Evolving Trends: How Quants Are Earning Alpha · Rishi Narang — Transcript
Full transcript
- 0:02Markets, speculation, and risk. This is
- 0:06the Chat with Traders podcast, hosted by
- 0:08Aaron Fifield.
- 0:11Hey traders, welcome. Thank you for
- 0:13joining me on episode 207.
- 0:17This week I had the pleasure of catching
- 0:19up with Rishi Narang. Longtime listeners
- 0:22of the show may recall I interviewed
- 0:24Rishi back in 2016
- 0:27on episode 54.
- 0:30By the way, Rishi's the founding
- 0:32principal of T2AM,
- 0:35a fund of funds investment manager with
- 0:37a sole focus of allocating to
- 0:39quantitative trading strategies. The
- 0:42fund was formed in 2005.
- 0:45The Rishi's time in markets extends more
- 0:47than 20 years as he co-founded
- 0:50distinguished quant shop Tradeworks in
- 0:531999.
- 0:55Rishi's also the author of an excellent
- 0:58book, Inside the Black Box, and an
- 1:01advisor to DARPA's financial markets
- 1:04vulnerabilities project.
- 1:07Yes, that is DARPA as in the government
- 1:09agency responsible for the robot dogs
- 1:12you've probably seen in those
- 1:14somewhat terrifying videos.
- 1:17So, the purpose of our catch-up was
- 1:20mostly to go over prominent trends he's
- 1:22seen develop in recent times specific to
- 1:25quant trading and investing.
- 1:28Without giving away too much, it
- 1:30involves talk of stat arb and index
- 1:33rebal strategies,
- 1:35an attraction to China's market,
- 1:38narrower trading models, use cases of
- 1:41alt data, and we wrap up discussing the
- 1:44rise of retail or DIY quants.
- 1:48I can appreciate these topics won't be
- 1:50of interest to everyone,
- 1:52but as I've said before, I think it is
- 1:54highly beneficial to have some insight
- 1:56on how the larger players are moving in
- 1:59markets and how they are earning alpha.
- 2:02Finally, please excuse the odd bird
- 2:05sound. I believe Rishi was wandering
- 2:07outside his home as we were talking.
- 2:10And after this first question, you
- 2:12actually won't hear from me for about 20
- 2:14minutes or so. My suggestion, think of
- 2:17this first part as more of a talk or a
- 2:20an interesting presentation from Rishi,
- 2:23and then an interview follows. Please
- 2:26enjoy. I present to you from Los
- 2:28Angeles, California, Rishi Narang.
- 2:34You know, speaking here with regard to
- 2:36research and strategy,
- 2:39what has stood out to you as notable
- 2:40trends in the field of quantitative
- 2:42trading and investing?
- 2:45Sure. Yeah, so there's there's a couple
- 2:47of few things to highlight. One One is
- 2:50the
- 2:51Well, I mean, there's a few things
- 2:52actually. So,
- 2:53one is that as the landscape has
- 2:55shifted, meaning
- 2:58you try to earn alpha in a context. It's
- 3:01not something that happens in a vacuum.
- 3:03And that context is sort of maybe best
- 3:06described as follows. Alpha comes from
- 3:10inefficiencies in asset prices
- 3:13that are themselves caused by either
- 3:17inefficiencies or relative indifferences
- 3:21on the part of various market
- 3:22participants.
- 3:24So, an old example would be that
- 3:27institutional traders, institutional
- 3:29investors were pretty sloppy in their
- 3:31handling of
- 3:33trading of single stocks.
- 3:35And so, they would do these large lumpy
- 3:37orders and and ignore market impact.
- 3:41And that basically created a whole game
- 3:44called stat arb, statistical arbitrage.
- 3:47And what would happen is there would be
- 3:49a price dislocation in some name.
- 3:53You would see that, but it's hard to
- 3:56know
- 3:57from the outside whether that came for
- 3:59good reason or not.
- 4:00And so, rather than just buy a name that
- 4:03that fell or short a name that went up a
- 4:05bunch,
- 4:07you would find something that reduces
- 4:09your risk.
- 4:10And so, you'd put on a hedge. So, you
- 4:13could imagine hedging with the index.
- 4:16But the problem with hedging with the
- 4:18index is there's a lot of basis risk.
- 4:20You know, if you're
- 4:21if you're buying Tesla and shorting the
- 4:23S&P 500, there's a lot of idiosyncratic
- 4:25risk in Tesla.
- 4:28Well, I maybe picked a really crappy
- 4:29example because I don't know who a good
- 4:31peer for Tesla is that's publicly listed
- 4:33today, but
- 4:34the point is that you would try to find
- 4:36something that reduces your risk. So, at
- 4:38least an automaker reduces your risk,
- 4:40right? So, maybe Ford or General Motors
- 4:43is a better hedge for Tesla
- 4:45than
- 4:46you know, the S&P 500 index. So,
- 4:50that game of
- 4:53finding related things and shorting the
- 4:56outperformer and buying the
- 4:57underperformer, which is also known as
- 4:59statistical arbitrage or relative mean
- 5:00reversion,
- 5:02existed in the '80s, '90s, early 2000s
- 5:05because of
- 5:07the sloppiness of a certain class of
- 5:08investors,
- 5:10mostly for that reason.
- 5:12Well, as those folks evolved,
- 5:15that inefficiency kind of went away.
- 5:17And it changed. So, relative mean
- 5:19reversion still exists, but it exists
- 5:21for a very different reason, and that
- 5:23reason also causes it to manifest in a
- 5:25different way. Which means if you're
- 5:26pursuing statistical arbitrage today,
- 5:29your model better not look just like it
- 5:31did
- 5:32in, say, 1997.
- 5:34Because otherwise, you're like not going
- 5:36to make any money and probably just lose
- 5:38money.
- 5:39There are a number of ways in which the
- 5:40landscape has shifted, really more than
- 5:41we have time to get into now, but you
- 5:43can just think about it for yourself how
- 5:45much the world has changed, say, in the
- 5:46last 15-20 years.
- 5:48There's been a gigantic move from
- 5:51passive to active,
- 5:52other way around. Wow. From active to
- 5:54passive, sorry.
- 5:56And a gigantic move from
- 6:00discretionary trading
- 6:03interaction with the markets to
- 6:06algorithmic interaction with the
- 6:07markets.
- 6:09At least within US equities, for
- 6:10example, almost all the volume is well,
- 6:13not almost all, a giant portion of the
- 6:16volume
- 6:17is now done market on close as opposed
- 6:20to throughout the day.
- 6:22And so, there's a bunch of these
- 6:23dynamics that
- 6:25are all intersecting with one another
- 6:27and changing the approach
- 6:30to making alpha.
- 6:32So, that's one thing that
- 6:35that we've seen that's changing the game
- 6:38around around us and
- 6:40forcing us to adapt and evolve.
- 6:43So, one example is
- 6:46the
- 6:47relative increase in prominence of index
- 6:50rebalance strategies. So, these are
- 6:52strategies that
- 6:53try to anticipate what names are going
- 6:55to be added to various indices.
- 6:58Why are we doing that? Well,
- 7:00or added and deleted, by the way. Why
- 7:03are we doing that? Well, we're doing
- 7:05that because there's
- 7:06a ton of money
- 7:08that gets invested on the back of what,
- 7:10say, Standard & Poor's puts in the S&P
- 7:12500.
- 7:14And that's true for MSCI indices and all
- 7:16kinds of other indices as well.
- 7:18So, all this passive money isn't exactly
- 7:21passive. It's just sort of sheep-like
- 7:24following
- 7:26the the the
- 7:29dictums.
- 7:30I guess that would be dicta
- 7:32of various index construction
- 7:34companies.
- 7:36And so, those index construction
- 7:38companies
- 7:39say, "Okay,
- 7:41we're including company X in the S&P
- 7:44500."
- 7:45And that announcement
- 7:47means that now all of the Vanguards and
- 7:49Fidelities and all of the other index
- 7:52managers and the the folks who
- 7:54administer the SPY ETF
- 7:57and so on, they all have to go out and
- 7:58buy company X in the right proportion.
- 8:02And it of course, with a fixed number of
- 8:05names type of index like the S&P 500, if
- 8:08a name is in, then a name is also out.
- 8:10And so, those folks then also have to
- 8:11sell
- 8:12that other name.
- 8:14So, that's interesting, too, right? So,
- 8:16you have something where there's going
- 8:17to be enormous buying pressure and
- 8:18enormous selling pressure.
- 8:20Pretty useful to know about that in
- 8:22advance if you could.
- 8:25You kind of can because a lot of them
- 8:27publish
- 8:28indicators of how or guide guidebooks or
- 8:32guidelines for how they think about
- 8:34adding and removing names for inclusion.
- 8:37So,
- 8:39that game
- 8:40really exists because of inefficiencies
- 8:42in the way passive folks invest. There's
- 8:44an indifference there.
- 8:46You know, if you're the S&P 500, you
- 8:49only care a little bit about the
- 8:50performance of the S&P 500. You're
- 8:52you know, certainly for an ETF manager,
- 8:55you don't really care.
- 8:56Cuz
- 8:58by definition, you get those stocks in
- 9:00those proportions and you just sit and
- 9:02hold them.
- 9:04Guess what return you're going to get?
- 9:05You're going to get the S&P 500 return.
- 9:07If you have adverse market impact,
- 9:10meh, it doesn't really matter. I mean,
- 9:12it goes into the price and whatever.
- 9:14So, there's not that much incentive to
- 9:16be that clever about
- 9:19managing market impact.
- 9:21I mean, there is to some extent, but
- 9:23it's limited.
- 9:24And so,
- 9:25there are a whole bunch of strategies
- 9:27that exist.
- 9:28Not a whole bunch, but there's a whole
- 9:30bunch of money being managed around this
- 9:32strategy of index rebalancing.
- 9:34And it's not just the S&P, of course.
- 9:36It's all these indices globally. So,
- 9:37that's one big example of how the world
- 9:39has changed is this active to passive
- 9:42thing. And by the way, even if you're a
- 9:43stat arb manager,
- 9:45you now kind of need to start accounting
- 9:47for index rebals in your price action.
- 9:51So, as an example,
- 9:53if your
- 9:54model is ignorant of the fact
- 9:57that company X has just been added and
- 9:59company Y has just been deleted from the
- 10:00index.
- 10:02And you're just watching price action
- 10:04relative to peers,
- 10:06then suddenly company X is soaring and
- 10:09company Y is reeling.
- 10:11And you may get in front of that
- 10:13mountain of flow that's coming your way.
- 10:16So, you kind of have to start to have
- 10:18almost a risk factor
- 10:20or a
- 10:22uh an adverse selection avoidance
- 10:24technique
- 10:25for index rebal. So, even if you're not
- 10:28trading it as an alpha, which you would
- 10:30do if you're doing index rebal, you
- 10:32still have to have index rebal
- 10:34if you're going to try to be
- 10:36good at the game.
- 10:38So, that's one thing. Another thing is
- 10:40uh like a big
- 10:42dispersion or um diaspora of
- 10:46the places and markets that quant
- 10:48strategies are are applied in. So, you
- 10:51know, Indian stat arb and emerging
- 10:53market stat arb and China stat arb China
- 10:56stat arb is like a huge business right
- 10:57now.
- 10:58Uh I don't even know the numbers
- 11:00offhand, but the amount of money going
- 11:02into both onshore money and offshore
- 11:05money going into quantitative
- 11:08equities and futures trading
- 11:10of mainland Chinese assets is just
- 11:12stupendous.
- 11:14Billions and billions and billions of
- 11:15dollars. Huge number.
- 11:17Um and the number of firms is is also
- 11:20pretty big. You know, there are a lot of
- 11:22folks who, for example,
- 11:24worked at US firms and
- 11:27then realized, well, there's this more
- 11:29or less frontier market where there's a
- 11:31lot more alpha, a lot more retail
- 11:32participation,
- 11:34and a lot more inefficiencies in China.
- 11:37It's also harder in a bunch of ways to
- 11:39trade offshore
- 11:41or if you're an offshore entity,
- 11:42especially up until quite recently.
- 11:45And so,
- 11:47maybe, you know, if you were someone who
- 11:48was here on H-1B visa in the US
- 11:52and worked for Two Sigma or D. Shaw or
- 11:56whatever quant firm,
- 11:58maybe just go home and set up a shop
- 12:01there
- 12:02uh in China and have a bunch of local
- 12:04advantages over the US and and UK firms
- 12:08that are trying to play that game.
- 12:10So, that's definitely been happening. Um
- 12:12is Oh, and and the application of quant
- 12:14techniques into all kinds of other
- 12:15areas, by the way.
- 12:17Venture capital,
- 12:18uh crypto,
- 12:20credit, peer-to-peer lending,
- 12:23um municipal bond market making. I mean,
- 12:26all kinds of seen so many interesting
- 12:28things that are not just
- 12:30basic equity stat arb in the developed
- 12:32markets or basic futures strategies in
- 12:35the developed markets.
- 12:37And I guess the third big trend
- 12:40is maybe the most interesting,
- 12:43which is the
- 12:46there's been a shift in the balance
- 12:48between taking over-fitting risk and
- 12:50under-fitting risk.
- 12:52And just philosophically, I think it's
- 12:54really interesting.
- 12:56In the old days, well, pretty dirty
- 13:00insult for a quant was that they've
- 13:01overfit.
- 13:03But now we're seeing folks do a variety
- 13:06of things
- 13:07that
- 13:09their actions are telling you that
- 13:11they're at least as afraid of
- 13:12under-fitting as of over-fitting.
- 13:15That looks like a couple different
- 13:16things. So, first of all, in the old
- 13:19days, for the purposes of largely
- 13:20cross-validation and statistical
- 13:22significance,
- 13:23models were typically universal, meaning
- 13:26for some universe of many securities,
- 13:30basically without exclusion,
- 13:32uh
- 13:33this model applies to all of them. So,
- 13:35if there are
- 13:37uh all US stocks, but then
- 13:39realistically, you can only trade the
- 13:40top, say, 3,000 of them if if that's
- 13:43what you determine from a liquidity
- 13:44standpoint,
- 13:46then you have one model with one set of
- 13:47parameters,
- 13:49one set of factors,
- 13:51uh and one set of weights for all 3,000
- 13:54stocks, meaning the drivers of your
- 13:56forecast for Walmart uh are the same as
- 14:00for Pfizer, are the same as for Amazon,
- 14:03are the same as for Google and Facebook,
- 14:06and Tesla. All those companies are
- 14:09forecasted with the exact same set of
- 14:11factors, the same weights, the same
- 14:12parameters.
- 14:15Um we're now seeing industry-specific
- 14:18and even company-specific models, or if
- 14:20you're talking about futures, individual
- 14:22market models.
- 14:24And what you're seeing with those is
- 14:28a decision that your prior
- 14:32might be a lot stronger on a single
- 14:34company forecast forecasting model.
- 14:37Like, I can know what the KPRs KPIs are
- 14:40for some specific company.
- 14:43But um
- 14:45I might not have a lot of data history
- 14:48to prove that out in a backtest, but I'm
- 14:50willing to trust my prior more
- 14:53uh and not worry as much about having
- 14:55statistical significance.
- 14:58Uh another easy example would be
- 15:00something like inventory data. So, you
- 15:02know, on US companies' quarterly
- 15:05reports, there's a standard line item
- 15:07for inventory. Well,
- 15:09if you're a bank, what is inventory?
- 15:12If you're a
- 15:14uh you know, an
- 15:16an audit and consulting firm like Price
- 15:18Waterhouse Coopers or Ernst & Young or
- 15:20whatever,
- 15:21what what is that? You know?
- 15:25Uh what is inventory? But if you're a
- 15:28a manufacturer of television sets, well,
- 15:30now we know what inventory is, right?
- 15:33So, but that is important information
- 15:35for the latter kind of company and not
- 15:36for the former.
- 15:38And so,
- 15:39you don't want to throw out that data
- 15:41potentially.
- 15:42Uh but that does mean you have a
- 15:44different model for those industries
- 15:46where
- 15:48inventory is a thing and a different
- 15:50model for those sectors and industries
- 15:52where it's not.
- 15:53So, that's one expression of this shift
- 15:57in the balance between over- and
- 15:58under-fitting risk.
- 16:00The second is
- 16:02the increased use of alternative data
- 16:05sources.
- 16:06And this is related to the first thing,
- 16:10which we'll call narrower models as
- 16:12opposed to more broad models.
- 16:14Um models are applied in a applied in a
- 16:16more narrow set of um of securities. For
- 16:20the second thing,
- 16:21um this is now models that are driven by
- 16:24data that are only available for certain
- 16:27kinds of companies. So, a great example
- 16:29is uh is credit card data. So, I'm
- 16:33guessing that your listeners will know
- 16:35that their credit card spending is being
- 16:39tracked by various companies,
- 16:42and whether individualized or
- 16:45aggregated, that data is being sold
- 16:47to hedge funds,
- 16:50uh who use it mostly quantitatively, but
- 16:53not always, uh to make decisions about
- 16:56what to buy and sell. And basically,
- 16:58what they're doing is
- 17:00looking for a faster or more accurate
- 17:04estimation of, say, revenues for
- 17:07uh certain kinds of companies. Well, if
- 17:09it's credit card data,
- 17:11it's interesting cuz it's you can't buy
- 17:13everything and you don't buy everything
- 17:14on credit cards. And in any case,
- 17:16there's only a subset of things
- 17:19that go into
- 17:22uh the category of like
- 17:25things that humans, like regular people,
- 17:27buy. So, as an example, it doesn't tell
- 17:29you anything about uh how Boeing is
- 17:32doing on its aerospace sales. Like,
- 17:34credit card data does not apply to
- 17:36Boeing.
- 17:37Probably doesn't apply much to auto
- 17:39manufacturers, either, cuz most people
- 17:41are not buying cars on their credit
- 17:43card.
- 17:44They're either sending a wire or writing
- 17:46a check or writing a check and then
- 17:48having a financing line.
- 17:50And that's a different thing.
- 17:52It tells you a lot about what they're
- 17:54doing with Amazon
- 17:56and what they're doing with Macy's and
- 17:59whatever else. Like, so it's really
- 18:01useful for retail and consumer and
- 18:03hospitality-type companies.
- 18:05Um and a few other categories, but, you
- 18:08know, credit card data is really
- 18:09interesting for
- 18:11a few hundred companies in the US.
- 18:14Not very interesting for the other
- 18:17two and a half, three and a half
- 18:18thousand easily tradable companies.
- 18:22So, that's the second thing is this sort
- 18:24of trend towards alternative data,
- 18:26which is very much hand in hand with the
- 18:30trend towards narrower models.
- 18:33The third thing is
- 18:36the increased use of machine learning
- 18:39or artificial intelligence uh
- 18:41strategies. So, that's the other really
- 18:43big piece in all of this.
- 18:46So, uh you know, these things, by
- 18:48definition,
- 18:49are looking at data
- 18:52to tell you how what the forecasting
- 18:56variables should be and how much they
- 18:57should matter.
- 18:59Uh and we can talk about the different
- 19:01use cases of ML and AI if you want, but
- 19:03um there are some firms
- 19:06that claim to be making forecasts
- 19:07without priors, without saying, "Oh, I
- 19:09have this economic model of how
- 19:10companies'
- 19:11asset prices work, what drives a stock
- 19:14to go up, what drives a stock to go
- 19:15down.
- 19:17I'm just going to let the machine figure
- 19:18it out."
- 19:19And that's inherently a a more fitting
- 19:23kind of an exercise.
- 19:25So, here you've
- 19:26obviously outlined three different
- 19:31trends that you've that you've noticed
- 19:33uh in space of quant trading and
- 19:35investing. Going right back to the
- 19:37start, and I just have a few questions
- 19:39about each of these.
- 19:41You spoke about knowing in advance
- 19:44when a symbol may be added to an index.
- 19:49Can you maybe go into that a little bit
- 19:50more about how you may know in advance.
- 19:54Sure. Yeah, so
- 19:56in many cases and this will just be one
- 19:58example for the sake of time.
- 20:00In many cases most of the determination
- 20:03is simply a market capitalization based
- 20:05cutoff.
- 20:06So, we're going to take like the Russell
- 20:093000 Russell 2000.
- 20:11There's a certain ranking of market caps
- 20:14that happens.
- 20:16And if it's
- 20:17company 1001
- 20:20through 3000
- 20:22in terms of the ranking by market cap
- 20:26kind of presumably
- 20:28that company is supposed to be in the
- 20:31Russell 2000.
- 20:33And they have certain dates that they
- 20:35announce these um inclusions and
- 20:37exclusions. They announce the new
- 20:39composition of the index and then dates
- 20:42at which those changes are made
- 20:44effective. And there's some time in
- 20:46between there. So, there's a range of
- 20:48strategies that you can do. Similar to
- 20:51the way that people treat merger arb,
- 20:52you can
- 20:54look at like the history of all mergers
- 20:56and say, "Well, what kinds of companies
- 20:59get bought?"
- 21:00And then I'm going to go out and screen
- 21:02the world for those kinds of companies.
- 21:04And maybe buy some or all of them.
- 21:07And
- 21:09then some of them are going to be
- 21:11targets for for takeovers.
- 21:13So, there you're trying to get an ahead
- 21:15of the
- 21:17uh the acquirer. Here you're trying to
- 21:19get ahead of the market cap
- 21:22um based index
- 21:24solution or index provider. And then all
- 21:27of the people and firms that have to
- 21:30follow suit and either bring in the new
- 21:32names or get rid of the old ones.
- 21:34It's not exactly front running, but all
- 21:36alpha is in a sense probabilistic front
- 21:39running. You're trying to know before it
- 21:41happens
- 21:42that other people are going to buy this
- 21:44stock in aggregate or sell this stock or
- 21:46this asset
- 21:48in in advance. Um
- 21:50And you're doing that with a bunch of
- 21:51hard work and not with any kind of
- 21:53actual knowledge of their trades.
- 21:55But you do have to get there before they
- 21:57do. Otherwise, it's not alpha. Then it's
- 22:01you know, something else that doesn't
- 22:03work, you know.
- 22:04I wonder what the name for that is.
- 22:06Doing uh doing all the right trades, but
- 22:09doing them after the fact. I mean, I
- 22:11guess that's trend following, but
- 22:14This information, I mean, is that
- 22:16publicly available to everybody? Yeah.
- 22:19The information about what's going to be
- 22:21included into an index? Some of it is.
- 22:23So, as an example, um the index
- 22:26providers have gotten smart about as as
- 22:29all capitalist companies generally are,
- 22:31but how to extract money for everything.
- 22:34So, just as exchanges are now charging a
- 22:36huge amount for data
- 22:38that
- 22:39you know, is a major part of their
- 22:41revenue streams
- 22:42even though they're really not in the
- 22:44business of selling data, they're in the
- 22:45business of
- 22:47processing transactions of securities
- 22:50and providing a safe environment for the
- 22:51transaction of securities. And so, they
- 22:53charge fees for that, too. But then they
- 22:56they know you have to have that data.
- 22:58If you're like a high frequency trader,
- 23:00you have to have the data of every
- 23:01single exchange
- 23:03for various reasons in the US. And so,
- 23:07um
- 23:08those exchanges have you know, extracted
- 23:10a toll for for that data. Well, just so
- 23:12the index providers are
- 23:14charging
- 23:15for index composition data.
- 23:17And they are
- 23:20and some of it's not cheap. I mean, to
- 23:22get like Russell, S&P, MSCI
- 23:26that is some hundreds of thousands of
- 23:28dollars a year.
- 23:30Just for those three and not even for
- 23:32all regions.
- 23:35Wow, really? Mhm. That much?
- 23:37anyone like it's not like you have to
- 23:39know somebody who knows somebody to get
- 23:40the data, but you do have to have
- 23:42hundreds of thousands of dollars of data
- 23:43of budget for the data. Yeah.
- 23:46That's true for a lot of stuff, man.
- 23:48And this you're talking about the data
- 23:50around index rebalancing. Specifically
- 23:53that, yes. And all the No, no, just to
- 23:55be clear, this is not index rebal data.
- 23:57This is index composition data. This is
- 23:59to tell you just what is in the S&P 500
- 24:02right now. What tickers in at what
- 24:04weights. That's it. That's all that
- 24:06hundreds of thousands of dollars gets
- 24:07you.
- 24:08But you can imagine if you're doing
- 24:10index rebal,
- 24:11it's kind of an important ingredient in
- 24:14in doing the trade right.
- 24:16So, what what are some of the trades you
- 24:18see around this? Like is it I mean, I
- 24:22presume that they're probably a little
- 24:23bit more sophisticated than just buying
- 24:25a symbol which you think is going to be
- 24:28uh added into an index.
- 24:31Whoa. Why would you presume that? It's
- 24:33That's it. That's the trade.
- 24:36Now, the sophistication may be in why do
- 24:39you think that it's going to be added?
- 24:41But um and there are a range of
- 24:43strategies around that. So, there are
- 24:45folks who traffic in
- 24:47like the lower probability names in
- 24:49index rebal versus those that traffic in
- 24:52the names that have already been
- 24:53announced, but it hasn't completely been
- 24:56um implemented yet.
- 24:59So, there's like a range of certainty
- 25:01about whether or not this name is going
- 25:03to be included. And as a result of that
- 25:06range of certainty, there are different
- 25:08players with different risk appetites
- 25:10and different models. So, let's say for
- 25:12example,
- 25:13that you're talking about the Russell
- 25:151000, which if I if memory serves and if
- 25:18my knowledge is correct, is simply the
- 25:201000 largest market caps in the US
- 25:23equity market.
- 25:25So, now let's say that um there's a
- 25:27rebal coming up, an announcement coming
- 25:29up about the rebal.
- 25:31Uh and when and when I say coming up,
- 25:33like how far in advance you trade is an
- 25:35open question here. So, but let's just
- 25:37pretend that the answer is 3 months
- 25:39before the rebal, I'm going to start
- 25:41putting on positions.
- 25:43Okay?
- 25:44So, it's uh I think their rebals are
- 25:45announced in March, but I don't
- 25:47remember. Could be November. I have no
- 25:49idea. Whatever. I don't remember the
- 25:50frequency. But let's say it's March. So,
- 25:52I'm going to start putting on my
- 25:53positions in December, January.
- 25:56At that time,
- 25:57there's a bunch of companies
- 26:00that are say ranked from 900 or 800 to
- 26:061200 in market cap, which over the
- 26:10course of 3 months
- 26:13might move up or down enough in market
- 26:15cap to either get them excluded from the
- 26:17top 1000 or included in the top 1000.
- 26:19So, this is interesting cuz this isn't
- 26:21just about is this company going to be
- 26:22good?
- 26:23It's is it going to be good at It's is
- 26:25it going to be better or worse
- 26:27than the companies around it in this
- 26:29specific weird peer group of companies
- 26:31that may have nothing to do with each
- 26:32other otherwise.
- 26:34But they're all not really competing,
- 26:35but you can think of them as competing
- 26:37like horse racing
- 26:39to be in that top 1000 so that they get
- 26:41included. Now again, maybe the companies
- 26:43don't care if they're included in the
- 26:44Russell 1000 or not, but their stock
- 26:46prices care.
- 26:47Now let's say that I'm really good at
- 26:49forecasting 3 months price outcomes for
- 26:51companies. Well, I better use that
- 26:53information. And then I'll
- 26:57have a little bit more conviction to buy
- 26:59bigger
- 27:01those names that I'm going to hold for 3
- 27:03months-ish and are likely to be included
- 27:06cuz then when that announcement comes,
- 27:07there's a big pop.
- 27:09And as the certainty builds, there's a
- 27:11big pop. So, there's this kind of weird
- 27:14short-term momentum thing that happens
- 27:16in those in that category of names that
- 27:19has nothing to do with say the top 10
- 27:21companies in the Russell 1000 cuz
- 27:23there's almost a zero probability
- 27:25that in 3 months those companies are uh
- 27:29are going to drop out of the index.
- 27:31Right? So, it's a little bit like
- 27:33options theory. If you're the number one
- 27:35ranked market cap in the US or in any
- 27:38market,
- 27:39and you know, you're far and away the
- 27:41biggest, which happens, it would be like
- 27:45um what is the value of a deep
- 27:48out of the money put
- 27:49on that company
- 27:51with a 3 month expiry. The value of that
- 27:53is like very close to zero.
- 27:55It's basically zero.
- 27:59If you're talking about something where
- 28:01it's a 3 month expiry for very near the
- 28:04money options,
- 28:05well, now that's fun, right? And if you
- 28:08have information that gives you a clue,
- 28:10then you should use that information.
- 28:13Does that make sense?
- 28:15It does make sense, yeah.
- 28:17And what's more, there's a
- 28:20there's an asymmetry here.
- 28:23If you're company number 999,
- 28:26the last time or 1000, the last time the
- 28:29Russell 1000 was constituted,
- 28:32and you're still kind of 999, 1000,
- 28:36there's a chance you get excluded, which
- 28:38means there's a whole ton of money that
- 28:39has to come out of you.
- 28:42If you're company 1001,
- 28:46and you remain company 1001, there's not
- 28:48a lot of downside for that price because
- 28:51it's already excluded and it's still
- 28:52excluded.
- 28:54So, for those companies that are outside
- 28:56the index looking in,
- 28:58there is more upside than downside. And
- 29:00for those companies that are barely in
- 29:02the index, there's more downside than
- 29:03upside. So, it's just very interesting
- 29:05that dynamic and what it means when
- 29:08you're building a model.
- 29:11So, ultimately you're trying to
- 29:13buy early enough predicting that a
- 29:16company may be added to an index um in
- 29:19order to essentially
- 29:21front run all the flow which is going to
- 29:22be coming into that stock over the
- 29:24coming weeks and months.
- 29:26Yeah, and again, this word front run has
- 29:28like actual legal meaning and we don't
- 29:31mean it that way, but yes.
- 29:32Of course.
- 29:33it that. You want to get in front of it.
- 29:35Which is a lot like front running, but
- 29:37but the legal definition of front
- 29:38running is about, you know, having
- 29:39knowledge of someone like actual
- 29:41knowledge of someone's order and then
- 29:42going in front of it. Like if you call
- 29:44your broker and say, "I I to buy a
- 29:45billion shares of IBM.
- 29:47And then your broker's like, "Cool. Let
- 29:50me go buy a
- 29:51million shares of IBM right before that,
- 29:54and then I'll do your order." Because
- 29:56that way I'll profit from all your
- 29:57market impact. Let's just say you're
- 29:59you're trying to be the first to act.
- 30:01Yeah, or or yeah, certainly not the
- 30:03last. You're trying to be early to act.
- 30:05Early, yeah.
- 30:07Yeah.
- 30:07The second point you made was about
- 30:10China, which I thought was quite
- 30:12interesting. You said there's a lot of
- 30:13startup opportunities there right now.
- 30:17Um how come that is the case?
- 30:20It's a combination of factors. So,
- 30:22there's onshore money
- 30:24and offshore money. There's onshore
- 30:26firms and offshore firms. Like you can't
- 30:30have onshore money
- 30:31if you're an offshore firm.
- 30:33That's like not a thing.
- 30:35Uh you can have offshore money if you're
- 30:38an onshore firm. So, there's a supply
- 30:41thing and a demand thing that's
- 30:42happening. From the supply side, the
- 30:45ability to
- 30:47trade
- 30:49China as an offshore investor, that's
- 30:50recently been a little bit um that's
- 30:53been made a little bit easier recently.
- 30:55And so that's that's one thing, right?
- 30:57So, if you're like Two Sigma or PDT or
- 31:01some other US or whatever non-Chinese
- 31:03firm, you can now access China
- 31:06relatively easily actually. There's a
- 31:08bunch of like international brokers like
- 31:11Morgan Stanley and UBS and so on that
- 31:14can give you access to that market uh
- 31:16simply very simply. Um and there are
- 31:19rules around shorting and in shorting
- 31:21individual names and shorting the index
- 31:23and so on. So, you can even be market
- 31:26neutral.
- 31:28If I understand correctly, and I haven't
- 31:29studied this too too much, but um if I
- 31:32understand correctly, if you're an
- 31:33onshore
- 31:35firm managing onshore money,
- 31:37uh you cannot short
- 31:40individual names.
- 31:42I'm not even sure if you can short the
- 31:44index, but maybe you can.
- 31:46But that's not Chinese money.
- 31:48Um that's been going on a little bit of
- 31:50time,
- 31:51but there's also a different supply
- 31:53dynamic there,
- 31:55which is that you
- 31:57uh as I said, have a lot of new supply
- 32:00of quants who used to work at
- 32:04non-Chinese firms
- 32:06and have decided, "Let's set up shop in
- 32:08China, where we live, where there's a a
- 32:10less efficient market."
- 32:12That second thing, where there's a less
- 32:14efficient market, that statement
- 32:16is what's driving the demand side of
- 32:19this. Which is to say, again, as some of
- 32:22the more obvious and straightforward
- 32:25markets to access
- 32:26and bigger markets have filled up. If
- 32:28you look at China, it's a really big
- 32:30market. It's a really big economy, only
- 32:32set to get bigger.
- 32:34And, you know, there's a rush to get in
- 32:37there before it gets too crowded and be
- 32:40early and gather data and all those
- 32:42things that give you an early mover
- 32:44advantage.
- 32:45Um access to talent, access to data,
- 32:48storing your own data is super valuable,
- 32:51but this you know, you can't do that
- 32:52until you start doing it. So, like go
- 32:54start doing it. So, that there's been
- 32:56like a lot of that.
- 32:58Have you allocated to any managers who
- 33:01uh trading Chinese markets, like in
- 33:04order for for your fund to have exposure
- 33:06there?
- 33:07Yeah, we have offshore two offshore
- 33:10Chinese well,
- 33:12money with two firms that are both based
- 33:13in mainland China, but are offering
- 33:16offshore strategies to offshore
- 33:18investors.
- 33:19And yeah, we've we've picked a couple of
- 33:21them and we've looked at dozens.
- 33:23Okay.
- 33:24Is that just a recent thing?
- 33:27Pretty recent. Um
- 33:29both this year. Uh
- 33:31I want to say the first one
- 33:33went live mid-year, but we've probably
- 33:34you know, that means we've been talking
- 33:35to them for a little while. Yeah, okay.
- 33:39You also brought up the point about uh I
- 33:41might get this slightly wrong, but it
- 33:43was something to the extent of things
- 33:46which are used in quant being applied to
- 33:50other areas outside of financial
- 33:52markets. I know you mentioned venture
- 33:54capital, and there were a couple other
- 33:55things as well. Could you speak to that
- 33:58a little bit more?
- 33:59Yeah, and it's again the same drive for
- 34:01diversification and and access to
- 34:04markets that are a little weirder and
- 34:06more complicated to to So, there's some
- 34:09more barriers to entry. And if you think
- 34:11about this like just from a basic
- 34:12economic principle standpoint, it's the
- 34:13same thing as saying, "Well, there are
- 34:15barriers to entry to making this kind of
- 34:17widget,
- 34:18but if I happen to have some edge in
- 34:19making that widget, then I should
- 34:21probably go do that because I'm going to
- 34:22enjoy a competitive advantage for a
- 34:25while."
- 34:26So, and being early is its own
- 34:28competitive advantage because that means
- 34:29I have a head start on others.
- 34:32But you know, there's something you have
- 34:33to risk. You have to put up a bunch of
- 34:34upfront costs to get there. So, credit
- 34:37um
- 34:38you know, this is like debt issued by
- 34:40companies that
- 34:41uh that are So, it's not government
- 34:42debt, it's usually uh you know, it's
- 34:45it's it's debt from private companies.
- 34:48That debt trades.
- 34:50So, it's not just that the company
- 34:52issues
- 34:53debt and you decide, "I will lend to
- 34:55this company."
- 34:56You can buy and sell that debt. And
- 34:58there is a market for it.
- 35:00It's in the US
- 35:02corporate credit market, not
- 35:04um fully exchange uh traded, and in
- 35:07fact, it's mostly not.
- 35:09Um but, you know, some of the credit
- 35:11default swaps are exchange-traded, and
- 35:13it's a changing market, an evolving
- 35:14marketplace.
- 35:16Um
- 35:17it it has very different challenges, and
- 35:19one of the most obvious is is data.
- 35:22Um like it's not that complicated to
- 35:24think about what would cause a bond to
- 35:26be more attractive relatively and what
- 35:28would cause it to be less attractive.
- 35:30What's complicated is asking
- 35:33like,
- 35:34"How do I even track the history of
- 35:37bonds?" So, unlike stocks, bonds expire.
- 35:40And on top of that, they're issued at
- 35:41various coupons at various times and
- 35:44various durations. And so,
- 35:46managing that data set, it's like it's
- 35:48just a different problem
- 35:50from uh managing normal equity data set
- 35:53or even futures data set. You know,
- 35:55futures expire too, but there's far
- 35:56fewer of them
- 35:58than there are companies and credits.
- 36:01So, yeah, there's like this crazy data
- 36:03wrangling problem. There's systematic
- 36:06strategies
- 36:07uh in peer-to-peer lending. So, this is
- 36:09like Lending Club and Lending Tree,
- 36:11where if you're just some random person
- 36:14and you want to borrow $5,000 for
- 36:16whatever random reason,
- 36:18you can apply for a loan,
- 36:21and other random people who have 5,000
- 36:24extra dollars sitting around can lend
- 36:25that to you and earn on it like their
- 36:28local savings account, but then they're
- 36:29taking credit risk on you as an
- 36:32individual, the random borrower. And so,
- 36:35um there are
- 36:36now quant strategies out there
- 36:39that are pouring over credit data and
- 36:41pouring over characteristics of
- 36:43borrowers
- 36:44and saying, "Well,
- 36:46this borrower might not have the
- 36:47greatest credit score." I don't know if
- 36:48you guys have the equivalent in
- 36:50Australia or elsewhere, wherever
- 36:52listeners may be, but in the US, you
- 36:54know, you have a credit score that a
- 36:56credit agency uh has a list of factors
- 36:59that
- 37:00uh describe your creditworthiness. And,
- 37:04you know, like a in a very popular such
- 37:06scheme, a really high credit score, I
- 37:08think the maximum is 850.
- 37:10850. I don't know why it's not a
- 37:12thousand or why it's not a hundred, but
- 37:14it's 850.
- 37:15So, that's what it is.
- 37:16And um you know, you're considered as
- 37:19having good credit, I think, if your
- 37:21credit's above 750, let's call it. And
- 37:24so, you know, if you're a an 820 credit
- 37:27score person, it's very easy to get
- 37:29money fairly cheaply. If you're a 600
- 37:31credit score person, it's a lot less
- 37:33easy. But it might be that that's just
- 37:35because of something stupid. Because
- 37:37these credit these credit ratings are
- 37:39not like
- 37:41God telling you this is a good lender,
- 37:43or good borrower, I mean. This is like a
- 37:45a flawed universally applied model
- 37:47telling you that. Maybe some random
- 37:49circumstance happened three years ago
- 37:51beat up your credit score, but it turns
- 37:53out you're actually a really good
- 37:54borrower who pays everything back.
- 37:56So, they look for those kinds of people
- 37:59to try to find the ones that
- 38:01are better borrowers.
- 38:04And but because their credit score
- 38:06sucks, they know they're going to have a
- 38:08higher interest rate.
- 38:10Does that make sense? You try to find
- 38:11like these good value loans to make,
- 38:14where you get a high yield, but the risk
- 38:16isn't that bad.
- 38:18It's not not in line how high the yield
- 38:20is.
- 38:21Um so, yeah, there's systematic
- 38:22strategies for that, and you can tell
- 38:23there's a whole different set of data
- 38:25there and a whole different kind of
- 38:26problem there than there is when you're
- 38:28trying to forecast the price of IBM.
- 38:32Gotcha. That's really quite fascinating.
- 38:37And some of this stuff isn't new. Some
- 38:39of it's been going on for a while.
- 38:41Um you know, systematic credit trading's
- 38:43been going on
- 38:4510, 15 years now.
- 38:47But at the same time, the prominence of
- 38:49it is much higher. Like there's at least
- 38:51two
- 38:52stand-alone quant funds
- 38:54that are doing just that.
- 38:57And then the magnitude of that activity
- 38:59within some of the larger multi-strats
- 39:01is much larger than it used to be. And
- 39:03it's the same for quant trading in
- 39:05China. Like that's been happening for a
- 39:06minute, but it's just it's now like on
- 39:09the map, you know.
- 39:11So, the third point
- 39:14uh I wanted to ask you a question on
- 39:16that uh with regard to underfitting and
- 39:19overfitting a model. Do you have any
- 39:21tips for how to get the balance right?
- 39:25Oof, that's actually quite legitimately
- 39:28really hard. I mean, there's so many
- 39:29factors.
- 39:31And
- 39:32the reason that this change is happening
- 39:35is once again both from a supply
- 39:38perspective and from a demand
- 39:40perspective. And I know I'm mapping
- 39:42things to supply and demand. This isn't
- 39:44like a clean mapping, but I'll explain
- 39:46what I mean. So, on the supply side,
- 39:48we have an increasing number of people
- 39:51who are
- 39:52in jobs in quant firms
- 39:55that have training in machine learning
- 39:56and artificial intelligence. And some of
- 39:58those folks used to work at places like
- 40:01Facebook, and they're just sick of
- 40:03selling advertisements to people.
- 40:05Right? Like they thought, "Oh, cool. I
- 40:07get to work for this really cool
- 40:08company.
- 40:09It's a tech company. I'm not selling my
- 40:11soul. Look, it's social media. I'm like
- 40:13connecting families and friends and
- 40:15former colleagues and alumni."
- 40:18But then it turns out what what I'm
- 40:19really doing is I'm selling people ads.
- 40:21And that gets at some point maybe
- 40:22grating on your soul.
- 40:24And so, you quit that job and you go
- 40:25work for a different soul-crushing place
- 40:27that's just trying to make rich people
- 40:28richer. Cool.
- 40:30Nonetheless,
- 40:31um that is one
- 40:34um
- 40:35source of increased supply. The other is
- 40:37the field itself has advanced. I mean,
- 40:39the problem of forecasting asset prices
- 40:43is a really hard problem. Uh this These
- 40:46are incredibly noisy processes.
- 40:49Uh very close to random walks.
- 40:52Uh your out-of-sample R squareds, for
- 40:54those of you who know what that that is,
- 40:55are like low single digits.
- 40:58So, like 0.03 is a successful R squared.
- 41:02And, you know, anyone in any other sign
- 41:04most other scientific fields would look
- 41:06at that as a forecast out-of-sample R
- 41:08squared and go,
- 41:10"You're basically noise. Like you don't
- 41:12have anything better than a a coin toss
- 41:14here.
- 41:15So, why are we doing this?"
- 41:17But, that's a that's a successful That's
- 41:20a successful outcome in my world, right?
- 41:22So,
- 41:24when you have that noisy of a process,
- 41:26the techniques to
- 41:28uh and by the way, very limited data.
- 41:30Um
- 41:31you know, if you're doing self-driving
- 41:33cars, you can basically generate more
- 41:34data just by So, like let's say they
- 41:37have a self-driving car and, you know,
- 41:39you find that it's not performing well
- 41:41in rain.
- 41:42Well, let's go to Portland and Seattle
- 41:44for a while
- 41:45and train more cars there. Like you
- 41:47could just get more data.
- 41:49For the capital markets, you can't get
- 41:51more data. The data are non-stationary,
- 41:54and there just isn't that much of it.
- 41:56You know, if you compare like the number
- 41:57of Facebook posts, the number of
- 41:59Facebook users to like the total amount
- 42:02of data about US equities.
- 42:05And I mean, like the daily data in the
- 42:07one versus the total cumulative data in
- 42:10the other. It one like they're just
- 42:12orders of magnitude different, you know?
- 42:15Uh
- 42:16we're talking about billions of data
- 42:17points per day, if not more,
- 42:19and on in the kind of social media
- 42:22world, and we're talking about maybe
- 42:26billions of data points in the whole
- 42:27cumulative history of everything in US
- 42:29equities.
- 42:31And so,
- 42:33you know, you can see that that's Those
- 42:35are just different problems.
- 42:37There are other differences, too, but
- 42:38just for the sake of time, let's just
- 42:40acknowledge that these are very
- 42:41different problems.
- 42:42You may need different and better
- 42:44techniques and more powerful computing
- 42:46clusters to be able to tackle these
- 42:50differing challenges.
- 42:51And so, that's the other aspect of the
- 42:53supply side.
- 42:55But again, the demand side is there.
- 42:57So,
- 43:00as conventional old-school techniques
- 43:01have become
- 43:03more heavily competed, as the
- 43:05philosophies and concepts and ideas and
- 43:09theories that under pin those strategies
- 43:11have proliferated and been leaked out
- 43:13into
- 43:15competitive firms by virtue of employees
- 43:17changing jobs or leaked out into the
- 43:19academic journals and textbooks and so
- 43:22on, you know, those old ideas are harder
- 43:24and harder to make money on. And so,
- 43:27just as some people are deciding, "Okay,
- 43:29well, the hell with it. I'm going to go
- 43:30venture far afield and and trade China,
- 43:33even though it's a pain in the neck."
- 43:35Uh
- 43:36maybe
- 43:37uh some of them are saying, "Well, I'm
- 43:39going to venture further afield in my
- 43:40techniques."
- 43:42I happen to be really good at this
- 43:43machine learning thing, and so I'm going
- 43:45to do that, where it's harder to compete
- 43:47with me.
- 43:48So, there's that happening, too.
- 43:50That there's a desire
- 43:52to take
- 43:54more fitting risk because the
- 43:56lower-hanging fruit, which doesn't have
- 43:57fitting risk, you know, it's not easy
- 43:59making money that way anymore.
- 44:02I feel as though a lot of what we've
- 44:03been talking about here is very much
- 44:06applicable to, you know, the large quant
- 44:09funds, uh and very much the professional
- 44:11space.
- 44:12I'd like to ask you,
- 44:14what have you seen in the trend among
- 44:17DIY quants? So, the at-home retailer who
- 44:22has, you know, might have some machine
- 44:24learning skills, you know, has learned
- 44:26how to program, has wrote some sort of
- 44:28systematic strategies, cuz I think when
- 44:31we last spoke in
- 44:32uh 2016, you know, the past 4 years has
- 44:35been a big rise
- 44:37uh in this space.
- 44:39Is there anything you've been seeing
- 44:40here which is particularly interesting?
- 44:45Yeah,
- 44:46uh for sure. So, there's a few dynamics
- 44:48there. It's a good question. And my
- 44:50perspective on retail is mostly US and
- 44:53then like random smatterings Asia. I
- 44:56don't actually, weirdly, have any
- 44:57impression at all of
- 44:59Australia or Europe as far as retail
- 45:01investing, but maybe you can tell me and
- 45:03and we can reflect on it a little bit.
- 45:04But, as far as
- 45:07the US goes,
- 45:08um retail investors were a steady kind
- 45:12of part of that market
- 45:14for a very long time. I can tell you
- 45:16like when I was a child, I distinctly
- 45:18remember my dad going through
- 45:21the newspaper, like the physical print
- 45:22newspaper,
- 45:24and going through like which stocks he
- 45:26owned and how they did yesterday. And
- 45:27this is like the next morning finding
- 45:29out how each stock he did. No clue how
- 45:31his portfolio was like because you
- 45:33didn't have to like sit there and do
- 45:35that math.
- 45:36But, you could certainly look at the
- 45:38Wall Street Journal or the local paper,
- 45:40and it would have a printout of
- 45:42thousands and thousands of ticker in
- 45:43very fine print,
- 45:45and what their last close was, what the
- 45:47change was yesterday, and what the
- 45:48volume was, and a few other columns.
- 45:51You know, folks like my dad have been
- 45:53doing that for a long time.
- 45:55Uh
- 45:56when the internet kind of boom happened
- 45:59in the late '90s,
- 46:01and commensurate with that, tech stock
- 46:04sort of IPOs boomed.
- 46:06And, you know, pets.com and
- 46:09and all that were out there. Um
- 46:11So, now defunct company that sold like
- 46:15pet care products online.
- 46:17You know, there were a bunch of
- 46:18companies that went public,
- 46:20and
- 46:22a lot of interest on those, but also
- 46:23just in general on the markets because
- 46:25the market was rallying really hard, and
- 46:27it was pretty easy to kind of {quote}
- 46:28{unquote} look like a genius. You bought
- 46:30five internet companies and just were up
- 46:32like 100% a week or whatever the hell.
- 46:35And so, that dynamic really did lead to
- 46:39a
- 46:39pretty significant increase in the
- 46:41participation of retail investors.
- 46:43When that crashed,
- 46:45that definitely took a a a big
- 46:48chunk of retail participation with it
- 46:50because those folks realized, "Well, you
- 46:52know, it turns out this is hard, and I'm
- 46:53kind of shitty at it."
- 46:55And so, they stopped.
- 46:57That took another leg down just a few
- 47:00years later when the second that, you
- 47:01know, 2008 crisis happened.
- 47:04You know, the first crisis ended in
- 47:052002,
- 47:07late 2002, I think, maybe even early
- 47:092003.
- 47:11Uh and then the second crisis started in
- 47:13November of 2007. So, you didn't even
- 47:15have 5, 6 years between those two
- 47:18crises. And that was two 50% drawdowns
- 47:21in US equities in the space of 10 in the
- 47:23space of 10 years. It's pretty crazy.
- 47:26And unsurprisingly, that turned off a
- 47:27lot of people
- 47:29to investing in stocks at all.
- 47:31And if they were going to invest in
- 47:32stocks, it was just going to be passive.
- 47:35So, that kicked off that passive trend.
- 47:36It also kicked off a de-retailization
- 47:39trend in the US.
- 47:41Uh in some other markets, retail
- 47:43investors have been more present uh for
- 47:45longer and more consistently, like in
- 47:48places like China and Singapore and so
- 47:49on.
- 47:50Um Korea.
- 47:52Uh but, where there's And I don't know
- 47:55how to say this without it sounding
- 47:56politically incorrect, but like folks
- 47:57just gamble. Like you go to the casino,
- 47:59that's a completely normal thing to do.
- 48:01In the US, that's not like a completely
- 48:03normal thing to do. Casinos are like
- 48:05illegal in most of the states in
- 48:06America, which is a funny thing.
- 48:09There's just less of that kind of
- 48:11culture of just making bets and punting
- 48:12around.
- 48:13Um
- 48:15which, for better or worse, is what it
- 48:16is. Now, we've had this crazy bull
- 48:19market,
- 48:20and on top of that, a lot more tech is
- 48:23available than has ever been before.
- 48:25That's just the nature of things. That
- 48:27same state statement will be true like 2
- 48:29years from now, unless we
- 48:31truly devour ourselves.
- 48:34So, the platforms out there,
- 48:36and I don't mean Robinhood, which is, to
- 48:38me, just sort of a joke.
- 48:40Uh I mean more
- 48:42uh like Quantopian and other similar
- 48:45platforms where anyone can get out there
- 48:48and just
- 48:49code up trading systems.
- 48:52And then there's brokers like
- 48:53Interactive Brokers, where if you're
- 48:54like semi-pro, you have a few hundred
- 48:56thousand dollars,
- 48:58you can set up an account there and
- 49:00have some of the same tools and
- 49:04access as professionals have, although
- 49:07it's
- 49:08definitely a far cry, frankly, cost-wise
- 49:10especially.
- 49:12Yeah, it's, you know, it's kind of cool.
- 49:14So, you can
- 49:15you can set up like a one-person quant
- 49:17shop, and I know lots of folks who do.
- 49:19In fact, some of the folks we've
- 49:20allocated to
- 49:22come exactly from that lineage, where
- 49:24they didn't go They weren't like
- 49:26ex-employees of Two Sigma. Maybe they've
- 49:28had some professional experience, but
- 49:30what they really did was they're just
- 49:32homegrown
- 49:33hobbyists who are really good at it and
- 49:35go, "You know what? We're going to We're
- 49:37This is I'm good at this. I want to do
- 49:38this professionally and scale up.
- 49:40Uh and we've had some success with that
- 49:42model. I mean, but you know, we've also
- 49:44had some failure with that. It's not
- 49:45easy.
- 49:46But yeah, there's been this
- 49:47kind of
- 49:49uh combination or confluence of things
- 49:52where we have this protracted bull
- 49:53market
- 49:55that's encouraged retail participation
- 49:57and a lot of really specific brand
- 49:59excitement around names like Apple and
- 50:01Tesla.
- 50:02And then at the same time
- 50:05the availability of technology
- 50:07to actually make it easy to code up
- 50:10strategies and backtest them and
- 50:12implement them and enter competitions.
- 50:15Um and those things kind of combined to
- 50:19to make it pretty
- 50:21It's been Yeah, there's been like a big
- 50:23surge in that world.
- 50:25Yeah. I remember last time when we
- 50:27spoke, you you mentioned you'd allocated
- 50:30to someone who was essentially just a
- 50:32one-man band working from his apartment.
- 50:35Still have that money with that same
- 50:36guy.
- 50:37Yeah, I think you'd mentioned he was
- 50:39managing about $15 million. I'm not sure
- 50:41what that is today, but 70.
- 50:4470 million, one guy from his apartment.
- 50:47Mhm. That's incredible. That's not 70
- 50:50million from you. That's you're just a
- 50:52portion of that. No, it's all from us.
- 50:53All from you. Yeah. Wow.
- 50:58That's that's uh quite incredible. Let
- 51:00me ask you this question
- 51:02and we'll wrap things up here. But let's
- 51:04say there is someone listening to this
- 51:05who kind of fits into that category of
- 51:08DIY quant,
- 51:10you know, retailer working at home.
- 51:13They've developed a quantitative
- 51:14strategy. They've been having a little
- 51:15bit of success with it, trading it for
- 51:17some time, real money.
- 51:20They like the idea of
- 51:21running more capital through that
- 51:23strategy, scaling up. What would this
- 51:26trader need to do to be ready for
- 51:28investment?
- 51:30That's a great great question, Aaron.
- 51:33So,
- 51:34it depends is the answer.
- 51:37Um if
- 51:38they are willing to simply license their
- 51:40signal, meaning
- 51:42they would once a day or on whatever
- 51:44rate the right frequency is for that
- 51:46strategy, send a list of trades that
- 51:49their model comes up with
- 51:51to someone who has all the
- 51:52infrastructure and compliance and
- 51:54operational aspects to do that.
- 51:56Um to to put on those trades and do it
- 51:59correctly.
- 52:00Uh then the bar is a lot lower because
- 52:03they don't really need much of anything.
- 52:06If they want to actually like set up a
- 52:09fund
- 52:10and or a firm that does this and, you
- 52:13know, you'll have to register with the
- 52:14relevant authorities as a as an
- 52:17investment advisor or whatever the
- 52:18equivalent is in your jurisdiction. Um
- 52:21you'll have to show to investors that
- 52:23you can actually process trades and all
- 52:26kinds of other things. Um
- 52:29You know, that hurdle's a lot higher,
- 52:30like a lot higher. And it's expensive
- 52:33and not likely to work. So, just to be
- 52:35clear.
- 52:36Um you know, I think that the
- 52:39far more sensible path is to contract
- 52:43with someone who's already got a
- 52:45substantial operation
- 52:47and
- 52:49license your signals out.
- 52:51And then if it turns out you're really
- 52:52good at this in the long run, A, you'll
- 52:54just make money doing it. You may not
- 52:55have to quit your day job.
- 52:58But B, if you decide you really want to
- 52:59do it, at least you'll have some track
- 53:01record
- 53:02of what your signals have done. And then
- 53:05when you try to go out and get unreal
- 53:06size, you know, and then it at least
- 53:09when you try to go out and get capital
- 53:11one of the big questions that investors
- 53:13ask, which is is this thing scalable?
- 53:16You know, if I run a $500 billion
- 53:18pension fund or a $10 million fund of
- 53:21hedge funds or family office or
- 53:23whatever.
- 53:24Um I'm not looking to make
- 53:26a $10 million investment in a hedge fund
- 53:29that might return 10% per year. Like
- 53:31making a million dollars of P&L per year
- 53:34for someone at that size is like not
- 53:36worth the effort.
- 53:39Not worth the marginal cost. And so, you
- 53:41got to really demonstrate that you are
- 53:44uh able to scale. And if you can
- 53:46demonstrate that, oh no, my signal
- 53:48generated 20 million of P&L
- 53:50admittedly it wasn't my firm that that
- 53:52put the trades on.
- 53:54Then yeah, there's there's something
- 53:56there. But the more realistic thing is
- 53:58that if you're a hobbyist, stay a
- 53:59hobbyist, frankly, and license your
- 54:01signals to people who
- 54:03are doing this for real and
- 54:06um
- 54:07and you get paid.
- 54:08Uh and that's kind of cool. What would
- 54:10be an example of someone who may be
- 54:12interested in
- 54:13in licensing signals though? Like would
- 54:16that be
- 54:17a fund manager or
- 54:19Yeah, so um and I I think they've
- 54:21stopped doing this now, but WorldQuant
- 54:24used to do it and globally. Like they
- 54:25had
- 54:27farmers in Vietnam who were contributors
- 54:29of signals. Right. Um but I think
- 54:32they've shut that practice down. I'm not
- 54:34100% sure.
- 54:36Yeah, we know there's there's there's a
- 54:38few out there. I mean
- 54:39some of them are looking for more like
- 54:41professional CVs, like people who used
- 54:44to work
- 54:45at at firms that are doing this
- 54:47professionally and maybe have
- 54:50gone off on their own.
- 54:52Yeah, there there are a few out there.
- 54:54I'm not comfortable naming names to be
- 54:56honest just because No, I wouldn't
- 54:58expect you to. Yeah, I don't have But
- 55:00yeah, these are professional quant hedge
- 55:01funds. Okay.
- 55:03Would you have any luck, you know, maybe
- 55:05with some smaller family offices? Do you
- 55:07think that's something they may be
- 55:08interested in?
- 55:10Sure, if you have a relationship. But
- 55:12you know, as with so many things, like
- 55:14these days there's so many people
- 55:15unemployed, especially in the US.
- 55:17That
- 55:18their resume is going into the inboxes
- 55:20of every person that's every employer
- 55:22that's got a job opening.
- 55:23It's not quite hopeless
- 55:26if you randomly send your resume into a
- 55:28vacuum, but you definitely know it's a
- 55:30lot better if you have a connection.
- 55:31Well, how much more so for something
- 55:33like this where you're asking for
- 55:34millions of dollars or tens of millions
- 55:36of dollars
- 55:37of risk capital and a budget.
- 55:40Um that you know, that's not just a
- 55:41normal job, right? That's a job with
- 55:44some real strings attached. So, if you
- 55:45have like a good relationship with
- 55:46someone, then sure, it's I think worth
- 55:48barking up that tree, but
- 55:50I I honestly think it's
- 55:52like lower probability than a lottery
- 55:54ticket if you're just like casting your
- 55:57net out into the ether.
- 55:58Trust is a big part of this, you know?
- 56:01For sure. Yeah.
- 56:03I know you've got to run. I was going to
- 56:05ask you really quickly
- 56:08about your involvement with DARPA. Is
- 56:09that something you'd be interested in
- 56:11talking about?
- 56:13Yeah, but I'm not like under NDA or
- 56:14anything and I can talk about it really
- 56:16briefly.
- 56:17Um
- 56:18It So, it was really cool and fun and
- 56:20interesting. So, basically uh
- 56:22some years ago I was approached by
- 56:25um a professor who's a friend of mine
- 56:28who is friends with some folks who are
- 56:30at DARPA
- 56:31or were at the time and said, "Hey,
- 56:33would you be interested in connecting
- 56:34with
- 56:35They were looking at uh financial
- 56:37warfare kind of things.
- 56:40And so, this is part of the Department
- 56:41of Defense, the US Department of
- 56:42Defense. And this was not an exercise in
- 56:46looking at, you know, aggressive
- 56:48tactics, rather in defensive tactics and
- 56:51strategies for
- 56:54protecting American markets and
- 56:57financial systems from
- 57:00various parties that for either
- 57:01political or economic reasons
- 57:04may want to
- 57:07cause problems, right? So, causing
- 57:09outages at exchanges or causing um
- 57:13outages with, you know, payroll
- 57:15processing systems or
- 57:17any number of other vectors of threat.
- 57:19And there are of course continuously
- 57:21both governmental and non-governmental
- 57:23parties that are very aggressive in
- 57:27in that stuff. And so, it's, you know,
- 57:28it was this super fun intellectual
- 57:31exercise of like what are the vectors of
- 57:33threats
- 57:35detect them early, prevent them
- 57:37mitigate the spread of the damage.
- 57:39Um so, you know, we do things like war
- 57:41games and
- 57:42um
- 57:44and table exercises to to sort of flesh
- 57:47out some of these things a little bit.
- 57:49And the part of the goal was to see is
- 57:51there something
- 57:53measurable and quantifiable and within
- 57:54DARPA's remit to actually focus on
- 57:58to uh
- 58:00to work on this problem. So, the way
- 58:01DARPA works is it starts as a seedling,
- 58:03which is all this thing has been so far.
- 58:05And then it only becomes a program
- 58:08if the seedling shows that there is
- 58:11not only something to do here, but
- 58:12something for DARPA specifically to do
- 58:15because DARPA isn't, you know
- 58:16doesn't have a completely
- 58:18carte blanche remit. It has a specific
- 58:20remit. And so, if it if there's stuff to
- 58:23do, but it's better for like the
- 58:24Treasury Department to do, well, then
- 58:26that's fine. Like hand that off to the
- 58:28Treasury Department. They can decide
- 58:29what to do.
- 58:30You know, if it's it's for DARPA, then
- 58:32it has to fit some criteria. And
- 58:35uh so yeah, they're still in the
- 58:37evaluation phase. It does appear that
- 58:40there's enough for some stuff to be done
- 58:42as a program. But that has not been
- 58:45finalized.
- 58:47It's super fun, really interesting. It
- 58:50does sound like an interesting project
- 58:51to be involved with. So, you're just
- 58:53sort of consulting on it at the for the
- 58:55time being?
- 58:57Yeah, and I mean I'm I'm like a
- 58:58volunteer. Uh so, I'm consulting
- 59:01informally, no security clearances, no
- 59:03NDAs.
- 59:05I literally had one condition, which is
- 59:06that I get to put it on my LinkedIn cuz
- 59:09how cool is that? Yeah, yeah.
- 59:11Exactly. That's why I asked you about
- 59:12it.
- 59:13Yeah.
- 59:14And I I wasn't really sure the link
- 59:16between DARPA cuz I just know them as
- 59:18creating the the robot dog.
- 59:21And other crazy robots. I wasn't really
- 59:22sure the link between that and financial
- 59:24markets, but but yeah.
- 59:26It was a little bit surprising, but you
- 59:28can see that it's
- 59:29you know, economic defense and military
- 59:31defense are sort of more interlinked
- 59:33than ever, you know. Yeah. Yeah.
- 59:34Many wars have been fought over economic
- 59:36interests, I think far more than people
- 59:38generally realize. Right.
- 59:41For sure. Well, Rishi, I know you need
- 59:43to shoot off. Uh I just want to say uh
- 59:45it was really nice to catch up with you
- 59:47and uh thank you very much for coming on
- 59:49the podcast and um share a bit of an
- 59:51update on some of the things you've been
- 59:52seeing and what's been developing in
- 59:55markets. So, if someone wants to follow
- 59:58you, uh would you like to share your
- 1:00:00Twitter handle and maybe your website
- 1:00:02and also mention your book, too?
- 1:00:04Sure, Aaron. Thanks a lot for having me
- 1:00:06again. Again, good to catch up. Uh
- 1:00:09I'm on Twitter, not super actively, but
- 1:00:11I am on Twitter
- 1:00:13uh at Rishi K. Narang.
- 1:00:15And
- 1:00:18and I did write a book, the second
- 1:00:20edition of which came out in 2013,
- 1:00:23called Inside the Black Box. That should
- 1:00:25be readily available on places like
- 1:00:27Amazon.
- 1:00:28Uh what was the other thing I was
- 1:00:30supposed to share?
- 1:00:31Uh your website.
- 1:00:33I don't really have one. Okay.
- 1:00:36The T2 website, no?
- 1:00:38Uh yeah, there's nothing on there. But
- 1:00:40yeah, my company is www.t2am.com.
- 1:00:44Okay. And Inside the Black Box is a
- 1:00:47fantastic book. I've read it a
- 1:00:49several times. Uh I have a copy myself.
- 1:00:51Uh would definitely recommend it if
- 1:00:52you're interested in this type of thing.
- 1:00:54So, yeah, until next time, Rishi.
- 1:00:57Uh we'll we'll chat soon.
- 1:00:59Thanks, Aaron. Stay safe, man. Be good.
- 1:01:03You've reached the end of this episode
- 1:01:05of Chat with Traders. But rest assured,
- 1:01:07there are more episodes loaded with real
- 1:01:10market insight and zero hype on the way
- 1:01:13soon. So, to stay updated with each
- 1:01:15great new release, subscribe to the
- 1:01:17podcast on iTunes. And we'd love it if
- 1:01:19you'd leave a rating and review. We'll
- 1:01:21catch you next time on Chat with
- 1:01:23Traders.
About this transcript
This page contains the full transcript of Evolving Trends: How Quants Are Earning Alpha · Rishi Narang by Chat With Traders, generated from the public captions YouTube serves with the video. The transcript has 9,954 words across 1,730 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.