The Exchanges

Q So one of the things you've talked about in this process is protecting those 80 to 90% downsides. So after you've gone through the process of looking at the cross-sectional moving average, picking decent projects to invest in and kind of screening out the other stuff, how do you go about a strategy that captures the upside and then tries to protect you on the downside?

A Yeah, so operationally, what happens is these models flow into a factor dashboard that we've got, and it's my job to basically manage risk in looking at these models on a day-to-day basis and tweaking the beta of the portfolio relative to where we are in the trend. And a lot of that risk management comes down to after these markets go through big adoption cycles, You want to be really weary of continuing to hold a ton of beta. Now, our models are not designed to get out before the top. They're inherently designed to get out on the other side of the top because, because of the momentum characteristics of the market, these bull markets, these adoption cycles can go on way longer than we'll expect. And we've seen that in the past and people marvel at how far these things have run. So we'll always be late to get out, but they've also given pretty good indications that it's just exhausted, and so we look at other data associated with number of page views to the CeFi exchanges, inflows and outflows of coins from those exchanges, and all sorts of sentiment measures of, is the market continuing to see that adoption cycle progress, or has it ended? And when those models start to roll over, we have to start to hedge. And what usually happens is we'll hedge with some super liquid assets that we can do very cheaply, specifically perpetual futures contracts in Ethereum, Bitcoin, and a coupl…

AI assessment note: “when those models start to roll over, we have to start to hedge.”

Q So Lee, twice you mentioned earlier on going down the rabbit hole, going down the rabbit hole in crypto, before that going down the rabbit hole in quant. I mean, people are trying to dive in and learn more about crypto. What does going down the rabbit hole mean to you?

A So I think No matter where your investments in crypto are and how you look at the asset class from a personal portfolio standpoint, or even if you're at a large asset manager and you're investing for the asset manager, I think the only way to learn about this stuff is to own some of it and experiment. And so for me, the place that everybody should start, well, there's really two. One is Kevin Rose at the New York Times wrote a really great full overview of almost everything in crypto. It's actually really good. I wouldn't expect the New York Times to produce something that was so well thought out, but they did. And so I would start there. It's a really great primer on all the different pieces of crypto. It doesn't have a Dogmatic view on stuff. I think it's pretty open to everything, yet credulous at certain points. After that, you have to just create a wallet and start buying stuff and playing around with it and going through the process of doing these transactions. Small amounts of money doesn't need to be significant at all, but you need to be willing to experiment because that is the only way you will actually understand where we are with this technology, which is still really, really early. The UX of this stuff is terrible. This is not a scalable thing yet at all. And you need to understand that to understand why there's so much volatility and why the adoption cycles look …

AI assessment note: “the only way to learn about this stuff is to own some of it and experiment.”

Q What was the kernel of the idea for Estimize?

A Well, at Geller, we ran a whole bunch of event-driven earnings-related strategies. Specifically, we were attempting to pull apart the inefficiency of the Thompson Reuters IBIS estimates data set in the context of momentum and trend. So when a stock has been in a pretty specific bullish trend, you usually see upward revisions to estimates, you see positive surprises. What we were trying to figure out was where is the inefficiency In the sell side estimates and when can we get ahead of that inefficiency and then take advantage of the turning points or just giving us more confidence in the continuation of a trend. The concept behind Estimize was that we had to do that at Geller with models, but we started to see people at StockTwits make estimates for fundamentals just on the stream and on structured formats. And the basic idea here was marrying the two concepts of what if we could just get these people, these hundreds of thousands, millions of people to do this in a structured way, to build a structured estimates data set that was crowdsourced, that theoretically, if you set it up the right way from first principles In how a good crowdsourcing project should be done. You look back at James Surwicky's work, wisdom of crowds, that if we built this from scratch and we tightly controlled the heuristics around how, when, and why people entered their estimates, theoretically, we felt t…

AI assessment note: “marrying the two concepts of what if we could just get these people”

Q So Lee, how did you get from spending all of your career on quantitative and fundamentally driven estimates and long short equity to the crypto world?

A So back in 2013, I got interested in Bitcoin. I bought a little bit of it really because I enjoy kind of living a couple of years in the future in terms of all the new technology. I try not to live a decade in the future, like some people I know do, because I still have to be anchored to what markets will discount. And they're not really discounting a decade out, but they are discounting two, three, four or five years out. And the technology looked interesting. And I guess the thesis was that you need to play around with stuff to understand it. And if you don't, if you get stale If you become a bit of a Luddite and say, I only like what I use, you will not understand any of the new stuff. So you got to own stuff to understand it. You got to give yourself an incentive to do that. I wrote about it on my blog at the time back in 13. In 15, I think I gained a lot more confidence in what was going on with the technology and in the psychology of that market. And the basic thesis was that this stuff was just religious proselytization as a financial asset. Like if you could own a digital religion, would you want to? Yes, absolutely. The first global digital religion that is basically a Ponzi scheme But you can own it. And that seems like a great investment. And it was. The funny thing is, if you look back at the two posts that I wrote, the one in 13 and the one in 15, while I got the c…

AI assessment note: “So back in 2013, I got interested in Bitcoin. I bought a little bit”

Q On both sides of the turns, once, as you're reducing risk and things are rolling over and then trying to get back in when things bottom out, is there a general strategy you found that's better to either pick the spot and come out in chunks or gradually work your way out, gradually work your way in?

A So the way that we use these models is we use them as an ensemble. So different models will tell you different things at different times. A time series model will tell you different things than a pure trend breakout model will tell you different things than a moving average crossover model. And so what we try and do is we try and look at all of these in conjunction and we do try and step back in, but our timeframes are not stepping back in over the course of two months. It's stepping back in over the course of A week or two. Because we can be relatively tactical about removing that beta if we need to, and there are false starts to new bull markets, and trend models have a history of getting shaken around at the bottom quite a bit. And so the theory is that at the top, you want to be much more discerning about how quickly you get rid of your beta, and at the bottom, After a big drawdown, like we've had, you want to be a little bit looser with being able to sustain some losses in order to be involved at the bottom, because these momentum models tend to get shaken around quite a bit at that point.

AI assessment note: “we do try and step back in, but our timeframes are not stepping back in over”

Q What's the most surprising thing that you saw in all these hedge funds? And I'm thinking about this from sort of an outsider allocator's perspective who might be invested in a bunch of these. What did you see from the inside that you think people on the outside might not be privy to?

A The biggest thing that coming into this world, even being a Geller, That I learned at Estimize in talking to so many of these firms about their process for generating trading ideas and fundamental estimates and all of this stuff is really that at the end of the day, almost all of these firms on the long, short equity side, they're gunslingers still, and they're playing with beta. They are gambling with beta. You gamble with beta long enough, and eventually you're likely to lose. And it definitely made me somewhat jaded regarding that model, because while there are some really great managers, the probability that you're going to pick one of those managers in long, short equity, which is just really, really tough, really tough world these days, is likely to be low.

AI assessment note: “almost all of these firms on the long, short equity side, they're gunslingers still”

Q So where did you take it from owning some Bitcoin to Starkiller today?

A So in early 17, I got very interested in what was going on with Ethereum. And The idea was that I was looking for, okay, where is the utility in this stuff actually going to be, right? Because sure, this can just be kind of a financial Ponzi scheme for a while, but at some point there has to be some kind of utility to this, if it's going to be anything long-term. And Ethereum really looked like it was that thing. It was programmable money. It was exactly what they left a hole in the original internet for the transfer value and smart contracts and everything. And if you look at what Bitcoin is, it's a good first crack at solving a pretty fundamental technology problem, but it's not actually designed very well for all of the things that even the earliest people on the internet wanted to be able to do. And so it looked like this technology and what Vitalik was saying about it and trying to build was going to be the real stuff. And I went way down that rabbit hole. The other thing that happened was I sat down with a friend who runs a pretty large quantitative asset manager, and we had a hypothesis about the crypto market. And the hypothesis was basically that because the crypto market was largely retail and fund flows driven, there's a lot of leverage in the system, and there's relatively little intrinsic value to any of these assets. And I don't mean that in a pejorative sense. I …

AI assessment note: “in early 17, I got very interested in what was going on with Ethereum.”

Q With everything going on in the space, what is your team like to follow all this stuff, particularly, you know, I guess on the qualitative side?

A So I'm the CIO and the PM for the beta portfolio. We also, on top of the beta portfolio, we run yield farming strategies on all of our assets at all times. We've got a really great DeFi yield portfolio manager, and he's actually really the one doing a lot of the deeper diligence work on these protocols beyond the fundamental data that I look at a lot. Without him, you know, we likely could not do this because there's a lot of inside baseball that goes on in understanding the safety of these protocols and the entrepreneurs behind them. And then we've got a great quantitative researcher and another quantitatively oriented investment analyst. And what I basically tell all of them is that while they each have their independent role to play, the space is so nascent still. We're still so early in the development of this technology that while we have to focus on our strategy and running it well, we have to be inquisitive to what's going on in the space, even outside of our specific niche here and their specific roles. And so I ask everybody to bring stuff. What's interesting today? What are you seeing that's new? Like what's piquing your interest? And there are a couple of core things that I've learned in crypto over the years. One of them is the most alpha is usually found in the things that are annoying and hardest to do. Meaning if you have to set up a new wallet, if you got to get…

AI assessment note: “We've got a really great DeFi yield portfolio manager, and he's actually really the one”

Q an awakening. If that's not how you would describe your investment management tech, request a demo at ridgeline.ai. And now, back to the show. So you did do some of it yourself, a difference between a backtest and out of sample. Why don't we dive through just your investment process? So there's lots of tokens out there, and how are you deciding what goes into the filter of the portfolio?

A So at our size, we have to be cognizant of liquidity. The liquidity in this market is definitely not evenly distributed across the top 300 coins. There's a lot of liquidity in the top 10, 15, and then it falls off relatively quickly. So I'd say, yeah, our universe that we can really touch in any significant way is something like 354 hundred coins today. That's growing pretty quickly because the liquidity is growing relatively quickly. And we will take positions across all of the different sectors, whether it be L ones, L twos, metaverse stuff, game stuff, infrastructure stuff, which is some of my favorite DeFi. This is an entire market, just like the equity market is comprised of a lot of different types of protocols and businesses. So we start there. We don't trade shit coins mostly because while we do see that those coins exhibit outsized momentum factor effects for sure, they operate a little bit differently in terms of that momentum effect than what we call productive assets. So businesses with real users and real cash flows and tokenomics that we like. So we kind of weed those out, and those don't go into our models. Interestingly, we're actually working on a completely separate shitcoin trading strategy that is fully systematic, that is much more high velocity than the core strategy that we run. It's exactly what you think it is. It's tied directly to just volume spikes o…

AI assessment note: “So at our size, we have to be cognizant of liquidity.”

Q So how do you bring in that, let's call it the fundamental side to turn this quantitative momentum approach into a quantum mental approach?

A The first layer of the asset selection is the cross-sectional momentum models. And the theory behind that is basically that These things are all still experiments. They're all experiments. Even Ethereum is still an experiment. It may not be scalable. And so you have to treat these things like series A to D kind of startups. And when you're treating it like that, you're not looking for a 40% gain. You're looking for an adoption cycle. You're looking for a five, 10 X on these things, which is why these momentum models are so well designed to capture the growth in this market. But after that, there are a lot of other things that we have found that will limit your risk of stepping into pure frauds, rugs, things that have no intrinsic value at all. And it's basically two layers. The first layer is we're looking at the underlying fundamental data that's on chain. So the nice part about crypto relative to equities is, whereas in equities, we get one report a quarter, and if you want any fundamental information between the quarters, you got to go grab all the unique data to try and piece together what's going on. But in crypto, we can literally see real time. How is this business going? How many transactions on that chain are in that protocol? How much locked value in the protocols? The velocity of those transactions, the growth, like the second order effects in all of those different …

AI assessment note: “The first layer is we're looking at the underlying fundamental data that's on chain.”

Q So when you first got into crypto, you talked about it as really enjoying looking a few years into the future. And if you're looking a few years into the future today, what are you seeing?

A So there are, I think, two different punch bowls that you can drink in crypto. One of them is this techno libertarian thing. And that is, it's not how Bitcoin started, but it is the thing that launched Bitcoin for sure. The techno libertarians who felt that this could be a new global reserve currency, that it was an inflation hedge, that you could extricate yourself from holding fiat government currency and things like that. I think a lot of people drank that Kool-Aid, and it's wrong. I just think it's straight up wrong. I don't think any of that's gonna happen. The other punch bowl is basically the technology punch bowl, is what can this stuff do from a technology standpoint? And here's kind of how I look at it. In a scaled fashion, this technology has the opportunity to provide transparency in a pseudonymous way, To every financial transaction that takes place, and that's buying of goods, it's loans, it's purchase, it's everything. And the economic theory here is that the history of economic growth is basically the history of debt. It's the history of expanding leverage in an economy, all the way back to literally people making loans to each other with cowrie shells. That is the history of economic growth. The idea here is that, and all of our other financial crashes happened, basically because we didn't know where the leverage was, and somebody blew up, and we couldn't hedge…

AI assessment note: “this technology has the opportunity to provide transparency in a pseudonymous way”

Q How do you think about taking those tried and true models and figuring out in this world where, as you said, you're not presuming there's intrinsic value attached to the things you're trading, which either time series or duration of models are optimal for the strategy, and then how do you evolve it over time?

A Yeah, so this is purely about what is the goal of the portfolio over a long horizon. And the way that we designed this was when we looked at the market, you basically have two ways to invest in crypto today. It's changing a little bit now with us and a couple other firms, but basically You could just buy some coins yourself and hodl, which is what some of the more technology oriented participants have done since the very beginning, and they've done very well. The problem is that it's very hard behaviorally to see a massive amount of wealth draw down 70 or 80% over the course of a six month period and not, like, lose your lunch at the bottom and then puke the beta up at the very wrong time. So you can do that with a couple thousand dollars. You may even be able to do it with a couple 100,000 dollars, but good luck doing that with tens of millions of dollars or hundreds of millions or billions of dollars, right? So it's just no pension fund, no, no big asset management firm can suffer a 80% drawdown and live to tell that tale. Just not going to happen with their LPs. The other thing that you can do is you can give it to a crypto VC that's investing in illiquid stuff at an early stage, and they've done incredibly well as well. Although, as Cliff Asnes would say, there's, I guess, behavioral alpha in not being able to market every month. Whether any of us agree with that or not, Cl…

AI assessment note: “when we looked at the market, you basically have two ways to invest in crypto”