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 00:00:35,820 --> 00:00:45,071 Welcome and welcome back to this week's edition of the Systematic 10 00:00:45,071 --> 00:00:52,640 Investor series with Richard Brennan and I, Niels Kaastrup-Larsen, 11 00:00:52,640 --> 00:01:02,732 where each week we take the pulse of the global markets through 12 00:01:02,732 --> 00:01:07,778 the lens of a rules-based investor. And let me just also say 13 00:00:39,018 --> 00:00:45,947 a warm welcome if today is the first time you're joining us, and 14 00:00:45,947 --> 00:00:51,810 if someone who cares about you and your portfolio recommended that 15 00:00:51,810 --> 00:00:59,805 you tune into the podcast, let me just say a big thank you for sharing 16 00:00:59,805 --> 00:01:05,668 this episode with your friends and colleagues. It really does mean 17 00:01:05,668 --> 00:01:07,800 a lot to us. Rich, it is wonderful to be back with you this 18 00:00:55,800 --> 00:01:07,788 week. How are things down under? 19 00:01:08,760 --> 00:01:31,479 Very good, Niels. Always good to be back on this podcast, but the 20 00:01:12,250 --> 00:01:16,089 weather over here's been a bit inclement lately. It's trying to 21 00:01:16,089 --> 00:01:20,626 work out whether it should still be the tail end of summer or 22 00:01:20,626 --> 00:01:24,814 the beginning of winter. It doesn't know what to do. So, we're 23 00:01:24,814 --> 00:01:28,653 getting these very heavy storms at the moment in a fairly 24 00:01:28,653 --> 00:01:31,445 cold climate. So, it's very unusual, very interesting. 25 00:01:32,260 --> 00:01:40,996 Yes, well I see you have dressed up in the winter TTU merch. 26 00:01:40,996 --> 00:01:49,732 I had to change to the summer edition for this week because it's 27 00:01:49,732 --> 00:01:57,796 actually getting pretty hot in Europe, at least in parts of Europe. 28 00:01:57,796 --> 00:02:07,204 So, there we are. I know this is an audio podcast for the most 29 00:02:07,204 --> 00:02:16,612 part, so people have to really catch us on social media for a quick 30 00:02:16,612 --> 00:02:21,316 sneak of what I'm referring to here. Anyways, we have a really important 31 00:01:37,336 --> 00:01:47,488 but also a wonderful topic to discuss today, many facets of it, 32 00:01:47,488 --> 00:01:59,332 and so I'm excited about that. But as you know, I always would like 33 00:01:59,332 --> 00:02:11,176 to start by just catching up on what sort of spin there is on 34 00:02:11,176 --> 00:02:21,328 your radar lately before we get into the main topics of today. 35 00:02:22,160 --> 00:03:13,856 Okay Neil, so what's been on my radar? So, I've been thinking 36 00:03:13,856 --> 00:04:09,860 about something for the past few weeks and it started when I listened 37 00:04:09,860 --> 00:05:01,556 to your conversation with Hari Krishnan and Cem Karsan on the Rise 38 00:05:01,556 --> 00:05:40,328 of Passive Investing. Great podcast, so I'd recommend listeners 39 00:05:40,328 --> 00:05:57,560 go and find that. So, Hari and Cem, they're working at a level of 40 00:02:40,890 --> 00:03:05,239 macro and quantitative detail that I do not work at, and what they 41 00:03:05,239 --> 00:03:29,588 say is really worth listening to in their own words. But it got 42 00:03:29,588 --> 00:03:48,318 me thinking about the same territory from a different angle. 43 00:03:48,318 --> 00:04:10,794 Not from the macro side and not from the mathematical side, but 44 00:04:10,794 --> 00:04:33,270 from the lens I've been working on for the past two years 45 00:04:33,270 --> 00:04:52,000 - complex adaptive systems. Because once I started thinking about 46 00:04:52,000 --> 00:05:10,730 passive through that lens, something became very clear to me 47 00:05:10,730 --> 00:05:33,206 that I'd not seen quite so sharply before. And that's the thing 48 00:05:33,206 --> 00:05:57,555 I want us to walk through into the rest of the conversation today. 49 00:02:22,160 --> 00:03:08,556 So, I just want to set up the picture for anyone who's been tracking 50 00:03:08,556 --> 00:03:48,324 the numbers of passive. So, for the past two decades, the share 51 00:03:48,324 --> 00:04:28,092 of US equity capital held by passive vehicles has grown from a 52 00:04:28,092 --> 00:05:01,232 small minority to a dominant force in the ownership structure. 53 00:05:01,232 --> 00:05:37,686 So, the most commonly cited figure puts passive at around about 54 00:05:37,686 --> 00:05:57,570 60% of US equity fund assets. And when you include institutional 55 00:02:49,775 --> 00:03:39,482 capital, that benchmarks to passive indices, you know, pension 56 00:03:39,482 --> 00:04:29,189 funds, sovereign wealth funds, endowments that effectively hug their 57 00:04:29,189 --> 00:05:24,419 benchmarks, the share of price insensitive ownership in US equities 58 00:05:24,419 --> 00:05:57,557 is meaningfully higher than that 60%. So, I want to flag something 59 00:02:39,620 --> 00:03:08,720 important before we go further. The figures I've just cited 60 00:03:08,720 --> 00:03:46,550 are US equity figures, but the story is much broader. The rise of 61 00:03:46,550 --> 00:04:12,740 passive is not simply an equity-only phenomenon. The growth 62 00:04:12,740 --> 00:04:41,840 of ETFs has spread passive vehicles into fixed income, commodities, 63 00:04:41,840 --> 00:05:05,120 currency exposure, multi-asset products. So, capital that allocates 64 00:05:05,120 --> 00:05:37,130 by index weight rather than price judgement is now a structural 65 00:05:37,130 --> 00:05:57,500 presence in markets well beyond US equities. I'll use the 66 00:02:32,582 --> 00:03:07,322 US equity figures throughout this episode because the data is 67 00:03:07,322 --> 00:03:31,640 cleanest, but the architectural argument extends across 68 00:03:31,640 --> 00:04:13,328 the whole landscape; all of the assets. So, the story is usually 69 00:04:13,328 --> 00:04:55,016 told as, you know, a victory of common sense. You know, passive 70 00:04:55,016 --> 00:05:33,230 is cheaper. Active managers have a lousy track record of beating 71 00:05:33,230 --> 00:05:57,548 their benchmarks. Investors make the rational choice. And to 72 00:02:28,314 --> 00:03:11,392 be clear, there is a degree of truth in that. The case for passive, 73 00:03:11,392 --> 00:03:45,239 at the individual investor level, is quite well made and largely 74 00:03:45,239 --> 00:04:22,163 correct. So, I'm not going to argue against that today, but what 75 00:04:22,163 --> 00:04:59,087 I want to do today is something different. I want to ask, 76 00:04:59,087 --> 00:05:32,934 what has passive's growth done to the agent ecology of markets? 77 00:05:32,934 --> 00:05:57,550 Because that's the question my framework points at. And the answer 78 00:02:38,729 --> 00:03:33,959 changes how I think about almost everything else that's happening 79 00:03:33,959 --> 00:04:51,281 in the markets right now. So, here's the move that I want to make. 80 00:04:51,281 --> 00:05:52,034 Markets on the view I work from, Niels, are complex adaptive 81 00:05:52,034 --> 00:05:57,557 systems. 82 00:05:58,260 --> 00:06:06,432 All right, okay, so you've used that phrase twice now in your 83 00:06:06,432 --> 00:06:13,242 intro: complex adaptive systems. And I think, for those who 84 00:06:13,242 --> 00:06:24,138 may not… I think we did a deep dive a few times on this that people 85 00:06:24,138 --> 00:06:32,991 should also go and listen to. But I think it's important to actually 86 00:06:32,991 --> 00:06:42,525 maybe to spend a little bit of time explaining it in kind of words 87 00:06:42,525 --> 00:06:50,016 and languages that the broad audience can follow, if you don't 88 00:06:50,016 --> 00:06:50,697 mind. And by the way, I like when you use the word ecology because 89 00:06:08,426 --> 00:06:18,592 it reminded me that when I left high school, and I had biology 90 00:06:18,592 --> 00:06:30,322 as my main major, we wrote a paper on ecology and we have to stand 91 00:06:30,322 --> 00:06:38,924 in Copenhagen Zoo the whole winter observing a big black monkey. 92 00:06:38,924 --> 00:06:49,090 So, I was reminded of that. But I'm sure that's not where we're 93 00:06:49,090 --> 00:06:50,654 going today. 94 00:06:51,440 --> 00:09:00,884 No, but, you know, biology's a good start, Niels. In this quant 95 00:09:00,884 --> 00:11:31,902 world, we're about as far from biology as we can get. But I think 96 00:11:31,902 --> 00:13:41,346 we need to move closer to biology to see what we're talking 97 00:13:41,346 --> 00:14:02,920 about here. When I say a market is a complex adaptive system, 98 00:08:19,490 --> 00:10:05,150 I mean something simple. It's a system made up of many participants, 99 00:10:05,150 --> 00:11:24,395 each reacting to partial information, each adapting to what 100 00:11:24,395 --> 00:12:52,445 others are doing, and each changing the environment that everyone 101 00:12:52,445 --> 00:14:02,885 else must respond to. So that's the picture. No participant 102 00:07:09,800 --> 00:08:59,960 has the full view. Everyone is responding to a situation that everyone 103 00:08:59,960 --> 00:10:50,120 else is also changing. And the behavior of the whole system; the 104 00:10:50,120 --> 00:12:21,920 prices, the trends, the cascades, they emerge from all of 105 00:12:21,920 --> 00:14:02,900 these interactions, not from any one decision or any one participant. 106 00:06:51,440 --> 00:08:23,389 So, the other phrase I'm going to be using today is agent ecology. 107 00:08:23,389 --> 00:09:48,265 So that's just shorthand for the mix of participants in the market: 108 00:09:48,265 --> 00:10:51,922 value investors, passive index funds, momentum traders, trend followers, 109 00:10:51,922 --> 00:11:55,579 market makers, pension funds, volatility targeting funds (we love 110 00:11:55,579 --> 00:12:59,236 that word), options dealers, retail buyers, hedge funds, everyone 111 00:12:59,236 --> 00:14:02,893 whose actions in that market help shape overall price. So, the 112 00:07:14,762 --> 00:09:46,355 agent ecology is just the cast of participants who are at the table 113 00:09:46,355 --> 00:11:19,643 in the markets. Different ecologies produce different market 114 00:11:19,643 --> 00:13:39,575 behavior, and that's the whole reason it matters about who is at 115 00:13:39,575 --> 00:14:02,897 that table. So let me develop that a little more carefully. A market 116 00:07:54,723 --> 00:08:52,253 is not an object that exists independently of its participants. 117 00:08:52,253 --> 00:10:12,795 It is not a reflection of an underlying economy. It is not a representation 118 00:10:12,795 --> 00:11:16,078 of fundamental value, however that is measured. A market is constituted 119 00:11:16,078 --> 00:12:13,608 by the interactions of the population of agents, each holding 120 00:12:13,608 --> 00:13:16,891 a partial model of the situation, each acting on that model, 121 00:13:16,891 --> 00:14:02,915 and each affecting the conditions the others observe. So, 122 00:06:59,900 --> 00:08:32,960 the behavior of the market; the prices, the trends, the cascades, 123 00:08:32,960 --> 00:10:14,480 the periods of calm and the periods of violence in the market, 124 00:10:14,480 --> 00:11:22,160 they emerge from those interactions. Nothing about market 125 00:11:22,160 --> 00:13:03,680 behavior is given in advance. It's produced in real time by the 126 00:13:03,680 --> 00:14:02,900 agents in the system and their behavior. So, there is a corollary 127 00:07:17,745 --> 00:08:15,616 I want to make explicit, because it cuts against an assumption 128 00:08:15,616 --> 00:09:13,487 that almost everybody I speak to carries about this market. They 129 00:09:13,487 --> 00:10:11,358 tend to think the market sits downstream of other systems; the 130 00:10:11,358 --> 00:11:14,490 firm, the economy, the banking system, and that its job, the market's 131 00:11:14,490 --> 00:12:22,883 job, is to reflect what these other systems are doing. But I believe 132 00:12:22,883 --> 00:13:20,754 that assumption is incorrect, or at least incomplete. Each of those 133 00:13:20,754 --> 00:14:02,842 is actually a system in their own right. So, the assumption is 134 00:07:19,276 --> 00:08:28,866 that they correlate. So, these systems, they correlate with the 135 00:08:28,866 --> 00:09:52,374 market, they constrain it in some ways, but they are not what 136 00:09:52,374 --> 00:11:08,923 I call antecedent to it. They are loose autonomous systems that 137 00:11:08,923 --> 00:12:25,472 interact with the market. The market is not their direct reflection, 138 00:12:25,472 --> 00:13:35,062 it's its own thing: a loose semi-contained system of participatory 139 00:13:35,062 --> 00:14:02,898 agents using partial models. So, to say the market should reflect 140 00:08:03,351 --> 00:09:46,081 fundamental value is to assume a binding correlation between two 141 00:09:46,081 --> 00:11:18,538 autonomous systems that, on evidence, don't always agree. They 142 00:11:18,538 --> 00:13:01,268 run on their own dynamics. They influence each other without 143 00:13:01,268 --> 00:14:02,906 being directly reducible to each other. So, if that is right, 144 00:07:19,825 --> 00:08:33,626 the question you can therefore ask about what a market is, is not 145 00:08:33,626 --> 00:09:36,073 what its fair value is, it's not what its macroeconomic backdrop 146 00:09:36,073 --> 00:11:01,228 is, it is not what the central bank is going to do. The most important 147 00:11:01,228 --> 00:12:09,352 question you can ask is what kind of agents populate that system 148 00:12:09,352 --> 00:13:11,799 and what are the properties of the interactions between them? Because 149 00:13:11,799 --> 00:13:57,215 that's what actually determines what a market actually 150 00:13:57,215 --> 00:14:02,892 does. And the property of an agent that matters most, in a complex 151 00:10:36,560 --> 00:13:25,400 adaptive market, is whether the agent conditions its behavior 152 00:13:25,400 --> 00:14:02,920 on price. So, some agents do. For instance, a value investor sees 153 00:08:04,570 --> 00:09:25,013 a stretched valuation and they push back, a counter trend trader 154 00:09:25,013 --> 00:10:52,769 fades the move, a market maker fades the extreme orders, a pension 155 00:10:52,769 --> 00:12:20,525 fund rebalances out of an asset that has run too far. These 156 00:12:20,525 --> 00:13:26,342 are what I call price sensitive agents. Their behavior 157 00:13:26,342 --> 00:14:02,907 responds to where prices are. But other agents, they push in the 158 00:07:42,631 --> 00:08:55,761 other direction. A momentum algorithm buys what is rising, not 159 00:08:55,761 --> 00:10:08,891 what is cheap. A volatility targeting fund reduces exposure when 160 00:10:08,891 --> 00:11:14,708 volatility rises, not when prices fall. A stop-loss executes 161 00:11:14,708 --> 00:12:35,151 at a threshold, not a value. These agents are price sensitive 162 00:12:35,151 --> 00:13:40,968 too, but they're in the opposite direction. They amplify 163 00:13:40,968 --> 00:14:02,907 rather than absorb. But then there is this third category which 164 00:07:57,816 --> 00:09:37,380 has become the dominant force in markets today, and which I think 165 00:09:37,380 --> 00:11:25,241 is a structural story of the past 20 years that the macro conversation 166 00:11:25,241 --> 00:12:56,508 is not quite focused on properly, or at least in complex 167 00:12:56,508 --> 00:14:02,884 adaptive systems terms. They are the passive investors. So, a 168 00:07:01,358 --> 00:08:05,825 passive vehicle buys what the index tells it to buy in the proportion 169 00:08:05,825 --> 00:09:00,374 the index dictates at whatever price the market happens to offer. 170 00:09:00,374 --> 00:10:04,841 Price is not an input into its decision function. It does not buy 171 00:10:04,841 --> 00:11:04,349 more when something is cheap. It does not buy less when something 172 00:11:04,349 --> 00:12:03,857 is expensive. It does not push back when a valuation gets stretched. 173 00:12:03,857 --> 00:13:03,365 It does not step in when a valuation gets crushed. It simply 174 00:13:03,365 --> 00:13:52,955 executes whatever inflow it receives in proportion to the index's 175 00:13:52,955 --> 00:14:02,873 current composition. So passive is not price sensitive. And 176 00:07:50,660 --> 00:09:06,800 because it's not price sensitive, it does not participate 177 00:09:06,800 --> 00:10:39,860 in price discovery. So, the active work of figuring out what 178 00:10:39,860 --> 00:12:21,380 something is worth is price discovery. And that is not what passive 179 00:12:21,380 --> 00:13:45,980 is doing. Passive is doing something else entirely. It's allocating 180 00:13:45,980 --> 00:14:02,900 by rule. 181 00:14:03,880 --> 00:14:13,132 Okay, so just first of all, very interesting when you list all 182 00:14:13,132 --> 00:14:21,613 these different things and you can tell that it's all behavioral. 183 00:14:21,613 --> 00:14:30,094 We've always said that trend following is rooted in human behavior. 184 00:14:30,094 --> 00:14:40,888 This takes it to another level when you kind of break it down and 185 00:14:40,888 --> 00:14:50,140 explain like that. So, I really like that. But let me just 186 00:14:50,140 --> 00:14:53,224 ask you something here. When it comes to passive investors, either 187 00:14:09,277 --> 00:14:19,300 they might say (and I don't think we could argue that they are 188 00:14:19,300 --> 00:14:26,239 necessarily wrong) that the whole point of passive investing 189 00:14:26,239 --> 00:14:35,491 is precisely to avoid making bets on price. Right? So, it's the 190 00:14:35,491 --> 00:14:43,972 discipline, it's what works. Are you going to argue that owning 191 00:14:43,972 --> 00:14:53,224 index funds, to some degree, might be the wrong thing to do? 192 00:14:53,240 --> 00:17:22,948 Not really, Niels. So, I need to be clear about this, because the 193 00:17:22,948 --> 00:19:41,140 argument I'm making is not at the level of the individual investor. 194 00:19:41,140 --> 00:21:47,816 So, at the individual level, owning a low-cost index fund is, 195 00:21:47,816 --> 00:23:42,976 for most people, a perfectly rational decision. I'm not arguing 196 00:23:42,976 --> 00:24:06,008 against that. What I am making is a structural argument about what 197 00:16:50,850 --> 00:19:11,982 happens to the agent ecology when a growing share of capital becomes 198 00:19:11,982 --> 00:21:21,353 price insensitive by design. So, that is a separate question from 199 00:21:21,353 --> 00:22:55,441 whether any individual investor should hold index funds. 200 00:22:55,441 --> 00:24:06,007 Both can be true at once. So, index funds can be a sensible choice 201 00:16:29,368 --> 00:18:17,512 for the individual. And the aggregate consequence of millions 202 00:18:17,512 --> 00:20:17,672 of those sensible choices can have changed the structural behavior 203 00:20:17,672 --> 00:22:05,816 in markets in ways the individual choice never anticipated. 204 00:22:05,816 --> 00:24:05,976 So, that's the move I want to talk about today. So let me develop 205 00:15:51,424 --> 00:18:31,430 what actually follows. In a complex adaptive system, or CAS for 206 00:18:31,430 --> 00:21:11,436 short, I want to talk about passive not being price sensitive. 207 00:21:11,436 --> 00:24:05,988 So, because passive is not price sensitive, we get a few consequences. 208 00:14:53,240 --> 00:16:30,788 The first is the most obvious one, the population of price sensitive 209 00:16:30,788 --> 00:18:08,336 agents in the market, the ones who are buying and selling responds 210 00:18:08,336 --> 00:19:45,884 to where prices are, has therefore shrunk as a share of the 211 00:19:45,884 --> 00:21:07,174 total. Not in absolute terms, there's still, you know, trillions 212 00:21:07,174 --> 00:22:52,851 of dollars of active capital. But as a share on the table, the 213 00:22:52,851 --> 00:23:57,883 price sensitive participants have displaced by price insensitive 214 00:23:57,883 --> 00:24:06,012 ones. The second consequence follows from the first. And this 215 00:15:56,213 --> 00:17:27,174 is where the CAS lens does its real work. In any complex adaptive 216 00:17:27,174 --> 00:18:37,144 system, the structure of the behavior that emerges is determined 217 00:18:37,144 --> 00:19:54,111 by the interactions between the agents. And there are two kinds 218 00:19:54,111 --> 00:20:50,087 of feedback which run constantly through those interactions. 219 00:20:50,087 --> 00:21:53,060 There's what I call reinforcing feedback, and this is 220 00:21:53,060 --> 00:22:49,036 where the feedback amplifies movements. And there's balancing 221 00:22:49,036 --> 00:23:59,006 feedback, that's where the feedback absorbs and slows down price 222 00:23:59,006 --> 00:24:06,003 movement. So, the interplay between those two determines whether 223 00:15:49,216 --> 00:16:59,186 the system trends, ranges, or breaks. So reinforcing feedback comes 224 00:16:59,186 --> 00:18:02,159 mostly from agents whose behavior is conditioned on price 225 00:18:02,159 --> 00:19:12,129 in motion. Momentum buyers chase a rising move stop, losses 226 00:19:12,129 --> 00:20:15,102 triggering into a falling one, volatility targeting reducing exposure 227 00:20:15,102 --> 00:21:25,072 as volatility rises, index inclusion forcing buying as a stock 228 00:21:25,072 --> 00:22:35,042 joins a benchmark. Algorithmic flow, in general, follows this process. 229 00:22:35,042 --> 00:23:52,009 All of that pushes prices further in the direction that they're 230 00:23:52,009 --> 00:24:06,003 already going. The balancing feedback, however, comes mostly from 231 00:15:50,143 --> 00:17:03,304 agents whose behavior is conditioned on price relative to 232 00:17:03,304 --> 00:18:24,594 what they can perceive as value. Value Investors stepping in 233 00:18:24,594 --> 00:19:37,755 when something looks cheap. Sellers of overpriced assets, countertrend 234 00:19:37,755 --> 00:20:59,045 funds. The market maker who fades extremes. That balancing feedback 235 00:20:59,045 --> 00:22:12,206 is what stops reinforcing loops from running unchecked. It's 236 00:22:12,206 --> 00:23:25,367 what produces the equilibrium seeking behavior that classical finance 237 00:23:25,367 --> 00:24:06,012 assumes has always been present. So, what passive has done, 238 00:15:40,890 --> 00:17:44,780 by design, not by accident, is to reduce the share of the market 239 00:17:44,780 --> 00:19:20,080 doing the balancing work. The reinforcing agents are still there, 240 00:19:20,080 --> 00:20:55,380 the algorithms are still there. The forced flows around index 241 00:20:55,380 --> 00:22:30,680 inclusion and rebalancing have actually grown as passive has grown. 242 00:22:30,680 --> 00:24:05,980 But what has shrunk is the population that pushes back. So, 243 00:15:06,095 --> 00:17:53,210 let me put that in the language I use in my own practice. 244 00:17:53,210 --> 00:20:01,760 So, in systematic trading we talk about divergent and convergent 245 00:20:01,760 --> 00:21:57,455 traders. Divergent traders ride moves in the direction price 246 00:21:57,455 --> 00:23:40,295 is already going. Trend followers, momentum traders, they 247 00:23:40,295 --> 00:24:06,005 are divergent. So, convergent traders position for prices to come 248 00:17:25,728 --> 00:20:17,277 back towards fair value: value investors, mean reversion traders, 249 00:20:17,277 --> 00:23:27,887 counter trend funds. They're convergent. So, map that onto the 250 00:23:27,887 --> 00:24:06,009 feedback architecture. Reinforcing feedback comes from divergent 251 00:16:14,530 --> 00:18:24,594 traders and from passive long-only investors. Passive joins 252 00:18:24,594 --> 00:21:39,690 that divergent side because its flows buys more of what is already 253 00:21:39,690 --> 00:24:06,012 large. Its rule of allocation amplifies the existing distribution. 254 00:14:53,240 --> 00:15:47,312 Balancing feedback comes from the convergent population. And the 255 00:15:47,312 --> 00:16:53,400 convergent population is the part of the ecology that is actually 256 00:16:53,400 --> 00:17:53,480 being diluted by passive. The residue doing the balancing work 257 00:17:53,480 --> 00:18:47,552 is therefore shrinking. The reinforcing side, which is divergent 258 00:18:47,552 --> 00:19:59,648 plus passive, that has grown. And that is a consequence for the 259 00:19:59,648 --> 00:21:05,736 structure of market behavior itself. So, we see trends now run 260 00:21:05,736 --> 00:21:53,800 further before they meet resistance, because the resistance 261 00:21:53,800 --> 00:22:35,856 is structurally smaller. Reinforcing feedback operates in 262 00:22:35,856 --> 00:23:47,952 an ecology where the balancing feedback that used to slow it has 263 00:23:47,952 --> 00:24:05,976 now thinned out. The system is more prone to extended directional 264 00:16:09,480 --> 00:17:35,250 moves, more prone to overshoot, more prone to cascade, 265 00:17:35,250 --> 00:19:10,550 not because the reinforcing forces have grown stronger, but because 266 00:19:10,550 --> 00:20:55,380 the balancing forces have grown weaker. So, there is a third 267 00:20:55,380 --> 00:22:40,210 consequence which I'm going to plant here but develop properly later 268 00:22:40,210 --> 00:24:05,980 in this episode when we get into the topics. That's what I call 269 00:18:14,248 --> 00:24:06,012 the float that does respond to price. So, the active price sensitive 270 00:15:15,175 --> 00:16:07,819 capital, that is still doing the work, is operating in a structurally 271 00:16:07,819 --> 00:16:51,689 thinner pool because the price sensitive buyers have now dominated. 272 00:16:51,689 --> 00:17:39,946 So, most of the outstanding shares sit in entities that will 273 00:17:39,946 --> 00:18:36,977 not sell into a rising price, they will not buy into a falling 274 00:18:36,977 --> 00:19:29,621 price, they will not adjust on valuation. So, when the price sensitive 275 00:19:29,621 --> 00:20:22,265 capital does move, its impact on price is larger than it would 276 00:20:22,265 --> 00:21:10,522 have been 20 years ago. (There's a bit of a counterintuitive 277 00:21:10,522 --> 00:21:58,779 statement here, which I need to explain later.) So, the market 278 00:21:58,779 --> 00:22:47,036 is structurally inelastic in a way that the old framework didn't 279 00:22:47,036 --> 00:23:30,906 anticipate. And the size of that inelasticity is more profound 280 00:23:30,906 --> 00:24:06,002 than almost anyone realizes, which we'll get to. So, that's what's 281 00:15:28,523 --> 00:17:37,894 on my radar. One of the biggest structural changes in the 282 00:17:37,894 --> 00:19:47,265 architecture of global markets over the past two decades has not 283 00:19:47,265 --> 00:21:33,114 been algorithmic trading, has not been high frequency trading, 284 00:21:33,114 --> 00:23:54,246 has not been the dominance or the mega caps. They're all downstream 285 00:23:54,246 --> 00:24:06,007 effects. The structural change is the transfer of a growing share 286 00:17:11,430 --> 00:19:29,620 of market ownership from price sensitive agents to price insensitive 287 00:19:29,620 --> 00:21:47,810 ones. And the consequence is that the feedback architecture of 288 00:21:47,810 --> 00:24:06,000 markets, the part of the system that produces trends, cascades. 289 00:14:53,240 --> 00:18:04,580 The fat-tailed behavior, that any honest practitioner has observed 290 00:18:04,580 --> 00:21:37,180 for decades, is operating in a structurally different ecology from 291 00:21:37,180 --> 00:24:06,000 the one market theory was built for. So let me put it one more 292 00:15:49,317 --> 00:17:25,449 way, because I want to make sure the listener can't mishear it. 293 00:17:25,449 --> 00:19:09,592 The issue is not whether passive is good or bad for the person 294 00:19:09,592 --> 00:20:37,713 buying the index fund. The issue is what happens when passive 295 00:20:37,713 --> 00:21:57,823 becomes so large that fewer participants are left asking whether 296 00:21:57,823 --> 00:23:33,955 the price itself makes sense. Both of those things can be true 297 00:23:33,955 --> 00:24:05,999 at the same time. Index funds can be a perfectly reasonable choice 298 00:16:33,744 --> 00:18:26,811 for any individual and the aggregate consequence of millions 299 00:18:26,811 --> 00:20:19,878 of those reasonable choices can change the structural behavior 300 00:20:19,878 --> 00:22:38,071 of the system. The individual level and the structural level are 301 00:22:38,071 --> 00:24:06,012 different questions and they've got different answers. So, 302 00:15:03,111 --> 00:17:11,434 the question is not whether passive is good or bad for the individual 303 00:17:11,434 --> 00:19:00,015 investor. At that level it might be entirely sensible. The structural 304 00:19:00,015 --> 00:20:48,596 question is different. What happens when a growing share of capital 305 00:20:48,596 --> 00:22:37,177 no longer asks whether price makes sense? What happens when fewer 306 00:22:37,177 --> 00:24:06,016 participants are left to push back against that movement? That 307 00:15:03,476 --> 00:16:56,072 is a question that leads directly into the topics that we'll 308 00:16:56,072 --> 00:18:48,668 be discussing later today. Why markets trend, why they have always 309 00:18:48,668 --> 00:20:41,264 trended, and why the modern agent ecology may make those trends 310 00:20:41,264 --> 00:22:13,388 actually structurally present going forward, more so rather than 311 00:22:13,388 --> 00:24:05,984 less so. So that's where I'll leave it over to you. 312 00:24:07,891 --> 00:24:33,655 I mean, your radar must have been over flooded. That was probably 313 00:24:33,655 --> 00:24:55,125 the most comprehensive answer to that question, which is great. 314 00:24:55,125 --> 00:25:18,742 It's an important one and it's a great framing, of course. There's 315 00:24:08,848 --> 00:24:21,289 one thing I was thinking of when you were kind of talking about 316 00:24:21,289 --> 00:24:31,816 convergent, divergent, and the changes in the structure. And I don't 317 00:24:31,816 --> 00:24:45,214 know if you've thought about this and that is we do hear a lot 318 00:24:45,214 --> 00:25:00,526 about it, and I know Cem talks a lot about it as well, a lot about 319 00:25:00,526 --> 00:25:08,182 kind of structured products, issuance of structured products, 320 00:25:08,182 --> 00:25:18,709 and how they suppress volatility, and so on, and so forth. 321 00:24:07,891 --> 00:24:18,909 And I wonder, to some extent when, if they have the kind of build 322 00:24:18,909 --> 00:24:29,140 up, a little bit, of the role that other sort of convergence strategies 323 00:24:29,140 --> 00:24:38,584 may have filled. I agree with you that passive is winning, and 324 00:24:38,584 --> 00:24:48,028 is taking over, and that is changing the structure. But I wonder 325 00:24:48,028 --> 00:24:58,259 if the blow is being softened a little bit with this enormous amount 326 00:24:58,259 --> 00:25:07,703 of issuance of structured products, where a lot of it is trying 327 00:25:07,703 --> 00:25:16,360 to keep certain price spikes from happening. Don't know if you 328 00:25:16,360 --> 00:25:18,721 thought about that. 329 00:25:19,471 --> 00:25:49,732 Look, I suppose we shouldn't perhaps wish for too much because 330 00:25:49,732 --> 00:26:22,744 what I'm actually going to be saying, later in these topics, is 331 00:26:22,744 --> 00:26:55,756 this rise in passive is great for trend, so it's fantastic for 332 00:26:55,756 --> 00:27:20,515 us. So, this rise of structured products, however, might 333 00:27:20,515 --> 00:27:23,266 not… So, the more different ecologies you have, the broader the 334 00:25:32,921 --> 00:25:46,371 ecology, the more the different forces at play create structural 335 00:25:46,371 --> 00:25:57,131 diversity, that creates a foundation of strength. However, 336 00:25:57,131 --> 00:26:13,271 the less you get, when things start dominating, you get a concentration 337 00:26:13,271 --> 00:26:30,756 effect. And a good example, let's say a calm market. A calm market 338 00:26:30,756 --> 00:26:46,896 starts off where volatility is very low, and we start seeing strategies 339 00:26:46,896 --> 00:26:59,001 adapting to that calm market. They start deliberately operating 340 00:26:59,001 --> 00:27:13,796 within that calm regime. That's the only way that they can 341 00:27:13,796 --> 00:27:23,211 make their edge from that calm regime. And what you typically find, 342 00:25:28,066 --> 00:25:46,975 because it's a calm regime, they start leveraging up or adopting 343 00:25:46,975 --> 00:26:05,884 structured products that allow them to leverage up in their calm 344 00:26:05,884 --> 00:26:24,793 market. So, if you could imagine the market is actually building 345 00:26:24,793 --> 00:26:40,264 up risk in that calm environment because it's getting 346 00:26:40,264 --> 00:27:00,892 a concentration of strategy in it. We're getting a loss of diversity. 347 00:27:00,892 --> 00:27:21,520 The loss of diversity made the entire market fabric much more structurally 348 00:27:21,520 --> 00:27:23,239 robust. Now we've got this calm market where this concentration 349 00:25:29,155 --> 00:25:40,991 occurs, and this buildup of risk occurs. And because of this 350 00:25:40,991 --> 00:25:49,599 leverage that's introduced into this calm environment, because 351 00:25:49,599 --> 00:26:01,435 investors have got to make the return somehow, and there's only 352 00:26:01,435 --> 00:26:12,195 small volatility in this calm market, they naturally leverage up. 353 00:26:12,195 --> 00:26:24,031 But that makes the system more sensitive, more sensitive, more sensitive. 354 00:26:24,031 --> 00:26:34,791 It's actually making the system more fragile. So, when you 355 00:26:34,791 --> 00:26:47,703 get a small deviation out of that, that structural calm, a small 356 00:26:47,703 --> 00:26:56,311 deviation, you suddenly get this nonlinear amplification and 357 00:26:56,311 --> 00:27:08,147 this massive breakout that occurs out of that because the risk 358 00:27:08,147 --> 00:27:19,983 was building up, it wasn't dormant, it was there. That's how 359 00:27:19,983 --> 00:27:23,211 I'm seeing it. 360 00:27:23,850 --> 00:27:29,296 Yeah, and to some extent, the way I… And again, I have no evidence 361 00:27:29,296 --> 00:27:33,575 of this, it's just an observation, is that there was certainly 362 00:27:33,575 --> 00:27:38,632 a period of time, maybe a few years, where perhaps these pod shops, 363 00:27:38,632 --> 00:27:42,133 multi strat funds were dominating and they were growing 364 00:27:42,133 --> 00:27:45,634 like crazy and they were essentially probably often trading 365 00:27:45,634 --> 00:27:49,135 against the trend followers and keeping markets from breaking 366 00:27:49,135 --> 00:27:54,192 out. But then, as you say, when the markets then do break out, 367 00:27:54,192 --> 00:27:58,860 they find themselves on the wrong side of that trade. And then 368 00:27:58,860 --> 00:28:03,139 that actually amplifies the outcome in the end, even though you 369 00:28:03,139 --> 00:28:08,585 have to be a little bit more patient. So yeah, I think we're saying 370 00:28:08,585 --> 00:28:09,752 the same thing. 371 00:28:09,791 --> 00:28:18,020 If we think of, say, passive, this rise of passive, what it's leading 372 00:28:18,020 --> 00:28:23,717 to is theoretically an extinction level event of passive. 373 00:28:23,717 --> 00:28:28,781 Everyone's going passive, a concentration into passive. What 374 00:28:28,781 --> 00:28:35,744 happens when suddenly they're met with this explosion out of a 375 00:28:35,744 --> 00:28:42,074 market regime or a behavior that satisfies? You suddenly get 376 00:28:42,074 --> 00:28:49,037 this massive uncoiling of this system that has been building up 377 00:28:49,037 --> 00:28:55,367 risk. And this is what I call extinction level events. So, the 378 00:28:12,731 --> 00:28:28,901 way these complex adaptive markets work, you know, you get these 379 00:28:28,901 --> 00:28:39,191 growth, growth, growth, growth, extinction; another one, 380 00:28:39,191 --> 00:28:49,481 growth, growth, growth, extinction. It’s continually adapting, 381 00:28:49,481 --> 00:28:55,361 going through these different… 382 00:28:55,531 --> 00:28:58,816 And actually, what works against passive, to some degree, 383 00:28:58,816 --> 00:29:03,196 is demography. And at some point there will be a change. The 384 00:29:03,196 --> 00:29:07,211 challenge with some of these concepts that we talk about often 385 00:29:07,211 --> 00:29:11,956 is that you and I may only experience one side of that, and 386 00:29:11,956 --> 00:29:16,336 then the other side of that will be for the next generation. 387 00:29:16,336 --> 00:29:20,351 And so, it's kind of frustrating that you only see the 388 00:29:20,351 --> 00:29:23,271 headwind, not the tailwind. But there we are. 389 00:29:24,231 --> 00:29:25,101 Our kids will get it. 390 00:29:25,811 --> 00:35:38,159 Exactly, as long as our kids stay trend followers, we'll be fine. 391 00:29:25,811 --> 00:32:18,685 Anyways. All right. So, I mean, as I said, wow. What a radar. 392 00:32:18,685 --> 00:34:58,261 And I feel my observations on my radar are absolutely not very 393 00:34:58,261 --> 00:35:38,155 important, having heard. Nevertheless, I couldn't help noticing 394 00:30:17,526 --> 00:32:42,328 a story that was reported. I didn't see it on CNBC, but it was 395 00:32:42,328 --> 00:34:46,444 reported with the source of being CNBC, which, I guess, we still 396 00:34:46,444 --> 00:35:38,159 think of as being credible. Anyways, it seems like, and I don't 397 00:29:57,983 --> 00:31:06,923 mean to make fun of this, but it is just so unusual that I couldn't 398 00:31:06,923 --> 00:32:06,671 help mentioning it. And that is that there is now a new mandate 399 00:32:06,671 --> 00:32:57,227 if you want to be a Russian central banker. Because according 400 00:32:57,227 --> 00:33:52,379 to CNBC, it says, Russia has passed a law authorizing its central 401 00:33:52,379 --> 00:34:33,743 bank and other financial institutions to repel drone attacks 402 00:34:33,743 --> 00:35:24,299 with their own defense systems as the country struggles to defend 403 00:35:24,299 --> 00:35:38,087 against Ukraine strikes. The law passed by Russia's lower house 404 00:29:51,361 --> 00:30:31,511 of parliament on Tuesday will allow staff at Russia's central bank 405 00:30:31,511 --> 00:31:15,311 to be armed and to operate systems used to down unmanned aerial 406 00:31:15,311 --> 00:31:48,161 vehicles, UAVs, or drone attacks without the involvement of 407 00:31:48,161 --> 00:32:35,611 special forces. And it goes on. I'm not going to read it anymore. 408 00:32:35,611 --> 00:33:23,061 But I mean, the point I'm trying to make here is just what 409 00:33:23,061 --> 00:34:14,161 a crazy world we live in. I mean, you know, it's not good for 410 00:34:14,161 --> 00:34:50,661 anyone and it's just completely crazy. So, I'm going to 411 00:34:50,661 --> 00:35:38,111 leave that. I'm not going to even ask for a comment on it. However, 412 00:29:29,357 --> 00:30:11,909 on a more related note, I noticed in the latest Spring 2026 413 00:30:11,909 --> 00:30:43,823 Investor Sentiment Survey, from our friends over at SocGen, 414 00:30:43,823 --> 00:31:22,829 that CTAs, and specifically CTA trend, has moved up two places 415 00:31:22,829 --> 00:32:05,381 from the autumn of 2025, although not quite back to the spring 416 00:32:05,381 --> 00:32:58,571 of ‘25, not quite back to the fall or spring of ‘24 in terms of 417 00:32:58,571 --> 00:33:41,123 popularity, but a little bit better than last year where I guess 418 00:33:41,123 --> 00:34:23,675 people were a little bit stung by the drawdowns by mid-year. So, 419 00:34:23,675 --> 00:35:02,681 they probably asked people just after June of last year, what 420 00:35:02,681 --> 00:35:38,141 do you think of trend? And the response wasn't great. But anyways, 421 00:29:38,875 --> 00:30:50,727 oddly enough, I think that this survey shows how many investors, 422 00:30:50,727 --> 00:32:15,643 as a percent, are interested in a strategy and the winner is not 423 00:32:15,643 --> 00:33:07,899 trend, unfortunately. But with 59% respondents, it's discretionary 424 00:33:07,899 --> 00:34:39,347 global macro. I have to say it must be difficult to be a discretionary 425 00:34:39,347 --> 00:35:38,135 trader in this environment. So, good luck to them. Number two, 426 00:29:32,285 --> 00:29:58,181 maybe not surprising, commodities. Okay, that makes sense, 427 00:29:58,181 --> 00:30:33,788 43%. 42% equity market neutral stat arb. Okay, there's a convergent 428 00:30:33,788 --> 00:31:12,632 guy for you. The same with the next one. Equity long/short. Then 429 00:31:12,632 --> 00:31:48,239 comes quantitative global macro. Okay, I guess we can, we understand 430 00:31:48,239 --> 00:32:17,372 that, 37%. Multi strategy fundamental, not sure exactly what 431 00:32:17,372 --> 00:32:56,216 that is but that's up one place, but with 31% interested. Then 432 00:32:56,216 --> 00:33:38,297 comes event driven, and then I don't even know what place it is, 433 00:33:38,297 --> 00:34:10,667 probably around number 8, anyways, CTA trend with 29% interested, 434 00:34:10,667 --> 00:34:49,511 and 28% interested in CTA non-trend. So at least we beat the 435 00:34:49,511 --> 00:35:28,355 non-trend guys by a place and by a percent. Anyways, that's what 436 00:35:28,355 --> 00:35:38,066 it looks like. And further down than the last couple of surveys, 437 00:30:31,511 --> 00:31:59,111 actually, the lowest of the last five surveys, in terms of ranking, 438 00:31:59,111 --> 00:33:12,111 is actually (as we just talked about) multi strategy systematic. 439 00:33:12,111 --> 00:34:54,311 So, I guess that means the pod shops are not as popular as they 440 00:34:54,311 --> 00:35:38,111 were only about a year ago. Anyways we've got a long program 441 00:31:02,945 --> 00:33:28,646 today with this speed definitely going turtle speed today, 442 00:33:28,646 --> 00:35:38,158 pardon the pun. The pun was probably intended. Anyways, a trend 443 00:29:38,084 --> 00:30:27,176 barometer update, it doesn't look good coming to the end of the 444 00:30:27,176 --> 00:31:20,359 month. It's down to 39% which it hasn't touched for quite a while. 445 00:31:20,359 --> 00:32:17,633 So, we are seeing a little bit of weakness in the trend space. It 446 00:32:17,633 --> 00:32:50,361 doesn't mean performance, necessarily, will reflect this. This 447 00:32:50,361 --> 00:33:39,453 is the shorter-term time frames but it is coming down a little 448 00:33:39,453 --> 00:34:24,454 bit. And, I guess, looking at the markets this month, mid-month 449 00:34:24,454 --> 00:35:13,546 everything looked okay, but in the last week, 10 days, we've seen 450 00:35:13,546 --> 00:35:38,092 a lot of it, isn't it? Yeah, I mean I see the oil markets the last 451 00:30:14,801 --> 00:31:18,488 week is down more than 5%. Anyway, we've given up some of the 452 00:31:18,488 --> 00:32:17,276 performance but I think the month, when I look at the numbers, 453 00:32:17,276 --> 00:33:16,064 which I will quote now even though they do not include yesterday 454 00:33:16,064 --> 00:34:19,751 which was a down day I think for most managers. Yeah, just slightly 455 00:34:19,751 --> 00:35:28,337 up. So, I'd love to hear your thoughts but let me just run through 456 00:35:28,337 --> 00:35:38,135 the numbers. BTOP 50 up 79 basis points as of Tuesday night, 457 00:30:36,061 --> 00:32:14,411 up 10.43 for the year. SocGen CTA index up 72 basis points in May, 458 00:32:14,411 --> 00:33:52,761 up 11% for the year. SocGen Trend up 1.07, up 11.31% as far as 459 00:33:52,761 --> 00:35:17,061 I remember. SocGen Short Term Traders Index up 0.25%, up 5.5% so 460 00:35:17,061 --> 00:35:38,136 far this year. MSCI World is, however, having a good month still 461 00:30:19,001 --> 00:31:41,741 up 3.8%, up almost 10% for the year. The S&P US Aggregate Bond Index 462 00:31:41,741 --> 00:32:58,571 completely flat for the month, and pretty flat for the year up 28 463 00:32:58,571 --> 00:34:21,311 basis points. And the S&P 500 up 4.43% in May, up now 10.38% so 464 00:34:21,311 --> 00:35:38,141 far this year. It’s probably at a new all-time high, I would imagine. 465 00:29:25,811 --> 00:33:48,635 Any thoughts, any takeaways, from your side, in terms of the trend 466 00:33:48,635 --> 00:35:38,145 environment or what you've seen? 467 00:35:38,161 --> 00:35:42,391 Your TTU barometer I think, I'm actually finding, because I'm 468 00:35:42,391 --> 00:35:47,467 reporting on it with you each week or whatever, I'm finding it's 469 00:35:47,467 --> 00:35:52,543 a very good leading indicator. I'm finding that you know the moves 470 00:35:52,543 --> 00:35:56,773 we're experiencing in our fund, your indicator tends to sort 471 00:35:56,773 --> 00:36:01,426 of almost be the lead indicator for that. So, you know, 472 00:36:01,426 --> 00:36:03,118 we're seeing that now. 473 00:36:05,241 --> 00:36:05,481 The lead indicator… 474 00:36:05,481 --> 00:36:09,716 It’s faster, it's shorter term, it picks up these things I 475 00:36:09,716 --> 00:36:14,336 think a bit quicker than our lagging funds do but I'm finding 476 00:36:14,336 --> 00:36:16,261 it's a good directional indicator. 477 00:36:16,281 --> 00:36:19,531 Yeah, I was going to say, maybe I could license it to Andrew, 478 00:36:19,531 --> 00:36:20,781 he can replicate the index. 479 00:36:21,341 --> 00:36:23,980 Oh yeah, that's not a bad idea. 480 00:36:23,981 --> 00:36:26,620 I'll ask him in a couple of weeks when I talk to him. 481 00:36:26,621 --> 00:36:27,581 A few errors in it. 482 00:36:28,061 --> 00:38:50,051 Exactly. Anyways, actually my observations over the many, many 483 00:36:37,933 --> 00:36:55,209 years, I think it came about to life in 2007, so it's almost 20 484 00:36:55,209 --> 00:37:12,485 years old now and we have not changed the formula here but I will 485 00:37:12,485 --> 00:37:28,527 say it's not very good as a trading tool, in my opinion. But 486 00:37:28,527 --> 00:37:47,037 I think what it does do well at month end, if you just look at 487 00:37:47,037 --> 00:38:01,845 it at month end, it actually has an extraordinary way of giving 488 00:38:01,845 --> 00:38:16,653 you an indication of what was the month like, partly because of 489 00:38:16,653 --> 00:38:30,227 its lookback period. So, that's really what I think it's useful 490 00:38:30,227 --> 00:38:48,737 for is just to, say, yeah, was it a good environment, was it a bad 491 00:38:48,737 --> 00:38:49,971 environment? And the reason it came about, in 2007, was there was 492 00:36:48,884 --> 00:37:13,493 a period of time where trend followers didn't do so well and people 493 00:37:13,493 --> 00:37:34,316 couldn't really explain it. That's when it was born. And even 494 00:37:34,316 --> 00:37:55,139 some of our very large competitors, who have all been guests 495 00:37:55,139 --> 00:38:19,748 on Top Traders Unplugged, came and asked if they could use it in 496 00:38:19,748 --> 00:38:38,678 their marketing material, which we of course back then were 497 00:38:38,678 --> 00:38:50,036 very honored to have that request. Anyways, enough about that. 498 00:36:36,289 --> 00:37:07,144 We do have a question. We have a question all the way, if I'm not 499 00:37:07,144 --> 00:37:29,771 mistaken, from Costa Rica, from Jose, whom I get many emails 500 00:37:29,771 --> 00:37:56,512 from in support of what we do. So, I really appreciate that. And 501 00:37:56,512 --> 00:38:19,139 Jose is asking you, Rich, he's asking whether systematic trend following 502 00:38:19,139 --> 00:38:41,766 pays too much attention to correlation and not enough to variance 503 00:38:41,766 --> 00:38:49,994 at the instrument level. And the point he's making is that correlation 504 00:36:47,645 --> 00:37:12,125 tells you how stocks spread between markets, but variance might 505 00:37:12,125 --> 00:37:41,501 be an earlier and more direct signal of where stress is becoming 506 00:37:41,501 --> 00:38:05,981 acute within a single instrument. A position can look diversified 507 00:38:05,981 --> 00:38:28,013 on correlation matrix but still become a disproportionate source 508 00:38:28,013 --> 00:38:50,045 of portfolio stress when its own volatility rises sharply. So 509 00:36:37,527 --> 00:38:02,721 that's his concern essentially. So, I'm sure you have 510 00:38:02,721 --> 00:38:50,051 a professorial answer to share. 511 00:38:50,771 --> 00:39:21,113 Okay, so it is a sharp question from Jose. So, he is pointing 512 00:39:21,113 --> 00:39:44,453 at a real distinction. So, correlation is a between instrument 513 00:39:44,453 --> 00:40:07,793 property, it tells me how shocks travel across the portfolio. 514 00:40:07,793 --> 00:40:28,799 But variances are within instrument property, it tells me 515 00:40:28,799 --> 00:40:49,805 where shock is becoming locally intense. Both matter. They 516 00:40:49,805 --> 00:40:59,141 both matter to me. But this is where I'd slightly push back on the 517 00:39:03,871 --> 00:39:18,281 assumptions in the question. The premise of the question is that 518 00:39:18,281 --> 00:39:27,451 trend following overweight correlation across instruments and 519 00:39:27,451 --> 00:39:39,241 underweights variance within instruments. But it often looks that 520 00:39:39,241 --> 00:39:51,031 way in the commentary. We're always talking about correlations, 521 00:39:51,031 --> 00:40:02,821 etc., across instruments because, you know, that's what we 522 00:40:02,821 --> 00:40:15,921 talk about. But the engine itself, the trend following engine, 523 00:40:15,921 --> 00:40:29,021 is absolute momentum style that runs on variance. So, every 524 00:40:29,021 --> 00:40:43,431 position, for example, in my program is sized by average true 525 00:40:43,431 --> 00:40:59,151 range (ATR), which is a direct measure of the instrument's own variance. 526 00:38:50,771 --> 00:39:12,166 So, when a market's variance rises, my position size shrinks automatically, 527 00:39:12,166 --> 00:39:31,616 no discretion, no judgment call. So, the program is already 528 00:39:31,616 --> 00:39:56,901 taking less of it before the next bar prints. So, the same applies 529 00:39:56,901 --> 00:40:16,351 to stops, to portfolio level volatility targeting when it's applied 530 00:40:16,351 --> 00:40:39,691 properly (Niels, you know how I feel about whether this is done 531 00:40:39,691 --> 00:40:59,141 well), and to draw down management at the position level. 532 00:38:50,771 --> 00:39:14,311 This is all what I call variance driven. So, the framing 533 00:39:14,311 --> 00:39:37,851 I'd offer to Jose is this. Correlation is the diagnostic, it 534 00:39:37,851 --> 00:40:03,531 tells me what the system is doing. Variance is the control, it 535 00:40:03,531 --> 00:40:35,631 tells me how to size, when to cut, where to back off. A trend follower 536 00:40:35,631 --> 00:40:59,171 needs both of these things at different levels of the architecture. 537 00:38:50,771 --> 00:39:38,915 The system, I feel, tells you the weather, and variance tells you 538 00:39:38,915 --> 00:40:23,047 whether to put on another coat right now, effectively. So, thanks 539 00:40:23,047 --> 00:40:59,155 for the question, Jose. It was a good one. 540 00:41:05,311 --> 00:41:13,397 Now we are already 40 minutes in and we've only served the appetizer, 541 00:41:13,397 --> 00:41:21,483 Rich. How about that? So, I know you are ready now for delivering 542 00:41:21,483 --> 00:41:30,191 the main course, I feel. But it is, as you say, it's a continuation 543 00:41:30,191 --> 00:41:38,277 of the framing you just said. And I actually think, I mean, these 544 00:41:38,277 --> 00:41:45,741 are really important topics and, as you pointed out to me before 545 00:41:45,741 --> 00:41:51,961 we press record, you've noticed that more people talk about 546 00:41:51,961 --> 00:41:58,803 complex adaptive systems and I would not disagree with that. So, 547 00:41:58,803 --> 00:42:04,401 let's dive in, let's see where it takes us. I'm excited about it 548 00:41:12,267 --> 00:41:33,135 and I think this will be incredibly useful for those who just 549 00:41:33,135 --> 00:41:55,742 sit down, relax, take a few notes because there's a lot of information 550 00:41:55,742 --> 00:42:04,437 in that three word concept. 551 00:42:05,331 --> 00:42:41,997 All right, so here we go. I'm going to start with an image. So, 552 00:42:41,997 --> 00:43:13,425 I want you to picture a murmuration of starlings over a winter 553 00:43:13,425 --> 00:43:42,234 field, late afternoon. The light is dimming and there are thousands 554 00:43:42,234 --> 00:44:18,900 of birds in the sky moving as one. Not a single bird is leading. 555 00:44:18,900 --> 00:44:45,090 There is no choreographer, there's no plan. There's no bird 556 00:44:45,090 --> 00:45:13,899 that has the full picture of where the flock is going. The flock 557 00:42:14,311 --> 00:42:50,231 turns, expands, contracts, ripples, splits, recombines and produces 558 00:42:50,231 --> 00:43:30,641 this coherent moving shape of extraordinary complexity. And it 559 00:43:30,641 --> 00:44:20,031 does all of that with no central control whatsoever. So how 560 00:44:20,031 --> 00:45:13,911 does it work? Each bird is following three simple rules - heuristics. 561 00:42:05,331 --> 00:42:31,588 You might be familiar with the trend followers heuristics: cut losses 562 00:42:31,588 --> 00:42:53,071 short, let profits run. However, in the starling’s world, 563 00:42:53,071 --> 00:43:16,941 stay close to your nearest neighbors, match their speed, avoid 564 00:43:16,941 --> 00:43:38,424 collision. That's their simple local heuristics. That's it. Three 565 00:43:38,424 --> 00:44:07,068 local rules applied by each bird to the few birds nearest them. 566 00:44:07,068 --> 00:44:35,712 Nothing in the rules specifies the shape of the flock. Nothing in 567 00:44:35,712 --> 00:44:59,582 the rules specifies the turn. Nothing in individual birds behavior 568 00:44:59,582 --> 00:45:13,904 predicts the structure of the whole. But from those three local 569 00:42:18,071 --> 00:42:43,551 rules, applied in parallel by thousands of birds, the collective 570 00:42:43,551 --> 00:43:06,483 produces a coherent structure that flows, turns, reorganizes itself 571 00:43:06,483 --> 00:43:29,415 in real time. So that structure is emergent structure. 572 00:43:29,415 --> 00:43:57,443 It does not exist in any individual bird. It's not built, 573 00:43:57,443 --> 00:44:28,019 it is not planned. It's not the sum of the birds intentions. 574 00:44:28,019 --> 00:44:50,951 It emerges from the interactions between the birds. The 575 00:44:50,951 --> 00:45:13,883 structure lives in the interactions, not in the components. 576 00:42:05,331 --> 00:42:45,329 So that's a picture I want to place to the listener to hold on 577 00:42:45,329 --> 00:43:16,756 to for the rest of this episode because markets work the 578 00:43:16,756 --> 00:43:42,469 same way. So, every participant in the market, every 579 00:43:42,469 --> 00:44:05,325 pension fund rebalancing its allocation, every momentum algorithm 580 00:44:05,325 --> 00:44:33,895 adjusting its position, every retail investor responding to a headline, 581 00:44:33,895 --> 00:44:59,608 every options dealer hedging an exposure, they're following local 582 00:44:59,608 --> 00:45:13,893 rules based on local information. No participant has the 583 00:42:14,311 --> 00:42:34,516 full picture. No participant intends the collective outcome. But 584 00:42:34,516 --> 00:42:54,721 the interactions between them produces structure: trends that persist 585 00:42:54,721 --> 00:43:12,681 far beyond any single catalyst, crashes that accelerate 586 00:43:12,681 --> 00:43:37,376 beyond any single piece of bad news, volatility that clusters in 587 00:43:37,376 --> 00:43:59,826 ways no random process ever could, concentration in a handful 588 00:43:59,826 --> 00:44:20,031 of names that no rational allocation framework can explain. 589 00:44:20,031 --> 00:44:42,481 That structure is all emergent. It doesn't exist in any 590 00:44:42,481 --> 00:45:02,686 single participant. It emerges from the interactions between them. 591 00:45:02,686 --> 00:45:13,911 The market is the flock. 592 00:45:15,041 --> 00:45:19,177 Yes. Okay. I mean, it's interesting when you paint that picture 593 00:45:19,177 --> 00:45:23,689 because I remember a paper that had that picture on it, written 594 00:45:23,689 --> 00:45:28,201 by one of our good Dutch friends in this industry. So, markets 595 00:45:28,201 --> 00:45:33,089 are really just a murmuration or a flock? Is that how it is? 596 00:45:33,601 --> 00:46:09,625 Exactly. So, here's what I want the listener to hold from that 597 00:46:09,625 --> 00:46:36,643 image. Because it changes everything about how they should 598 00:46:36,643 --> 00:47:12,667 be thinking about what markets actually are. So, most of the last 599 00:47:12,667 --> 00:47:48,691 50 years, finance has taught markets as if they were a different 600 00:47:48,691 --> 00:48:00,699 kind of object entirely. The dominant intellectual framework, 601 00:45:42,385 --> 00:46:08,737 the one taught in every sort of MBA program, embedded in every 602 00:46:08,737 --> 00:46:24,109 institutional allocation model, written into every textbook, 603 00:46:24,109 --> 00:46:50,461 they've treated markets as if they were governed by the same kinds 604 00:46:50,461 --> 00:47:12,421 of laws as physical systems: particles in motion, forces pushing 605 00:47:12,421 --> 00:47:29,989 prices towards a stable equilibrium, deviations from fair 606 00:47:29,989 --> 00:47:51,949 value being treated as noise around that equilibrium, random short-lived 607 00:47:51,949 --> 00:48:00,733 and unprofitable to exploit. That's the equilibrium model. It's 608 00:46:01,896 --> 00:46:52,827 elegant, it's mathematically tractable. And as a description of 609 00:46:52,827 --> 00:47:43,758 how real markets actually behave, the wrong picture entirely. 610 00:47:43,758 --> 00:48:00,735 So, here's why. So, the equilibrium model treats markets 611 00:45:44,917 --> 00:46:09,435 as though they are clocks, not flocks. A clock is governed by physical 612 00:46:09,435 --> 00:46:30,181 laws. The parts don't observe each other. The pendulum does not 613 00:46:30,181 --> 00:46:52,813 change its behavior based on what the gears are doing. The springs 614 00:46:52,813 --> 00:47:19,217 do not panic when the hands move too fast. A clock is a kind 615 00:47:19,217 --> 00:47:41,849 of system where you can describe the whole by adding up the 616 00:47:41,849 --> 00:48:00,709 parts. But a market is not that kind of system. Market participants 617 00:45:38,059 --> 00:46:00,349 observe each other constantly. They imitate, they panic when others 618 00:46:00,349 --> 00:46:18,181 panic, they anchor to reference points other participants 619 00:46:18,181 --> 00:46:40,471 have just set, they build models of what other participants 620 00:46:40,471 --> 00:47:04,990 will do and act on those models, which changes what other 621 00:47:04,990 --> 00:47:31,738 participants do, which changes the models. So, the agents in a market, 622 00:47:31,738 --> 00:47:51,799 they're not particles, they are responsive. And their responsiveness 623 00:47:51,799 --> 00:48:00,715 is the entire story. So, when classical finance treats markets 624 00:45:48,559 --> 00:46:10,996 as equilibrium-based seeking systems whose behavior can be derived 625 00:46:10,996 --> 00:46:35,926 from the rational decisions of independent agents, it's making a 626 00:46:35,926 --> 00:47:08,335 category error. It's treating a flock as if it were a clock. And 627 00:47:08,335 --> 00:47:30,772 the consequences of that category error show up everywhere 628 00:47:30,772 --> 00:47:58,195 in how markets are described, modelled, and explained in the mainstream 629 00:47:58,195 --> 00:48:00,688 conversations. 630 00:48:02,651 --> 00:48:07,955 Okay, so let's just take a quick pause here because, I mean, 631 00:48:07,955 --> 00:48:11,933 this is quite interesting because my understanding is that 632 00:48:11,933 --> 00:48:16,795 that's pretty much how we've taught everyone in terms of finance, 633 00:48:16,795 --> 00:48:22,099 that there is this equilibrium model. And if we are now proposing 634 00:48:22,099 --> 00:48:26,519 that that's wrong, I'm surprised. And maybe you can explain 635 00:48:26,519 --> 00:48:31,823 that. How has that survived for so long if we think actually…? 636 00:48:32,181 --> 00:48:47,306 It's been a useful model, Niels, even though it's been wrong. 637 00:48:47,306 --> 00:49:03,806 So, the equilibrium model gives you a way to talk about prices, 638 00:49:03,806 --> 00:49:13,431 value securities, construct portfolios, convince investors. It 639 00:49:13,431 --> 00:49:24,431 produces numbers for a spreadsheet. The alternative, however, 640 00:49:24,431 --> 00:49:38,181 is admitting markets are not equilibrium seeking, that there is 641 00:49:38,181 --> 00:49:54,681 no fair value that prices are converging on, that the system is 642 00:49:54,681 --> 00:50:05,681 constituted by reflexive interactions. That makes it incredibly 643 00:50:05,681 --> 00:50:19,431 difficult. So, that doesn't give you a spreadsheet, Niels, that 644 00:50:19,431 --> 00:50:35,931 gives you a research program. And no one wants to hear that. So, 645 00:48:33,587 --> 00:48:44,835 the equilibrium model has survived because it's convenient, 646 00:48:44,835 --> 00:49:00,301 not because it's right. And here's the deeper problem with it. 647 00:49:00,301 --> 00:49:12,955 The equilibrium model has a particular assumption built into 648 00:49:12,955 --> 00:49:31,233 it, that once you see it, you can't stop seeing it. The assumption 649 00:49:31,233 --> 00:49:48,105 is that price movement is driven by news. So, the mental model 650 00:49:48,105 --> 00:50:04,977 is (this is how they view it), information enters a system, rational 651 00:50:04,977 --> 00:50:16,225 agents process the information, price adjusts to reflect 652 00:50:16,225 --> 00:50:31,691 the new reality: cause, then effect, clean and linear. Every move 653 00:50:31,691 --> 00:50:35,909 has a story. And that's the story CNBC tells us every day. The 654 00:48:53,521 --> 00:49:14,861 market moved because of inflation data, because of Fed minutes, 655 00:49:14,861 --> 00:49:34,067 because of an earnings surprise. External event causes price 656 00:49:34,067 --> 00:49:51,139 change. The world delivers information, the market reflects 657 00:49:51,139 --> 00:50:18,881 it. That's how they view that model. And it is, in most cases, 658 00:50:18,881 --> 00:50:35,953 much less of the story than it sounds. So, news can matter as a 659 00:48:44,355 --> 00:49:08,703 trigger, but the size and duration of the move are often determined 660 00:49:08,703 --> 00:49:35,080 less by the news than by the internal feedback that the news activates 661 00:49:35,080 --> 00:50:01,457 in the system. So, the match or the trigger starts the fire. The 662 00:50:01,457 --> 00:50:23,776 fire's shape and extent depends on the dryness of the wood, 663 00:50:23,776 --> 00:50:35,950 not the size of the match. 664 00:50:37,561 --> 00:50:43,741 Okay, fair enough. But people listening to that explanation might 665 00:50:43,741 --> 00:50:52,393 want you to dig a little bit deeper and explain that a bit more. 666 00:50:52,393 --> 00:50:59,191 Even though I think we have talked about these concepts before. 667 00:50:59,191 --> 00:51:06,607 And of course, this is also what our friend Dave Dredge likes 668 00:51:06,607 --> 00:51:14,641 to talk about. And I'm hoping to bring you and Dave together on 669 00:51:14,641 --> 00:51:17,731 a conversation very, very soon. So, if you're listening, Dave, 670 00:50:45,931 --> 00:51:11,041 you are warned. But let's just back it up a little bit and go into 671 00:51:11,041 --> 00:51:17,737 the microstructure of this. 672 00:51:18,551 --> 00:52:29,930 Okay, so we're going to talk about microstructure here. So, there 673 00:52:29,930 --> 00:53:34,820 is a substantial body of this microstructural research, 20 years 674 00:53:34,820 --> 00:54:33,221 of it. So, let's look precisely at this question. The likes 675 00:51:23,673 --> 00:51:46,722 of Jean-Philippe Bouchaud, Gabay and Cohen, they've looked at 676 00:51:46,722 --> 00:52:09,771 this microstructural research, and the answer keeps converging on 677 00:52:09,771 --> 00:52:35,381 the same thing. The vast majority of price movement can't 678 00:52:35,381 --> 00:53:03,552 be attributed to identifiable news. Prices move often on days with 679 00:53:03,552 --> 00:53:31,723 no news. They move in ways that bear no proportional relationship 680 00:53:31,723 --> 00:54:05,016 to the news that does arrive. So, some of the largest moves in 681 00:54:05,016 --> 00:54:25,504 market history have occurred without any corresponding catalyst 682 00:54:25,504 --> 00:54:33,187 of equivalent magnitude. So, if news were the primary driver, 683 00:51:49,519 --> 00:52:24,911 the relationship between information and price change would 684 00:52:24,911 --> 00:53:22,423 be consistent. But it's not. So, if news is not the primary driver 685 00:53:22,423 --> 00:54:28,783 of price, then what is it? And the answer is what I call the endogenous 686 00:54:28,783 --> 00:54:33,207 engine. Most price movement is not driven by anything coming from 687 00:51:44,851 --> 00:52:11,151 outside the system. It's driven by the system processing itself. 688 00:52:11,151 --> 00:52:32,191 Price changes behavior, behavior changes price, and around 689 00:52:32,191 --> 00:53:03,751 the loop it goes. A small breakout lifts price. That lift draws 690 00:53:03,751 --> 00:53:32,681 in momentum systems that buy what is rising. Those systems alter 691 00:53:32,681 --> 00:53:51,091 volatility. The volatility change triggers volatility targeting 692 00:53:51,091 --> 00:54:09,501 and rebalancing. The rebalancing shifts positioning across 693 00:54:09,501 --> 00:54:30,541 portfolios. The position shifts force further, each price 694 00:54:30,541 --> 00:54:33,171 adjustment. Each step in that sequence happens because of the price 695 00:51:59,101 --> 00:52:35,596 action, not because of any external information. The original 696 00:52:35,596 --> 00:53:16,146 spark, whatever it was, perhaps news, perhaps just an order 697 00:53:16,146 --> 00:53:44,531 imbalance, becomes increasingly irrelevant as the cascade 698 00:53:44,531 --> 00:54:29,136 develops. And the cascade is what matters. And the cascade is 699 00:54:29,136 --> 00:54:33,191 internal. So, this is what I mean by the endogenous engine. Markets 700 00:51:58,206 --> 00:52:30,651 don't simply reflect the world, they process themselves. Each 701 00:52:30,651 --> 00:53:06,701 price sensitive participant is watching price, acting on it and 702 00:53:06,701 --> 00:53:46,356 changing it for the next price sensitive participant. And the picture 703 00:53:46,356 --> 00:54:29,616 matters here. So, the engine runs on the agents who respond to 704 00:54:29,616 --> 00:54:33,221 price. And as we said earlier in the introduction to this podcast, 705 00:52:19,733 --> 00:53:15,353 passive participants, they're not in that loop. They are present 706 00:53:15,353 --> 00:54:22,097 in the market, but they're not the ones whose actions drive the 707 00:54:22,097 --> 00:54:33,221 recursive cycle. The whole system of price sensitive participation 708 00:51:42,883 --> 00:52:17,643 is running a continuous recursive loop where the output of 709 00:52:17,643 --> 00:52:59,355 the system becomes the input to the system's next state. So, external 710 00:52:59,355 --> 00:53:34,115 events may trigger cascades, but the engine that sustains them, 711 00:53:34,115 --> 00:54:05,399 amplifies them, and determines their ultimate magnitude. And that's 712 00:54:05,399 --> 00:54:33,207 internal. The system moves itself, so to speak. 713 00:54:33,791 --> 00:54:38,708 And which is probably what Soros wrote about many, many, many 714 00:54:38,708 --> 00:54:44,072 years ago. Right? Reflexivity. I think Cem talks about that as well, 715 00:54:44,072 --> 00:54:48,989 in the conversations that I have with him. Would that be…? 716 00:54:49,311 --> 00:55:22,983 Yes, that's exactly… Soros, I think, had it right. So, price sensitive 717 00:55:22,983 --> 00:55:45,431 participants, they don't simply observe markets and react 718 00:55:45,431 --> 00:56:10,685 to them. Their reactions reshape the markets, which reshape 719 00:56:10,685 --> 00:56:38,745 the next observation, which reshapes the next reaction. So, the 720 00:56:38,745 --> 00:57:06,805 loop runs continuously among the agents who respond to price. 721 00:54:49,311 --> 00:55:18,263 It's the operational consequence of this endogenous engine, 722 00:55:18,263 --> 00:55:54,453 the mechanism by which the system is processing itself. And 723 00:55:54,453 --> 00:56:45,119 once you have that picture in your mind, a lot of things that seem 724 00:56:45,119 --> 00:57:06,833 mysterious about markets stop being mysterious. So why do trends 725 00:55:02,103 --> 00:55:34,083 persist far beyond what any single catalyst can explain? Because 726 00:55:34,083 --> 00:56:06,063 the endogenous engine keeps running long after the trigger has 727 00:56:06,063 --> 00:56:31,647 faded. Why do markets overshoot fundamental value? Because 728 00:56:31,647 --> 00:57:06,825 there is no force in the system pulling them toward it. And 729 00:54:51,221 --> 00:55:12,231 as we established in that ‘what's on the radar’ segment, fundamental 730 00:55:12,231 --> 00:55:35,151 value belongs to a separate system that does not have to agree 731 00:55:35,151 --> 00:55:58,071 with the markets. So why do the largest moves often occur without 732 00:55:58,071 --> 00:56:11,441 corresponding catalysts? Because the engine produces arbitrarily 733 00:56:11,441 --> 00:56:28,631 large moves when the architecture is sensitized. The trigger 734 00:56:28,631 --> 00:56:53,461 is the match, the system is the fuel. And here's the move that 735 00:56:53,461 --> 00:57:06,831 changes how I see my own craft. So, the picture I've just 736 00:55:03,636 --> 00:55:32,286 drawn: markets as a flock, the system processing itself through 737 00:55:32,286 --> 00:55:52,341 reflexive feedback, prices emerging from interactions rather 738 00:55:52,341 --> 00:56:20,991 than from rational pricing, it's not a strange theoretical claim. 739 00:56:20,991 --> 00:56:46,776 It's the description any honest practitioner has been operating 740 00:56:46,776 --> 00:57:06,831 inside of, intuitively, for their entire career. So, the trader 741 00:54:55,377 --> 00:55:17,619 who has watched a stop-run knows price is moving because price 742 00:55:17,619 --> 00:55:39,861 is moving, not because anyone new information. The trader who has 743 00:55:39,861 --> 00:56:02,103 felt a market squeeze, they know the cascade is internal. We've 744 00:56:02,103 --> 00:56:26,367 known this for decades. The CAS framework is not adding new information. 745 00:56:26,367 --> 00:56:48,609 It's giving us a language to say, effectively, what we already 746 00:56:48,609 --> 00:57:06,807 knew, but it's articulating it in a better way. 747 00:57:09,251 --> 00:57:13,907 So, based on sort of what you're sharing with us today, you 748 00:57:13,907 --> 00:57:18,563 know, let's use your terms. So, markets are “flux”, so to speak. 749 00:57:18,563 --> 00:57:22,831 And the model that we've been all taught about equilibrium, that's 750 00:57:22,831 --> 00:57:27,099 not right anymore. And prices are not necessarily relating to fundamentals, 751 00:57:27,099 --> 00:57:31,755 which is something I spoke about the last couple of times. You 752 00:57:31,755 --> 00:57:36,023 just look at the month of April, you wouldn't expect markets 753 00:57:36,023 --> 00:57:40,679 to behave the way they did, so I’m completely on-board with that. 754 00:57:40,679 --> 00:57:45,335 But of course it leaves the question for all investors, and that 755 00:57:45,335 --> 00:57:48,439 is, so how do we deal with that? 756 00:57:48,951 --> 00:57:59,121 Well, that's a great question. That's exactly where we're going 757 00:57:59,121 --> 00:58:11,325 to go in section two. Because once you accept that markets are 758 00:58:11,325 --> 00:58:21,495 a complex adaptive system, that the endogenous engine is what 759 00:58:21,495 --> 00:58:30,648 drives most price movement, that reflexivity is the operational 760 00:58:30,648 --> 00:58:43,869 mechanism, once you have all of that in your hand, the next question 761 00:58:43,869 --> 00:58:56,073 is what is the architecture of the engine? What are the feedback 762 00:58:56,073 --> 00:59:05,226 loops actually doing? How do reinforcing and balancing forces 763 00:59:05,226 --> 00:59:15,396 interact to produce the patterns we observe? And what does 764 00:59:15,396 --> 00:59:26,583 that mean for a practitioner like us trying to operate within 765 00:59:26,583 --> 00:59:35,736 that system? That's the architecture of feedback. And that's 766 00:59:35,736 --> 00:59:46,923 where we're going to go next in the next sectional topic. But 767 00:57:50,444 --> 00:58:09,853 before we get there, I just want to leave the listener with one 768 00:58:09,853 --> 00:58:27,769 final image because everything in the next section is going to come 769 00:58:27,769 --> 00:58:47,178 back to it. So, the flock is not chaos, it is highly structured. 770 00:58:47,178 --> 00:58:57,629 The starlings produce coherent, complex, beautiful patterns 771 00:58:57,629 --> 00:59:17,038 in the sky. And they do it without anyone in charge. The same 772 00:59:17,038 --> 00:59:31,968 is true of markets. The patterns are real, the structure 773 00:59:31,968 --> 00:59:46,898 is real, trends are not anomalies, cascades are not aberrations, 774 00:57:48,951 --> 00:58:36,143 concentration is not coincidence. These are emergent properties 775 00:58:36,143 --> 00:59:35,133 of a complex adaptive system populated by agents whose interactions 776 00:59:35,133 --> 00:59:46,931 produce structure. So, the fact that news is more often a trigger 777 00:58:03,331 --> 00:58:20,587 than a driver does not make the structure less real. It makes 778 00:58:20,587 --> 00:58:32,091 a structure something different from what classical finance 779 00:58:32,091 --> 00:58:50,785 said it was. And once you see that structure clearly, you can begin 780 00:58:50,785 --> 00:59:08,041 to align your craft with it, which is what trend following has 781 00:59:08,041 --> 00:59:23,859 been doing for 40 years now, even when its practitioners might 782 00:59:23,859 --> 00:59:43,991 not have had the language to say so. So, let's get into that next 783 00:59:43,991 --> 00:59:46,867 second topic. 784 00:59:47,651 --> 00:59:56,836 Absolutely. And by the way, I love the talk about architecture. 785 00:59:56,836 --> 01:00:07,691 I was sharing with you, before we press record, I listened to a 786 01:00:07,691 --> 01:00:14,371 wonderful conversation with Norman Foster, the very well-known 787 01:00:14,371 --> 01:00:23,556 famous architect, and actually he was talking in language that I 788 01:00:23,556 --> 01:00:32,741 thought, wow, there could be some analogies useful in trend following. 789 01:00:32,741 --> 01:00:44,431 And here we are. Well, I don't know, but here we are talking about 790 01:00:44,431 --> 01:00:45,266 architecture. Anyways, obviously, you painted a picture. 791 00:59:54,431 --> 01:00:05,731 So far, we've talked about complex adaptive systems, kind of 792 01:00:05,731 --> 01:00:14,771 this endogenous engine, and we've now introduced reflexivity 793 01:00:14,771 --> 01:00:28,331 as well. And now you're adding architecture. So, what is the engine 794 01:00:28,331 --> 01:00:41,891 actually made of? And how do we know, empirically, that this is 795 01:00:41,891 --> 01:00:45,281 what's going on? 796 01:00:46,321 --> 01:01:02,081 Okay, so let's be concrete about what feedback actually means, 797 01:01:02,081 --> 01:01:17,841 because the word gets thrown around loosely in financial commentary. 798 01:01:17,841 --> 01:01:35,177 It's almost lost its meaning. So, in any complex adaptive system, 799 01:01:35,177 --> 01:01:50,937 feedback is a mechanism by which local actions become global 800 01:01:50,937 --> 01:02:08,273 structure. It's how the flock turns. It's how the cascade builds. 801 01:02:08,273 --> 01:02:24,033 It's the operational machinery that lets the system process itself. 802 01:02:24,033 --> 01:02:39,793 And in markets, feedback operates in two distinct modes. The 803 01:02:39,793 --> 01:02:50,825 interplay between them determines everything about what 804 01:02:50,825 --> 01:02:57,129 the market actually does. So, the first mode, the reinforcing feedback, 805 01:01:02,386 --> 01:01:27,631 it amplifies movement. A small buy order nudges price upward. That 806 01:01:27,631 --> 01:01:48,286 nudge activates trading systems by what is rising. Models 807 01:01:48,286 --> 01:02:02,056 recalibrate, institutions rebalance, performance chasers arrive 808 01:02:02,056 --> 01:02:25,006 late but with size, each action reinforces the previous one. 809 01:02:25,006 --> 01:02:47,956 The local behavior forms a directional pattern, and this is 810 01:02:47,956 --> 01:02:57,136 the fuel of trends. But the second mode is balancing feedback, 811 01:01:01,588 --> 01:01:25,579 it absorbs movement. Market makers quote both sides of the book, 812 01:01:25,579 --> 01:01:47,389 value investors step in when prices fall below perceived worth, 813 01:01:47,389 --> 01:02:07,018 risk managers reduce exposure when volatility rises beyond tolerance. 814 01:02:07,018 --> 01:02:24,466 These responses counter extreme movement and create temporary 815 01:02:24,466 --> 01:02:46,276 zones of calm. Without them, every reinforcing loop would grow 816 01:02:46,276 --> 01:02:57,181 unchecked until the system collapsed. So, the two run simultaneously 817 01:01:01,536 --> 01:01:28,923 against each other. When reinforcing loops dominate, the system 818 01:01:28,923 --> 01:01:50,224 becomes directional and transforms. When balancing loops 819 01:01:50,224 --> 01:02:20,654 dominate, the system becomes range bound. When the influence shifts 820 01:02:20,654 --> 01:02:54,127 rapidly from one to the other, phase transitions occur and volatility 821 01:02:54,127 --> 01:02:57,170 surges. The market breathes through the opposing loops, never 822 01:01:15,401 --> 01:01:51,751 still, always adapting to the consequences of its own actions. 823 01:01:51,751 --> 01:02:24,466 And that's the architecture: two kinds of feedback continuously 824 01:02:24,466 --> 01:02:49,911 interacting, determining whether the market trends, ranges, 825 01:02:49,911 --> 01:02:57,181 or breaks. 826 01:02:58,311 --> 01:03:02,694 And so, the architecture actually shifts because, what you 827 01:03:02,694 --> 01:03:07,564 said earlier, that passive now contributes to reinforcing the feedback 828 01:03:07,564 --> 01:03:12,921 and diluting the balancing agent, so to speak, if that's correctly 829 01:03:12,921 --> 01:03:13,408 understood. 830 01:03:13,511 --> 01:03:58,868 Spot on, spot on. So, that's why I want to land that connection 831 01:03:58,868 --> 01:04:30,269 properly here, in this segment, because it's a structural 832 01:04:30,269 --> 01:05:05,159 fact that drives almost everything we're going to talk about. 833 01:05:05,159 --> 01:05:36,560 So, the reinforcing agents amplify price movement. The momentum 834 01:05:36,560 --> 01:06:14,939 chaser chasing the move. The stop loss selling into the fall. 835 01:06:14,939 --> 01:06:42,851 The volatility targeting fund reducing exposures as volatility 836 01:06:42,851 --> 01:07:14,252 rises. The passive fund, allocating by market cap weight, 837 01:07:14,252 --> 01:07:56,120 buying more of what is already large. The options dealer who must 838 01:07:56,120 --> 01:08:37,988 delta hedge into the direction of the move. The forced index inclusion 839 01:08:37,988 --> 01:09:19,856 buying as a stock joins a benchmark. All of these things amplify 840 01:09:19,856 --> 01:09:23,345 direction. Some are price sensitive: the momentum trader. Some 841 01:04:33,479 --> 01:06:03,443 are price insensitive: the passive fund responding to index 842 01:06:03,443 --> 01:07:53,399 weight rather than price. Either way, they take what the market 843 01:07:53,399 --> 01:09:23,363 is already doing and add more force to it. So, the balancing agents, 844 01:03:28,015 --> 01:03:57,023 however, absorb price movement: the value investor stepping 845 01:03:57,023 --> 01:04:36,909 in when something looks cheap, the selling of an overpriced asset, 846 01:04:36,909 --> 01:05:09,543 the market maker, fading extreme orders, the counter trend 847 01:05:09,543 --> 01:05:49,429 trader, all push back against price in motion. So, what passive 848 01:05:49,429 --> 01:06:36,567 does (and this is a point I want every listener to hold), it 849 01:06:36,567 --> 01:07:05,575 doesn't eliminate balancing agents. The value investor still 850 01:07:05,575 --> 01:07:41,835 exists, the counter trend trader still exists, the market maker 851 01:07:41,835 --> 01:08:25,347 still exists. But what passive does is dilute their share of the 852 01:08:25,347 --> 01:09:01,607 ecology. They reduce their weight relative to the capital that 853 01:09:01,607 --> 01:09:23,363 no longer pushes back on price. So, the reinforcing agents, 854 01:03:47,915 --> 01:05:13,925 they haven't been diluted. The forced flows around index inclusion 855 01:05:13,925 --> 01:06:39,935 and rebalancing have actually grown. But what has been diluted 856 01:06:39,935 --> 01:08:14,546 is the proportional weight of the agents conditioned on price relative 857 01:08:14,546 --> 01:09:23,354 to value. The balancing feedback has been thinned. So, the 858 01:03:35,927 --> 01:04:54,383 structural consequence, the architecture, reinforcing loops now 859 01:04:54,383 --> 01:06:46,463 run further before they meet resistance. The system is therefore 860 01:06:46,463 --> 01:08:27,335 more prone to extended directional moves, more prone to 861 01:08:27,335 --> 01:09:23,375 overshoot, more prone to cascade. Look what's happening for 862 01:03:44,987 --> 01:05:27,284 example, with the rise and rise of the S&P 500, when the world 863 01:05:27,284 --> 01:07:01,712 about it is falling down. This is this passive injection which is 864 01:07:01,712 --> 01:08:12,533 reinforcing feedback. But as it progresses, any relationship it 865 01:08:12,533 --> 01:09:23,354 might have to fundamental value is being progressively widened. 866 01:03:13,511 --> 01:04:07,446 So, the structural consequence, how reinforcing loops 867 01:04:07,446 --> 01:05:32,201 now run further before they meet resistance. The system is more 868 01:05:32,201 --> 01:06:41,546 prone to these extended directional moves, more prone to 869 01:06:41,546 --> 01:07:50,891 overshoot, more prone to cascades. Not because the reinforcing 870 01:07:50,891 --> 01:09:00,236 forces have grown stronger, but because the balancing forces 871 01:09:00,236 --> 01:09:23,351 have grown weaker. And there is a further amplifier I need to 872 01:04:08,987 --> 01:05:10,627 plant here. There's been research over the past few years 873 01:05:10,627 --> 01:06:12,267 on what is called market inelasticity, and the findings are 874 01:06:12,267 --> 01:07:26,235 striking. You had a podcast, I believe, with Al Alan, talking about 875 01:07:26,235 --> 01:08:46,367 this effect from Gabay and Cohen. So, let me set this up carefully 876 01:08:46,367 --> 01:09:23,351 because the picture's a little counterintuitive. So, when the majority 877 01:03:48,735 --> 01:05:34,407 of shares are held by passive vehicles; index funds and long term 878 01:05:34,407 --> 01:07:11,273 holders who do not respond to price (price insensitive), the effective 879 01:07:11,273 --> 01:08:39,333 pool available to absorb active buying or selling (the price 880 01:08:39,333 --> 01:09:23,363 sensitive agents) is structurally thin. So, think of it 881 01:03:29,247 --> 01:04:12,521 as a price discovery pool. The price insensitive holders, they own 882 01:04:12,521 --> 01:04:55,795 the shares, the bulk of the shares, they're not transacting on 883 01:04:55,795 --> 01:05:46,937 the basis of price. So, when an active buyer or seller comes into 884 01:05:46,937 --> 01:06:34,145 the market, they have to find the other side of the transaction 885 01:06:34,145 --> 01:07:21,353 with a much smaller fraction of the available float. And the fraction, 886 01:07:21,353 --> 01:08:00,693 that's the fraction still responding to price, the price sensitive 887 01:08:00,693 --> 01:08:43,967 fraction, it's operating on a thinner float. So, the impact of 888 01:08:43,967 --> 01:09:23,307 the price movement, from the interaction, is greater, not smaller. 889 01:03:13,511 --> 01:05:21,051 Here's the consequence that surprises people. Each unit of active 890 01:05:21,051 --> 01:07:54,099 flow has to move price further to find a willing counterparty because 891 01:07:54,099 --> 01:09:23,377 the pool absorbing the flow is smaller. So, US$1 of net buying doesn't 892 01:03:42,707 --> 01:04:50,831 push the price up by US$1 of value, it pushes the price up by 893 01:04:50,831 --> 01:05:54,089 multiple of that. And the best work in this space from Gabay, Cohen 894 01:05:54,089 --> 01:06:42,749 and, I think, Bouchaud's worked on this as well, suggests 895 01:06:42,749 --> 01:07:50,873 the multiplier is on the order of 5. US$1 of net buying can increase 896 01:07:50,873 --> 01:08:20,069 aggregate market capitalization by approximately US$5. 897 01:08:20,069 --> 01:09:08,729 US$1 of selling can destroy approximately US$5. So, that's what 898 01:09:08,729 --> 01:09:23,327 economists call inelasticity. An inelastic market is one where 899 01:04:35,705 --> 01:07:47,491 price moves a lot per unit of flow. What the rise of passive has 900 01:07:47,491 --> 01:09:23,384 done is make markets structurally less elastic. You know, 901 01:03:25,839 --> 01:04:39,807 I talked about in car markets we get price sensitivity. We get 902 01:04:39,807 --> 01:05:41,447 markets being more sensitive. This is this effect. It's becoming 903 01:05:41,447 --> 01:06:36,923 more inelastic. Fewer price sensitive participants in the active 904 01:06:36,923 --> 01:07:38,563 pool means less capacity to absorb flow without price moving. 905 01:07:38,563 --> 01:08:46,367 So, the same dollar of buying produces more price movement than 906 01:08:46,367 --> 01:09:23,351 it would have 20 years ago. Now, counterintuitively, fewer price 907 01:03:49,595 --> 01:05:19,805 sensitive players in the active mix means more violent price 908 01:05:19,805 --> 01:07:17,078 action per unit of flow, not less. So, the agents who are not 909 01:07:17,078 --> 01:08:47,288 in the active pool, they're simply present, holding their share, 910 01:08:47,288 --> 01:09:23,372 not absorbing the trade. So, the feedback architecture does not 911 01:04:38,867 --> 01:06:46,901 just sustain trends, Niels, the inelasticity ensures that the 912 01:06:46,901 --> 01:09:23,387 trends it produces are larger than the flows that created them. 913 01:09:24,271 --> 01:09:27,560 Yeah, I mean, I had heard about the multiplier before. Not 914 01:09:27,560 --> 01:09:31,746 sure I had heard how big it was. But five is a pretty striking 915 01:09:31,746 --> 01:09:35,633 number, though it does ring a bell. How can that be proved? Can 916 01:09:35,633 --> 01:09:36,530 it be proved? 917 01:09:38,061 --> 01:09:47,771 Well, according to these studies, they seem to have demonstrated 918 01:09:47,771 --> 01:09:59,423 it, but yeah, yeah, good piece of research, but it definitely warrants 919 01:09:59,423 --> 01:10:08,162 further investigation. So, I'll stand behind this claim. The 920 01:10:08,162 --> 01:10:17,872 exact multiplier itself, I suppose, depends on which study you 921 01:10:17,872 --> 01:10:29,524 look at and how the methodology is set up. So, the range 922 01:10:29,524 --> 01:10:42,147 in the literature that I've seen runs from roughly 3 to 1 at 923 01:10:42,147 --> 01:10:54,770 the conservative end to closer to 7 to 1 in the more aggressive 924 01:10:54,770 --> 01:10:55,741 estimates. So, the point that I want the listener to hold is not 925 01:09:57,489 --> 01:10:13,679 the specific number, but the structural fact. The multiplier is 926 01:10:13,679 --> 01:10:31,488 large, whichever way you view it. It's much larger than classical 927 01:10:31,488 --> 01:10:49,297 finance assumed. And it's a direct consequence of the agent ecology 928 01:10:49,297 --> 01:10:55,773 we've just been describing. So, the price insensitive holders 929 01:09:44,876 --> 01:10:03,958 are now so large such a share of the float that the marginal price 930 01:10:03,958 --> 01:10:18,951 sensitive dollar is doing the price discovery work in a structurally 931 01:10:18,951 --> 01:10:33,944 thinner pool. Each unit of active flow moves price more than 932 01:10:33,944 --> 01:10:50,300 it would have 20 years ago. Much more. So that's the architecture 933 01:10:50,300 --> 01:10:55,752 as it stands today. So, I want to walk you through the evidence. 934 01:10:14,871 --> 01:10:55,771 And this isn't a theory, this can be empirically demonstrated. 935 01:10:56,501 --> 01:10:59,801 Okay, let's do that. It's exciting. Just be mindful of the 936 01:10:59,801 --> 01:11:03,101 time. But, but it's very exciting. So, let's see where we're 937 01:11:03,101 --> 01:11:04,001 going with this. 938 01:11:04,341 --> 01:11:07,953 Okay, so you remember in our last podcast I discussed this research 939 01:11:07,953 --> 01:11:09,458 series, the Fractals of Finance? 940 01:11:09,461 --> 01:11:10,981 Yeah, absolutely. 941 01:11:10,981 --> 01:11:46,471 About a year ago I worked through a second phase of this research, 942 01:11:46,471 --> 01:12:13,771 Niels, and that was where we're talking about the statistical 943 01:12:13,771 --> 01:12:43,801 fingerprint of feedback driven markets across 40 years and 68 markets. 944 01:12:43,801 --> 01:13:08,371 That was the test I conducted this research on. So, in this second 945 01:11:22,433 --> 01:11:53,926 phase, a colleague and I extended the work in a particular 946 01:11:53,926 --> 01:12:25,419 direction. We built what we call a Minimal Agent Based Model 947 01:12:25,419 --> 01:13:02,638 of a financial market. So, the model had only two kinds of agents 948 01:13:02,638 --> 01:13:08,364 in it. The first kind was what we call chartists. They're like us, 949 01:11:39,053 --> 01:11:59,469 divergent traders, agents whose behavior is conditioned on 950 01:11:59,469 --> 01:12:22,437 recent price action. Rising prices attracted them as buyers. 951 01:12:22,437 --> 01:12:47,957 Falling prices attracted them as sellers. They were the reinforcing 952 01:12:47,957 --> 01:13:08,373 or the divergent agents in our simple model. The second kind were 953 01:11:19,077 --> 01:11:37,293 fundamentalists. They were convergent agents who traded on an 954 01:11:37,293 --> 01:11:59,557 estimate of intrinsic value. And when price strayed too far from 955 01:11:59,557 --> 01:12:21,821 value, they pushed back. They were the balancing agents. So, the 956 01:12:21,821 --> 01:12:44,085 model we generated had one knob. You can turn the proportion 957 01:12:44,085 --> 01:13:08,373 of chartists in the population up or down, from 0% to 100%. And 958 01:11:13,991 --> 01:11:47,101 we asked a simple question. What happens to the statistical behavior 959 01:11:47,101 --> 01:12:17,201 of this simulated market as the proportion of price sensitive 960 01:12:17,201 --> 01:12:50,311 chartists change? So, the result was so striking. So, I want 961 01:12:50,311 --> 01:13:08,371 to walk through this with you. So, with zero chartists in the population, 962 01:11:28,460 --> 01:11:48,436 meaning all agents were fundamentalists, no reinforcing feedback 963 01:11:48,436 --> 01:12:15,903 in the system, the simulated market behaved exactly the way classical 964 01:12:15,903 --> 01:12:35,879 finance said real markets should behave: independent returns, 965 01:12:35,879 --> 01:13:05,843 Gaussian tails, no memory from one day to the next, no volatility 966 01:13:05,843 --> 01:13:08,340 clustering. 967 01:13:14,311 --> 01:13:19,399 All right, Rich, let's just take a quick pause here because it 968 01:13:19,399 --> 01:13:24,063 seems like you are about to spread some statistical terms here. 969 01:13:24,063 --> 01:13:29,575 I know we spoke about one of them, the hurst exponent, I think 970 01:13:29,575 --> 01:13:34,663 you and I talked about not that long ago, kurtosis. And let's 971 01:13:34,663 --> 01:13:39,327 just maybe, again, to make sure we have everyone listening on 972 01:13:39,327 --> 01:13:43,991 board here. What is kind of the non-quant version of that? 973 01:13:44,341 --> 01:14:58,863 Okay, so the hurst exponent is just a measure of memory in a time 974 01:14:58,863 --> 01:16:02,739 series. So, a reading near 0.5 tells you that today's price move 975 01:16:02,739 --> 01:16:50,646 tells you almost nothing about tomorrow's. The series therefore 976 01:16:50,646 --> 01:17:06,615 has no memory. The further the reading drifts above 0.5, the more 977 01:14:37,882 --> 01:15:49,270 memory the series has, the more the past is leaving a trace 978 01:15:49,270 --> 01:17:06,607 in what comes next. That's all it is. It's a measure of memory. 979 01:13:44,341 --> 01:15:11,029 But kurtosis, that's a measure of fat tails. So, in plain English, 980 01:15:11,029 --> 01:16:30,493 it asks whether extreme moves happen more often than the textbook 981 01:16:30,493 --> 01:17:06,613 bell curve says they should. So excess kurtosis near zero says 982 01:14:13,237 --> 01:15:06,213 the world is behaving like the textbook. High excess kurtosis says 983 01:15:06,213 --> 01:15:59,189 the world is producing extreme events that the textbook said were 984 01:15:59,189 --> 01:17:01,797 almost impossible. That's all it is, a measure of how fat the tails 985 01:17:01,797 --> 01:17:06,613 are. So, here's the result we got with our research, with zero 986 01:14:19,101 --> 01:14:53,861 chartists in the simulation, the Hurst exponent sat at about 0.54. 987 01:14:53,861 --> 01:15:19,141 That was statistically indistinguishable from a random walk 988 01:15:19,141 --> 01:15:53,901 - no memory. Excess kurtosis was negative 0.2. Tails were thinner 989 01:15:53,901 --> 01:16:28,661 than the textbook, not fatter. Five sigma events, there were zero. 990 01:16:28,661 --> 01:17:03,421 Every assumption of efficient markets held in this world of zero 991 01:17:03,421 --> 01:17:06,581 chartists. But when we turn the knob and raise the proportion 992 01:14:47,551 --> 01:15:44,440 of chartists or divergent participants in that market, the 993 01:15:44,440 --> 01:16:35,008 behavior didn't drift gradually. It changed through what 994 01:16:35,008 --> 01:17:06,613 I call a phase transition. So, a phase transition is where behavior 995 01:14:06,811 --> 01:14:35,701 changes suddenly rather than gradually, like water turning to 996 01:14:35,701 --> 01:15:17,431 steam. You can heat water from 20 degrees to 90 degrees, it stays 997 01:15:17,431 --> 01:15:52,741 watered the whole way, it just gets warmer. But somewhere between 998 01:15:52,741 --> 01:16:24,841 99 and 100, the same ingredients suddenly produce a different 999 01:16:24,841 --> 01:17:00,151 kind of behavior. The water becomes steam. That's what a phase 1000 01:17:00,151 --> 01:17:06,571 transition is. And we got that in our research when we turned the 1001 01:14:16,582 --> 01:14:54,685 knob. But we got that at a figure where we had around about 1002 01:14:54,685 --> 01:15:21,064 25% of the model being chartists or divergent practitioners. 1003 01:15:21,064 --> 01:15:50,374 The balance being convergent. It wasn't a 50/50 transition. This 1004 01:15:50,374 --> 01:16:25,546 was we only needed to put 25% chartists, who are price sensitive, 1005 01:16:25,546 --> 01:16:54,856 working in the reinforcing direction to create this phase transition 1006 01:16:54,856 --> 01:17:06,580 in the market structure. So, above this 25% chartists, everything 1007 01:14:22,267 --> 01:15:19,156 changed at once. Kurtosis erupted, the hurst exponent leapt 1008 01:15:19,156 --> 01:16:16,045 above 0.85. So, memory appeared, fat tails appeared, volatility 1009 01:16:16,045 --> 01:17:06,613 clustering appeared. Every statistical signature of real minds. 1010 01:13:44,341 --> 01:14:34,911 Those signatures, they've been documented across 40 years of data, 1011 01:14:34,911 --> 01:15:35,595 and across every asset class we looked at, and it appeared at 1012 01:15:35,595 --> 01:16:26,165 once at the same threshold for each market. That's interesting. 1013 01:16:26,165 --> 01:17:06,621 So, we ran the comparison, the simulated market. Once feedback was 1014 01:13:59,140 --> 01:14:38,604 introduced above this critical threshold, it became statistically 1015 01:14:38,604 --> 01:15:23,001 indistinguishable from the 68 real futures markets spanning eight 1016 01:15:23,001 --> 01:16:12,331 asset classes, six continents, 41 years of trading history. The 1017 01:16:12,331 --> 01:17:06,594 simulated market and the real markets had exactly that same fingerprint. 1018 01:17:06,791 --> 01:17:12,551 Yeah, that's kind of completely crazy, right? So, you 1019 01:17:12,551 --> 01:17:18,311 take a simulation where there's no news, no fundamentals, 1020 01:17:18,311 --> 01:17:24,711 no central banks, no geopolitics either, and you get the 1021 01:17:24,711 --> 01:17:31,111 same statistical behavior as a real market, across four decades 1022 01:17:31,111 --> 01:17:38,791 of time span, just by having, let's call it, a critical proportion 1023 01:17:38,791 --> 01:17:43,911 of agents whose behavior is conditioned on price? 1024 01:17:44,281 --> 01:18:40,666 They're reinforcing the feedback. We started with the fundamental 1025 01:18:40,666 --> 01:19:43,316 investors, they're convergent traders. And when we had 100% of 1026 01:19:43,316 --> 01:20:39,701 them, it exhibited convergent properties, which is very similar 1027 01:20:39,701 --> 01:21:04,761 to the Gaussian properties. But then when we got to 25% participation 1028 01:18:20,737 --> 01:18:57,193 of chartists or divergent traders (effectively the reinforcing 1029 01:18:57,193 --> 01:19:42,763 agents, not the balancing agents), we suddenly got this phase 1030 01:19:42,763 --> 01:20:23,776 transition. And the market signature; the signature across every 1031 01:20:23,776 --> 01:21:04,789 market we examined was identical to real market structure. 1032 01:17:44,281 --> 01:18:42,317 So, the implication is this is just feedback alone that's doing 1033 01:18:42,317 --> 01:19:35,077 this, without any of the conventional narrative which is said 1034 01:19:35,077 --> 01:20:17,285 to drives markets. We reproduce the statistical signatures 1035 01:20:17,285 --> 01:21:04,769 of real markets without any of that conventional narrative. The 1036 01:17:47,622 --> 01:18:24,373 absence of the chartists produced a world that has never existed. 1037 01:18:24,373 --> 01:18:54,442 So, the controlled experiment shows that feedback is sufficient 1038 01:18:54,442 --> 01:19:24,511 to generate the statistical signatures we observe. That's much 1039 01:19:24,511 --> 01:19:54,580 stronger than a loose correlation. It's evidence that feedback 1040 01:19:54,580 --> 01:20:34,672 is not a minor mechanism sitting at the edge of market behavior. 1041 01:20:34,672 --> 01:21:04,741 It's the evidence that feedback IS the central mechanism. 1042 01:17:44,281 --> 01:18:38,582 So, what I want to get the listener to hold, about this phase 1043 01:18:38,582 --> 01:19:28,706 transition, is 20% to 25% of chartists, that's not a high bar. 1044 01:19:28,706 --> 01:20:10,476 A modest minority of price sensitive participants in the reinforcing 1045 01:20:10,476 --> 01:21:00,600 direction is a enough to push the entire system across that critical 1046 01:21:00,600 --> 01:21:04,777 boundary. So, let's now set passive to one side since we've established 1047 01:18:22,968 --> 01:18:54,621 that passive participants are price insensitive by design. So, 1048 01:18:54,621 --> 01:19:29,791 among the rest of the market, virtually every active participant 1049 01:19:29,791 --> 01:20:08,478 is price sensitive to some degree: stop loss orders of feedback, 1050 01:20:08,478 --> 01:20:36,614 margin calls are feedback, volatility targeting is feedback, 1051 01:20:36,614 --> 01:21:04,750 options hedging is feedback, performance chasing is feedback. 1052 01:17:44,281 --> 01:18:33,181 Whatever fraction of the market is doing active price discovery, 1053 01:18:33,181 --> 01:19:22,081 the proportion of that fraction that responds to price is 1054 01:19:22,081 --> 01:20:15,871 not at 25%. It's far above that level. We've just demonstrated 1055 01:20:15,871 --> 01:21:04,771 with passive was represented as 60% of the equity market. So, 1056 01:17:48,736 --> 01:18:24,376 the real markets comprise these price sensitive reinforcing 1057 01:18:24,376 --> 01:19:13,381 agents that far exceed that 25% threshold. In fact, they've exceeded 1058 01:19:13,381 --> 01:20:11,296 it over the last 40 years. There has been no decade where we've 1059 01:20:11,296 --> 01:20:55,846 observed this statistical fingerprint is lost. This is a really 1060 01:20:55,846 --> 01:21:04,756 important conclusion. So, we did these rolling five-year windows, 1061 01:18:11,266 --> 01:19:01,381 across the 40 years of data, and we saw that the hurst exponent 1062 01:19:01,381 --> 01:19:47,641 never once fell to the random walk baseline of 0.5. It fluctuated 1063 01:19:47,641 --> 01:20:26,191 between 0.62 and 0.82, responding to both crises and regime 1064 01:20:26,191 --> 01:21:04,741 changes. It never approached the level that would indicate independence. 1065 01:17:44,281 --> 01:18:34,402 The fingerprint, for instance, survived Black Monday, the transition 1066 01:18:34,402 --> 01:19:24,523 to electronic trading, the global financial crisis, the rise 1067 01:19:24,523 --> 01:20:25,782 of high frequency trading, the COVID crash. It's been present continuously 1068 01:20:25,782 --> 01:21:04,765 since the first day of available data. 1069 01:21:04,791 --> 01:21:11,181 In other words, Rich (and I know we have to kind of wrap it up 1070 01:21:11,181 --> 01:21:15,441 maybe and continue on our next conversation), but essentially what 1071 01:21:15,441 --> 01:21:18,849 you're saying is that markets can't stop trending. 1072 01:21:19,761 --> 01:22:34,497 Yes, and the evidence points to it's going to increase and get 1073 01:22:34,497 --> 01:23:49,233 better. So, I'll just finish off this topic and then we'll conclude. 1074 01:23:49,233 --> 01:25:10,197 I won't go into the final topic and I'll leave this for an 1075 01:25:10,197 --> 01:26:06,249 exciting installment at one of our next meetings together. So, the 1076 01:21:29,003 --> 01:22:24,455 data that we looked at started in 1985. The fingerprint is there 1077 01:22:24,455 --> 01:23:10,665 from the first observation, from 1985 going forward, across every 1078 01:23:10,665 --> 01:24:01,496 asset that we looked at: 68 futures markets, 8 asset classes. 1079 01:24:01,496 --> 01:24:56,948 There is no era of available market data in which markets behave 1080 01:24:56,948 --> 01:25:43,158 like the equilibrium model said they should. That world doesn't 1081 01:25:43,158 --> 01:26:06,263 exist in that historical record. So, here's a move that I 1082 01:21:36,291 --> 01:22:09,351 think how a practitioner like us should see their craft. If feedback 1083 01:22:09,351 --> 01:22:39,656 driven behavior has been present for at least 40 years, never 1084 01:22:39,656 --> 01:23:04,451 weakened, appeared in every market on every continent, across 1085 01:23:04,451 --> 01:23:34,756 every asset class, then what trend following has been doing for 1086 01:23:34,756 --> 01:23:59,551 those 40 years is not exploiting a temporary inefficiency, 1087 01:23:59,551 --> 01:24:29,856 it's not catching a window that might close, it is harvesting 1088 01:24:29,856 --> 01:25:00,161 a structural feature of how markets actually work. So, the directional 1089 01:25:00,161 --> 01:25:27,711 moves that trend followers capture are the natural output of 1090 01:25:27,711 --> 01:25:49,751 the endogenous engine running through the feedback’s architecture. 1091 01:25:49,751 --> 01:26:06,281 They are the murmuration in motion. And given what we said about 1092 01:21:52,197 --> 01:22:35,445 passive diluting balancing agents while leaving the reinforcing 1093 01:22:35,445 --> 01:23:24,099 ones intact. The architecture is not weakening, it's hardening: 1094 01:23:24,099 --> 01:24:18,159 the float is thinner, the multiplier on flow has increased, 1095 01:24:18,159 --> 01:25:12,219 the system is producing the same fingerprint with more violence 1096 01:25:12,219 --> 01:26:06,279 per unit of flow than it did 20 years ago. There's one more empirical 1097 01:21:31,825 --> 01:22:14,049 finding before I go that I want to share with you. So, we ran 1098 01:22:14,049 --> 01:22:53,257 one further test, and instead of looking at markets one at a time, 1099 01:22:53,257 --> 01:23:26,433 we asked whether 68 markets across the world were behaving as 1100 01:23:26,433 --> 01:23:56,593 68 independent systems or as one permanently coupled system? So, 1101 01:23:56,593 --> 01:24:20,721 the methodology assembled every market's rolling feedback state 1102 01:24:20,721 --> 01:24:56,913 into a single cross-sectional view and asked if, at any given date, 1103 01:24:56,913 --> 01:25:27,073 what fraction of markets were sharing the majority feedback sign? 1104 01:25:27,073 --> 01:25:54,217 So, if markets were truly independent, that fraction should 1105 01:25:54,217 --> 01:26:06,281 have averaged around 0.55. But what we found was that the fraction 1106 01:22:04,705 --> 01:22:55,267 averaged 0.6 persistently across the entire 40 year sample. 1107 01:22:55,267 --> 01:24:02,683 So, in no year since 1987 did the cross-market coupling fail to 1108 01:24:02,683 --> 01:24:58,863 the level that independence would predict. During the most coordinated 1109 01:24:58,863 --> 01:26:06,279 regimes, the fraction sharing a majority sign ran as high as 0.82. 1110 01:21:19,761 --> 01:22:44,934 So, the implication is that the world's futures markets are not 1111 01:22:44,934 --> 01:23:46,878 68 independent feedback engines. They are one permanently 1112 01:23:46,878 --> 01:24:41,079 coupled feedback engine with different surface manifestations 1113 01:24:41,079 --> 01:25:58,509 in different asset classes, all running on the same underlying 1114 01:25:58,509 --> 01:26:06,252 architecture. And the reason for the coupling? When you look at 1115 01:22:07,511 --> 01:23:04,811 it, it's not mysterious. Many of the same investors apply the same 1116 01:23:04,811 --> 01:23:57,336 models across many of the same markets. That's what's creating the 1117 01:23:57,336 --> 01:24:45,086 coupling - the portfolio diversification that we all run as 1118 01:24:45,086 --> 01:25:28,061 practitioners. Because we're never investing in just a single 1119 01:25:28,061 --> 01:26:06,261 market. We're investing our models across many markets. So, this 1120 01:21:30,371 --> 01:22:12,811 means we're deploying correlated models across asset classes. 1121 01:22:12,811 --> 01:23:21,776 And when the models produce a buy signal in one market, it often 1122 01:23:21,776 --> 01:24:20,131 produces a similar signal in another. The coupling is a structural 1123 01:24:20,131 --> 01:25:07,876 consequence of how systematic capital is being globally allocated. 1124 01:25:07,876 --> 01:26:00,926 And the feedback signature is occurring across the entire coupled 1125 01:26:00,926 --> 01:26:06,231 network. The coupling shifts position depending on the macro backdrop. 1126 01:22:22,653 --> 01:23:25,545 So, during crisis episodes, the system collectively tilts posture 1127 01:23:25,545 --> 01:24:35,425 towards positive feedback back and trends emerge. But during quantitative 1128 01:24:35,425 --> 01:25:31,329 easing, suppressed regimes, the system collectively tilts towards 1129 01:25:31,329 --> 01:26:06,269 negative feedback and oscillation emerges. But the coupling 1130 01:21:32,040 --> 01:22:04,784 itself doesn't disappear because the underlying reason for 1131 01:22:04,784 --> 01:22:53,900 it doesn't disappear. So, what that means (and this is where I'll 1132 01:22:53,900 --> 01:23:34,830 conclude) is that trend following is not a strategy that 1133 01:23:34,830 --> 01:24:23,946 captures opportunities in some markets some of the time, it is a 1134 01:24:23,946 --> 01:25:00,783 process that aligns with a global feedback architecture that 1135 01:25:00,783 --> 01:25:37,620 is present simultaneously across every market that humans trade. 1136 01:25:37,620 --> 01:26:06,271 The architecture is not local, it's structural. 1137 01:26:08,472 --> 01:26:19,266 You know, it's funny, I mean, I wish I knew all of this, like, 1138 01:26:19,266 --> 01:26:29,289 37 years ago when I came into this industry, because it would make 1139 01:26:29,289 --> 01:26:38,541 me so much more confident that I'd made the right career choice, 1140 01:26:38,541 --> 01:26:48,564 but I didn't. But what I did observe, starting out as a government 1141 01:26:48,564 --> 01:26:57,816 bond trader, is that market moves up and down. And I thought 1142 01:26:57,816 --> 01:27:07,068 that was the most common sense to actually embrace a strategy that 1143 01:27:07,068 --> 01:27:15,549 would be agnostic to direction and just basically, you know, chase 1144 01:27:15,549 --> 01:27:22,488 direction rather than something that you hope for will 1145 01:27:22,488 --> 01:27:31,740 stay still but is clearly not when you look at the charts. So, 1146 01:26:10,555 --> 01:26:37,634 my very basic non quant brain came to a conclusion, but would have 1147 01:26:37,634 --> 01:27:00,547 certainly, how should I say, probably made the journey a little 1148 01:27:00,547 --> 01:27:21,377 bit more convincing, initially, had I known all of this 1149 01:27:21,377 --> 01:27:31,792 research that you've shared today. So, impressive work and I 1150 01:26:14,882 --> 01:26:31,548 know we're not even done, but I have a feeling people know where 1151 01:26:31,548 --> 01:26:48,214 we're going with this in terms of how to deal with this complex 1152 01:26:48,214 --> 01:27:03,598 adaptive system world. And so, I think the story only gets better. 1153 01:27:03,598 --> 01:27:22,828 And if we can get Dave to join you next time, or the time after, 1154 01:27:22,828 --> 01:27:31,802 we'll probably explore it in many directions. 1155 01:27:32,852 --> 01:27:33,732 Good discussion. 1156 01:27:34,212 --> 01:29:11,082 So, that would be cool. I mean Rich, we could have gone on for a 1157 01:27:46,792 --> 01:27:59,372 little while longer. I know you had prepared another section, 1158 01:27:59,372 --> 01:28:15,726 but I think we've done a pretty good job today and I really 1159 01:28:15,726 --> 01:28:32,080 do appreciate all the time and the preparation in all of this, especially 1160 01:28:32,080 --> 01:28:50,950 a week where I still feel a bit jet lagged from trip to HQ last 1161 01:28:50,950 --> 01:29:08,562 week. I don't know why but I wake up at the weirdest times at 1162 01:29:08,562 --> 01:29:11,078 the moment. So, appreciate all the prep work for this and if people 1163 01:27:53,935 --> 01:28:17,244 are listening and feel that they, as I do every time I speak 1164 01:28:17,244 --> 01:28:36,967 with Rich, learned something from this, please go to your favorite 1165 01:28:36,967 --> 01:28:58,483 podcast platform and leave a rating and review so that more people 1166 01:28:58,483 --> 01:29:11,034 can learn from the professor as well. Next week I actually have 1167 01:27:42,022 --> 01:28:00,766 another very smart person, as I have every week, I should say, 1168 01:28:00,766 --> 01:28:21,072 but Katy Kaminski will be joining me next week and we can have 1169 01:28:21,072 --> 01:28:39,816 a very special guest, someone who has not been on the podcast. 1170 01:28:39,816 --> 01:29:03,246 So, I'll keep the suspense for a week, but I think we will be tackling 1171 01:29:03,246 --> 01:29:11,056 some of his latest work. But also, if you want to have both of 1172 01:27:58,422 --> 01:28:28,012 them tackle some of your questions, you can, as always, email 1173 01:28:28,012 --> 01:28:38,772 them to info@toptradersunplugged.com and 1174 01:28:38,772 --> 01:29:11,052 I will do my best to bring them up in the conversation. So, 1175 01:27:36,902 --> 01:28:09,182 from Rich and me, thanks so much for listening. We look forward 1176 01:28:09,182 --> 01:28:44,152 to being back with you next week and until next time, as always, 1177 01:28:44,152 --> 01:29:11,052 take care of yourself and take care of each other. 1178 01:29:13,322 --> 01:30:46,387 Thanks for listening to Top Traders Unplugged. If you feel you 1179 01:29:18,562 --> 01:29:31,662 learned something of value from today's episode, the best way 1180 01:29:31,662 --> 01:29:48,692 to stay updated is to go on over to your favorite podcast platform 1181 01:29:48,692 --> 01:30:05,722 platform and follow the show so that you'll be sure to get all 1182 01:30:05,722 --> 01:30:20,132 the new episodes as they're released. We have some amazing guests 1183 01:30:20,132 --> 01:30:37,162 lined up for you and to ensure our show continues to grow, please 1184 01:30:37,162 --> 01:30:46,332 leave us an honest rating and review. It only takes a minute and 1185 01:29:34,796 --> 01:30:24,902 it's the best way to show us you love the podcast. We'll see you 1186 01:30:24,902 --> 01:30:46,376 next time on Top Traders Unplugged. This podcast expresses 1187 01:29:19,136 --> 01:29:44,330 the views of its hosts and the guests appearing on the podcast as 1188 01:29:44,330 --> 01:30:09,524 of the date of its recording, and such views are subject to change 1189 01:30:09,524 --> 01:30:28,904 without notice. Noticed Top Traders Unplugged do not have any 1190 01:30:28,904 --> 01:30:46,346 duty or obligation to update the information contained herein. 1191 01:29:13,322 --> 01:29:29,212 Furthermore, Top Traders Unplugged make no representation 1192 01:29:29,212 --> 01:29:56,452 to its accuracy and it shall not be assumed that past investment 1193 01:29:56,452 --> 01:30:16,882 performance is an indication of future results. Moreover, wherever 1194 01:30:16,882 --> 01:30:41,852 there is a potential for profit, there is also the possibility 1195 01:30:41,852 --> 01:30:46,392 of loss. This content is made available for educational purposes 1196 01:29:54,682 --> 01:30:46,382 only and should not be used for any other purpose. The information 1197 01:29:18,352 --> 01:29:46,017 contained in this podcast does not constitute and should not be 1198 01:29:46,017 --> 01:30:16,197 construed as investment advice or an offer to sell or a solicitation 1199 01:30:16,197 --> 01:30:38,832 hesitation to buy any securities or related financial instruments 1200 01:30:38,832 --> 01:30:46,377 in any jurisdiction. Certain information contained herein concerning 1201 01:29:25,567 --> 01:29:47,608 economic trends and performance is based on or derived 1202 01:29:47,608 --> 01:30:07,200 from information provided by independent third party sources. 1203 01:30:07,200 --> 01:30:31,690 Top Traders Unplugged may believe that the sources from which 1204 01:30:31,690 --> 01:30:46,384 such information are obtained are reliable. However, Top Traders 1205 01:29:17,753 --> 01:29:31,046 Unplugged cannot guarantee the accuracy of such information and 1206 01:29:31,046 --> 01:29:45,816 has not independently verified the accuracy or completeness of such 1207 01:29:45,816 --> 01:30:00,586 information or the assumptions on which such information is based 1208 01:30:00,586 --> 01:30:12,402 this podcast, including the information contained herein, may 1209 01:30:12,402 --> 01:30:27,172 not be reproduced, copied, republished or posted in whole or 1210 01:30:27,172 --> 01:30:44,896 in part in any form, without the prior written consent of top 1211 01:30:44,896 --> 01:30:46,373 traders.