,210 --> 00:30:46,978 fading the move at each way, trying to bring things back to what 403 00:30:46,978 --> 00:31:37,658 they consider an equilibrium-based position. Then 404 00:26:07,514 --> 00:26:50,954 there is the price sensitive, they're price sensitive. And then 405 00:26:50,954 --> 00:27:34,394 there's another price sensitive agent who I call the divergent 406 00:27:34,394 --> 00:28:17,834 trader. They're the ones that amplify the move. Passive actually 407 00:28:17,834 --> 00:29:09,962 fits in that category. It's not that they are responding to price, 408 00:29:09,962 --> 00:29:53,402 they are price insensitive, they're responding to a rule. But 409 00:29:53,402 --> 00:30:45,530 that rule is invest in the biggest things. So, what that does 410 00:30:45,530 --> 00:31:28,970 is it puts an amplifying pressure or directional force into 411 00:31:28,970 --> 00:31:37,658 that market. So, with this agent-based model, what I did was 412 00:27:01,058 --> 00:28:05,378 I looked at just applying random participants in that model 413 00:28:05,378 --> 00:29:28,994 first and what I got was a Gaussian outcome. No fat tails, no 414 00:29:28,994 --> 00:30:26,882 memory, no volatility clustering. It was exactly what the 415 00:30:26,882 --> 00:31:24,770 outcomes you'd expect from the efficient market hypothesis, the 416 00:31:24,770 --> 00:31:37,634 Gaussian premise. But as soon as I started putting divergent participants 417 00:26:35,885 --> 00:27:26,775 in it, amplifying the move, I got to a position of about 25% and 418 00:27:26,775 --> 00:28:03,125 above of participants that are divergent. All of the signatures 419 00:28:03,125 --> 00:28:43,110 you see in the real market automatically come back into the 420 00:28:43,110 --> 00:29:23,095 market. You get fat tails, you get volatility clustering, you get 421 00:29:23,095 --> 00:30:03,080 the memory - market memory. All of these signatures come back 422 00:30:03,080 --> 00:30:39,430 in once the agent-based model integrates a small percentage, only 423 00:30:39,430 --> 00:31:15,780 about 25% of these divergent participants. And passive, of course, 424 00:31:15,780 --> 00:31:37,590 is a divergent participant as well. So, what I'm saying is that 425 00:26:35,540 --> 00:27:51,070 over the last 40 years this has been the case and this is why 426 00:27:51,070 --> 00:28:28,835 markets exhibit these properties: fat tails, volatility 427 00:28:28,835 --> 00:29:22,785 clusters that maybe the orthodoxy has sort of refused to 428 00:29:22,785 --> 00:30:16,735 believe persists and they're regarding them as anomalies as opposed 429 00:30:16,735 --> 00:31:10,685 to a fundamental feature of market structure. So, that's what 430 00:31:10,685 --> 00:31:37,660 I found in my modeling. And it was very interesting that it wasn't 431 00:26:50,954 --> 00:28:08,603 a gradual sort of bring back of the signatures of the real market. 432 00:28:08,603 --> 00:29:26,252 It was a transition event. It was once we achieved this 25% influence 433 00:29:26,252 --> 00:30:31,955 of divergent participants in there, it was a threshold and we 434 00:30:31,955 --> 00:31:31,685 got the typical phase transition as opposed to a gradual 435 00:31:31,685 --> 00:31:37,658 change. So, before we get to that 25%, I can populate this simulated 436 00:27:20,354 --> 00:28:11,810 agent-based model with convergent traders – Gaussian, random 437 00:28:11,810 --> 00:29:22,562 traders – Gaussian, random and convergent - Gaussian. As soon as 438 00:29:22,562 --> 00:30:20,450 I start putting those divergent population in there of 439 00:30:20,450 --> 00:31:37,634 25% and above, boom, all the signatures come back into the market. 440 00:26:03,170 --> 00:27:23,151 So, I found that a very fascinating bit of research that 441 00:27:23,151 --> 00:28:50,403 just tells me that over the history of these markets, this is 442 00:28:50,403 --> 00:29:55,842 a reality and it's this amplifying position, that's what's 443 00:29:55,842 --> 00:31:01,281 creating the tails, that's what's creating the memory, that's 444 00:31:01,281 --> 00:31:37,636 what creates the volatility clusters. 445 00:31:38,470 --> 00:31:42,617 Yeah, another reality of markets is that, of course, a lot 446 00:31:42,617 --> 00:31:47,518 of the time it's small moves, noise, and then from time to time 447 00:31:47,518 --> 00:31:52,042 we get the really big moves. So, I'm curious, from your perspective, 448 00:31:52,042 --> 00:31:56,189 Dave, where do these actually come from, these sudden, bigger moves? 449 00:31:56,189 --> 00:31:58,074 How do they suddenly occur? 450 00:32:01,360 --> 00:32:34,192 Broadly speaking, it's leverage, but more simply speaking, 451 00:32:34,192 --> 00:33:15,232 it's loss absorbing capital. And so, people, because they're following 452 00:33:15,232 --> 00:33:48,064 very simplified rules of Gaussian linearity, ergotic single 453 00:33:48,064 --> 00:34:37,312 slices of time, they are taking more risk than they have the 454 00:34:37,312 --> 00:35:14,248 loss appetite for or the capital to absorb losses. And so, 455 00:32:08,640 --> 00:32:48,680 when that mistake gets exposed because one of Rich's divergent traders 456 00:32:48,680 --> 00:33:28,720 comes in or something happens moving it, triggers stop losses for 457 00:33:28,720 --> 00:34:05,120 the guys who miscalculated the risk. Now, obviously, the biggest 458 00:34:05,120 --> 00:34:48,800 one, I'll argue, maybe in the whole system is the assumptions around 459 00:34:48,800 --> 00:35:14,280 stable correlation and what amounts to diversification. And so, 460 00:32:08,012 --> 00:32:41,272 the world, the investment universe says, we're happy with our 461 00:32:41,272 --> 00:33:21,184 risk appetite to take this much risk on an assumption of the 462 00:33:21,184 --> 00:33:44,466 diversification benefit created across these different assets, 463 00:33:44,466 --> 00:34:14,400 strategies, geographies, etc. And then when that correlation, for 464 00:34:14,400 --> 00:34:47,660 whatever reason, misbehaves, they've got more risk than they want. 465 00:34:47,660 --> 00:35:14,268 And so now they need to reduce risk. Now, of course, in that reflexive 466 00:32:17,896 --> 00:32:53,724 world that Rich speaks so well of, their reduction of risk in response 467 00:32:53,724 --> 00:33:15,772 to inconsistent correlation affects the correlation, which affects 468 00:33:15,772 --> 00:33:54,356 the risk appetite of the next guy who then also has to adjust his 469 00:33:54,356 --> 00:34:24,672 risk, which then triggers the divergent trader who is saying, wait 470 00:34:24,672 --> 00:34:54,988 a minute, correlation's moving. I want to go with that moving 471 00:34:54,988 --> 00:35:14,280 correlation which then affects the next guy. And so, again, the 472 00:32:11,928 --> 00:32:46,274 way we very much look at the world, the residual risk in most 473 00:32:46,274 --> 00:33:17,978 people's portfolios, I say all the time, most people do a pretty 474 00:33:17,978 --> 00:33:49,682 good job of managing the risk they manage, or managing the risk 475 00:33:49,682 --> 00:34:26,670 they measure. But what they don't do a very good job of is managing 476 00:34:26,670 --> 00:34:55,732 the risk they assume, the risk they don't measure. And generally 477 00:34:55,732 --> 00:35:14,226 speaking, in large investment portfolios, that's correlation. And 478 00:32:04,932 --> 00:32:33,508 it's where they've made simplifying assumptions around correlation 479 00:32:33,508 --> 00:33:09,228 and volatility, which are really the two things that everyone 480 00:33:09,228 --> 00:33:37,804 assumes constant in their simplified modern portfolio theory, 481 00:33:37,804 --> 00:34:09,952 Efficient Market Hypothesis Methodology is the thing that catches 482 00:34:09,952 --> 00:34:56,388 them out. And now they have more risk than the capital they had 483 00:34:56,388 --> 00:35:14,248 allocated to absorb the losses. That, of course, gets more 484 00:32:12,980 --> 00:32:36,220 and more extreme as different types, either explicit leverage (in 485 00:32:36,220 --> 00:33:01,784 the form of banks, balance sheets or hedge funds) get utilized, 486 00:33:01,784 --> 00:33:25,024 or the implicit leverage, obviously, the world that we’re most 487 00:33:25,024 --> 00:33:45,940 active in of volatility selling. Which creates then mispriced 488 00:33:45,940 --> 00:34:13,828 tails and the need to respond to things that you hadn't previously 489 00:34:13,828 --> 00:34:34,744 anticipated, which then leads to above normal standard deviation, 490 00:34:34,744 --> 00:34:57,984 what you would assume normal standard deviation, market size moves. 491 00:34:57,984 --> 00:35:14,252 That's simply how we see the world. 492 00:35:14,370 --> 00:35:20,893 Yeah, Rich, these endogenous forces, can they be measured in some 493 00:35:20,893 --> 00:35:21,486 way? 494 00:35:22,610 --> 00:35:52,406 So, I go back to Dave's forest fire analogy. The big moves that 495 00:35:52,406 --> 00:36:15,326 occur in the market, I believe, have an endogenous source 496 00:36:15,326 --> 00:36:40,538 as opposed to an exogenous source. It's not coming from news. 497 00:36:40,538 --> 00:37:01,166 News isn't actually creating the significant material moves. It's 498 00:37:01,166 --> 00:37:33,254 a state of the market. And the news might be the trigger, but it's 499 00:37:33,254 --> 00:37:56,174 the state of the market. Whether there's the undergrowth that's 500 00:37:56,174 --> 00:38:23,678 been building (as Dave loves to say of his forest fire analogy 501 00:38:23,678 --> 00:38:35,138 and I totally understand that). So, when you look at research 502 00:35:33,506 --> 00:35:49,850 like Jean-Philippe Bouchaud, you see when he actually assessed 503 00:35:49,850 --> 00:36:08,010 whether it was news that was driving market microstructure, he 504 00:36:08,010 --> 00:36:29,802 came out explicitly saying that 90% of the moves were from endogenous 505 00:36:29,802 --> 00:36:55,226 sources. And what he was doing there at, I think, the tick by tick 506 00:36:55,226 --> 00:37:17,018 or the minute by minute, he was looking at the correlation between 507 00:37:17,018 --> 00:37:36,994 news events and market price movements from all sources of news 508 00:37:36,994 --> 00:37:56,970 events and seeing if there was a strong correlation between the 509 00:37:56,970 --> 00:38:22,394 two. And he couldn't find it. And this is what I also observe in 510 00:38:22,394 --> 00:38:35,106 markets. It's the state of the market. When it comes from endogenous 511 00:35:37,420 --> 00:36:07,040 sources, it's not coming from nowhere, it's coming from within 512 00:36:07,040 --> 00:36:42,584 the market itself. And you find that it doesn't arrive out of 513 00:36:42,584 --> 00:37:03,318 nowhere. It builds progressively. So, activity picks 514 00:37:03,318 --> 00:37:29,976 up beforehand. Liquidity quietly thins out. The book therefore 515 00:37:29,976 --> 00:38:05,520 gets very fragile. And then bang, it goes with a trigger. It 516 00:38:05,520 --> 00:38:35,140 often takes, also, longer to settle afterwards. It's not instantaneous. 517 00:35:22,610 --> 00:35:41,480 So, the traditional orthodoxy was that news events were instantaneously 518 00:35:41,480 --> 00:36:02,237 accepted by the market, prices would adjust, but that's not the 519 00:36:02,237 --> 00:36:26,768 way. There's a lag. And this lag, I think Gabaix and Koijen, and 520 00:36:26,768 --> 00:36:51,299 a lot of research has been done on this that says that markets 521 00:36:51,299 --> 00:37:12,056 are very inelastic in their properties. And I think that it's 522 00:37:12,056 --> 00:37:32,813 this memory that exists in the market. Markets have memory. And 523 00:37:32,813 --> 00:37:53,570 you find this when you examine what this volatility clustering is. 524 00:37:53,570 --> 00:38:18,101 When you do the research, you find that calm begets calm, and volatility 525 00:38:18,101 --> 00:38:35,084 begets volatility. They're amplifying the state of each other. 526 00:35:22,610 --> 00:36:05,829 Now, what happens during these calm periods? You get a progressive 527 00:36:05,829 --> 00:36:56,906 increase in the risk building up in the structure, and it's a transition 528 00:36:56,906 --> 00:37:40,125 between the calm state and the volatile state that catches everyone 529 00:37:40,125 --> 00:38:27,273 naked. And we'll go into that in later questions. But that's what 530 00:38:27,273 --> 00:38:35,131 I found. In my models I found that with no news events, I didn't 531 00:35:55,142 --> 00:36:24,963 need any news events in those agent-based models I was talking 532 00:36:24,963 --> 00:36:52,073 about. All I needed was the divergent participant, the convergent 533 00:36:52,073 --> 00:37:13,761 participant, the random participant. That's what was creating 534 00:37:13,761 --> 00:37:46,293 all of those signatures. It wasn't news that was doing that in 535 00:37:46,293 --> 00:38:16,114 my agent-based models. It was just the impact that they're exerting 536 00:38:16,114 --> 00:38:35,091 with their models. That's what I found. 537 00:38:37,341 --> 00:38:42,465 Now going into a little bit of deep water on my side because both 538 00:38:42,465 --> 00:38:47,223 of you from time to time bring up this word ergodicity. And I'm 539 00:38:47,223 --> 00:38:51,615 always kind of, what exactly is it we're talking about now? But 540 00:38:51,615 --> 00:38:56,007 anyways, from memory, I think it has something to do with ensemble 541 00:38:56,007 --> 00:38:59,301 averages versus time averages, something along those lines. Anyway, 542 00:38:59,301 --> 00:39:03,693 Dave, why is this important for people who have money at risk 543 00:39:03,693 --> 00:39:04,791 in the markets? 544 00:39:06,371 --> 00:39:12,275 I think it's important for a lot of things and certainly important 545 00:39:12,275 --> 00:39:17,687 for money at risk. And you're right, something that is ergodic, 546 00:39:17,687 --> 00:39:21,131 basically, time doesn't matter, sequencing doesn't matter. 547 00:39:21,131 --> 00:39:26,051 The ensemble average equals the time average. From a portfolio 548 00:39:26,051 --> 00:39:30,971 perspective it would be the arithmetic mean equals the geometric 549 00:39:30,971 --> 00:39:31,463 mean. 550 00:39:45,011 --> 00:39:45,187 Butinsomethingthatisnon-ergodic,thatis notthecase. 551 00:39:45,187 --> 00:39:45,587 And itturnsoutthatthepaththroughtimemakesanenormousdifference. Andso,theeasyoneforeveryonetothinkofisthe 552 00:39:45,587 --> 00:39:45,811 usualproblem,thata50% drawdownisn'tcorrectedwitha50%recovery.. 553 00:39:45,811 --> 00:39:46,291 True. 554 00:39:46,371 --> 00:40:29,700 You're not back. You’ve got to get 100% recovery to get back. It's 555 00:40:29,700 --> 00:41:06,363 our, again, our race car example where it's a multi-lap race 556 00:41:06,363 --> 00:41:36,360 and if you crash, you don't finish, it matters. Nassim always 557 00:39:48,989 --> 00:40:06,006 tells the story about people going to the casino and if you give 558 00:40:06,006 --> 00:40:24,332 30 people money to go play one casino game and they all come back, 559 00:40:24,332 --> 00:40:41,349 well, they're going to get the average. And if the 19th guy loses 560 00:40:41,349 --> 00:40:57,057 his money, it doesn't affect what happened to the 20th guy. But 561 00:40:57,057 --> 00:41:18,001 if you send one guy to go play 30 games and he loses all his money 562 00:41:18,001 --> 00:41:36,327 on the 19th game, he's out. He doesn't get to play the 20th game. 563 00:39:46,371 --> 00:40:05,566 And so that's the nature of the investment process because it's 564 00:40:05,566 --> 00:40:21,271 a path through time. Sequencing matters is, again, another 565 00:40:21,271 --> 00:40:38,721 great Nassim quote. The average of the condition of your 566 00:40:38,721 --> 00:40:59,661 trousers is the same whether you launder them first or iron them 567 00:40:59,661 --> 00:41:20,601 first. But the end outcome is very, very different. Which way around 568 00:41:20,601 --> 00:41:36,306 you do it? Sequencing matters. It's a non-ergodic thing. And I say 569 00:39:49,704 --> 00:40:05,258 it all the time, and I'm sure you guys have either seen me write 570 00:40:05,258 --> 00:40:19,701 or say it. I think Ole Peters, who's sort of the most prominent 571 00:40:19,701 --> 00:40:31,922 voice in the world of ergodicity right now in his book 572 00:40:31,922 --> 00:40:41,921 Ergodicity Economics, should be mandatory reading at every school 573 00:40:41,921 --> 00:40:51,920 level. In mathematics, in physics, in finance, in economics, 574 00:40:51,920 --> 00:41:03,030 in business, in everything. It is the most beautiful mathematical 575 00:41:03,030 --> 00:41:13,029 description of common sense, of bounded rationality, of tacit 576 00:41:13,029 --> 00:41:27,472 knowledge that you will ever read. And the fact that it is not 577 00:41:27,472 --> 00:41:36,360 universally accepted and universally known is a disgrace. 578 00:41:37,161 --> 00:41:41,517 Well said. Now Rich, you also worked with some examples in this 579 00:41:41,517 --> 00:41:46,962 and I seem to remember one of them was where you took like a four 580 00:41:46,962 --> 00:41:50,592 decades worth of returns and changed nothing except the order 581 00:41:50,592 --> 00:41:54,585 of those returns. Tell us a little bit about what happened. 582 00:41:56,291 --> 00:41:59,331 This is such a… 583 00:42:00,410 --> 00:42:54,350 I totally agree with Dave, here. Mandatory reading, it should 584 00:42:54,350 --> 00:43:59,078 be because geometric wealth is what we should all be aspiring to. 585 00:43:59,078 --> 00:44:58,412 And all of the traditional metrics, or many of the traditional 586 00:44:58,412 --> 00:45:52,352 metrics we find in finance, they avoid that path dependency. 587 00:42:00,410 --> 00:42:43,016 So, the popular statistics you'll hear about Sharpe, Sortino, 588 00:42:43,016 --> 00:43:30,356 variance at risk, standard deviation, even what you find in 589 00:43:30,356 --> 00:44:12,962 the capital asset pricing model with modern portfolio theory, 590 00:44:12,962 --> 00:45:05,036 no matter what order of returns you get, it doesn't change 591 00:45:05,036 --> 00:45:52,376 the outcome. You still get the same order, invariance, statistic. 592 00:42:00,410 --> 00:42:46,092 But what you find with path dependency is that the order is essential, 593 00:42:46,092 --> 00:43:24,746 the sequence of returns is essential. And what I found when 594 00:43:24,746 --> 00:44:03,400 I looked at 40 years of history, reshuffled it, because wealth 595 00:44:03,400 --> 00:44:35,026 production is a multiplicative process. So, you're dealing with 596 00:44:35,026 --> 00:45:06,652 something multiplied by an accumulated balance at each step 597 00:45:06,652 --> 00:45:38,278 along the way. And that accumulated balance really defines 598 00:45:38,278 --> 00:45:52,334 what will favor that. What happens to that accumulated balance? 599 00:42:36,092 --> 00:43:17,721 Does it deteriorate significantly, therefore the impact 600 00:43:17,721 --> 00:43:59,350 of compounding is significantly compromised? Does it 601 00:43:59,350 --> 00:44:52,873 accelerate rapidly, which means the impact of compounding is 602 00:44:52,873 --> 00:45:46,396 significantly enhanced? What's happening to that wealth, that accumulated 603 00:45:46,396 --> 00:45:52,343 balance? And if things change along that path, stuff is taken away, 604 00:42:53,562 --> 00:43:46,714 taxes, all of these things, stuff is added, redemptions are added 605 00:43:46,714 --> 00:44:35,034 into that wealth path. It totally changes the path between 606 00:44:35,034 --> 00:45:33,018 the world of what I call additive returns and the world I 607 00:45:33,018 --> 00:45:52,346 call path dependent returns. So, what I found when I reshuffled 608 00:42:28,403 --> 00:43:12,392 those returns in the process, all of the standard metrics, Sharpe, 609 00:43:12,392 --> 00:43:48,383 Sortino, variance at risk, standard deviation, they all stayed 610 00:43:48,383 --> 00:44:44,369 the same. But what did change was the wealth. We either had a decline 611 00:44:44,369 --> 00:45:44,354 of 30% and it ranged up to a decline of 84% dependent on the trajectory 612 00:45:44,354 --> 00:45:52,352 you took. This isn't a spread we're talking about. This is the 613 00:42:31,330 --> 00:43:08,434 difference between a bad year and not being in business. So, it 614 00:43:08,434 --> 00:43:42,446 is such an important property. And Mandelbrot of course, was right 615 00:43:42,446 --> 00:44:19,550 to talk about the fractal nature of markets and the scale invariant 616 00:44:19,550 --> 00:44:59,746 nature of markets. But a very important sequel to that is the geometry 617 00:44:59,746 --> 00:45:33,758 of wealth production. Finance has ignored both of them for many 618 00:45:33,758 --> 00:45:52,310 years and they are so important. So, arithmetic returns 619 00:42:15,872 --> 00:43:12,566 are the world of additive returns. But what we're talking about, 620 00:43:12,566 --> 00:43:58,952 wealth, is a multiplicative process. It's a contingent, conditional 621 00:43:58,952 --> 00:45:05,954 process over the course of time. And the path you take has serious 622 00:45:05,954 --> 00:45:52,340 implications regarding the ultimate wealth you end up in. If 623 00:42:04,034 --> 00:42:40,274 you take arithmetic returns, I've achieved 100% return over 10 624 00:42:40,274 --> 00:43:23,762 years. Divide that by the 10 years, that's a 10% arithmetic return. 625 00:43:23,762 --> 00:44:18,122 The reality is that my path is never like that. I do not get the 626 00:44:18,122 --> 00:44:54,362 straight-line path. I get deviations from that path. And those 627 00:44:54,362 --> 00:45:23,354 deviations, the volatility starts creating a different trajectory 628 00:45:23,354 --> 00:45:52,346 between the geometric path and the arithmetic path. And as volatility 629 00:42:11,456 --> 00:42:51,958 increases, as Dave talks about volatility drag, you find that difference 630 00:42:51,958 --> 00:43:32,460 increases in a compounded manner. And so that you might have 631 00:43:32,460 --> 00:44:05,598 two very similar arithmetical or additive statistics saying it's 632 00:44:05,598 --> 00:44:42,418 the same thing, but the path has totally different trajectories 633 00:44:42,418 --> 00:45:15,556 based on this volatility principle that takes a geometric 634 00:45:15,556 --> 00:45:52,376 path away from the additive path of the arithmetic mean. 635 00:45:58,381 --> 00:46:03,451 Rich, you already mentioned it a little bit, but I'd like to maybe 636 00:46:03,451 --> 00:46:07,741 talk a little bit more about it from Dave's perspective, because, 637 00:46:07,741 --> 00:46:13,201 sadly, over here in Europe, at the moment, we have a lot of forest 638 00:46:13,201 --> 00:46:18,271 fires. I know you have them in Australia as well from during the 639 00:46:18,271 --> 00:46:22,171 summertime. But anyways, you have used forest fires very actively 640 00:46:22,171 --> 00:46:26,851 in your analogy. And it's a very important one because it kind 641 00:46:26,851 --> 00:46:31,141 of helps people understand kind of the dangers of things being 642 00:46:31,141 --> 00:46:36,601 calm, I think. So, can you talk us through it so that we fully 643 00:46:36,601 --> 00:46:37,771 embrace that concept? 644 00:46:38,511 --> 00:47:24,447 Yeah. So, it's based on a physics theorem, a subset of chaos 645 00:47:24,447 --> 00:47:58,899 theory known as self-organized criticality, as sort of advanced 646 00:47:58,899 --> 00:48:52,491 by a wonderful physicist by the name of Per Bak in, what I would 647 00:48:52,491 --> 00:49:34,599 say, is one of probably the most important book anybody will 648 00:49:34,599 --> 00:50:16,707 ever read called How Nature Works. And it's about how nature 649 00:50:16,707 --> 00:50:58,815 works. It turns out most things in the world are nature. And 650 00:46:42,126 --> 00:47:25,506 so, the forest fire is the simple analogy that most people are 651 00:47:25,506 --> 00:48:08,886 familiar with, that the fire risk grows as the forest goes longer 652 00:48:08,886 --> 00:48:48,651 and longer without a fire because you get more clogged brush 653 00:48:48,651 --> 00:49:28,416 between the trees. And so, as we spoke about earlier, diversification 654 00:49:28,416 --> 00:49:57,336 in your portfolio, diversification of your forest solves 655 00:49:57,336 --> 00:50:40,716 the problem of a lightning strike that hits only one tree. All 656 00:50:40,716 --> 00:50:58,791 your other trees are fine. The risk is when that fire spreads from 657 00:47:08,255 --> 00:47:49,153 that tree to the next tree because of the connectivity, closeness 658 00:47:49,153 --> 00:48:30,051 and clogged brush and dry brush between them. In the financial 659 00:48:30,051 --> 00:49:03,513 markets, that would be leverage. And so, correlation is 660 00:49:03,513 --> 00:49:40,693 the big problem because your diversification lets you down and 661 00:49:40,693 --> 00:50:29,027 the fire spreads from tree to tree to tree, again, in physics talk, 662 00:50:29,027 --> 00:50:58,771 what might be known as the percolation threshold. At what point 663 00:46:50,166 --> 00:47:36,786 does the density of trees, the density of the forest become dangerous? 664 00:47:36,786 --> 00:48:15,636 And you can think that everything behaves that way. Avalanche 665 00:48:15,636 --> 00:48:58,371 is another great example. An avalanche, the longer you go with 666 00:48:58,371 --> 00:49:44,991 snow building up, the more likely you are to get an avalanche. 667 00:49:44,991 --> 00:50:31,611 It doesn't mean there's not likely to be an avalanche, it means 668 00:50:31,611 --> 00:50:58,806 it's more likely to be an avalanche. And very much to Rich's 669 00:47:17,951 --> 00:48:36,831 point about endogenous risk, you don't need an exogenous shock. 670 00:48:36,831 --> 00:50:03,599 What causes the avalanche? The last snowflake, one tiny little snowflake 671 00:50:03,599 --> 00:50:58,815 is all you need and that's enough. Now financial markets very 672 00:46:59,335 --> 00:47:56,601 much behave this way because of this reflexive nature and the 673 00:47:56,601 --> 00:48:43,455 building, the waxing and waning of imbalances and corrections. 674 00:48:43,455 --> 00:49:56,339 And so that's very much how we think about it and talk about what 675 00:49:56,339 --> 00:50:53,605 we do when we're out helping people manage the uncertainty of 676 00:50:53,605 --> 00:50:58,811 risk. Again, people tend to be quite good at managing the risk that 677 00:47:22,503 --> 00:47:51,831 they're measuring, but the assumptions they're making about 678 00:47:51,831 --> 00:48:24,825 the stability of the correlation, the stability of the 679 00:48:24,825 --> 00:48:57,819 forest, because they've diversified well, they tend to leave 680 00:48:57,819 --> 00:49:41,811 that out. And they say, but Dave, you don't know which trees 681 00:49:41,811 --> 00:50:36,801 we own. And I say, I don't need to know. I just need to know 682 00:50:36,801 --> 00:50:58,797 where the most dry brush is. And as it turns out, conveniently, 683 00:47:02,913 --> 00:47:55,784 don't tell anybody this is a little bit of our secret sauce. The 684 00:47:55,784 --> 00:48:40,521 biggest fire risk, given the nature of how financial markets work, 685 00:48:40,521 --> 00:49:17,124 again, back to rational accounting men and price insensitive 686 00:49:17,124 --> 00:49:53,727 players, is where the insurance is the cheapest. Because 687 00:49:53,727 --> 00:50:34,397 again, things like value at risk are measuring the frequency, 688 00:50:34,397 --> 00:50:58,799 a short-term frequency of past fires. And it says, well, there hasn't 689 00:47:00,819 --> 00:47:45,435 been a fire. So, there's no risk. There hasn't been an avalanche, 690 00:47:45,435 --> 00:48:22,615 there's no risk. But that's actually generally where the risk 691 00:48:22,615 --> 00:49:18,385 is the highest. And so, this is just a great way to look at anything 692 00:49:18,385 --> 00:49:51,847 and everything. And the mistake that everybody makes is, 693 00:49:51,847 --> 00:50:40,181 oh, it's been calm for so long, it's been calm. Why do you 694 00:50:40,181 --> 00:50:58,771 think that's going to happen? Why do you think that having survived 695 00:47:03,473 --> 00:48:00,529 all the way from 30% of debt to GDP to 130% of debt to GDP and 696 00:48:00,529 --> 00:48:43,321 nothing's blown up. So, why should we worry about it? All the 697 00:48:43,321 --> 00:49:22,547 more reason why you should worry about it. Obviously, the imbalance, 698 00:49:22,547 --> 00:49:58,207 that fragility is building and that's exactly where strategies like 699 00:49:58,207 --> 00:50:33,867 ours and yours are looking for, protecting, capturing that tail 700 00:50:33,867 --> 00:50:58,829 that the flawed methodology doesn't recognize exists. 701 00:50:59,311 --> 00:51:03,571 Yeah, absolutely. Now, from the world of trend following, Rich, 702 00:51:03,571 --> 00:51:08,683 we think about position sizing and we have formulas for these things. 703 00:51:08,683 --> 00:51:14,221 What does this concept… What does that look like when we go into 704 00:51:14,221 --> 00:51:15,073 that world? 705 00:51:16,451 --> 00:52:00,261 The first thing I need to say is that when people experience a 706 00:52:00,261 --> 00:52:30,591 calm market, they think about equilibrium-based systems and they 707 00:52:30,591 --> 00:53:04,291 think about calmness being sort of almost at an equilibrium 708 00:53:04,291 --> 00:53:37,991 state, not much volatility occurring. But that's not what markets 709 00:53:37,991 --> 00:54:08,321 are. Markets are not closed systems that equilibrate, they're 710 00:54:08,321 --> 00:54:52,131 open systems and they sit at what I call the boundaries of chaos. 711 00:54:52,131 --> 00:55:29,201 They’re sitting right out in the frontier in the boundaries of 712 00:55:29,201 --> 00:55:32,571 chaos. New participants are always coming in, new capital's coming 713 00:51:49,859 --> 00:52:30,691 in, going out, participants are leaving, coming in. It's an open 714 00:52:30,691 --> 00:53:00,387 system. They're far from equilibrium-based systems. And when 715 00:53:00,387 --> 00:53:30,083 you experience calm in these non-equilibrium-based systems, this 716 00:53:30,083 --> 00:54:14,627 is not the absence of energy, this is the loose, wobbly rubber 717 00:54:14,627 --> 00:54:59,171 band being stretched. And what that is this buildup of risk that 718 00:54:59,171 --> 00:55:32,579 the metrics, the traditional metrics are not picking up. It's 719 00:51:20,110 --> 00:51:53,041 the undergrowth, as Dave's talking about, that's slowly progressively 720 00:51:53,041 --> 00:52:33,290 building up to this state of self-criticality. And this is what 721 00:52:33,290 --> 00:53:09,880 happens with this rubber band stretching. So quiet, markets are 722 00:53:09,880 --> 00:53:50,129 certainly not where risk is absent. They are being hidden. They're 723 00:53:50,129 --> 00:54:26,719 not being expressed by the traditional metrics. But what I've 724 00:54:26,719 --> 00:55:06,968 found in my research is they are the deepest coupled negative 725 00:55:06,968 --> 00:55:32,581 feedback states in the entire data set. And what I'm meaning there 726 00:51:38,531 --> 00:52:27,107 is that in these calm periods you get markets oscillating. It's 727 00:52:27,107 --> 00:53:20,099 not that there's an absence of energy, it's just that their oscillations 728 00:53:20,099 --> 00:54:13,091 are canceling out. Now, what happens in this state? And I'll explain 729 00:54:13,091 --> 00:55:10,499 why the risk builds up because let's look at some very popular models 730 00:55:10,499 --> 00:55:32,579 that are applied to markets. Let's look at volatility, position-based 731 00:51:38,921 --> 00:52:32,849 sizing or volatility targeted. They will have a period of calm and 732 00:52:32,849 --> 00:53:13,295 their models mechanistically increase the position size in that 733 00:53:13,295 --> 00:54:02,729 period of calm because they're saying, it's a safe market, it's 734 00:54:02,729 --> 00:54:47,669 not a volatile market. So mechanistically that builds up position 735 00:54:47,669 --> 00:55:32,609 size, the energy is going in, in, into the market. But also, behaviorally, 736 00:51:33,527 --> 00:52:30,447 because the markets haven't been expressing their trends or higher 737 00:52:30,447 --> 00:53:15,983 volatility, you find that people behaviorally are starting 738 00:53:15,983 --> 00:54:12,903 to leverage up in that environment because they want to 739 00:54:12,903 --> 00:55:04,131 create a living in this environment and, therefore, behaviorally 740 00:55:04,131 --> 00:55:32,591 they're encouraged to leverage up. Now, of course we know that volatility 741 00:51:38,865 --> 00:52:07,683 clusters, calm begets calm and volatility begets volatility. But 742 00:52:07,683 --> 00:52:49,309 what people assume is that the calm is going to… It often continues 743 00:52:49,309 --> 00:53:24,531 for a period of time where this mechanistic and behavioral tendency 744 00:53:24,531 --> 00:53:53,349 feeds leverage into the system, feeding leverage. So, when 745 00:53:53,349 --> 00:54:34,975 you do get the shock, it's a small perturbation out of that calm 746 00:54:34,975 --> 00:55:10,197 period. Suddenly you get an explosion out of that calm period. 747 00:55:10,197 --> 00:55:32,611 And that's where the risk events occur. So, what I found in 748 00:51:30,841 --> 00:52:02,499 my modeling is that the majority of risk events don't come 749 00:52:02,499 --> 00:52:34,157 out of volatile regimes, they come out from calm regimes where 750 00:52:34,157 --> 00:53:08,693 they have a volatility shock. And a good example is the Japanese 751 00:53:08,693 --> 00:53:43,229 bond market in, I think it was 2018, it had trailing volatility 752 00:53:43,229 --> 00:54:17,765 of a fraction. I think it was about 0.9% trailing volatility. And 753 00:54:17,765 --> 00:54:46,545 that authorized 16.4 times leverage. And therefore, when a move 754 00:54:46,545 --> 00:55:23,959 occurred that was only, I think, a 0.6% move, it took 9% of 755 00:55:23,959 --> 00:55:32,593 the capital away. And this is this impact of leverage coming out 756 00:51:40,463 --> 00:52:01,807 of quiet environments. They're the most devastating environments. 757 00:52:01,807 --> 00:52:33,823 So do not equate calmness with safety. That's one of the fundamental 758 00:52:33,823 --> 00:53:08,507 pitfalls, I think, of so many of the risk metrics that we find 759 00:53:08,507 --> 00:53:37,855 and that are accepted as rational reasons why we should increase 760 00:53:37,855 --> 00:54:04,535 our position size. The high Sharpe ratio, well that's actually 761 00:54:04,535 --> 00:54:31,215 authorizing a higher leverage. And that's exactly what you don't 762 00:54:31,215 --> 00:55:08,567 want to be doing. We'll get to that later. But this is where finance 763 00:55:08,567 --> 00:55:32,579 really needs to carefully look at what they're doing. 764 00:55:34,231 --> 00:55:40,705 Let's stay with leverage for a little while. And I don't know necessarily, 765 00:55:40,705 --> 00:55:48,175 I think I want to leave it as an open question to you, Dave, to 766 00:55:48,175 --> 00:55:54,151 maybe talk a little bit about what leverage exposes or the impact 767 00:55:54,151 --> 00:55:58,633 dangers, when it comes to leverage. You've obviously already 768 00:55:58,633 --> 00:56:05,107 given a great example of that with, I forget, was it SLK leverage? 769 00:56:05,131 --> 00:56:06,409 Hynix, S.K. Hynix. 770 00:56:06,491 --> 00:56:10,391 S.K. Hynix, right. Anyways, talk a little bit about leverage 771 00:56:10,391 --> 00:56:13,901 kind of free flow a little bit on that. 772 00:56:14,061 --> 00:56:59,651 Yeah. Again, back to what we've discussed. It's loss absorbing 773 00:56:59,651 --> 00:57:58,918 capital, right? Is sort of, in a sense, one of the many strange 774 00:57:58,918 --> 00:58:39,949 attractors that are pulling markets around. And so, leverage 775 00:58:39,949 --> 00:59:25,539 is obviously reducing the capital that's applied to a potential 776 00:59:25,539 --> 00:59:30,098 loss. And so, as that leverage builds because of the rules, the 777 00:56:42,056 --> 00:56:57,326 regulations, the misunderstanding or the misapplication 778 00:56:57,326 --> 00:57:20,231 of Gaussian risk measures, volatility is risk, that creates 779 00:57:20,231 --> 00:57:45,681 the fragility in the system. And through the active suppression 780 00:57:45,681 --> 00:58:06,041 of volatility, central banks, governments, policymakers, they want 781 00:58:06,041 --> 00:58:28,946 to take advantage. Rational accounting man, they want rational 782 00:58:28,946 --> 00:58:59,486 accounting man to lever up that which they want to grow: economies, 783 00:58:59,486 --> 00:59:19,846 house ownership, various assets, government bonds, and so 784 00:59:19,846 --> 00:59:30,026 they suppress the volatility. And then mechanistically rational 785 00:56:25,589 --> 00:56:57,291 accounting man, in sort of his form of non-price sensitive participant, 786 00:56:57,291 --> 00:57:31,875 goes and applies leverage to that and he's optimizing to the short-term 787 00:57:31,875 --> 00:57:52,049 returns, probabilistic returns, accounting returns, of a 788 00:57:52,049 --> 00:58:20,869 levered thing that reduces the capital protecting that risk. That 789 00:58:20,869 --> 00:58:55,453 then leads to the fragility that should the fire start to spread, 790 00:58:55,453 --> 00:59:30,037 that's where the problem is. And that's where fat tails come from. 791 00:56:14,061 --> 00:56:50,525 And so, I say all the time in our world you want to own and invest 792 00:56:50,525 --> 00:57:15,594 in things with thin tails, things that have high natural volatility 793 00:57:15,594 --> 00:57:40,663 where you're getting rewarded with the upside vol for the downside 794 00:57:40,663 --> 00:58:03,453 volatility risk you're taking. And then construct your hedges with 795 00:58:03,453 --> 00:58:28,522 things that have fat left tails, that have fat tails, that 796 00:58:28,522 --> 00:58:44,475 have artificially suppressed volatility and attract leverage. 797 00:58:44,475 --> 00:59:07,265 Leverage explicitly in the form of banking system leverage or 798 00:59:07,265 --> 00:59:30,055 repo leverage and implicitly in the form of volatility selling. 799 00:56:14,061 --> 00:56:39,345 And those things are the things that are going to have surprises 800 00:56:39,345 --> 00:57:06,736 and fat tails, fat left tails, some fat right tails as well, of 801 00:57:06,736 --> 00:57:32,020 course, that you want to go and position yourself, in our case 802 00:57:32,020 --> 00:57:50,983 through optionality and convexity, and in yours and Rich's 803 00:57:50,983 --> 00:58:16,267 case, by your signals in your trend thing saying, hey, this trend's 804 00:58:16,267 --> 00:58:39,444 pushing towards something that has that signal, that's got a fat 805 00:58:39,444 --> 00:59:04,728 tail potential. And the good news is, as Rich, again you guys 806 00:59:04,728 --> 00:59:30,012 talk about all the time, you only need a few fat tails. You don't 807 00:56:17,441 --> 00:56:39,411 need a whole bunch of winners, you just need a very few winners 808 00:56:39,411 --> 00:56:56,311 and then just constantly managing your losses to be small 809 00:56:56,311 --> 00:57:14,901 and that will dwarf all the other activities that you're involved. 810 00:57:14,901 --> 00:57:35,181 You guys were talking about cocoa, I think, the last time you 811 00:57:35,181 --> 00:57:55,461 were on the call. I suspect for years, and years, and years, 812 00:57:55,461 --> 00:58:17,431 none of you thought cocoa was a big deal in your trend following 813 00:58:17,431 --> 00:58:37,711 diversified portfolios. And then all of a sudden it was the big 814 00:58:37,711 --> 00:58:59,681 deal. That's exactly sort of the way we look at the world. And 815 00:58:59,681 --> 00:59:16,581 I'll argue that's driven by leverage, by imbalances created by 816 00:59:16,581 --> 00:59:28,411 inappropriately applied risk methodology, accounting rules, non-skin-in-the-game 817 00:59:28,411 --> 00:59:30,101 participants. 818 00:59:31,221 --> 00:59:37,441 Now you mentioned an important word, actually I think ‘optimal’ 819 00:59:37,441 --> 00:59:45,527 because I think a lot of people try to kind of be optimal 820 00:59:45,527 --> 00:59:53,613 in many things when it comes to investing. But before I let you 821 00:59:53,613 --> 01:00:01,699 loose here, Rich, and maybe you talk a little bit about the dangers 822 01:00:01,699 --> 01:00:09,163 of that word, when I think back, you also mentioned this thing, 823 01:00:09,163 --> 01:00:15,383 Dave, about the trend following world that we buy these 824 01:00:15,383 --> 01:00:23,469 breakouts, or we sell them, or whatever it is. But, and I think 825 01:00:23,469 --> 01:00:32,799 a lot of times (I'm going off a tangent here a little bit), I think 826 01:00:32,799 --> 01:00:39,641 a lot of times, when people think about trend following, they 827 01:00:39,641 --> 01:00:47,105 kind of think that's kind of the secret sauce, or we discovered 828 01:00:47,105 --> 01:00:53,325 this idea that you can buy a certain day breakout. But I would 829 00:59:33,711 --> 00:59:44,501 argue (and I have a feeling that you might agree, but I don't 830 00:59:44,501 --> 00:59:54,461 want to be assuming this), I actually think the whole secret that 831 00:59:54,461 --> 01:00:05,251 trend followers brought to the table back in the ‘70s is that they 832 01:00:05,251 --> 01:00:15,211 understood how to manage the risk and they understood how to size 833 01:00:15,211 --> 01:00:23,511 positions differently to what had been understood up until that 834 01:00:23,511 --> 01:00:34,301 time. So anyways, don't want to say that because it does tie into 835 01:00:34,301 --> 01:00:43,431 leverage and all of that stuff. But anyways, Rich, what I 836 01:00:43,431 --> 01:00:53,391 really meant to ask you is what's the problem of being optimal? 837 01:00:54,531 --> 01:01:40,408 Okay, so the problem of being optimal is that it foregoes the room 838 01:01:40,408 --> 01:02:22,756 for being wrong. So, let's look at the optimal bit. Let's look 839 01:02:22,756 --> 01:02:58,046 at the Kelly criterion. So, the Kelly criterion, it's looking 840 01:02:58,046 --> 01:03:29,807 for perfect information, it's asking a specific question. Given 841 01:03:29,807 --> 01:04:12,155 an edge, what fraction of your capital should you state to grow 842 01:04:12,155 --> 01:04:29,800 it as fast as possible? Now when you look at it, when you're 843 01:01:15,035 --> 01:01:40,665 looking at the optimum condition, it's putting you at a 844 01:01:40,665 --> 01:02:08,858 peak because if you go over that peak you become suboptimal. 845 01:02:08,858 --> 01:02:39,614 And what you find with the volatility drag, the going over that 846 01:02:39,614 --> 01:03:10,370 peak actually puts you into risk of ruin very quickly. If you 847 01:03:10,370 --> 01:03:41,126 under bet and you don't go quite to that Kelly limit, that's 848 01:03:41,126 --> 01:04:09,319 taking significant risk off the table. It's giving you room for 849 01:04:09,319 --> 01:04:29,823 error, giving you room for not being correct. Now this is the problem. 850 01:01:17,931 --> 01:02:00,051 With the Kelly criteria it's assuming perfect information and 851 01:02:00,051 --> 01:02:51,531 it's another sort of addition to this Gaussian world. And it's 852 01:02:51,531 --> 01:03:47,691 a false premise because what you find is that markets are non-stationary. 853 01:03:47,691 --> 01:04:29,811 You find that there's significant noise in the market. 854 01:00:54,531 --> 01:01:32,523 Now what people don't realize is that when they are choosing the 855 01:01:32,523 --> 01:02:07,349 optimal bet, they are applying an estimate; not the reality. It's 856 01:02:07,349 --> 01:02:35,843 an estimate. Given their sample size, what they're looking 857 01:02:35,843 --> 01:03:07,503 at, that's what they're placing as their optimal bet, their 858 01:03:07,503 --> 01:03:39,163 estimate. And as when you start introducing fat tails into 859 01:03:39,163 --> 01:04:17,155 estimates, your sample size, needing to make that a fair and reasonable 860 01:04:17,155 --> 01:04:29,819 sample size, explodes out. Now, the Gaussian world, it reduces 861 01:01:17,595 --> 01:02:07,567 it to a small sample size needing to be a degree of certainty. 862 01:02:07,567 --> 01:02:49,851 But when tails are included into that sample, it explodes the 863 01:02:49,851 --> 01:03:35,979 sample size out. So, you never have enough sample to be anything 864 01:03:35,979 --> 01:04:22,107 other than an estimate. And there is always room for error in 865 01:04:22,107 --> 01:04:29,795 that estimate. And so, because you're trading at that peak, there 866 01:01:07,527 --> 01:01:26,299 is no room for error because that's what you give up being optimal. 867 01:01:26,299 --> 01:01:46,515 To be optimal means you've got to give up that room for error. That's 868 01:01:46,515 --> 01:02:02,399 the biggest mistake. It's putting you at the knife edge. It's 869 01:02:02,399 --> 01:02:18,283 increasing leverage to the point where… And I've done this myself, 870 01:02:18,283 --> 01:02:34,167 I've looked at Kelly based position sizing models with our trend 871 01:02:34,167 --> 01:02:48,607 following approaches, Niels and Dave, and I've compared it to 872 01:02:48,607 --> 01:03:03,047 very simple position sizing rules that are regarded as equal 873 01:03:03,047 --> 01:03:14,599 weighted risk betting, fractional position sizing that we 874 01:03:14,599 --> 01:03:29,039 adopt in trend following world. And it outperforms the Kelly 875 01:03:29,039 --> 01:03:42,035 approach when you start introducing things such as regime 876 01:03:42,035 --> 01:03:55,031 changes, sufficient noise to your estimate, because the noise 877 01:03:55,031 --> 01:04:10,915 is often bigger than the signal that they're looking for to 878 01:04:10,915 --> 01:04:26,799 develop that optimal bet. And this is the problem long-term capital 879 01:04:26,799 --> 01:04:29,687 management face. So, they traded that, they used optimal trading 880 01:01:28,979 --> 01:02:20,651 positions for as long as the assumptions held. And they were very 881 01:02:20,651 --> 01:03:08,017 successful over the time the assumptions held. But when they went 882 01:03:08,017 --> 01:03:59,689 over that limit, they went quickly to risk of ruin. That's going 883 01:03:59,689 --> 01:04:29,831 over that peak I was talking about. That's the problem with optimal, 884 01:01:35,931 --> 01:03:15,291 to be optimal you've got to have perfect information and we never 885 01:03:15,291 --> 01:04:29,811 have that perfect information in these complex adaptive systems. 886 01:04:31,351 --> 01:04:48,514 There is, in podcasting, something called The Optimal Duo 887 01:04:48,514 --> 01:05:11,398 for having on a podcast show. And I think we're achieving that 888 01:05:11,398 --> 01:05:30,468 today. But there we are. That's a completely different concept 889 01:05:30,468 --> 01:05:38,096 - no Kelly here. Anyways, we're going to move on to something 890 01:04:35,247 --> 01:04:41,091 where I think over the years people have heard kind of analogies 891 01:04:41,091 --> 01:04:46,935 about football and should a trend follow, is he a striker, is 892 01:04:46,935 --> 01:04:52,779 he a defender, Is he a midfielder? But Dave, you and one 893 01:04:52,779 --> 01:04:58,136 of your colleagues, I think, Nicholas, have kind of taken this 894 01:04:58,136 --> 01:05:02,519 football analogy, and obviously with the recent World Cup, 895 01:05:02,519 --> 01:05:08,850 to a brand new level. And I know you published it this morning. 896 01:05:08,850 --> 01:05:13,720 (First of all, everybody should follow both Rich and Dave 897 01:05:13,720 --> 01:05:19,077 on LinkedIn or Twitter, wherever it is.) But this kind of 898 01:05:19,077 --> 01:05:25,895 takes it to, I think, a new level. I've only skimmed it, so I 899 01:05:25,895 --> 01:05:32,226 kind of need also, for selfish reasons, for you to take us through 900 01:05:32,226 --> 01:05:38,070 this new World Cup scoring system, I guess we could call it. 901 01:05:38,831 --> 01:05:41,399 Unfortunately, that was Nick who just left. We wanted to get rid 902 01:05:41,399 --> 01:05:41,827 of him. 903 01:05:41,831 --> 01:05:43,831 Yeah, I saw him leaving there. 904 01:05:43,831 --> 01:06:29,786 So, for forever we've used a football pitch picture in a lot of 905 01:06:29,786 --> 01:07:05,136 our presentations. If I speak publicly, I almost always (because 906 01:07:05,136 --> 01:07:51,091 I want to frame what I'm talking about and get it in [people’s] 907 01:07:51,091 --> 01:08:26,441 heads, because the world, the financial world, has so emblazoned 908 01:08:26,441 --> 01:09:05,326 in their mind the Gaussian distribution. So, we draw a football 909 01:09:05,326 --> 01:09:44,211 pitch and we draw the Gaussian distribution over it, and explain 910 01:09:44,211 --> 01:10:23,096 that that's a measure of frequency. That's a measure of where 911 01:10:23,096 --> 01:10:37,236 the ball spends time. On average, the ball's at the center 912 01:06:48,014 --> 01:08:47,211 of the pitch, like the average return of monthly S&P. It's the average. 913 01:08:47,211 --> 01:10:37,239 It means nothing. The average is meaningless. It is a wasted statistic. 914 01:05:43,831 --> 01:06:23,559 And then we draw what we call an entropy curve, a Shannon entropy 915 01:06:23,559 --> 01:06:54,119 curve, a measure of information value. And then we shade 916 01:06:54,119 --> 01:07:18,567 underneath that the long-term contribution to compounded returns 917 01:07:18,567 --> 01:08:01,351 of the S&P and show that the 2%iles of returns (the 10 worst months 918 01:08:01,351 --> 01:08:44,135 out of 480 months and the 10 best months out of 480 months) contribute 919 01:08:44,135 --> 01:09:23,863 almost all of the returns. And just like in a football match, it's 920 01:09:23,863 --> 01:09:57,479 what happens in the penalty boxes that matters, not what happens 921 01:09:57,479 --> 01:10:34,151 in the center of the pitch. Right? It's goal prevention and goal 922 01:10:34,151 --> 01:10:37,207 scoring. And so, we've been giving that presentation forever, 923 01:06:15,127 --> 01:07:02,071 certainly longer than you guys have known me, I've been giving that 924 01:07:02,071 --> 01:07:49,015 presentation. And so, Nick (who is who takes geek to a different 925 01:07:49,015 --> 01:08:39,871 level here) said, I'm going to build a model and we're going to 926 01:08:39,871 --> 01:09:30,727 track it through the World Cup and see if this story you've been 927 01:09:30,727 --> 01:10:13,759 telling forever, Dave, is true, that what happens in the penalty 928 01:10:13,759 --> 01:10:37,231 boxes is what drives the outcome. And we did it. And thankfully, 929 01:06:00,595 --> 01:06:22,947 nowadays, FIFA provides unbelievable data, and there's all 930 01:06:22,947 --> 01:06:56,475 kinds of improved data with the chip that's in the football now. 931 01:06:56,475 --> 01:07:30,003 And we came up with what we've called the convex goalkeeper theorem. 932 01:07:30,003 --> 01:07:57,943 And without going into all the mathematical complexities, the gist 933 01:07:57,943 --> 01:08:34,265 of it is it comes down to three layers. This is how people 934 01:08:34,265 --> 01:08:56,617 win a non-ergodic football tournament. Now it's non-ergodic 935 01:08:56,617 --> 01:09:30,145 because if you lose, you're out. You hit the ruin barrier, as 936 01:09:30,145 --> 01:10:00,879 Rich would say, the insolvency barrier, you're out. You’ve got to 937 01:10:00,879 --> 01:10:34,407 stay in to keep going. All right, so it's a very non-ergodic 938 01:10:34,407 --> 01:10:37,201 path. Layer one is defensive resilience. Right? Are you becoming 939 01:06:20,506 --> 01:07:09,406 more defensive as the ball moves from the midfield to your penalty 940 01:07:09,406 --> 01:08:14,606 box? And we measure it by the 18 yard box and by the 6 yard box. 941 01:08:14,606 --> 01:08:55,356 Now, obviously the ultimate defensiveness at the very end are 942 01:08:55,356 --> 01:09:40,181 saves by the goalkeeper. But we have other measures and metrics 943 01:09:40,181 --> 01:10:25,006 in there about how we're measuring sort of a positive convexity 944 01:10:25,006 --> 01:10:37,231 of defensive resilience. Likewise, the other part of the game 945 01:06:28,477 --> 01:07:25,879 is scoring goals. Are you becoming more effective, efficient 946 01:07:25,879 --> 01:08:42,415 in your attacking as you approach the other 18 yard box and 947 01:08:42,415 --> 01:10:05,329 6 yard box and goal? And it turns out layer one, most important, 948 01:10:05,329 --> 01:10:37,219 layer two next most important. But then we found a layer three. 949 01:06:03,578 --> 01:06:37,430 And layer three was individual tail risk, if you will, and individual 950 01:06:37,430 --> 01:07:05,640 performance above and beyond what could be recognized from simply 951 01:07:05,640 --> 01:07:31,029 the dominance in the respective penalty boxes. Messi making 952 01:07:31,029 --> 01:08:02,060 nine crosses in the last two minutes, right? Or Bellingham's last 953 01:08:02,060 --> 01:08:38,733 second two goals, etc. Think of that as a tail risk that, depending 954 01:08:38,733 --> 01:09:15,406 on which side of the game you're on, is either a negative tail 955 01:09:15,406 --> 01:09:52,079 or a positive tail. And so, staying in the game such that tail 956 01:09:52,079 --> 01:10:28,752 can be beneficial to you is very important. That can make a difference 957 01:10:28,752 --> 01:10:37,215 in the match. And so, the model is really fantastic. We put 958 01:06:45,238 --> 01:08:00,291 the paper up, a shortened paper. There's a full paper that 959 01:08:00,291 --> 01:09:22,167 goes through every single game over the whole course of the thing, 960 01:09:22,167 --> 01:10:37,220 we've put a shortened one up with a few key examples. And so, 961 01:05:50,499 --> 01:06:37,175 one of the best examples, and you guys will love this, is the England/Argentina 962 01:06:37,175 --> 01:07:17,183 match because you can divide that match into two matches. It was 963 01:07:17,183 --> 01:08:00,525 a very normal match until England went up 1 nil. And then England 964 01:08:00,525 --> 01:08:43,867 did, as they say in England, parked the bus. Right? They did what 965 01:08:43,867 --> 01:09:20,541 is so common in the investment industry. They thought they could 966 01:09:20,541 --> 01:09:53,881 reduce risk by capping the upside. By only playing defense. 967 01:09:53,881 --> 01:10:37,223 They did what in option space we might call a covered call strategy. 968 01:10:37,961 --> 01:10:39,161 Buy more bonds. 969 01:10:39,401 --> 01:11:02,963 They sold call options, they bought more bonds. They did 60/40. 970 01:11:02,963 --> 01:11:26,525 They reduced speed, they drove slower, and yet they're still exposed 971 01:11:26,525 --> 01:11:52,229 to the tail, 60/40 can still lose money. A covered call strategy 972 01:11:52,229 --> 01:12:17,933 has still got all of the downside. All you've done is forego. 973 01:12:17,933 --> 01:12:37,211 The ultimate objective of good defense is attacking. You can't separate 974 01:10:44,675 --> 01:11:02,255 (kid to riches is such beautiful story), you can't separate 975 01:11:02,255 --> 01:11:23,351 the parts. The whole is greater than the sum of the parts. 976 01:11:23,351 --> 01:11:42,689 The parts must function together. You must be convex to both 977 01:11:42,689 --> 01:12:02,027 sides of the magnitude where it matters. It's not the frequency, 978 01:12:02,027 --> 01:12:23,123 it's not what's happening in the middle of the pitch where the 979 01:12:23,123 --> 01:12:37,187 ball spends most of the time, that matters. And I found it wonderful 980 01:10:46,411 --> 01:10:57,627 watching. And obviously, unlike most Americans, I've played 981 01:10:57,627 --> 01:11:14,451 soccer (as we call it in America) my whole life and coached 982 01:11:14,451 --> 01:11:29,873 soccer when I was a younger man, some interesting teams. But 983 01:11:29,873 --> 01:11:46,697 even watching this World Cup, think how many goals (and Rich will 984 01:11:46,697 --> 01:12:04,923 love this), how many goals are the result of a mad scramble? So 985 01:12:04,923 --> 01:12:18,943 many, right? Very few are actually a smooth, perfectly orchestrated 986 01:12:18,943 --> 01:12:34,365 movement. So many are a rebound, a ricochet, a stumble, a 987 01:12:34,365 --> 01:12:37,169 fall, right? And the act of getting or preventing them is about 988 01:10:51,061 --> 01:10:59,223 this unknowable emergent property of defensive resilience 989 01:10:59,223 --> 01:11:10,883 and offensive effectiveness. And if you're just trying to play 990 01:11:10,883 --> 01:11:26,041 by the numbers, you're not going to find that. And so it's a 991 01:11:26,041 --> 01:11:40,033 really interesting model. I'm going to try to… In fact, I told 992 01:11:40,033 --> 01:11:56,357 some people today I'm going to speak at a CBOE event in Hong Kong 993 01:11:56,357 --> 01:12:12,681 later this year. I'm going to try to somehow form it into one of 994 01:12:12,681 --> 01:12:23,175 my presentations, tie it together with total portfolio approach 995 01:12:23,175 --> 01:12:36,001 and stuff. So, it's really fun. Everybody should go look at 996 01:12:36,001 --> 01:12:37,167 it. 997 01:12:37,321 --> 01:13:03,531 Absolutely. I completely agree. Absolutely. Now, you've already 998 01:12:39,067 --> 01:12:45,469 alluded to, Rich, that we are skeptical of Sharpe. Obviously, Dave 999 01:12:45,469 --> 01:12:52,453 has written tons about the Sharpe world that we live in. So, 1000 01:12:52,453 --> 01:12:59,437 if you have two portfolios with the same Sharpe, what do we 1001 01:12:59,437 --> 01:13:03,511 not know from looking at those numbers? 1002 01:13:04,171 --> 01:13:08,441 Look, before I get into that, I've just got to say, Dave, I can 1003 01:13:08,441 --> 01:13:11,491 see a sequel to Moneyball coming up. We're applying convexity 1004 01:13:11,491 --> 01:13:15,151 now in the game. So, looking forward to you doing that coach, 1005 01:13:15,151 --> 01:13:16,676 and seeing how that goes. 1006 01:13:19,021 --> 01:13:21,507 We talk about that all the time actually, Rich, that's spot 1007 01:13:21,507 --> 01:13:21,733 on. 1008 01:13:22,301 --> 01:13:56,083 Look, I also want to sort of add to Dave's point. And Niels, you'll 1009 01:13:56,083 --> 01:14:20,213 be interested in this because let's talk about trend following. 1010 01:14:20,213 --> 01:14:53,995 So, when I ran my models over 40 years, you find that October 2008, 1011 01:14:53,995 --> 01:15:22,951 that was a fantastic period for trend followers. It was the single 1012 01:15:22,951 --> 01:15:51,907 largest month in the return for the entire history of 40 years 1013 01:15:51,907 --> 01:16:20,863 that I looked at. It accounted for 3.4% of total portfolio wealth, 1014 01:16:20,863 --> 01:16:47,406 that one month, October 2008, 3.4%. But you know what happened 1015 01:16:47,406 --> 01:17:18,775 to the Sharpe ratio over that same period? It went from 0.914 to 1016 01:17:18,775 --> 01:17:40,492 0.931. So, it paid US$0.02 in the dollar, Niels. So, the Sharpe 1017 01:13:37,793 --> 01:14:34,597 penalizes trend followers. I just wanted to say that. It’s no 1018 01:14:34,597 --> 01:15:31,401 fault of the ratio. It's not deliberately picking us out, but 1019 01:15:31,401 --> 01:16:17,877 because when we're employing convexity, the Sharpe is perhaps 1020 01:16:17,877 --> 01:17:25,009 the worst measure to use to assess that. So, Sharpe, to me, fails 1021 01:17:25,009 --> 01:17:40,501 in two ways. The first thing it does is it doesn't let you know 1022 01:14:03,463 --> 01:14:52,109 the system's failure modes. And I want to give you a good sort 1023 01:14:52,109 --> 01:15:29,529 of example to consider. So, consider two portfolios, both with 1024 01:15:29,529 --> 01:16:18,175 a Sharpe ratio of 1.5. So, the first is a single concentrated position. 1025 01:16:18,175 --> 01:17:03,079 Levered it up using Kelly, resting on one estimate of one edge. 1026 01:17:03,079 --> 01:17:40,499 That's one of the portfolios producing a Sharpe of 1.5. The second 1027 01:13:28,039 --> 01:14:02,467 is a wide book of small positions that are only loosely related. 1028 01:14:02,467 --> 01:14:31,157 No individual position is carrying the load, unlike that single 1029 01:14:31,157 --> 01:15:05,585 levered estimate. That also has a Sharpe of 1.5 - same number. 1030 01:15:05,585 --> 01:15:40,013 So, the metric Sharpe itself calls them equal because by the time 1031 01:15:40,013 --> 01:16:08,703 it's calculated, the entire construction of Sharpe has been compressed 1032 01:16:08,703 --> 01:16:43,131 into two figures: an average return at the numerator, and a volatility 1033 01:16:43,131 --> 01:17:17,559 in the denominator. And every trace of how the portfolio is assembled 1034 01:17:17,559 --> 01:17:40,511 has been discarded apart from those two values. So, the number 1035 01:13:30,908 --> 01:14:05,336 records where you arrive, but it doesn't record how you got there. 1036 01:14:05,336 --> 01:14:36,895 And it's the second thing that decides whether you're still standing 1037 01:14:36,895 --> 01:15:05,585 when an assumption breaks. So, it fails. Sharpe fails, particularly 1038 01:15:05,585 --> 01:15:37,144 in two ways. It tells you, effectively, the load rating. And 1039 01:15:37,144 --> 01:16:11,572 so, consider a single strand over a canyon. We've got a single 1040 01:16:11,572 --> 01:16:43,131 strand rope that's holding a particular load that measures a Sharpe 1041 01:16:43,131 --> 01:17:08,952 of 1.5 compared to 1000 strands, all individual strands, 1042 01:17:08,952 --> 01:17:40,511 all loosely assembled, that has the same Sharpe over that canyon. 1043 01:13:22,301 --> 01:14:08,641 Now we've got one failure mode for the single cable but if we get 1044 01:14:08,641 --> 01:14:45,051 a single failure mode in the ensemble of loosely correlated estimates, 1045 01:14:45,051 --> 01:15:31,391 we only get a failure in one, and 999 is still keeping them up. 1046 01:15:31,391 --> 01:16:04,491 So, this is where Sharpe doesn't reveal this property. And 1047 01:16:04,491 --> 01:16:27,661 it's only through understanding the construction, the 1048 01:16:27,661 --> 01:16:54,141 architecture of the construction of the portfolio, that's 1049 01:16:54,141 --> 01:17:27,241 what's telling you: what are the foundations of that architecture, 1050 01:17:27,241 --> 01:17:40,481 what are its dependencies? Is there one failure mode or is there 1051 01:14:02,805 --> 01:14:48,372 potential for many failure modes where your portfolio won't 1052 01:14:48,372 --> 01:15:49,128 come crashing down. You don't get that out of that Sharpe estimate. 1053 01:15:49,128 --> 01:16:49,884 You get that from a knowledge of the construction. And that's something 1054 01:16:49,884 --> 01:17:40,514 none of these metrics is going to give to you. This is why I believe 1055 01:13:38,641 --> 01:14:17,857 Dave and I are what I call engineers. We are architecting convexity. 1056 01:14:17,857 --> 01:14:47,269 We are architecting this geometrical sort of progression that 1057 01:14:47,269 --> 01:15:16,681 mitigates our downside risk and exploits opportunity. This is 1058 01:15:16,681 --> 01:15:55,897 this brake, the brake system that Dave talks about in his formula 1059 01:15:55,897 --> 01:16:28,577 one. It's mitigating the volatility drag and getting the beneficial 1060 01:16:28,577 --> 01:16:54,721 opportunity, exploiting the opportunity because you've got good 1061 01:16:54,721 --> 01:17:27,401 brakes. That's exactly this principle. So, that's all I've got 1062 01:17:27,401 --> 01:17:40,473 to add to that. 1063 01:17:41,241 --> 01:17:46,532 Sure. Now, we've talked a lot about risk and how to manage how 1064 01:17:46,532 --> 01:17:52,230 to mitigate that, and I think. I can't say that for sure, but I 1065 01:17:52,230 --> 01:17:56,707 think a very popular way of thinking about mitigating risk is 1066 01:17:56,707 --> 01:18:01,184 just diversification. So, my question to you, Dave, is, does diversification 1067 01:18:01,184 --> 01:18:02,405 really protect you? 1068 01:18:03,401 --> 01:18:20,561 Yes, sometimes, just not when it matters the most. Right? So again, 1069 01:18:20,561 --> 01:18:30,571 traditional diversification is making assumptions about correlation, 1070 01:18:30,571 --> 01:18:40,581 making assumptions about ignoring the connectivity between 1071 01:18:40,581 --> 01:18:59,171 the trees. Now, I wrote one of my papers once years ago, pointing 1072 01:18:59,171 --> 01:19:13,471 out that diversification across a whole bunch of things, that 1073 01:19:13,471 --> 01:19:33,491 any one of them can cost you all of your capital is probably not 1074 01:19:33,491 --> 01:19:50,651 a good idea. Right? Because then any one position can wipe you 1075 01:19:50,651 --> 01:20:07,811 out. Any of them can go to infinity. Infinity loss will wipe 1076 01:20:07,811 --> 01:20:10,671 you out. On the other side, in what we're doing, the more diversification, 1077 01:18:23,113 --> 01:18:42,825 the better, because everything that we do, as best as possible, 1078 01:18:42,825 --> 01:19:02,537 we want infinity as a potential benefit. So, I only need 1079 01:19:02,537 --> 01:19:25,833 to get to Rich's point and your point. You only need one month 1080 01:19:25,833 --> 01:19:41,961 to make an enormous contribution. Right? You only need 1081 01:19:41,961 --> 01:20:03,465 one trade. If cocoa goes to infinity and you're on it, you've 1082 01:20:03,465 --> 01:20:10,633 made your budget forever. And so, the same thing with us, a big 1083 01:18:18,881 --> 01:18:42,961 part of what we do and how we construct the convexity for our investors, 1084 01:18:42,961 --> 01:19:05,321 as much as we possibly can we always want infinity on the other 1085 01:19:05,321 --> 01:19:25,961 side. I don't know if your listeners or you guys are familiar 1086 01:19:25,961 --> 01:19:44,881 with the St. Petersburg paradox, which is a coin toss game 1087 01:19:44,881 --> 01:20:05,521 that has an expected return of infinity but a very low probability 1088 01:20:05,521 --> 01:20:10,681 of achieving it. But if I can, at the right price, get someone to 1089 01:18:23,113 --> 01:18:46,409 commit to let me play the game for an unlimited number of hands, 1090 01:18:46,409 --> 01:19:06,121 and I've always got infinity on the other side, well, eventually 1091 01:19:06,121 --> 01:19:31,209 I got a pretty good chance of him begging me to unwind the commitment 1092 01:19:31,209 --> 01:19:54,505 that I can play forever and ever, and ever, and ever. And that's 1093 01:19:54,505 --> 01:20:10,633 a big part of how we think about things. So, diversification 1094 01:18:09,461 --> 01:18:42,791 needs to be understood for what it is. There are beneficial 1095 01:18:42,791 --> 01:19:04,001 diversifications, there are faux benefits of diversification, 1096 01:19:04,001 --> 01:19:34,301 and there are dangers of diversification where you don't have 1097 01:19:34,301 --> 01:20:01,571 bounded, understood downside across a range of diversified and 1098 01:20:01,571 --> 01:20:10,661 possibly reflexive items. 1099 01:20:11,081 --> 01:20:17,093 Yeah, Rich, you talk about this using words. The fact that diversification 1100 01:20:17,093 --> 01:20:23,105 is kind of more the engine that shields you. Talk us through 1101 01:20:23,105 --> 01:20:24,107 that idea. 1102 01:20:24,841 --> 01:22:10,969 So, Niels, what I found in my research is that diversification 1103 01:22:10,969 --> 01:23:37,801 for us classic trend followers, it's giving us exposure 1104 01:23:37,801 --> 01:25:04,633 to more right tail opportunities. To me, using diversification 1105 01:25:04,633 --> 01:27:10,057 to mitigate risk, as Dave says, can be good, it can be bad. 1106 01:20:24,841 --> 01:21:16,791 Here's an example. We've got those thousand strands going across 1107 01:21:16,791 --> 01:22:03,546 that canyon, it looks diversified, but imagine if they're 1108 01:22:03,546 --> 01:23:05,886 all connected to a single anchor at each end of that canyon. 1109 01:23:05,886 --> 01:24:03,031 There is still one failure mode there, but it looks diversified. 1110 01:24:03,031 --> 01:25:05,371 And this is the problem. You might borrow short and lend long 1111 01:25:05,371 --> 01:26:02,516 across a dozen markets during calm regimes and think it's fantastically 1112 01:26:02,516 --> 01:26:49,271 diversified, but you're exposed to a single interest rate 1113 01:26:49,271 --> 01:27:10,051 bet, effectively, through that. So, I think correlations… what 1114 01:20:51,496 --> 01:21:50,137 I've found in the research, and this is really interesting to 1115 01:21:50,137 --> 01:22:54,109 me, it's research I've done, I haven't really promoted it. But what 1116 01:22:54,109 --> 01:23:58,081 I've found is that most people think that markets during calm regimes, 1117 01:23:58,081 --> 01:24:56,722 ordinary regimes, tend to move independently of each other and then 1118 01:24:56,722 --> 01:25:39,370 during periods of stress, suddenly they become correlated. 1119 01:25:39,370 --> 01:26:48,673 Well, I beg to differ in that opinion because I think that markets 1120 01:26:48,673 --> 01:27:09,997 are always permanently coupled. Now imagine this. So, you 1121 01:20:51,856 --> 01:21:45,886 know how I was talking about ecosystems of participants. Now, 1122 01:21:45,886 --> 01:22:45,319 the individual agents in that system, they are rarely holding a 1123 01:22:45,319 --> 01:23:33,946 single position in anything. They are diversified across products. 1124 01:23:33,946 --> 01:24:27,976 Now, that model that they are applying, diversified across products, 1125 01:24:27,976 --> 01:25:22,006 is creating this permanent coupling. So, as long as diversification 1126 01:25:22,006 --> 01:26:32,245 exists in the market, and most investors apply it, I think we will 1127 01:26:32,245 --> 01:27:10,066 find that the markets are permanently coupled. But what does 1128 01:20:56,830 --> 01:23:04,786 change is that the phase of where it's sitting in that coupling. 1129 01:23:04,786 --> 01:25:02,079 And what you find during calm regimes, we get this oscillation 1130 01:25:02,079 --> 01:26:48,709 effect canceling effect. This is what creates this apparent benefit 1131 01:26:48,709 --> 01:27:10,035 of diversification. Diversification works during calm 1132 01:20:50,165 --> 01:21:53,475 environments, but it works specifically at the wrong time. You 1133 01:21:53,475 --> 01:23:15,778 want it to work during periods of crisis, but when you get into 1134 01:23:15,778 --> 01:24:12,757 periods of crisis, the coupling goes from oscillating to 1135 01:24:12,757 --> 01:25:09,736 trend, and you suddenly find that everything becomes correlated. 1136 01:25:09,736 --> 01:25:54,053 This is this permanent coupling through diversification, 1137 01:25:54,053 --> 01:26:57,363 through agent based models, all sort of spreading out across 1138 01:26:57,363 --> 01:27:10,025 asset classes. A commodity trader is spread out across its commodities. 1139 01:21:31,144 --> 01:22:44,814 A global macro trader is spread out across currencies, bonds, 1140 01:22:44,814 --> 01:23:58,484 all of these things, these forces, this coupling of extremes 1141 01:23:58,484 --> 01:25:19,521 and where the correlation is, a correlation is an expression of 1142 01:25:19,521 --> 01:26:40,558 where the coupling sits. It's always on, but people assume it's 1143 01:26:40,558 --> 01:27:10,026 off during good times. So, the problem I face is that you've got 1144 01:21:37,777 --> 01:23:15,025 to use architecture here. So, go back to that example. A thousand 1145 01:23:15,025 --> 01:24:52,273 strands, one anchor. We see the problems of that. But if we've 1146 01:24:52,273 --> 01:26:05,209 got a thousand strands connected to a thousand separate 1147 01:26:05,209 --> 01:27:10,041 anchors, we've got a totally different proposition there. So, 1148 01:20:31,376 --> 01:21:17,121 we're looking for not necessarily correlation properties, 1149 01:21:17,121 --> 01:22:29,006 we've got to know the architecture, we've got to know what 1150 01:22:29,006 --> 01:23:34,356 are the dependencies that exist in our portfolio. You're not 1151 01:23:34,356 --> 01:24:52,776 going to get that from the statistics, you're not going to get 1152 01:24:52,776 --> 01:25:45,056 that from correlation properties. You've got to understand 1153 01:25:45,056 --> 01:26:50,406 how the architecture, how it's created, what are the dependencies 1154 01:26:50,406 --> 01:27:10,011 in that architecture? Another really interesting point in my research 1155 01:20:59,015 --> 01:21:52,717 I found that diversification doesn't thin the tails. So, if you 1156 01:21:52,717 --> 01:22:36,655 imagine everyone's assumption is that, oh, with increased diversification, 1157 01:22:36,655 --> 01:23:40,121 the impact of tails dilutes. But what I found is that with increased 1158 01:23:40,121 --> 01:24:28,941 diversification we get an initial drop in kurtosis and then 1159 01:24:28,941 --> 01:25:27,525 a ramping up of kurtosis. Because what you find is that the 1160 01:25:27,525 --> 01:26:11,463 tail properties, because they're fractal in nature, as you 1161 01:26:11,463 --> 01:27:10,047 get bigger and bigger samples, you get exposed to more large things. 1162 01:20:24,841 --> 01:21:06,754 This is unlike the proposition that with increased diversification 1163 01:21:06,754 --> 01:22:02,638 the large things dilute. No, not in fractal structures. If I get 1164 01:22:02,638 --> 01:23:12,493 a big tree and I put a sample over that tree, a small sample, more 1165 01:23:12,493 --> 01:24:03,720 than likely I'm going to get leaves and small branches, nothing 1166 01:24:03,720 --> 01:24:50,290 material. Increase that sample, I get more exposure to the 1167 01:24:50,290 --> 01:25:46,174 trunks, to the large branches. Increase it further, I get the entire 1168 01:25:46,174 --> 01:26:37,401 tree, I get the major dominant fractal structures of that entire 1169 01:26:37,401 --> 01:27:10,000 tree. This is what happens with diversification. So, what I've 1170 01:20:36,097 --> 01:21:13,617 found in my studies, as you increase diversification, you get 1171 01:21:13,617 --> 01:21:58,641 more exposure to the right tail. This is, I'm not using diversification 1172 01:21:58,641 --> 01:22:39,913 here as a solve-all for the correlation argument. I'm not saying 1173 01:22:39,913 --> 01:23:17,433 it's going to significantly reduce your volatility, but I am 1174 01:23:17,433 --> 01:23:54,953 saying that you get more exposure to the fractal properties 1175 01:23:54,953 --> 01:24:43,729 of the market and because of this you get more exposure to the 1176 01:24:43,729 --> 01:25:28,753 right tail. And an example, who would have been sitting in cocoa? 1177 01:25:28,753 --> 01:26:02,521 Most traders don't even consider cocoa. But because we're 1178 01:26:02,521 --> 01:26:43,793 maximally diversified across all of these, we had a position in 1179 01:26:43,793 --> 01:27:10,057 cocoa, that was the thing that exploded. Now those people with a 1180 01:21:14,256 --> 01:23:12,852 small subset small portfolio which had the majors in it, the currencies, 1181 01:23:12,852 --> 01:25:21,331 the bonds, that sort of stuff, they had no exposure to cocoa. We 1182 01:25:21,331 --> 01:26:50,278 had that idiosyncratic exposure to cocoa because we are 1183 01:26:50,278 --> 01:27:10,044 maximally diversified. And this is what I find this leptokurtic 1184 01:21:17,985 --> 01:22:24,415 property of markets, the fractal nature of the markets means 1185 01:22:24,415 --> 01:23:44,131 that as you increase your sample size, you find that the large 1186 01:23:44,131 --> 01:24:57,204 things actually get larger. And that's why your drawdown is always 1187 01:24:57,204 --> 01:26:16,920 bigger ahead of you, as the greater returns will always be ahead 1188 01:26:16,920 --> 01:27:10,064 of you as you increase your sample size. And I've taken 125 years 1189 01:20:51,856 --> 01:21:56,692 of market data from the Dow Jones, combining that with the S&P 1190 01:21:56,692 --> 01:23:06,931 500, and when I look at the major material moves there, you find 1191 01:23:06,931 --> 01:24:06,364 that the major material moves, the tail properties that Dave is 1192 01:24:06,364 --> 01:25:05,797 convincing us about, and I totally agree with him, they are 1193 01:25:05,797 --> 01:25:59,827 the major drivers. That's where all the information is. The 1194 01:25:59,827 --> 01:26:53,857 rest of the movement is just noise, it's perturbations. That's 1195 01:26:53,857 --> 01:27:10,066 all it is. These drivers, as you extend the sample, you get access 1196 01:21:21,121 --> 01:22:23,029 to the drivers of the system in this fractal structure. Because 1197 01:22:23,029 --> 01:23:24,937 fractals are scale invariant, you can go down and the structure 1198 01:23:24,937 --> 01:24:32,473 always stays there. You can go up and the structure always stays 1199 01:24:32,473 --> 01:25:45,637 there. But as you go up, you get access to bigger, bigger, bigger 1200 01:25:45,637 --> 01:26:36,289 structure. This is what happens with sample increase and 1201 01:26:36,289 --> 01:27:10,057 diversification, at least in my research. 1202 01:27:11,350 --> 01:27:18,810 Sure, super interesting. Now we've got a couple questions left 1203 01:27:18,810 --> 01:27:29,254 that I love to talk to you about. The next one is pretty important, 1204 01:27:29,254 --> 01:27:36,714 I think, because what people will have hopefully gathered from 1205 01:27:36,714 --> 01:27:44,920 our conversation so far is that asymmetry is important in the 1206 01:27:44,920 --> 01:27:53,126 way we view opportunities, markets, etc. But you guys come at 1207 01:27:53,126 --> 01:27:56,856 it in a different way. So, Dave, you buy this asymmetry. And 1208 01:27:20,205 --> 01:27:36,650 of course Rich and other trend followers, we kind of built it into 1209 01:27:36,650 --> 01:27:53,095 what we do. So, let's start with you Dave. Why do you prefer 1210 01:27:53,095 --> 01:27:56,890 this implementation methodology? 1211 01:27:58,411 --> 01:28:33,163 Well, sort of two things that, I would say, differentiates how we 1212 01:28:33,163 --> 01:29:05,019 go about it. One, I'm creating the trend following type benefit 1213 01:29:05,019 --> 01:29:36,875 for our investors as opposed to running it ourselves. So, our 1214 01:29:36,875 --> 01:30:05,835 investors, we are freeing up capital that they have traditionally 1215 01:30:05,835 --> 01:30:34,795 tied up in poor diversifiers by creating explicit risk mitigation 1216 01:30:34,795 --> 01:31:15,339 brakes so that they can go out and put that capital to work participating 1217 01:31:15,339 --> 01:31:47,195 in the positive drift or long-term trend of asset price inflation 1218 01:31:47,195 --> 01:31:55,883 in compounding assets. So, the positively correlated side we're 1219 01:28:19,369 --> 01:29:04,778 allowing them to carry more of on their own books while we focus 1220 01:29:04,778 --> 01:29:32,722 only on the negatively correlated convexity asymmetry that 1221 01:29:32,722 --> 01:30:07,652 we need to constantly work to maintain sensitivity and attachment 1222 01:30:07,652 --> 01:30:46,075 as their assets grow in size, compounding, and price moves through 1223 01:30:46,075 --> 01:31:14,019 time through this drift and appreciation, asset appreciation 1224 01:31:14,019 --> 01:31:52,442 through time. So, we need to then actively manage the negative 1225 01:31:52,442 --> 01:31:55,935 correlation. So, one, we're trying to create very much what you 1226 01:28:34,401 --> 01:29:17,589 guys are doing, but the combination of what they do for themselves 1227 01:29:17,589 --> 01:30:04,376 and what we then provide them creates the long gamma, if you will, 1228 01:30:04,376 --> 01:30:47,564 on their combined portfolio as opposed to us running both the positive 1229 01:30:47,564 --> 01:31:30,752 and negative correlation in our books as you guys would as more 1230 01:31:30,752 --> 01:31:55,945 of a standalone strategy. So, that's one. Two, now this one, to 1231 01:28:20,006 --> 01:29:20,472 me, this is my big thing because it's what I do. All of that 1232 01:29:20,472 --> 01:29:55,024 asymmetry, all of those non-Gaussian distributions that you 1233 01:29:55,024 --> 01:30:33,895 guys capture the movement of the actual underlying realization 1234 01:30:33,895 --> 01:31:21,404 of, we're in the business of buying that implicit leverage that 1235 01:31:21,404 --> 01:31:55,956 allows us to overlay that on the mispricing. So, we're buying 1236 01:28:05,395 --> 01:28:28,675 long-dated term periods of the mispricing of the underlying distribution 1237 01:28:28,675 --> 01:28:54,283 which itself is driven by the supply of volatility dominated by 1238 01:28:54,283 --> 01:29:05,923 non-price sensitive participants, predominantly structured 1239 01:29:05,923 --> 01:29:29,203 product vol supply that is itself mispriced relative to its 1240 01:29:29,203 --> 01:29:54,811 own distribution. And so, you have this sort of double whammy, 1241 01:29:54,811 --> 01:30:20,419 I'll argue triple whammy, of a price for the flawed distribution 1242 01:30:20,419 --> 01:30:46,027 of the underlying over a long-term period of playing for that 1243 01:30:46,027 --> 01:31:06,979 itself is mispriced, that is constructed in product construction 1244 01:31:06,979 --> 01:31:37,243 that has inherent in it an added level of asymmetry - the non-linearity 1245 01:31:37,243 --> 01:31:55,867 of explicit option pricing or volatility type products. And so, 1246 01:28:09,721 --> 01:29:23,236 you get sort of this triple whammy. And then, again, to my St. 1247 01:29:23,236 --> 01:30:19,786 Petersburg paradox example, I'm not just getting to play the 1248 01:30:19,786 --> 01:31:33,301 game one time, I'm committing to play it for three years, at the 1249 01:31:33,301 --> 01:31:55,921 same price, every time. And so not only might I get to win the large 1250 01:28:20,741 --> 01:28:45,101 tail one time, but that one time almost certainly raises to the 1251 01:28:45,101 --> 01:29:01,341 realization of somebody who's provided this non-recourse leverage 1252 01:29:01,341 --> 01:29:29,761 that they've mispriced it. So, they may not want me to get to play 1253 01:29:29,761 --> 01:29:54,121 the game at that price, again, and again, and again, every day, 1254 01:29:54,121 --> 01:30:22,541 with infinity on the other side, for the next 2, 3, 5, 10 years. 1255 01:31:57,961 --> 01:31:58,361 The. 1256 01:31:58,521 --> 01:32:04,281 Recognition of the goalkeeping efficiency or the breaks that allows 1257 01:32:04,281 --> 01:32:07,161 then our investors to go out and be much more aggressive. And 1258 01:32:07,161 --> 01:32:10,921 so exactly the conversation we were just having about diversification. 1259 01:32:10,921 --> 01:32:14,481 And this is again another dredgism that goes back to my banker's 1260 01:32:14,481 --> 01:32:18,201 trust days running banker's trust businesses during the Asian 1261 01:32:18,201 --> 01:32:21,731 crisis out here, which was a pretty wild time, more wild than 1262 01:32:21,731 --> 01:32:24,451 anything I've ever seen, and I've seen almost everything that 1263 01:32:24,451 --> 01:32:29,211 was wild, that when you're positively convex, more risk is less 1264 01:32:29,211 --> 01:32:32,731 risk. Exactly what Rich was saying. When we're positively convex, 1265 01:32:32,731 --> 01:32:36,651 we want the whole tree, not just a branch. And we're happy to 1266 01:32:36,651 --> 01:32:41,651 have a room full just exactly like you guys are of opposing views. 1267 01:32:41,651 --> 01:32:45,011 We don't have actually any views. If somebody can create great 1268 01:32:45,011 --> 01:32:49,061 convexity that is positively directional and somebody else wants 1269 01:32:49,061 --> 01:32:52,821 to create direct convexity that is negatively directional, that 1270 01:32:52,821 --> 01:32:55,821 is bearish or bullish, the more the merrier. 1271 01:32:57,181 --> 01:33:23,571 And so that's how we're going about it in hopes that we create 1272 01:33:23,571 --> 01:33:41,841 the recognition of the goalkeeping efficiency, or the brakes, 1273 01:33:41,841 --> 01:34:08,231 that allows then our investors to go out and be much more aggressive. 1274 01:34:08,231 --> 01:34:26,501 And so exactly the conversation we were just having 1275 01:34:26,501 --> 01:34:30,561 about diversification. And this is, again, another Dredgism 1276 01:33:03,079 --> 01:33:13,892 that goes back to my Bankers Trust days running Bankers Trust 1277 01:33:13,892 --> 01:33:24,705 businesses during the Asian crisis out here, which was a pretty 1278 01:33:24,705 --> 01:33:35,518 wild time, more wild than anything I've ever seen, and I've 1279 01:33:35,518 --> 01:33:44,365 seen almost everything that was wild. ‘When you're positively 1280 01:33:44,365 --> 01:33:55,178 convex, more risk is less risk’. Exactly what Rich was saying. 1281 01:33:55,178 --> 01:34:05,991 When we're positively convex, we want the whole tree, not just 1282 01:34:05,991 --> 01:34:18,770 a branch. And we're happy to have a room full, just exactly like 1283 01:34:18,770 --> 01:34:29,583 you guys are, of opposing views. We don't have actually any 1284 01:34:29,583 --> 01:34:30,566 views. If somebody can create great convexity that is positively 1285 01:33:15,451 --> 01:33:33,721 directional and somebody else wants to create direct convexity 1286 01:33:33,721 --> 01:33:51,991 that is negatively directional, that is bearish or bullish, 1287 01:33:51,991 --> 01:34:12,291 the more the merrier. Right? We're reducing risk and capturing 1288 01:34:12,291 --> 01:34:30,561 more tails, all the better and go for it. So, it comes back down 1289 01:33:01,626 --> 01:33:13,183 to, then, obviously, and this is what Cem and I talk about all 1290 01:33:13,183 --> 01:33:22,962 the time, Niels, when we're together, is price. What's the price? 1291 01:33:22,962 --> 01:33:31,852 That price again is heavily, heavily driven by non-price sensitive 1292 01:33:31,852 --> 01:33:42,520 participants. And flaws, if you will, or I don't know if you 1293 01:33:42,520 --> 01:33:54,966 want to call them flaws. They just are what they are in the application 1294 01:33:54,966 --> 01:34:03,856 of regulatory risk and accounting rules on the main market 1295 01:34:03,856 --> 01:34:11,857 participants that allow them to recognize enhanced return, these 1296 01:34:11,857 --> 01:34:19,858 repackaged yields of captured option premium, and not necessarily 1297 01:34:19,858 --> 01:34:27,859 capitalize the tails. And they don't have skin-in-the-game. It's 1298 01:34:27,859 --> 01:34:30,526 not their money. And so, they're not thinking about the possibility 1299 01:33:05,213 --> 01:33:14,249 of what could happen. They're measuring the regulatory probability 1300 01:33:14,249 --> 01:33:26,297 of it. They're happy to… Another example I use all the time, 1301 01:33:26,297 --> 01:33:38,345 Rich, is we're going to a casino and because the croupier is 1302 01:33:38,345 --> 01:33:50,393 getting 10 cents of every dollar he makes, he's trying to attract 1303 01:33:50,393 --> 01:34:02,441 people to his roulette table because by offering 100 to one payout. 1304 01:34:02,441 --> 01:34:18,505 But what I'm trying to get him to do is commit to that 100 to 1 1305 01:34:18,505 --> 01:34:30,553 payout for a thousand spins, not just one spin at a time. And 1306 01:32:59,849 --> 01:33:34,533 when he figures out that he's got a problem, he's likely to try 1307 01:33:34,533 --> 01:34:09,217 to buy me out of my original agreement. And that's where the big 1308 01:34:09,217 --> 01:34:30,561 asymmetry comes from - volatility or volatility convexity. 1309 01:34:30,741 --> 01:34:35,141 Yeah, that's a great way of explaining it. Share some thoughts 1310 01:34:35,141 --> 01:34:39,541 on this, Rich, and why trend followers do it slightly differently. 1311 01:34:41,401 --> 01:35:47,173 We do it slightly differently, but I think we're playing, Dave and 1312 01:35:47,173 --> 01:36:36,502 myself particularly, we're playing the same game. So, we're 1313 01:36:36,502 --> 01:37:25,831 engineers. We're trying to architect the optimal geometry of 1314 01:37:25,831 --> 01:38:15,160 returns to survive and to capitalize, and exploit opportunities. 1315 01:34:41,401 --> 01:34:57,325 So, if you imagine when you look at the market structure, the 1316 01:34:57,325 --> 01:35:10,595 markets themselves have very little skew in them. But through 1317 01:35:10,595 --> 01:35:27,846 the processes that both Dave and a trend follower do is that they 1318 01:35:27,846 --> 01:35:39,789 truncate that left tail with their return distribution. That's 1319 01:35:39,789 --> 01:35:58,367 the brakes. They truncate that left tail. As we do it with our stops, 1320 01:35:58,367 --> 01:36:16,945 we do it with our small bets. Dave does it with a small premium 1321 01:36:16,945 --> 01:36:32,869 you pay for the option. We're truncating that left tail, but we're 1322 01:36:32,869 --> 01:36:47,466 leaving that right tail open - the positive skew. We're leaving 1323 01:36:47,466 --> 01:37:00,736 an unbounded opportunity because no one can predict what these 1324 01:37:00,736 --> 01:37:14,006 markets can do. But the fractal structure tells you that 1325 01:37:14,006 --> 01:37:27,276 there are things that we've never experienced before which we're 1326 01:37:27,276 --> 01:37:43,200 going to see and they can be exploited. This is the accelerator 1327 01:37:43,200 --> 01:37:59,124 on the racetrack - if we predict our downside, if we protect 1328 01:37:59,124 --> 01:38:15,048 that volatility drag of the down side and we exploit that opportunity. 1329 01:34:41,401 --> 01:35:15,331 Dave and our solution; his options and our trend following, 1330 01:35:15,331 --> 01:35:59,440 it's doing the same thing. If you look at a long position with 1331 01:35:59,440 --> 01:36:40,156 a stop, that's what we adopt. That's a synthetic call. A short 1332 01:36:40,156 --> 01:37:20,872 position with a stop is a synthetic put. You get the shape 1333 01:37:20,872 --> 01:37:58,195 of an option in trend following, but we're just not paying 1334 01:37:58,195 --> 01:38:15,160 an explicit premium for it. But there is a premium we are paying 1335 01:35:01,281 --> 01:35:28,616 for it. It's psychological. It's the losses we incur. It's a 1336 01:35:28,616 --> 01:36:00,921 foregone growth that we lose. There is a premium that a trend follower 1337 01:36:00,921 --> 01:36:33,226 pays. They often say, oh, but we don't pay the premium like the 1338 01:36:33,226 --> 01:37:03,046 long vol specialists out there paying for their options. We do pay 1339 01:37:03,046 --> 01:37:35,351 it. We do pay it. And we're paying it through the stops that 1340 01:37:35,351 --> 01:38:02,686 get triggered out to forego future opportunities. We do get those. 1341 01:38:02,686 --> 01:38:15,111 And it's a psychological cost. So, Dave gets a line item, where 1342 01:35:01,351 --> 01:35:38,401 he's got the premium to pay for his option. That's his line item. 1343 01:35:38,401 --> 01:36:09,751 We don't necessarily have that. We can't see that line item. 1344 01:36:09,751 --> 01:36:43,951 But it's there, trust me. It's there in how we experience things. 1345 01:36:43,951 --> 01:37:15,301 We experience lots of losses, whipsaws along the way. That's the 1346 01:37:15,301 --> 01:37:46,651 price we're paying. That's our premium we're paying to keep that 1347 01:37:46,651 --> 01:38:15,151 truncated left tail and to exploit the right tail opportunity. 1348 01:34:41,401 --> 01:35:08,428 So, I see both of our approaches as architecting a profile 1349 01:35:08,428 --> 01:35:30,541 to optimize geometry because path dependence is so critically 1350 01:35:30,541 --> 01:35:52,654 important and that's what we've expressed over this podcast. 1351 01:35:52,654 --> 01:36:19,681 So, I can only see similarities here. I might see a 1352 01:36:19,681 --> 01:36:49,165 slight difference in that I think Dave trades what I call discontinuity. 1353 01:36:49,165 --> 01:37:21,106 In other words, when we put a stop in, there's no guarantee that 1354 01:37:21,106 --> 01:37:55,504 stop is going to be triggered. We might get a gap through that stop. 1355 01:37:55,504 --> 01:38:15,160 We'll still have small bets with maximum diversification. Dave 1356 01:34:44,455 --> 01:35:14,995 has a guarantee with the explicit premium, so he immediately 1357 01:35:14,995 --> 01:35:54,697 gets that guarantee. We get it through what I call sort of duration 1358 01:35:54,697 --> 01:36:28,291 over the course of time. Very similar, but slightly different approaches. 1359 01:36:28,291 --> 01:37:04,939 So, yeah, I can only see similarities. In fact, what I see 1360 01:37:04,939 --> 01:37:32,425 is a benefit of perhaps combining both approaches together 1361 01:37:32,425 --> 01:38:12,127 to get the both of both worlds. But that's how I see it, 1362 01:38:12,127 --> 01:38:15,181 Niels. 1363 01:38:16,991 --> 01:38:17,711 Absolutely. 1364 01:38:17,791 --> 01:38:26,129 We advocate that to everybody we work with. There's path dependency. 1365 01:38:26,129 --> 01:38:34,467 And again, any way you can add positive convexity, add positive 1366 01:38:34,467 --> 01:38:44,321 convexity. And there are a whole bunch of ways to go about it. 1367 01:38:44,321 --> 01:38:52,659 And the fact is that the opportunity to capture cocoa, we 1368 01:38:52,659 --> 01:39:00,997 didn't have cocoa because nobody came to us with a volatility 1369 01:39:00,997 --> 01:39:09,335 structure and volatility supply that suits how we do it. Now, 1370 01:39:09,335 --> 01:39:16,915 we might have things that nobody ever thought anybody would 1371 01:39:16,915 --> 01:39:23,737 do, but that's because there was a volatility supply. But we're 1372 01:38:19,137 --> 01:38:27,213 limited in what we can do because there needs to be efficient 1373 01:38:27,213 --> 01:38:36,635 volatility markets for us to do it in, and there needs to be efficient 1374 01:38:36,635 --> 01:38:42,692 volatility products, even above and beyond the simplicity of 1375 01:38:42,692 --> 01:38:50,095 what might be exchange traded and stuff. So, they're very complementary. 1376 01:38:50,095 --> 01:38:56,825 There are different parts of the distribution. There are different 1377 01:38:56,825 --> 01:39:04,228 types of scenarios. The breadth of what you guys can do, 1378 01:39:04,228 --> 01:39:11,631 we can't match in terms of markets. The attachment points are 1379 01:39:11,631 --> 01:39:19,707 going to be different. Every way you can add convexity, add convexity. 1380 01:39:19,707 --> 01:39:23,745 That's the answer to everyone's problem. 1381 01:39:25,302 --> 01:39:33,414 Yeah, I mean, there are a lot of similarities. I was smiling when 1382 01:39:33,414 --> 01:39:41,526 I read your June paper where you have this chart where you show 1383 01:39:41,526 --> 01:39:49,638 what happens when you add 50/50 long vol and the S&P and, of 1384 01:39:49,638 --> 01:39:57,750 course, you don't just get an average, you kind of almost get the 1385 01:39:57,750 --> 01:40:05,238 sum. And I use exactly the same chart for trend following. If 1386 01:40:05,238 --> 01:40:13,974 you add the S&P, and 50%, and 50%, on capital, it's not just the 1387 01:40:13,974 --> 01:40:20,214 average, it's almost the sum of the two. It's incredible. Now 1388 01:39:26,188 --> 01:39:37,706 this conversation will air on Saturday, so a day and a half from 1389 01:39:37,706 --> 01:39:50,996 now. And what I'm hoping for is that there are a lot of CIOs out 1390 01:39:50,996 --> 01:40:00,742 there who are going to say, whoa, this is interesting stuff. 1391 01:40:00,742 --> 01:40:13,146 So, I guess my final question, unless I think of more, what do you 1392 01:40:13,146 --> 01:40:20,234 want them to do differently Monday morning?... Dave. 1393 01:40:21,482 --> 01:40:39,379 It's the same thing I will always say, add convexity, right? 1394 01:40:39,379 --> 01:40:58,903 So, the first thing that needs to change in the fiduciary investment 1395 01:40:58,903 --> 01:41:16,800 universe is a change in the incentive structure. So, we have 1396 01:41:16,800 --> 01:41:33,070 to somehow shift the incentive to being this geometric compounding 1397 01:41:33,070 --> 01:41:49,340 path. That just doesn't exist. Everybody's getting paid on calendar 1398 01:41:49,340 --> 01:42:03,983 year, arithmetic returns, time ignoring ergodic processes. So, you 1399 01:42:03,983 --> 01:42:21,880 need to change the incentive structure and get people to understand 1400 01:42:21,880 --> 01:42:34,896 that the objective is the growth of wealth. Expected return 1401 01:40:27,160 --> 01:40:58,389 is a nonsensical number and it's destroyed more wealth for capital 1402 01:40:58,389 --> 01:41:29,618 owners than maybe any other simple thing. Now, arguably, it may 1403 01:41:29,618 --> 01:42:00,847 have transferred wealth to fiduciaries and so there may be some 1404 01:42:00,847 --> 01:42:34,915 reason why they stick with it, but that's just me being cynical. 1405 01:40:21,482 --> 01:40:58,389 So, you need to add convexity. How do you add convexity? Rich and 1406 01:40:58,389 --> 01:41:32,457 I have talked about optimizing isn't the right answer. It's a ‘learn 1407 01:41:32,457 --> 01:42:00,847 by doing’ because you're now building resilience to the divergences 1408 01:42:00,847 --> 01:42:29,237 from the expectation as opposed to optimizing to a backward 1409 01:42:29,237 --> 01:42:34,915 looking expectation. So, you can't optimize to the unknown future 1410 01:40:33,082 --> 01:40:54,832 path. All you can do is put brakes on your car and go out and 1411 01:40:54,832 --> 01:41:13,682 learn how to drive. Go out and explore how to go faster, aerodynamics, 1412 01:41:13,682 --> 01:41:23,832 stronger engine, different tires, different steering wheel. 1413 01:41:23,832 --> 01:41:39,782 But before you start exploring that, you better have brakes, right? 1414 01:41:39,782 --> 01:41:58,632 So, you need to start freeing capital. (This is sort of my current 1415 01:41:58,632 --> 01:42:13,132 theme and presentations I give.) You need to start freeing 1416 01:42:13,132 --> 01:42:30,532 capital from the ballast that you've been using to try to just 1417 01:42:30,532 --> 01:42:34,882 slow you down. And you want to then use that capital to invest in 1418 01:40:39,599 --> 01:41:01,010 brakes, or free the capital by getting brakes, and then use it to 1419 01:41:01,010 --> 01:41:19,127 go out and find ways to participate in more positive market 1420 01:41:19,127 --> 01:41:32,303 environments, positive upside participating type assets. And so, 1421 01:41:32,303 --> 01:41:53,714 learning by doing, trial and error, is how you learn how to drive 1422 01:41:53,714 --> 01:42:11,831 with brakes, as opposed to trying to estimate the average lap 1423 01:42:11,831 --> 01:42:28,301 speed. Start trying to extend separation over multiple laps and 1424 01:42:28,301 --> 01:42:34,889 it's ‘learn by doing’. 1425 01:42:35,432 --> 01:42:39,260 Yeah, I've got a follow-up question on that in a second, Dave. 1426 01:42:39,260 --> 01:42:42,769 But Rich, what do you want people to do Monday morning? 1427 01:42:43,412 --> 01:43:40,568 Well, I want them to pay less attention to those damn statistics 1428 01:43:40,568 --> 01:44:18,672 and more get into the underlying, foundational assumptions 1429 01:44:18,672 --> 01:45:11,065 of the system you're assessing or look at the dependencies. But 1430 01:45:11,065 --> 01:46:08,221 specific things I'd like them to do? Here's one - cap leverage 1431 01:46:08,221 --> 01:46:22,510 in absolute terms. Now, what I'm doing here, there's no statistical 1432 01:43:24,172 --> 01:44:25,312 reason for this apart from what we're doing is we are breaking 1433 01:44:25,312 --> 01:45:21,357 the automatic link between a quiet market and a large position. 1434 01:45:21,357 --> 01:46:22,497 So, if we cap leverage in absolute terms, we break that link. 1435 01:42:43,412 --> 01:43:42,632 That's an important one because most of the capital destruction, 1436 01:43:42,632 --> 01:44:53,696 you see, is coming out of quiet markets with too high position 1437 01:44:53,696 --> 01:45:52,916 size, too much leverage. Leverage is the destruction of compounding. 1438 01:45:52,916 --> 01:46:22,526 This is what I've found. The second one is size every market to 1439 01:43:07,756 --> 01:43:38,186 the same risk rather than tilting toward whichever sort of 1440 01:43:38,186 --> 01:44:11,659 edge looks strongest. Now this comes back to the selection argument. 1441 01:44:11,659 --> 01:44:45,132 There's selection bias in our selection of things that have recently 1442 01:44:45,132 --> 01:45:15,562 performed, etc. By selecting or increasing our position size to 1443 01:45:15,562 --> 01:45:45,992 a particular asset because of its recent performance, what we're 1444 01:45:45,992 --> 01:46:13,379 doing is we're striving towards that single cable between 1445 01:46:13,379 --> 01:46:22,508 the canyon again. We are losing the breadth that we had through 1446 01:43:18,008 --> 01:43:56,448 the all of the different cables tied to separate anchors. 1447 01:43:56,448 --> 01:44:46,420 And we don't find it in the noise that exists in that market. 1448 01:44:46,420 --> 01:45:28,704 We're always estimating the return. But it's not enough to give 1449 01:45:28,704 --> 01:46:14,832 us the optimal bet (as we talked about with Kelly), this is 1450 01:46:14,832 --> 01:46:22,520 the problem. So, size every market with the same risk rather 1451 01:42:57,803 --> 01:43:12,194 than tilting towards whichever edge looks currently strongest. That 1452 01:43:12,194 --> 01:43:31,382 way you miss cocoa, that way you miss all of these explosions 1453 01:43:31,382 --> 01:43:52,169 that we talk about. It's not a refusal to think, it's actually the 1454 01:43:52,169 --> 01:44:08,159 opposite. You still choose your markets, you still choose your 1455 01:44:08,159 --> 01:44:25,748 horizons and your rules. That's the whole craft of what we 1456 01:44:25,748 --> 01:44:44,936 do as trend followers. But what you decline is a step afterwards, 1457 01:44:44,936 --> 01:44:59,327 ranking those markets by apparent strength and betting the 1458 01:44:59,327 --> 01:45:15,317 difference. So, the difference between edges is smaller than the 1459 01:45:15,317 --> 01:45:34,505 noise involved in measuring them. So, a ranking is a signal too 1460 01:45:34,505 --> 01:45:53,693 faint to trust. Trust in my sort of understanding of how these 1461 01:45:53,693 --> 01:46:08,084 markets work. And tilting towards the leaders quietly rebuilds 1462 01:46:08,084 --> 01:46:22,475 that single cable that we've got to worry about. And the third 1463 01:42:51,947 --> 01:43:28,932 one is grow by widening the book, not by scaling the bet. So, 1464 01:43:28,932 --> 01:44:03,072 we are avoiding leverage here. So, when you want more return, the 1465 01:44:03,072 --> 01:44:37,212 instinct is to increase size on what you already hold. Now, what 1466 01:44:37,212 --> 01:45:08,507 I'm saying is, I'm not advocating that. I'm saying add markets, 1467 01:45:08,507 --> 01:45:25,577 add different systems, different diversification. Expand 1468 01:45:25,577 --> 01:45:48,337 your diversification, don't increase concentration in your single 1469 01:45:48,337 --> 01:46:19,632 bets. That way you're reducing this reliance on this estimate once 1470 01:46:19,632 --> 01:46:22,477 again. So, it's the difference between a heavier load on one cable 1471 01:43:02,849 --> 01:43:24,053 and rather than look for more cables to that bridge, more sort 1472 01:43:24,053 --> 01:43:43,490 of unrelated things that won't bring your portfolio down, that gives 1473 01:43:43,490 --> 01:44:06,461 you more exposure to the tails of the fractal markets that we live 1474 01:44:06,461 --> 01:44:29,432 in. And the last one is enter small. So, always enter small and 1475 01:44:29,432 --> 01:44:48,869 let the market build the position out of open profit. This 1476 01:44:48,869 --> 01:45:13,607 is where I say let the market do the heavy work in lifting your 1477 01:45:13,607 --> 01:45:31,277 performance. Don't use size, or positioning, or leverage, to try 1478 01:45:31,277 --> 01:45:54,248 and sort of push your will upon the market. Let the market itself 1479 01:45:54,248 --> 01:46:15,452 do the lifting power. Keep your bet sizes small on any individual 1480 01:46:15,452 --> 01:46:22,520 estimate that you make. 1481 01:46:23,572 --> 01:46:32,493 Okay, so while you've been talking, I've thought of a question 1482 01:46:32,493 --> 01:46:45,469 and when you do things sort of on the fly, it may not come out as 1483 01:46:45,469 --> 01:46:56,823 clearly as you would like. But let me try anyways. So back in the 1484 01:46:56,823 --> 01:47:04,933 ‘90s we started talking about, in this industry, about portable 1485 01:47:04,933 --> 01:47:14,665 alpha. Now, that really never came too much but I certainly clearly 1486 01:47:14,665 --> 01:47:23,586 remember this and the concept of, say, having a equity portfolio 1487 01:47:23,586 --> 01:47:34,129 who, on top of it, would add trend following. I don't think I 1488 01:47:34,129 --> 01:47:43,050 talked about long vol back then. It probably didn't really exist 1489 01:47:43,050 --> 01:47:54,404 in the area that I was working in back then. But, of course, more 1490 01:47:54,404 --> 01:48:04,947 recently it's coming back with a vengeance. We see a number of managers 1491 01:48:04,947 --> 01:48:12,246 launching these portable alpha return stack products, etc., etc. 1492 01:48:12,246 --> 01:48:22,789 So, this is what I'd like to for you to help investors understand 1493 01:48:22,789 --> 01:48:32,521 or maybe deal with. It's the challenge where, on one side, as 1494 01:48:32,521 --> 01:48:39,009 Dave says, you need to free up capital. Now you can do that, in 1495 01:46:29,590 --> 01:46:41,626 a sense, by doing a portable alpha product because you can replicate 1496 01:46:41,626 --> 01:46:54,665 the index, and you can add trend following or you can add long 1497 01:46:54,665 --> 01:47:07,704 vol, and you still have money freed. But that line item is, essentially, 1498 01:47:07,704 --> 01:47:19,740 more efficient. But if you look at it from a traditional perspective, 1499 01:47:19,740 --> 01:47:31,776 you would say, well, I've got 100% equities, and now I'm adding 1500 01:47:31,776 --> 01:47:44,815 another 100% trend, or another 100% vol, that, to me, is a leveraged 1501 01:47:44,815 --> 01:47:58,857 product. So, on one hand we say don't go down the route of just 1502 01:47:58,857 --> 01:48:10,893 adding too much leverage. At the same time we're saying this is 1503 01:48:10,893 --> 01:48:21,926 so much more efficient. And when you combine these two, specifically 1504 01:48:21,926 --> 01:48:30,953 long vol and trend following with your underlying traditional 1505 01:48:30,953 --> 01:48:38,977 market, you actually get a much better product. So, how do investors 1506 01:46:34,412 --> 01:47:12,352 who may not be spending all their time, like we do, in this world, 1507 01:47:12,352 --> 01:47:47,582 how do they square the risk of too much leverage with the benefit 1508 01:47:47,582 --> 01:48:17,392 that we're arguing, saying, well actually in this case you should 1509 01:48:17,392 --> 01:48:39,072 do it, if you know what I mean. 1510 01:48:39,092 --> 01:48:55,273 Exactly. And it comes back to concavity and convexity. So, most 1511 01:48:55,273 --> 01:49:08,512 people are levering concave risks. So, they're levering risks 1512 01:49:08,512 --> 01:49:26,164 where their correlation is low to good markets and rising to bad 1513 01:49:26,164 --> 01:49:39,403 markets. So, they've got volatility drag in that dynamic. 1514 01:49:39,403 --> 01:49:54,113 Right? The volatility has a negative impact on their geometric 1515 01:49:54,113 --> 01:50:10,294 path and adding leverage to that will create disasters, the disasters 1516 01:50:10,294 --> 01:50:25,004 that we talk about in terms of what drives markets. If you're positively 1517 01:48:42,830 --> 01:48:55,290 convex, if you have explicitly negatively correlated risk that is 1518 01:48:55,290 --> 01:49:04,012 growing, so your overall portfolio's increasing correlation 1519 01:49:04,012 --> 01:49:15,226 to positive markets and reducing correlation to bad markets, 1520 01:49:15,226 --> 01:49:26,440 and think of it as non-recourse leverage, where you've 1521 01:49:26,440 --> 01:49:41,392 gone out and bought an option that allows you to participate in 1522 01:49:41,392 --> 01:49:53,852 the upside and give the downside to somebody else, well, 1523 01:49:53,852 --> 01:50:10,050 that is risk reducing as you grow that position. So back to what 1524 01:50:10,050 --> 01:50:23,756 I said before. When you're positively convex, more risk is less 1525 01:50:23,756 --> 01:50:25,002 risk. And so, you can see when we play around with examples, well, 1526 01:49:04,520 --> 01:49:34,186 the more long vol, the more hedge we add, the less risk you have. 1527 01:49:34,186 --> 01:50:03,852 So, you go take more risk, and then guess what, you could add more 1528 01:50:03,852 --> 01:50:20,804 negatively correlating highly asymmetric protection. You go take 1529 01:50:20,804 --> 01:50:25,042 more risk. And so, you can keep going, and going, and going, 1530 01:49:06,252 --> 01:49:30,696 with this, ad infinitum. But adding leverage, recourse leverage, 1531 01:49:30,696 --> 01:49:55,140 to concave positions is exactly what creates the volatility 1532 01:49:55,140 --> 01:50:22,300 drag and destroys compounding paths and creates the tail risks, 1533 01:50:22,300 --> 01:50:25,016 literally. 1534 01:50:25,282 --> 01:50:27,757 Yeah. Do you want to add something to that? 1535 01:50:28,082 --> 01:50:39,535 I can't really add to Dave, apart from the fact that Dave works 1536 01:50:39,535 --> 01:50:48,345 in this world all the time recommending long vol componentry 1537 01:50:48,345 --> 01:50:56,274 to add to traditional portfolios. Fortunately, in my trend 1538 01:50:56,274 --> 01:51:06,846 following world I'm 100% trend, so I never have to face this 1539 01:51:06,846 --> 01:51:15,656 demon that we're talking about. But if it improves concavity, 1540 01:51:15,656 --> 01:51:23,585 that's a good move. That's how I'd see it. But really what we're 1541 01:50:31,122 --> 01:50:41,762 doing here is we're applying a band aid fix to a bad return stream 1542 01:50:41,762 --> 01:50:47,842 with that traditional portfolio. It's basically like a 1543 01:50:47,842 --> 01:50:56,962 band aid approach to fix it, I'd say. Ideally, resurrect the architect 1544 01:50:56,962 --> 01:51:03,802 and start from Brown principles and construct that portfolio 1545 01:51:03,802 --> 01:51:12,922 without having to have that demon to deal with that poor old 1546 01:51:12,922 --> 01:51:21,282 Dave has to deal with long vol integration into traditional portfolios 1547 01:51:21,282 --> 01:51:23,562 all the time. 1548 01:51:25,272 --> 01:52:04,444 I also remember from my early days, in the early ‘90s in this space, 1549 01:52:04,444 --> 01:52:38,020 that one of my colleagues was using this slogan saying, and of 1550 01:52:38,020 --> 01:53:08,798 course again, this was about trend following, and I'm sure that 1551 01:53:08,798 --> 01:53:42,374 could be said about long vol, is that ‘trend following allows you 1552 01:53:42,374 --> 01:54:15,950 to own equities’. And I'm just thinking all these years later, I 1553 01:54:15,950 --> 01:54:52,324 mean, it's probably as true as it ever has been, yet most investors 1554 01:54:52,324 --> 01:55:25,900 have not yet embraced any of the two strategies that they need 1555 01:55:25,900 --> 01:55:31,496 the most. But hopefully the last two hours with you, and thank 1556 01:52:03,162 --> 01:52:52,419 you for going long here, so to speak, will change or challenge some 1557 01:52:52,419 --> 01:53:26,520 of people's assumptions about their investments. So, very excited 1558 01:53:26,520 --> 01:54:11,988 about that and super grateful, by the way, for all the knowledge 1559 01:54:11,988 --> 01:55:01,245 that you guys have shared in the last couple of hours. And I'm 1560 01:55:01,245 --> 01:55:31,557 sure all our listeners will be as well. And maybe a little bit of… 1561 01:51:45,792 --> 01:52:30,252 what I like to do, actually, I’d like to do something a little 1562 01:52:30,252 --> 01:53:14,712 bit different here. First of all, I’d like to ask a small favor 1563 01:53:14,712 --> 01:53:55,752 of the listeners and that is to go and show their appreciation 1564 01:53:55,752 --> 01:54:36,792 to both Dave and Rich by heading over to your favorite podcast 1565 01:54:36,792 --> 01:55:14,412 platform, leave a rating and review because it really does help 1566 01:55:14,412 --> 01:55:31,512 with more people seeing this. But I like to up the stakes for the 1567 01:51:48,123 --> 01:52:18,591 first time here. Since people watch us on YouTube they will see 1568 01:52:18,591 --> 01:52:49,059 that I'm the one who forgot the Top Traders Unplugged merch here. 1569 01:52:49,059 --> 01:53:24,605 So, what I'd like to do, I'd like to make a little challenge here 1570 01:53:24,605 --> 01:53:57,612 and say there is a Top Traders Unplugged vest at stake here for 1571 01:53:57,612 --> 01:54:33,158 the best review that is posted in the next month. So, we give it 1572 01:54:33,158 --> 01:55:11,243 to the end of August, so the people have a chance to go out and 1573 01:55:11,243 --> 01:55:31,555 leave an amazing review for Rich and Dave. However, since I don't 1574 01:51:32,516 --> 01:51:54,248 have infinite time to monitor all these platforms, I need to ask 1575 01:51:54,248 --> 01:52:15,980 a favor. I need you to email the usual one info@toptraders unplugged.com. 1576 01:52:15,980 --> 01:52:34,090 I'm probably going to, hopefully, I'm going to regret this 1577 01:52:34,090 --> 01:52:57,633 with a large number of emails to email us what the review was, 1578 01:52:57,633 --> 01:53:22,987 where you posted it, when you posted it, so that one of my team 1579 01:53:22,987 --> 01:53:44,719 members will go and verify and then we will either by democratic 1580 01:53:44,719 --> 01:54:04,640 vote on the Top Traders Unplugged team, select the best and 1581 01:54:04,640 --> 01:54:33,616 get in touch to the email that you send us so that you can get your 1582 01:54:33,616 --> 01:55:04,403 merch or we may ask, if we can figure out a way to do it. We may 1583 01:55:04,403 --> 01:55:27,946 upload them to ChatGPT and ask ChatGPT who do you think should get 1584 01:55:27,946 --> 01:55:31,568 this vest? So, upping the stakes for the first time here, hopefully 1585 01:52:02,582 --> 01:52:47,354 that will get more people going on that. Anyways, as you can 1586 01:52:47,354 --> 01:53:28,395 tell I'm super excited about this conversation and I really do 1587 01:53:28,395 --> 01:54:20,629 hope that lots of people will tune in and listen to it. I mean 1588 01:54:20,629 --> 01:55:09,132 there are no better people than the two of you to explain these 1589 01:55:09,132 --> 01:55:31,518 things. So, exceptionally grateful for that. Next week I will 1590 01:51:32,968 --> 01:51:59,904 be joined by Harry Moore from Man AHL. That will also be an interesting 1591 01:51:59,904 --> 01:52:19,144 conversation, not least because I think we may touch also 1592 01:52:19,144 --> 01:52:46,080 on portable alpha as I have a feeling he's coming out with a paper 1593 01:52:46,080 --> 01:53:07,244 soon. So, we can put all of these conversations into perspective. 1594 01:53:07,244 --> 01:53:34,180 But in any event I'm going to give you the chance as I mentioned 1595 01:53:34,180 --> 01:53:59,192 before, go and follow Dave, go and follow Rich. They put out some 1596 01:53:59,192 --> 01:54:22,280 amazing information. But also, I want to leave room for both of 1597 01:54:22,280 --> 01:54:45,368 you to any final words, anything you… I wouldn't say we left 1598 01:54:45,368 --> 01:55:06,532 out. I think we've been pretty thorough…. But anyways, if there 1599 01:55:06,532 --> 01:55:31,544 is something that I forgot to bring up you can certainly add it. 1600 01:55:31,582 --> 01:55:38,830 I'll throw one last Dredgism out there. And so, this is something 1601 01:55:38,830 --> 01:55:45,474 that amongst my various other activities I'm involved in a lot 1602 01:55:45,474 --> 01:55:51,514 of risk councils, and risk conferences, and stuff, and people 1603 01:55:51,514 --> 01:55:59,366 often are waiting for me to give them some magic formula of what 1604 01:55:59,366 --> 01:56:06,010 is risk, and what is risk management, and what's the right 1605 01:56:06,010 --> 01:56:14,466 way to do it? And I end all of those presentations with this slide. 1606 01:56:14,466 --> 01:56:18,694 ‘Risk is about accountability’, and that there's 1607 01:56:18,694 --> 01:56:25,338 no (exactly what we're talking about here), there's no magic modeling 1608 01:56:25,338 --> 01:56:31,982 of risk, there's no magic formula of risk, there's no magic 1609 01:56:31,982 --> 01:56:38,626 regulatory model rule. The only way that risk can be appropriately 1610 01:56:38,626 --> 01:56:42,854 managed is through accountability, is through tacit 1611 01:56:42,854 --> 01:56:48,894 knowledge, is through recognition that if I crash I'm out 1612 01:56:48,894 --> 01:56:54,934 of the race. And until, somehow, that gets instilled into 1613 01:56:54,934 --> 01:56:59,766 all of these non-price sensitive rational accounting men, 1614 01:56:59,766 --> 01:57:03,390 the system will forever be fragile. And I say this to regulators 1615 01:55:41,782 --> 01:56:03,882 every time I talk to them since Nassim and I wrote notes saying 1616 01:56:03,882 --> 01:56:22,582 do not embed value at risk in BIS regulatory capital guidelines 1617 01:56:22,582 --> 01:56:39,582 and remove accountability from the guys actually making the decisions 1618 01:56:39,582 --> 01:56:59,982 or you'll get the GFC. And nobody's listened to me yet. GFC 1619 01:56:59,982 --> 01:57:03,382 or not. 1620 01:57:04,052 --> 01:57:08,324 Well, I also like the fact that you often remind everybody that 1621 01:57:08,324 --> 01:57:12,240 actually risk is everything you didn't think of, which in itself 1622 01:57:12,240 --> 01:57:13,308 is a problem. 1623 01:57:14,132 --> 01:57:15,172 Not in your backtest. 1624 01:57:16,452 --> 01:57:18,632 Yeah, but it's not in your backtest. That's right. Anything 1625 01:57:18,632 --> 01:57:20,158 on your side, Rich, you want to... 1626 01:57:20,172 --> 01:57:24,215 It's just been a pleasure for the famous Mr. Dredge. Big fan, as 1627 01:57:24,215 --> 01:57:27,325 I've already mentioned, but it was great chatting, great sharing 1628 01:57:27,325 --> 01:57:31,368 ideas and this is a great venue for it, Niels. So, thank you 1629 01:57:31,368 --> 01:57:32,923 very much for this opportunity. 1630 01:57:36,832 --> 01:57:40,576 Sure, absolutely. I'm sure we will do it again. Anyways, to all 1631 01:57:40,576 --> 01:57:44,632 of you out here, thank you so much for listening. From Dave, Rich 1632 01:57:44,632 --> 01:57:48,688 and me, we look forward to being back with you next week here 1633 01:57:48,688 --> 01:57:52,120 from Top Traders Unplugged. And until next time, as usual, take 1634 01:57:52,120 --> 01:57:54,928 care of yourself and take care of each other. 1635 01:57:57,582 --> 01:59:30,570 Thanks for listening to Top Traders Unplugged. If you feel you 1636 01:58:02,894 --> 01:58:16,174 learned something of value from today's episode, the best way 1637 01:58:16,174 --> 01:58:33,438 to stay updated is to go on over to your favorite podcast platform 1638 01:58:33,438 --> 01:58:52,030 and follow the show so that you'll be sure to get all the new 1639 01:58:52,030 --> 01:59:05,310 episodes as they're released. We have some amazing guests lined 1640 01:59:05,310 --> 01:59:22,574 up for you and to ensure our show continues to grow, please leave 1641 01:59:22,574 --> 01:59:30,542 us an honest rating and review. It only takes a minute and 1642 01:58:19,038 --> 01:59:09,102 it's the best way to show us you love the podcast. We'll see you 1643 01:59:09,102 --> 01:59:30,558 next time on Top Traders Unplugged. This podcast expresses 1644 01:58:03,516 --> 01:58:29,230 the views of its hosts and the guests appearing on the podcast as 1645 01:58:29,230 --> 01:58:54,944 of the date of its recording, and such views are subject to change 1646 01:58:54,944 --> 01:59:16,702 without notice. Top Traders Unplugged do not have any duty or 1647 01:59:16,702 --> 01:59:30,548 obligation to update the information contained herein. Furthermore, 1648 01:57:59,850 --> 01:58:22,530 Top Traders Unplugged make no representation to its accuracy and 1649 01:58:22,530 --> 01:58:47,478 it shall not be assumed that past investment performance is an 1650 01:58:47,478 --> 01:59:10,158 indication of future results. Moreover, wherever there is a potential 1651 01:59:10,158 --> 01:59:30,570 for profit, there is also the possibility of loss. This content 1652 01:58:07,914 --> 01:58:54,408 is made available for educational purposes only and should 1653 01:58:54,408 --> 01:59:30,570 not be used for any other purpose. The information contained 1654 01:58:05,331 --> 01:58:33,744 in this podcast does not constitute and should not be construed 1655 01:58:33,744 --> 01:59:04,740 as investment advice or an offer to sell or a solicitation to 1656 01:59:04,740 --> 01:59:30,570 buy any securities or related financial instruments in any jurisdiction. 1657 01:57:57,582 --> 01:58:15,173 Certain information contained herein concerning economic trends 1658 01:58:15,173 --> 01:58:40,303 and performance is based on or derived from information provided 1659 01:58:40,303 --> 01:59:00,407 by independent third-party sources. Top Traders Unplugged may 1660 01:59:00,407 --> 01:59:25,537 believe that the sources from which such information are obtained 1661 01:59:25,537 --> 01:59:30,563 are reliable. However, Top Traders Unplugged cannot guarantee 1662 01:58:06,294 --> 01:58:19,362 the accuracy of such information and has not independently 1663 01:58:19,362 --> 01:58:30,978 verified the accuracy or completeness of such information 1664 01:58:30,978 --> 01:58:46,950 or the assumptions on which such information is based. This podcast, 1665 01:58:46,950 --> 01:59:00,018 including the information contained herein, may not be reproduced, 1666 01:59:00,018 --> 01:59:17,442 copied, republished or posted in whole or in part in any form, 1667 01:59:17,442 --> 01:59:30,510 without the prior written consent of Top Traders Unplugged