Letter #123: David Magerman (2020)
Letter #123: David Magerman (2020)
Head of Production and Research Scientist at Renaissance Technologies and Founder of Differential Ventures | Discussing RenTec
Hi there! Welcome to A Letter a Day. If you want to know more about this newsletter, see "The Archive.” At a high level, you can expect to receive a memo/essay or speech/presentation transcript from an investor, founder, or entrepreneur (IFO) each edition. More here. If you find yourself interested in any of these IFOs and wanting to learn more, shoot me a DM or email and I’m happy to point you to more or similar resources.
Today’s letter is the transcript of an excerpt from David Magerman’s talk to a group of high school students for the First Generation Investors (FGI) Speaker Series. In these 2 excerpts, David talks extensively, in layman’s terms, about Renaissance Technologies’ approach to investing.
David is a Cofounder and Managing Partner of Differential Ventures, an AI-focused venture fund. Prior founding Differential, David spent nearly two decades over two stints at Renaissance Technologies, the legendary quantitative investment group founded by Jim Simons. David was initially the Head of Production where he developed and maintained software for trading electronically with brokers and assisted in the development of statistical models for predicting the movements of stock and commodity prices. He was also a member of the company’s management committee, and the highest-ranked technical employee. After a brief retirement, he returned as a Research Scientist in the futures and equities research groups, where he implemented robust estimation techniques for coefficients for quant trading for a variety of instrument types including bonds, currencies, stock index futures, metal futures, and more. He also developed and enhanced optimization algorithms for trading financial instruments. Before joining Renaissance in 1995, David had a few stints as a research scientist focused on natural language processing at Bolt Beranek and Newman, IBM, and SRI International (the latter two of which were while he was a PhD student at Stanford).
(Transcript and any errors are mine.)
Related Resources
Transcript
So I ended up stumbling into this really weird job where these guys who worked at a quantitative hedge fund were looking for someone with my computer skills to help them with computer science. It was founded by a famous mathematician who did some really complicated stuff I don't really understand in mathematical theory in the 70s and 80s. And he used his mathematical skills to build a company that studied how prices changed in markets and tried to build mathematical models to understand them. And basically, he had a lot of what he wanted to do already built. But he didn't have a lot of computer science engineers to help build it.
And at the time I joined the company, they were trading in some markets, they were trading in bonds and currencies and stock index futures. But they hadn't figured out how to trade in the stock markets. And part of the problem was that their software systems couldn't deal with trading thousands of stocks at the same time. They could trade individual instruments like individual bonds, they could trade dozens of futures contracts in bonds, they could trade five or six currencies at a time. But they didn't know how to scale their systems up to trade thousands of stocks together.
And so when I joined Renaissance, they were trying to do stock trading. And I joined the company, and they were failing. And I remember that the people who I worked with were annoyed at me and people I worked with, because there were four or five of us working in stock trading and we were managing like 5% of the company's capital. We were responsible for like -10% of the company's profits, because we were losing money, and we were paid like everybody else, so people were kind of annoyed with us. They thought we were overpaid. They thought we should be paying the company because we were losing money. But they had faith in us, and eventually, we validated that faith by producing a system that works.
So what we did was akin to what a casino does. If you think about a casino, let's say they have Blackjack, they have a Craps table, they have a slot machines. And they work off of the premise that all of their gaming tables have an edge. It might be 51-49. It might be that that every time someone gambles at their table, 51% of the time the house is going to make money and 49% of time the gambler is gonna make money. Some other things have 54% to 46%. But the only games that casinos will play are games where they have a statistical edge.
And there's a thing called the law of large numbers, which says that if, on average, you're going to make money 55% of the time, it doesn't matter if you go on a losing streak. Because that's just bad luck. But eventually, if you gamble enough times and you bet on your strategy enough times, if you're going to make money 55% of the time, in the long run, you're going to make money, as long as you don't go below zero. That's kind of a key thing, especially when you're leveraging, when you're borrowing money to double down on your bets, you have to make sure you don't go below zero. But the law of large numbers tells a mathematician that if you if you run your your operation like a casino, casinos are always going to make money.
And Renaissance was a casino. They were the house. They had mathematical models that they convinced themselves could have an edge over the rest of the market. And so if we knew that our models could make money, if we bet $1 that 55% of the time we're going to make money, and 45% we're gonna lose money, as long as we never went below zero, we could keep playing and eventually make more money.
And the cool thing about the stock market at the time, and eventually it got even better for us, we could get anywhere from four and a half to seven to one leverage...So if we went long and short in a balanced portfolio, we could actually be almost seven and a half to one levered. And so basically, if we could be like five or six to one levered, we can get to the law of large numbers five or six times faster, which would make it even more likely that we'll make money.
So our goal was to use computers and math and statistical modeling and smart people searching for new ways of predicting things, and we were going to try to build a system that we believed had as much of an edge over the rest of the market, and then we were just going to leverage to death and make as many bets as we could.
In the mid 90s, we started doing this, and there wasn't a lot of people doing computerized trading at the time. Again, pre internet, there was no Internet. There were a few platforms that did stock trading on computers, but the vast majority of stock trading was done in person at the New York Stock Exchange. We, with our little shop with maybe a few 10s of millions, maybe like $100-$200mn of capital under management, were trading four or five percent of the New York Stock Exchange volume. So for every 20 shares traded in the New York Stock Exchange, we were one of those shares. So it was really kind of a heady time for us, because we're this little group in Stony Brook University with a group of like a couple of programmers writing software, and we were trading that much volume.
And at the time, I was the only one who knew how the system works, because I was the guy who wrote the software that did the trading. And I basically wasn't allowed to go on vacation. Because if I went on vacation and something broke, then no one could fix it. And the truth is, in the early days, our systems failed, our programs crashed probably like every two or three hours at best. Sometimes, like five or six times a day. And I had to be there to pick up the pieces. At the time, we traded every 15 minutes. So when a program failed, I usually had like five or 10 minutes to figure out what went wrong, fix it, and get the programs up and running so we could get back to trading. So that was the early days.
Obviously, things got a lot more mature as we got more money under management, and eventually we're like a $5-$6bn fund, we were trading all around the world using similar models. But at the end of the day, our system worked because we had an edge over everybody else because our models predicted the markets, predicted what was going to happen well enough for us to make money like a casino.
And so I learned a couple of key lessons about stock trading from this. And also from my experience watching the world develop around me and other funds, trading, making money, blowing up, and doing different things.
The most important thing which I wanted to convey to you is, while it's great that you guys are learning how to invest in the stock markets, and you should use your intuition, and if you understand something about electric cars or about movies, bet with what you know. But the truth is that what I learned at Renaissance is that when we're ready to trade, information doesn't matter. Like you'd think we'd have hundreds of people pouring over news feeds, watching news, trying to figure out what happened in the world. The truth is that while some people do trade that way, and they can make fundamental bets on stocks that they want to hold for a long time, we were trading in and out of stocks pretty quickly. Not quite high frequency trading, but we were trading out, holding like, we say from like, a day to a week to a month. But usually on the shorter side.
And when we... generally, what we looked at, was not news. We just looked at price movements, because we assumed that any news we could see, as people, someone else had already seen that news, they already traded off of it, and whatever I needed to learn from that news was already encoded in some way in the prices in the stock market.
So all our models just looked at price and volume and trading volume and how much trading was happening on different parts of the spread and how deep the book was of orders that were waiting. We looked at the the actual people betting on the market, and we assumed that anything that could be known about stocks, especially because we're trading four or five or 10,000 different stocks at a time, that we couldn't possibly figure out what was going on in the world from the news. But we knew that there are people who are experts on every single stock, who were reading the news, and they were trading off of it. And so if we just looked at the stock prices, we could glean what we needed to know about stocks.
So it's important for you to understand, if you guys are thinking about doing trading on your accounts that you guys are managing, when you see news happen, you have to assume that everyone bigger than you already knows that. So you can't assume Oh, there's just been news about Disney, I'm gonna go trade because Disney is gonna go up because of that news. No. Everyone who trades Disney already heard that news minutes to hours before you did. They've already traded off of it. And if they are going to buy Disney stock based on that, the stock price has already moved up to where that news reflects the price should be based on that news. So you want to be thinking, not so much reacting to individual news or immediate news, but thinking long term about what it means.
A lot of times, people overbuy things, and stock prices go up more than they should based on news. And if you can think, Wow, they overreacted to this news, and tomorrow they're going to look at this and say, Wow, this was really dumb, we shouldn't have bought this stock, we shouldn't have bought Disney, then you can use that intuition to say, Hey, I'm gonna bet against Disney because it went up too much. And that's kind of the kind of thing we would look for, is people over-betting, and we would call them reversionary signals. If we saw a stock going up too much, we would sell it, especially looking at it relative to its industry. But basically, we learned that information news is not what you think it is.
Eventually, we figured out that there are ways in which news can be useful. Like I said, if you see some big news happen, and then you see a stock go up a lot, the combination of that news and the stock going up might mean it's going to go down. So positive news about a stock may actually be a sign a stock's gonna fall.
And so that's why using computers and math and statistical models to bet in the stock markets is sometimes better, because human beings think a certain way. And all the traders out there are human beings, well these days, not as much, but there are a lot of human beings behind the trading going on, and you want to be thinking ahead of them. And you can't get ahead of them with information, because they're going to be [eating] information. But you can get ahead of them by thinking about their psychology, and thinking how you can game their thinking, and then kind of reverse engineer what they're doing. So that's one thing that we learned about the markets that really were key to me understanding how I should view the markets.
I usually mention this at the end, but I don't own any publicly traded stocks. I probably have less than 1% of my overall net worth invested in publicly traded stocks. And the reason I do that, and I'm not recommending that... God forbid, you should learn about the markets, and most people in the world do invest in the stock market. It's a good thing for retirement and for long term planning to be invested in the stock markets.
But I understand from my experience at Renaissance, and also by seeing other companies like Renaissance blow up, that the values that companies have in the stock market, sometimes they're based on the value of a company, but a lot of times they're based on some market player choosing to buy a lot of a company or short a lot of a company for reasons that have nothing to do with that company.
A lot of times companies will use kind of what we call boring stocks... So let's call Microsoft, for instance, a boring technology company, or IBM. They're old companies, they're not so volatile, they're not going to double -- they've gone up a lot lately, but they're not typically going to double in price, they have low volatility, but they're representative of the technology industry.
So if the technology industry is going to go up 20%, Microsoft's probably gonna go up 20%, IBM is probably gonna go up 20%. And let's say you have some really cool, volatile new technology company that you want to buy. I mean, Tesla is a bad example, it's not really in the same industry. But let's say if a Tesla-like company that's new to the industry that's really a high flyer and you want to buy it, but you don't want to take the risk that technology stocks are going to tank, because you just think that the company you're betting on is going to go up a lot more than everybody else, not necessarily go up if technology goes down.
So what you might do is you might buy that stock and short IBM, or short Microsoft. And that way, you're covering your risk that technology stocks, the whole industry is gonna go down, but you're still gonna bet on your high flying stock that it's gonna go up. And so a lot of times, when you look at the values of stocks, sometimes they don't reflect the value of the company, they might reflect the fact that there's some really big hedge funds out there that are going long a bunch of volatile stocks in an industry, and they gotta short somebody. So they're going to deflate the value of these boring stocks in a way which is going to affect -- if you own those stocks, and they go down in value, you're going to still lose money if you have to sell the stock even if there's nothing fundamentally wrong with the company.
And we learned this because there was this period of time in 2008 where all of a sudden, literally, almost literally, every quantitative hedge fund in the world, started losing money. And not just like a little bit of money. Like Renaissance lost 20 times more than it ever lost in its history, and was losing money every day. And it was such a bad run that if it went on for another week, we would have gone out of business. And a bunch of funds did go out of business.
And long story short, what ended up happening was, there was some big hedge fund that was like Renaissance in some ways, that went out of business and they sold off their whole portfolio. And they did it in a really clumsy way that caused all of these boring companies that everyone was using to short to hedge their exposure, to go up in value because this fund was buying them all back. And so there was this huge mispricing of all these different instruments that happened for a few weeks, like a week or two weeks, that affected every quantitative hedge fund in the world that used similar models. And when it stopped, we made most of our money back. But there was a period of time we were losing like over a billion dollars a day. We were losing over a billion dollars a day. I mean, think about that. And it was all because of this phenomenon that hedge funds push prices of instruments around in ways which don't reflect their fundamentals.
And if a big fund decides to dump all its stock, the market's gonna have a problem. So that's the main reason why I don't feel comfortable investing in stocks, because I'm not looking -- if I was looking to be a longterm investor, it doesn't really matter when you get in. You hold stocks for 20 years, and then they're gonna move where they're gonna move. But I don't like to be an active trader in stocks, because I feel like their values are manipulated.
Question on what technologies David is most excited about investing in over the next decade where he references Renaissance again.
Yeah, so I'm... it's funny, because I invest in data science companies that do AI and machine learning, and I am the biggest skeptic about what the ceiling is about machine learning and AI. Computers don't think, computers don't understand. They mimic human understanding, they mimic learning. But ultimately, they're silicon, they're computers. And they they do what they're programmed to do. And so, the technologies I like to invest in are the ones that are taking solvable problems, problems that are solvable by AI, and building well-engineered versions of software that does that. An example is Renaissance Technologies. We were not trying to predict the future of where a stock was gonna go. We were just trying to use math to get a small edge on the rest of the world, really kind of modeling human psychology more than anything else, hedging out the risks that we couldn't model, and doing what we could, and we made a lot of money that way. So I think that the technologies that I'm most interested in are the ones that understand the limits of machine learning and AI, and try to solve a problem that is solvable.
Wrap-up
If you’ve got any thoughts, questions, or feedback, please drop me a line - I would love to chat! You can find me on twitter at @kevg1412 or my email at kevin@12mv2.com.
If you're a fan of business or technology in general, please check out some of my other projects!
Speedwell Research — Comprehensive research on great public companies including Copart, Constellation Software, Floor & Decor, Meta, RH, interesting new frameworks like the Consumer’s Hierarchy of Preferences (Part 1, Part 2, Part 3), and much more.
Cloud Valley — Easy to read, in-depth biographies that explore the defining moments, investments, and life decisions of investing, business, and tech legends like Dan Loeb, Bob Iger, Steve Jurvetson, and Cyan Banister.
DJY Research — Comprehensive research on publicly-traded Asian companies like Alibaba, Tencent, Nintendo, Sea Limited (FREE SAMPLE), Coupang (FREE SAMPLE), and more.
Compilations — “A national treasure — for every country.”
Memos — A selection of some of my favorite investor memos.
Bookshelves — Your favorite investors’/operators’ favorite books.
Great read, thanks for sharing.