This post was last updated on September 26th, 2019 at 03:02 pm
Welcome to this two-part series on crypto derivatives trading. I’m joined by Sam Bankman-Fried, the CEO of FTX, one of the fastest-growing crypto derivatives exchanges in the space. FTX launched in April of 2019, and at the time of this recording, does around $100 million dollars in daily volume.
Sam is also CEO of Alameda Research, a quantitative trading firm that provides liquidity to crypto markets worldwide. Alameda manages over $100 million in digital assets and trades $600 million to $1.5 billion dollars per day across thousands of products. They are responsible for approximately 5% of all (legitimate) crypto exchange volume.
This conversation is broken up into 7 chapters:
- Chapter 1: Quantitative trading
- Chapter 2: The relationship between Alameda Research and FTX
- Chapter 3: The current state of crypto derivatives trading
- Chapter 4: The similarities and differences between leveraged volume and spot market volume
- Chapter 5: The current state of FTX
- Chapter 6: Growth hacking and brand differentiation in the crypto derivatives exchange space
- Chapter 7: The future
In part 1 (episode #54), we cover chapters 1 through 3. In part 2 (episode #55), we’ll close the conversation by covering chapters 4 through 7.
Topics Discussed In These Episodes
- How FTX got started
- The opportunities in cryptoland versus traditional finance
- How quantitative trading works
- How the infrastructure for crypto investing has evolved over the past few years
- Blind spots to avoid when creating trading bots
- Why there are so few players in the derivatives and options platform space
- Rules, regulations, and norms for liquidity providers
- Why FTX wants more liquidity providers on their platform
- The 4 components (which Wall Street has) that are missing from crypto derivatives trading
- The difference between a future and an ETF
- How FTX handles outlier detection
- Sam’s strategy and approach with FTX’s FTT Token
- Why leveraged volume and spot market volume are essentially the same
- Why the FTX and Alameda Research teams are distributed throughout the world (Antigua, Tokyo, Hong Kong, etc.)
- How US regulators approach the crypto space differently from other countries
- Why emerging markets with less developed financial ecosystems are quickly adopting crypto
- What FTX is doing to differentiate their platform in the crypto derivatives exchange space
- How much revenue FTX is currently generating
- The importance of WeChat for building crypto communities in China
- Why co-location is not a big deal yet but will eventually be in the future
Links Relevant To These Episodes
- Cryptoinvestor Weekly Newsletter
- Clay Collins
- Sam Bankman-Fried
- FTX on LinkedIn
- FTX on Twitter
- Sam on Twitter
- Jane Street Capital
Quotes“Index products are all over traditional finance. From ETFs to futures to mutual funds, the whole industry basically runs on index products. These have been missing in crypto.” @SBF_Alameda Click To Tweet "There were like 5 crypto derivatives platforms & the median one was making hundreds of millions of dollars a year in revenue w/ a mediocre product. We saw a huge opportunity to enter this space & build something better."… Click To Tweet "Alameda Research is currently the biggest liquidity provider on FTX. However, the Alameda account on FTX is not privileged over any other accounts. They plugs into the API & have the same latency & fee schedule." @SBF_Alameda Click To Tweet "If you want to service customers from all over the world, your company will end up becoming decentralized." @SBF_Alameda Click To Tweet "A lot of exchanges try to make big waves by saying, 'We're setting up a data center in Aurora, Chicago right next to CME.' And then they end up just not even bothering to do it because no one wanted it." @SBF_Alameda Click To Tweet "Right now, what does crypto do? There are 2 obvious use cases: store value and means of transfer. The key thing here is that this is going to be a lot more useful in countries that have really bad alternatives." @SBF_Alameda Click To Tweet
Part 1 Transcript
Clay: Welcome to Flippening, the first and original podcast for full time, professional, and institutional crypto investors. I’m your host, Clay Collins. Each week, we discuss the cryptocurrency economy, new investment strategies for maximizing returns, and stories from the frontlines of financial disruption. Go to flippening.com to join our newsletter for cryptocurrency investors and find out just why this podcast is called Flippening.
Clay Collins is the CEO of Nomics. All opinions expressed by Clay and podcast guests are solely their own opinion and do [00:00:30] not reflect the opinion of Nomics or any other company. This podcast is for informational and entertainment purposes only and should not be relied upon as the basis for investment decisions.
Welcome to part 1 of this series on crypto derivatives trading. I’m joined by Sam Bankman-Fried. Sam is CEO of FTX, which is a crypto derivatives exchange and may be one of the fastest-growing derivatives platforms to date. Launched in April of 2019, [00:01:00] FTX does around $100 million dollar in volume daily at the time of this recording and has launched several features and products including leaderboards, spot orders, altcoin index futures, and much more.
Sam is also CEO of Alameda Research, a quantitative trading firm that provides liquidity to crypto markets worldwide. Alameda manages over $100 million in digital assets and trades $600 million to $1.5 billion per day across thousands of products. They are responsible for approximately 5% [00:01:30] of all legitimate crypto exchange volume as a percentage that’s up there with the high end of what a high-frequency trader would do in a US equities markets.
This conversation is broken up into 7 chapters: In Chapter 1, we discuss quantitative trading. In Chapter 2, we explore the relationship between Alameda Research and FTX. In Chapter 3, we dive into the current state of crypto derivatives trading. In Chapter 4, we cover the similarities and differences between leveraged volume and spot market volume. [00:02:00] In Chapter 5, we discuss the current state of FTX. In Chapter 6, we delve into growth hacking and brand differentiation in the crypto derivatives exchange space. Finally, in Chapter 7, we close our conversation by exploring what the future might look like
We’ll get right to this episode in just a second, but before we get started, I’d like to pause for a moment to tell you that this episode is brought to you by the Nomics API and CSV Data Export Service. If you need an Enterprise-Grade Crypto Market Data API For Your Fund, Smart Contract, or App, or if you need historical CSV dumps [00:02:30] of trading data from top exchanges or even obscure ones, then consider trying out the Nomics API or our historical data export service. Our API enables programmatic access to clean, normalized, and gapless primary source trade data across a number of cryptocurrency exchanges. Instead of having to integrate with multiple exchange APIs of varying quality, you can get everything through one screaming fast fire hose. If you’d like to order historical cryptocurrency market data as CSV exports from top exchanges, email us at firstname.lastname@example.org. [00:03:00]
OK, back to our regularly scheduled program. Here’s part 1 of my two-part conversation with Sam Bankman-Fried from FTX. Enjoy.
Sam, what’s the origin story of… first, your involvement in crypto and then second, FTX [00:03:30] as an exchange?
Sam: I graduated from MIT in 2014, went to work at Jane Street Capital in New York, was there for about three years trading international ETFs. Those are, for anyone who is interested, US-listed funds that own foreign stocks.
Clay: Is that global macro kind of stuff?
Sam: There are a lot of people who do global macro with them, but I was doing more of the quantitative trading side. So rather than saying, “Do I think that the Brazilian economy is strong? If so, I’m going [00:04:00] to buy EWZ,” which is a US-listed Brazil ETF, we’d be answering questions like, “Well, here are two different ETFs with the same or similar stocks in them. What’s the relative pricing?” We’d be looking for quantitative traits in them, arbitrage to do, have it an ETF and a future both on a similar, the same index or something like that. It was a lot of derivatives trading, a lot of modeling, short-to-medium time scale modeling.
Clay: I’ve heard a lot of traders that I [00:04:30] really respect say kind things about Jane Street, that they don’t do a lot of publicity of PR, but that they’re just a really solid organization. Then, I’ve heard a lot of great things about their contributions to the OCaml programming language as well.
Sam: It was a great place. They’re good at what they do, they’re good at trading, but they’re also really good to their employees. They were really good to me, and it’s the sort of thing that you can just really feel being out of place. When I was there, I talked to most of my friends who had other jobs at places that generally get [00:05:00] good raps. These weren’t people at places that everyone hates; these were people working at places like Google, and they all spoke uniformly blandly about their employers. “It’s fine, you know, they treat me fine. They haven’t wronged me. They pay me well. My job is reasonable. I don’t have any major complaints to file.”
But that was the general air of it, and it was actually rare to find a friend who spoke really strongly positively, but I did feel that way about Jane Street, that they [00:05:30] went way out of their way to be good to their employees. I don’t mean really nice bean bags or things like that—there were some cool bean bag rooms—but more I mean things like really respecting their employees, pushing them to accomplish as much as they can without being the person whipping everyone into shape. It wasn’t cruel, but it was very driving, and at the same time very nurturing and gave a lot of respect and [00:06:00] autonomy to people that were able to grasp it. So, it was a good place.
Clay: So, you’re there, you’re doing quantitative derivatives training. What’s the next chapter?
Sam: I was there through 2017, and then I left. I think there were a few things going on, but I think the biggest was that there were a lot of things I wanted to try doing with my life, and it sort of felt like a “now or never” type thing. I didn’t expect Jane Street to get worse. It’s not like I was going to stay there for another two years and be like, “Now’s the time to leave.” Either it was worth leaving, or it wasn’t. [00:06:30]
It was a tough call, but I think in the end, there were a lot of different things I wanted to do. Some of those were in the charity sector, some of those were getting more involved in public life, and some of those were trying to build my own thing from the ground up. It seemed worth trying a bunch of those out and seeing how they went, seeing what worked.
Anyway, I left Jane Street, and then I started up Alameda Research. Alameda Research, it’s a quantitative trading firm in crypto. So, where did that come from? Basically, I looked at Coin Market Cap, and [00:07:00] it’s like, “Either this data is all fake, or it’s completely insane.” It’s claiming that these two Bitcoin USD markets are trading 5% a part. For context, I think a lot of old timers in crypto hear that and think, “Yeah, the good old days. 5%.” But if you’re in traditional finance and you hear five basis points, you’re like, “Holy […], that’s a good trade.” You’re like, “Wow, I can just make five basis points on that trade, just a free five basis points? What’s happening?” Then you get to crypto and it’s like 5%; it’s literally 100 times bigger. That just doesn’t happen.
It’s like one day every four [00:07:30] years you get to do a trade, in fact in traditional finance, and then you’re just like, “All right, Bitstamp and Coinbase. Yeah, 5% a part.” This is either the best trading opportunity of the decade, or this data is just fraudulent, and I have no idea. It’s just impossible to tell from the outside. So, I decided to dive in and take a look, and as it turns out, yeah, about half the data was fake. That number has probably gone up over time. But it wasn’t all fake. In fact, there really was a big Coinbase/Bitstamp […] at some times. [00:08:00] There was an entire week where Litecoin was trading five to 40% rich on Coinbase, trading a billion dollars a day. And that is real data.
So it turned out that, yeah, it wasn’t all real, but enough of it was that that is a really exciting opportunity and sort of what was going on was, crypto was really taking off into the public eye in a way that it hadn’t before. It was going from a hobby and passion for a small number of people, to a global phenomenon in the course of [00:08:30] three months. It went from that weird Bitcoin thing to half of your friends are talking about trying to buy their first Bitcoin, in this really short period of time.
So, this unbelievably massive amount of customer demand for crypto coming from lots and lots of different places: the United States, Korea, Japan, China, Russia, and you pair that with a really immature ecosystem, something that was servicing a field 10% this size a year ago. The field wasn’t ready to support this amount of customer volume. There was [00:09:00] not enough liquidity provided, there wasn’t enough capital in the liquidity providers, the infrastructure wasn’t set up to support it. Everything was breaking and falling apart at the seams, and this created a lot of good training opportunities if you could sew everything back up together enough to get your money flowing through.
Clay: Okay. Let’s kick off chapter 1, which is about quantitative trading. When you talk about doing quantitative trading, is any of that operating or running real time trading environments, doing market-making [00:09:30] or high frequency trading? Were you writing bots that were doing the trading for you, or were you sitting behind the screen or 12, manually placing these orders?
Sam: It’s a combination, and especially in crypto, I think the way to describe it is, the vast majority of Alameda’s volume is done by automated trading systems, not by humans. But rather than thinking of these automated training systems as sentient AIs who have a mind of their own and they’re like, “Today, we’re buying Litecoin,” and so I guess we’re buying Litecoin. Who knows why, right? [00:10:00] Instead, they sort of just do what we tell them to. It’s just an algorithm with a bunch of different inputs, and we designed each one of those. Maybe we ran some studies, but the studies were human-legible studies.
Rather than thinking of it as a mind of its own, we sort of think of these bots as really dumb humans. They have 10,000 hands, they can operate really quickly, they can compute things in their heads really quickly, and they don’t […] up computations, ever. If you’re like, “Yes, here’s my algorithm, it’s averaging together these 50 factors [00:10:30] in real time,” a human can’t do that. A computer can do that, and it can do that on every marketing crypto at once. Even if it’s core trait is doing something that you as a human can understand.
Clay: In other words, you’re not doing machine learning. You’re not training models on raw data and then simply implementing that. There are some rules that are in place. It’s following those rules, it just can see quickly and see broadly.
Sam: Exactly. I think that there is some machine learning in finance, but there’s actually surprisingly [00:11:00] little. You might think that it dominates the trading sphere, but it’s just a surprisingly small fraction of it. The reason is that the number of variables that you could plausibly use is unbelievably massive. Literally, it’s like we’ve got in crypto alone, 200 exchanges with 200 order books each, spitting out 10 pieces of data a second, every second for years, and then you’re like, “All right, now run the algorithm on that,” you can’t do that.
Computers are powerful, but they’re actually not powerful enough to just take that and do something with it yet. [00:11:30] So instead what you need to do is, you need to direct it. You can do a study, and studies are useful, but really a lot of the key in doing a study is deciding which study to do. If you have some really compelling to reason to think that when it’s raining on Tuesdays EOS is going up, then you can run that study. You can just collect the data and see whether there is evidence that that EOS does go up on Tuesdays when it’s raining, and the study will just answer the question for you. You’ll have plenty of data for that. But if you don’t have that hypothesis to start with, there’s no chance your model is going to test that, [00:12:00] that a neural net is going to pick up on that. There are just too many hypotheses that you could test, unless you as a human give it some direction there.
There are studies backing a lot of this up, but a lot of these studies are like, we’re sitting there and are curious what we think might be going on. Bob wakes up sometime around noon each day. It varies a little bit. Right now, Bob just buys Bitcoins whenever he wakes up, every single time. So Bitcoin goes up whenever Bob wakes up, and we can kind of tell when Bob wakes up, because he starts out [00:12:30] by placing his daily way out of the market bids in Litecoin. So as soon as we see these way are the market bids on Litecoin, we know Bitcoin’s about to go up.
That’s not a real thing. It’s not exactly that, but you have things like that where you start to come up with a real, actual model for why you think this effect might exist. Then, you have that, it’s easy to run the “Look for Litecoin bids,” and see if that predicts Bitcoin going up study.
Clay: It also seems to me that you don’t even have to reach very [00:13:00] far to find some kind of edge in this space. You don’t need to be ingesting real-time order book data from 200 exchanges, most of which are faking volume simultaneously to find edge. I mean, there’s still pretty simple arbitrage loops within an exchange. You just don’t have to go there. Then, there’s people that add social signals or all kinds of things that are very cyclical or decay a lot. Why go there? You just don’t need to do that yet.
Sam: Yeah. The other thing that I’d say is, social [00:13:30] signals. I’m not saying there’s nothing there, but your goal in doing it is not just to have something, it’s to beat the pack, right? If you really think you’re going to become one of the world’s experts in trading Bitcoin on social signals, then great. There might be some amazing trades to do there. But if you just read a tweet on Twitter where some technical analysis person said Litecoin is going up today, you’ve got to do better than literally the easiest thing in the world to do in order to make a big trading business out of that. [00:14:00]
I really respect someone who’s like, “This social media signaling for crypto pricing, I’m going all in; I’m going to spend thousands of hours really getting as comprehensive a network here as possible.” I think that could be awesome. This is going down over time and things are getting more efficient, but especially two years ago, you see Litecoin trading 25% higher on Coinbase or GDocs at the time than on Binance, and it’s not like, “All right, time to go run my neural net.” You’re like, “No, I want to buy Bitcoin on Binance and send [00:14:30] it to Coinbase and sell it.” That’s the trade.
That weirdly was the easy part. At the beginning, the easiest part of doing quantrading crypto was finding the good trade, and the hardest part was actually doing it. Anyone who has tried to wrestle with 50 exchanges at once has, especially when you start sizing up, you start noticing all the cracks at the seams. So, why didn’t we make billions of dollars that week? Well, guess how much money we could withdraw every day from our Coinbase account. I think at the time it [00:15:00] was $10,000. That was how much we could cycle each day.
So now you’re like, “Oh God, wait. I thought I was going to make billions. Actually I made $3000.” The reason was withdrawal limits on some exchange rate, and then eventually you solve your withdrawal limit problem. “Great, I spent 200 hours, and I got my withdrawal limits taken up.” Then you’re like, “Aw, […]. I’ve got to wire money from Coinbase to my bank account, and I just sent $50,000 to Bank of America and they shut me down because it was coming from crypto exchange.” [00:15:30]
I don’t have a bank account, so I can’t get money off of Coinbase anymore. I now have high withdrawal limits but nowhere to put it, and you’re like, “All right. God. I’ve got to figure out this whole banking crypto thing when banks hate crypto.” And then you’re like, “Ah, I got this great […]. Wait, […], it takes a day to move around the blockchain.” So literally it’s going to take me a day to do this arbitrage, and it’s not there by the time it gets there.
And then you realize, “Wait, actually, this exchange hasn’t allowed withdrawals for this coin for the last year, and no one knows why.” That’s why it’s trading at 50% what, there it was. [00:16:00] You start piling all these things on top of each other, and what you realize is, in some cases 80% of the battle is navigating the infrastructure of the ecosystem, not just finding the good trade in the first place.
Clay: I know a lot of developers and engineers; I was one in a past life. The first thing that I hear a lot of them say is, “I’m going to create a bot,” or whatever. I often find that a couple of blind spots that exist is trading fees, so they don’t take trading fees into [00:16:30] consideration. Then also, depth of order books. It’s like, “Yeah, you back tested this, but you just can’t move that much volume.” Maybe you could make the first trade at that price, but the additional $100,000 of that trade is going to cost you a lot more, even if you distribute that across multiple exchanges and do some kind of meta order thing.
Sam: The tech is amazing until you realize that you’re paying stuff and basis points on each side of the trade. I think other things that run into, exchange latency. Again, this only matters if you’re doing high currency [00:17:00] trading, but some exchanges have relatively high latency in crypto; it’s not uncommon to see an exchange that has one second latency. Some exchanges have really good latency 80% of the time, but when markets move a lot, that goes up to a minute. And also maybe 150% of your edge was coming during those crazy times. Your entire strategy involved sending orders when no one could get an order through to an exchange.
You see cases where it’s like, “Oh my God, there’s this great arbitrage,” and then you realize one of the late series is [00:17:30] Korean won. But Korea doesn’t allow you to turn that won into dollars and withdraw it; it’s a restricted currency. So now, there’s this ticker KRW that trades on a lot of exchanges, but you can’t just use the Bloomberg exchange rates dollars to price it, because you can’t get it out of the country. It’s locked. So that’s another thing that you run into a lot.
The fact that you can’t short sell on most products. If you have some […]coin, you’re like, “I’m going to buy Grosso coin on this exchange [00:18:00] and sell it on the other exchange.” You definitely can’t short sell Grosso coin. Your strategy has to be is, you buy it on one exchange and then you send it over, wait for it to hit, and sell it. But if the actual trade here was just that someone bought way too much Grosso coin on one exchange, so it’s trading way too high for a few minutes. You just bought it a little bit too high on the other exchange, and by the time your Grosso coin gets over to that first exchange, there’s no chance it’s still trading there. So you see this arbitrage, but you can’t do the […], you can’t short sell. That’s another [00:18:30] common issue that you can run into.
Figuring out what stable coins are, right? What’s a tether worth? What’s a trust token worth? What’s a QC worth? There’s a hundred stable points now in crypto, and they all have different levels of fungibility with fiat that they claim to be related to. You can easily lose the whole edge of your trade if you don’t realize that there’s a 1% redemption fee for some stable coin. So that’s another common issue. You could be looking at [00:19:00] exchange that is just faking its data, and that’s half your trade.
Clay: Another thing we saw is, during surges or flash crashes, often some of these rest APIs, where there wouldn’t be a WebSocket available and you page through data based on timestamp, but they’ll only return like a hundred trades per millisecond. So when trade volume really increases, there’s just a blind spot. You just can’t even see the data, there’s a whole bunch that’s being missed.
Sam: Oh yeah, it’s a big problem. On some [00:19:30] exchanges, when trading gets busy you literally can’t get all of your fills. You’re told half your fills, and you have to try and piece it all together based on the other random evidence you get.
On the topic of APIs, which you have to use if you’re going to really be trading on a lot of venues at once, a lot of them are total […]. There’s this rest issue of, what happens if trades happen […]? There’s the, oh, this exchange doesn’t tell you your balance. That’s not an API endpoint they have. Some exchanges [00:20:00] have that issue.
How about withdrawals? There’s no API withdrawals, so you have to be manually type in all these addresses. First of all, it’s terrifying. You’re gonna make a typo at some point or mis-click or something. But second of all, it just limits the set of trading you could do that doesn’t involve that many transfers.
You can look at exchanges where the API calls are all per market, so if you want to figure out your open orders, you have to call, get open orders for the ETC/VTC PAIR, [00:20:30] and you have to call that separately for each market. Sure, you could call it for all of the markets, except that you’re rate limited to one API call per second. So you would have to eat up three minutes of API calls just to get your open orders.
Clay: And they’re using nonstandard symbols, like they’re using XPT or…
Sam: Oh, yeah. Exactly. It used to be a big problem of like, some exchange says that there’s BCC, and you have no idea if it’s a fraud or if it’s Bitcoin cash.
Clay: Last question on origins [00:21:00] in Alameda. It says on your website that you manage over a hundred million in digital assets and trade 600 million to 1.5 billion per day across thousands of products. Do you have a sense or an estimation of what percentage of average daily legitimate exchange volume you’re on one side of the trade of? Or is coming from you?
Sam: Yeah, it’s somewhere in the single digit percent. I think 5% is a decent guess. Somewhere between [00:21:30] 2% and 7%, maybe. You know, it varies a little bit from day to day, but that’s the ballpark figure.
Clay: That’s incredible. That’s definitely up there with the high end of any HFT trader would do in a US equities market. That’s really incredible. I guess one more question is, have you found that, given that volume, that you get reached out to by exchanges, and are [00:22:00] there nice maker/taker schemes that you take advantage of, as a market maker and such?
Sam: Yes. A lot of exchanges have, if nothing else, volume-based features, and often we can get in the top tier there, and that does help a fair bit, especially when it comes to eeking out a business plan. […] That’s nothing. You made one basis point. If you guess randomly, maybe moderately right, about what Bitcoin is going to do in the next two seconds, maybe you make a basis point. [00:22:30] But if you can actually do that on one billion dollars a day of volume, then you can do the math, that’s actually a lot of money. We can’t always do that, but it sort of does mean that at some volume scale, you really do get down to, everything matters. Every little edge matters. So there is some of that. A lot of exchanges reach out to us about being liquidity fighters. We do do that on some exchanges.
I’ve actually been dialing that back a little bit, and part of the reason is, I would say it’s a few things. First of all, it’s just a huge load on our dev team to connect to more and more exchanges, [00:23:00] but the second is that we are finally starting to see some centralization of volume. Not a lot. Crypto volume is still very decentralized across exchanges, but less so than a year ago. A year ago, there was this long tail of 100 spot exchanges, each of which had non-trivial amounts of real volume each day. Some of them were faking $5 billion, but they still had $30 million in real dollars. That long tail has come in from [00:23:30] 100 or so platforms to 40 platforms over the last year. We’re seeing that trend continuing, and that just means that there’s less and less happening on exchanges that never managed to find their role on the market.
Clay: Let’s transition to chapter 2 which is about the relationship between Alameda Research and FTX.
Sam: Alameda had been trading for a year-and-a-half, I guess, at that point, on markets. There’s a pretty stark difference between [00:24:00] spot and derivatives markets. To get a sense of that, you look at a place like Coinbase, like Binance, like BitRex, and I don’t want to say they’re perfect spot exchanges, but you actually do get the sense that they’re on the path towards that. If what you want is a spot BTC/USD order book with no margin, no nothing, and you see the Coinbase and you’re like, “I don’t know. That works.”
The matching engine’s fine; no problems with it. It’s easy to get the dollars in and out, [00:24:30] easy to get the Bitcoins on and out, and that’s sort of the whole thing. Because of that, there’s still a lot of spot exchanges that manage to massively […] it up, but there are also a lot that do a pretty good job. So we see the role for innovation in unlevered spot trading of the most liquid assets as not that big right now in crypto, all things considered. But derivatives are a totally different story.
In the derivatives space, there’s actually only a few large platforms, and a lot of them are really bad. They have had a lot of really high-profile [00:25:00] problems, from hundreds of millions of dollars of customer funds lost, to mishandled risk of other users, to match engines that just can’t take the order load, to just extremely clunky products where you have to individual collateralize every single futures contract you trade with different types of collateral. You have BSD collateralized, BSD futures, EOS collateralized, EOS futures. It’s a massive pain managing it; you can’t move your collateral back and forth, because these are all different assets. [00:25:30] You get liquidated in one contract despite having an extra million dollars sitting around in another that you can’t move back and forth, to just a general lack of innovation in this space.
Not only are there really big problems with the top derivatives exchanges, but what do they have? They have futures on the top six coins. That’s what they have. Which is great, but it’s not in exchange for being poorly built products. They’re really moving the goalpost forward in terms of product offerings. There’s a whole structured product space in finance [00:26:00] and a whole derivative seal then, and so far cryptos only sort of done step one, and it’s done it poorly. It’s just a huge contrast. With the spot markets, we saw this and we were like basically, “[…] it, we can do this better.”
We have a lot of experience with these platforms. We know exactly what we do and don’t want from them. We know how they work. We’ve written white papers on exactly what parameters we would change about the existing products incented to these exchanges. We have a matching engine in our back pocket. We have a wallet infrastructure [00:26:30] built. We actually have a lot of the pieces put together. We know how to do this. We know how to build these products. We have a lot of experience in traditional financial derivatives. We know how they work, we know what the range of products are, what reasonable directions to go would be. That’s one side.
The other side is that they’re only like five of these platforms and the median one was making hundreds of millions of dollars a year in revenue, on what seemed like kind of a mediocre product. So it just seemed like a huge opportunity to enter this space and [00:27:00] build something better.
We sort of saw all of that and just felt like, “Boy, we can do this.” So late 2018 we started development, and ended up rolling out FTX in April of 2019. It’s a redesigned crypto derivatives exchange that’s made all the changes that we’ve wanted other platforms to make for years.
So, what does that mean? It means first of all, a cross margin between all of the different products. It means that it has stable coin-based collateral which simplifies so many things. [00:27:30] You still need a bank account, but it means that you can actually just have normal, like a Bitcoin USD future, where your PNL is the number of Bitcoins you traded times the difference in price. You have all of them settling in same stablecoin collateral.
We have both quarterlies and perpetuals. We also have index futures, which are a big thing missing from this space. We have an altcoin index future, a […]coin index future, a mid-cap index future, which are pretty cool, useful products that just obviously should have existed. Because often people do feel like, [00:28:00] “I’m bullish on altcoins, I want to get long holds.” But they don’t want to go figure out how to margin long each one of them and manage all those separately. Now you can just get long, the allcoin futures contract. Or if you’re really bullish on Ripple you can buy a bunch of Ripple and hedge it by shorting the altcoin contract to hedge out your general altcoin sector risk.
The last product that we have right now is the leverage tokens, which are a cool thing. They’re basically, what we did is we tokenized levered futures positions. So these are ERC20 [00:28:30] tokens that you can buy to have plus or minus 3X leverage to swap. Then, you don’t have to worry about liquidations, or margin, or collateral, anything. You just trade them like tokens. But one of them is like a short 3X token.
So that’s what we have now, and we’re moving in not very shocking directions and we’re hoping to have options out within the next couple months. Then, also looking into things like hash rate futures, Bitcoin dominance futures, prediction markets, and a whole host of other things.
Clay: Why do you think there are so [00:29:00] few players in this space? There’s an explosion in crypto to crypto exchanges, but when it comes to derivatives and options platforms, especially crypto native ones, there just weren’t a lot.
Is there something harder about doing that? Because the demand seemed to be there, and you’re right, I did hear some dissatisfaction about some of the exchanges in that space. Is the barrier to entry somehow higher?
Sam: Yeah, it’s quite a bit higher and there’s two big reasons. [00:29:30] First, you need liquidity for derivatives in a way you don’t for spot. Liquidity is always good, but on the spot, even if you have an illiquid market, if someone comes in and buys a hundred million dollars of Bitcoin there, maybe they get a bad price, but you’re pretty happy with the exchange.
Clay: Hey! I wanted to pause for a second to let you know that this episode of the Flippening podcast is brought to you by the Nomics API and CSV Data Export Service. As a sponsor and producer of this podcast, I wanted to give you an announcement that I’m doing a webinar every week day, on [00:30:00] crypto data and how it works. You should join me. The webinar is called Crypto Market Data 101: Fake Volume, Exchange Spam, and How The Seedy Market Data Underworld Actually Works
On the webinar we discuss: (1) How exchanges use exchange volume spamming and ticker stuffing to spam CoinMarketCap and other aggregators, (2) What everyone is getting wrong about transparency and fake volume, (3) Why most price aggregators are displaying bad data, (4) The three types of pricing data and why everyone is using the wrong one, and finally, the two transitions you [00:30:30] must make in order to move from Inaccurate Crypto Data to Good Crypto Data, and much much more. To join me on a webinar, go to nomicswebinar.com. Okay, back to the show.
Sam: If you’re a futures exchange and you’re illiquid, and someone puts on a hundred million dollars position and then they get liquidated, it’s a […] show because you just don’t have people ready to take that position on. It’ll crash your market to zero, and then there’ll be a massive bankruptcy. There’s going to be probably like a $30 million [00:31:00] loss that someone has to bear. It’s either you, the exchange, or your customers.
We saw this happen with platforms where they got bigger than their liquidity could allow, and there’s millions of dollars a week being clawed back because people were getting beyond bankrupt and no one could take on their positions. So the first thing is just really high liquidity demands.
Obviously, Alameda gave us a leg up there. We had one of the top liquidity fighters in crypto in our back pocket, so it meant that from before the first trade [00:31:30] ever on FTX, it had probably the most liquid order books in crypto. So that’s one of the big issues.
The other hard thing is, you have to get the parameters right. There’s a bunch of fiddly things here of like, what’s your initial margin? What’s your maintenance margin? How does it scale with position size? Is it different for different futures? How do you liquidate people exactly? If you screw those parameters up, even if you have enough liquidity, you’ll end up just tanking markets with really poorly executed liquidations.
So that’s another thing where it’s not [00:32:00] impossible, but you just need to do it well. There are a bunch of different parameters that go into it.
Clay: I have a question about what boundaries exist between Alameda and FTX.
Sam: Clay: So it’s perfectly appropriate for an OTC desk to take one side of the trade, in two-sided marketplaces, like an exchange, where there’s these two-sided order books. If you are a liquidity provider and you’re filling some of those trades, you could have [00:32:30] privileged access to data faster than anyone else. Obviously, they’re on your servers.
What kind of rules and regulations exist there? And also, what are the norms? How do you think about this, the boundaries between these two things?
Sam: It’s a good question. Basically the way we think of it, is we think of Alameda as currently the biggest liquidity provider on FTX, and not as being a special account in other ways than that. Alameda plugs into the API in the same way that [00:33:00] all other accounts do. They don’t have lower latency or anything, they’re just shown the same API endpoints. They have the same fee structure that everyone else has. There’s no last look or anything like that. So really, the Alameda Research account on FTX is not privileged over any other accounts on the platform because of the relationship between the two companies, is extremely incentivized for FTX to do very well.
So because of that, it’s providing [00:33:30] a ridiculous amount of liquidity on the platform. Rather than trying to maximize its trading profits on the platform and only pride as much liquidity as you sort of gain […] optimal, in order to make as much as it can, on this spread, it’s trying to provide as much as it can without losing money on the platform, in order to make the platform as liquid as possible. If you think about the incentives here, this [00:34:00] short-term PNL versus long-term growth of FTX trade off is just hugely in favor of long-term FTX growth. So really the dominant factor here, is just what can Alameda do to help FTX grow as much as possible?
Clay: you could have people who, on one end of this debate, say it’s de facto unethical. Then, you could have people on the other extreme saying someone placed an order [00:34:30] and they’re just filling it. There’s nothing wrong about this whatsoever.
I think I could see both sides, but I guess I’m a bit more on the side of someone placed an order and you’re just filling. You know what I mean? No one’s forcing anyone’s hand. No one’s being swindled. So even if you did have privileged access, I can see why you would want to maybe dog food your own APIs. Just from a product perspective, you’d want to consume the platform the same way any other market maker would, just [00:35:00] because it forces you to make a better product. But I’m wondering if you have an ethical obligation to do that. What do you think?
Sam: I think our obligation is just to be transparent about what we’re doing. I think we want to make this a different type of thing. We could say, “No, this isn’t like a standard exchange. This is a single dealer platform, where the market maker, who is Alameda, gets last look. That’s how it works. We’re being totally upfront about it, that’s what it is.” If that’s the thing you want to use, use it. If you don’t like that don’t use it. But you’ve been [00:35:30] really obviously warned.
I think that would be okay as long as we were really, really, really clear and obvious about how we presented the product to people, so that they know what they’re using. As you said, that starts to look more like just an OTC desk. Alameda does also have an OTC desk that is more on that model, where you’re just trading against Alameda. It’s more of the typical, we have chat quotes and things like that, and obviously Alameda [00:36:00] sees your order because they’re talking to you in a telegram shot that, negotiating the trade.
It’s a different sort of business. But that’s not how we see FTX. FTX is a standard, public two-way platform that anyone can trade on against anyone else. As it happens right now, Alameda is on the other side of a lot of trades. It’s a lot of the liquidity there, but we’d be super excited for two people who aren’t Alameda, to trade against each other on it, because this is a fee-based model, [00:36:30] not a single dealer, liquidity based-model.
I think that’s a platform that has the higher growth potential. I think it’s the right long-term play. I think the important thing is just that we communicate correctly what it is, and that because we’re presenting it as that to people, as the standard two-way exchange where Alameda is not privileged, then I think we do have a pretty high ethical obligation to in fact make sure that that is the case. So I think that’s the big thing I’d say is just that we accurately [00:37:00] present the product as it is.
Clay: It would seem to me that in the short-term you could benefit from privileged access, but in the long-term, assuming the continued growth of the crypto marketplace and the huge need for derivatives exchanges to have liquidity, it does behoove you to have as many liquidity providers and market makers on the platform. Because at some point the numbers just get so large and you’re going to need that there. Then, if you burn those bridges and no one [00:37:30] else is winning beside you, yeah, you’re in a bad spot.
Sam: Exactly. We’re working to onboard more market makers. We have a few others on already. We’d love to have even more. We’d just love to bring as many people there and try and make Alameda not that important for it. That’s our goal, is for it to be totally fine if Alameda pulls out of the exchange. We’re not there right now but that’s the long-term vision for the product, is one that is, on which Alameda is just one liquidity provider of many. Or if we can eventually get away with it, [00:38:00] maybe even not trading there at all.
Clay: going back to the origins a little bit more, you guys have ramped up really fast. You might be experiencing just as fast of a growth trajectory as Binance. It’s hard to say. I don’t have the numbers, but it’s pretty incredible.
When I interviewed CZ, I think one of the most interesting things that came out of that interview was the fact that he had been doing trading systems for 10 plus years, and he actually had a platform or a [00:38:30] business that provided white labeled cloud exchanges for third parties. So they’d come in and they’d provide the business entity, and perhaps fiat on-ramps, and all that stuff. But they would run the matching engine, the wallets, the order book, and stuff.
So when they started, it was very easy for them to scale because they had just withstood that kind of pressure before. They weren’t, in the case of a lot of exchanges, where just everyone’s [00:39:00] excited about crypto and doing all this stuff the first time.
So it seems like one advantage you guys have, from a product perspective, is that you’re very sophisticated derivatives traders, so you understand the kinds of products that can exist there and also the needs of traders very well. So in some ways, you’re eating your own dog food, or at least building the kind of dog food that you would like to consume.
But can you tell me a little bit about the order book and the wallet? You said that you already had a wallet infrastructure. Is this a fork of some kind of existing codebase [00:39:30] that had been around for a while? Did you do it 100% from scratch?
Sam: It’s basically all from scratch and obviously we looked at existing products, but almost all of what we built, we built from scratch. It’s something that became more and more true over time. We, again and again had this experience of we use a third party product, it seemed kind of good, and then we realized we want to make some changes to make it better to suit our needs, and we can’t because it’s a third party product. It is what it is and eventually, we just get super frustrated [00:40:00] with it, throw it out, and built own thing.
At this point we’re like, “All right, we know how this is going to end.” So, it’s basically all built in-house. A lot of these things are things we’ve been prototyping for a few earlier products that we built. Before launching the exchange, we had built an OTC desk, which had a wallet system in it, it had an order entry system in it, it had a website-based GUI system in it, and it had an account system in it.
So that’s where it, some of the exchange infrastructure, [00:40:30] we’d developed it earlier for that. The wallets, also obviously we need to do that for our trading. We’re sending a lot of coins around. Then, on the matching engine side, that was more of a product that we just built from scratch. We looked at how the existing, especially derivatives matching engines were structured, because there’s this risk with derivatives engines that they need to do risk checks. Not just sanity checks on the markets, but do we need to liquidate and account checks?
You have to think about how you interleave that with actual order matching. [00:41:00] So we looked at what the existing products were doing, how they were structured, what the pros and cons of them were, and then built out infrastructure. We’ve made some substantial changes to it. Over the last nine months or so, as it became clear that some particular part of it was going to be the […] or main part of the matching engine, and that we had to find some way to parallelize that.
Building a core matching engine, if it doesn’t need to be fast, is not hard. The core product is pretty simple. The hard part of it is just, [00:41:30] how do you figure out exactly what needs to be computed, in which places, in real time versus what can be parallelized? How do you keep everything in sync and how’d you do this?
And you need to start doing this because you need to speed up. So you can’t just have one process that you run through for everything, but you have to make sure that you’re not executing orders based on stale balances data. So that’s the tricky thing. That’s something that’s come along bit-by-bit, as we’ve needed to split out parts of it.
Clay: Let’s move to [00:42:00] chapter 3 which is about the current state of derivatives trading. What was available when you were operating in the traditional markets that wasn’t available in crypto when you started FTX? And I guess, kind of flipping that, what opportunities and products do you think can exist for crypto derivatives traders that just can’t be in the traditional markets?
Sam: I’ll start with what were the things that Wall Street had [00:42:30] that crypto doesn’t have basically, in terms of derivatives. I think there are a few big things. The first thing I’ll say is indexes. Index products are all over traditional finance. From ETFs to futures to mutual funds, basically the whole industry runs on index products. Most of what people do is not trade a single stock. Most of what people do is trade an index of stocks, or commodities or something like that, and see what the highest volume products in traditional finance. [00:43:00] It’s not Google. It’s the S&P 500 index future.
Why is that? Well, I think the big reason is if you ask yourself, “What really are people trading? What are they expressing? What need or desire do they have to trade or to own things?” There’s a lot of different companies in the world and unless you’re going to be an expert on doing research on Apple, you might have some desire to own stocks or to trade stocks but you probably don’t know that much about any single one of them and you probably don’t want nothing [00:43:30] but a single one of them. You don’t want your entire portfolio to be Apple because then if it goes under, you’re broke.
And sort of the default assumption in traditional finance is if you’re buying stocks, you buy a lot. You just buy a little bit of every major stock, and you do that through one of a few different types of index products, whether it’s a Vanguard Mutual Fund, whether it’s an iShares ETF, maybe you’re trading futures on an index of stocks, but that sort of is the default. If you want to trade something in emerging markets, [00:44:00] you wanted to get some exposure to Korea, you would buy Korean ETF or Korean future or something like that. That has been missing in crypto.
The most obvious application of that is index futures of a basket of different coins. There are a lot of people who have opinions about […]coins. There are a lot of people who think […]coins are going to go down. They want to short them, but first of all, you can’t really short most of them. Second of all, it’s not that they’re like this one, like this is [00:44:30] the one that’s going to zero. They’ve never heard of most of them and that’s sort of the point. Their whole point is these things collectively have this $30 billion market cap and there’s barely any real companies there. But, they don’t want to end up shorting the only one they’ve heard of and that terms out to be the diamond in the rough.
So, the most natural thing you want to do is just short the whole lot of them. Short a basket of them. Or conversely, sometimes think it’s […]coins. When I say […]coin, I don’t mean this in the sort of, [00:45:00] “Wow, it’s a piece of […].” I mean this in the sense of taking over crypto, that’s the word for a speculative small market cap coin. I say that somewhat endearingly in that I’ve issued a token FTT, which is by some definitions a […]coin, if you go by market cap or daily volume. So, I certainly think there are some diamonds in the rough there.
Maybe you think, “Look, these are actually getting their act together. I don’t know which ones, but I think some of these are really good projects. Their market caps have [00:45:30] been going down for a while, and I think that a few of these are going to take off, a few of these are going to become $5 billion tokens. I have no idea which, there are so many of them. But I think enough of them to justify substantially higher valuation then the basket has right now, and I think it’s pretty plausible that at some points, you would want to get levered long this index.” Again, you might have no idea which coin you want. So, that’s sort of the most obvious product to translate over because they’re all over traditional finance, [00:46:00] then just making common sense in crypto.
Again, one other obvious use of these, which is, you found that one diamond in the rough, you want to be really long that, but you don’t want to just lose all your money if the whole […]coin industry goes to zero. So, you buy the token that you think is a really good project and you hedge it by selling, by shorting a […]coin future so that you’ve sort of hedged against overall industry movements, and really are just making a bet on the thing that you have an opinion on, which is the relevant performance of this one company, [00:46:30] this one token. That, I think, is the most obvious product to port over.
After that, options obviously are sort of a prominent one, there’s a lot of options trading traditional finance, and we’re just scraping the top of the barrel right now in crypto. There’s a few true platforms in the world that have options. They’re pretty illiquid. They’re only on bitcoin and Ethe, and there’s a lot more potential there. I could imagine a lot of people wanting to buy moderately out of the money upside [00:47:00] calls on a […]coin index future for instance. That seems like a bet a lot of people would have a real opinion on.
You could also imagine that in traditional finance, there are a few volatility products that are extremely liquid. A lot of the world’s options liquidity ends up centering on these products. So, if you want to just make a large liquid bet on volatility, you can do it. Right now, in crypto, you basically can’t. There basically doesn’t exist an option that you don’t lose a massive spread trading. So, that’s [00:47:30] another obvious product that a lot of people would have opinions on. What’s short-term volatility in crypto going to be like? What’s long-term volatility going to be like? That’s another obvious product.
The last thing that I’ll say, there’s a whole host of things but sticking to the things that I think are the next frontier for crypto. There’s one other really core piece of infrastructure, that traditional finance has that crypto basically doesn’t, and that’s this concept of a centralized book. That if you’re trading [00:48:00] in traditional markets, and you get long Apple stock, and you get short on the New York Stock Exchange, you get short Google on NYSE, it’s just assumed that you’re going to get better merchant treatment than if you’re long both of them. That someone in the world is going to be the master of your risk, is going to say, “Those are positively correlated stocks. You’re long one, short the other. That’s a less risky position than being long both or short both.”
So, yeah, obviously, you [00:48:30] should be able to lever up more on that than getting long both and getting long Apple on NYSE and short Apple on NASDAQ, you should be able to lever up basically infinitely. That’s just an arbitrage. Those are the same thing. Then you go to crypto, and it’s the complete opposite. There’s literally no risk relation at all between a Binance bitcoin and a coin-based bitcoin. You get short a Binance bitcoin and long a coin-based bitcoin, you don’t get any merchant relief at all. Then bitcoin goes up 30%, you get liquidated on Binance for your short position, [00:49:00] even though you had the bitcoin sitting on coin-based, you’re perfectly hedged. But Binance didn’t know that. In fact, no one knew that.
There’s no one who knows anything other than you’re really local book. It gets even worse. Even when it someone does, they don’t do anything with it. If you get long a bitcoin future on OKX and short an Ethe future on OKX, even that doesn’t give you any merchant relief because they’re totally different merchant systems, totally different liquidation systems, totally different risk systems. Even though it’s literally through the same company, [00:49:30] the same exchange, looking at both these positions, and they don’t net them out at all, what this means is that if you want to be a lot of trading or holding complicated sets of positions in crypto, it’s unbelievably capital intensive.
You ask why we need a big balance sheet if we’re doing arbitrage, it’s because we need to separately collateralize all of our long positions and all of our short positions, even if they exactly cancel out. This creates unbelievable amounts of friction in the market. You ask why there sometimes exists a huge arbitrage, it’s because people [00:50:00] didn’t happen to have an enormous amount of capital […] of the right type of coin sitting on the right exchange at the right time ready to sell.
Now all of a sudden, you go from needing $1 million to needing $1 million of every coin on every exchange. And, oh boy, that just multiplied your capital requirements to put on, to be able to do the same thing by 100.
So, that just really key piece of infrastructure in traditional finance, which is basically the clearing firm. [00:50:30] It’s the firm that manages your book, and NYSE doesn’t manage liquidation risk really. NYSE just says you’re clearing through like JP Morgan Chase. Obviously they’ve got a trillion dollars. Obviously they’re not about to default on this $50 Apple position. They’re like, “Look, JP Morgan Chase, this is your deal now. You’ve got to make sure your clients are okay and you have to be good for them if they’re not.” And then JP Morgan Chase is like, “All right, I manage your book. You trade through me through all these exchanges, I can just see your [00:51:00] entire book here and now I decide netting all your positions together how much you should be able to put on.”
Again, this is a really key part of infrastructure in traditional finance that really greases the wheels and it just doesn’t exist in crypto.
Clay: I haven’t given a lot of thought to that there’s companies that are coming up that say they’re meta-exchanges, or there’s organizations like SFox that’ll distribute your order across multiple entities so you don’t plow through all the order books. [00:51:30] But that’s something that just kind of exists for free if you have an E-Trade or Scottrade or whatever account. You’re not trading directly on these exchanges. You’re trading on an interface that allows you to place orders on exchanges and then whoever your broker is sees the positions you have, and can extend you credit or margin based on what they see.
Sam: It has a big impact, even if you look at these multi-router type things in crypto. First of all, they don’t do any [00:52:00] risk management, but even if they did, there’s another problem they don’t solve, which you’re like, “I want to sell a bitcoin on Binance.” They’re like, “But your Binance account, it doesn’t have any bitcoins. You can’t.” It successfully has hooked up to a lot of exchanges, but it’s not giving you free capital. It’s giving you risk capital, maybe, but not physical bitcoins to play around with. So, you still need to store a bitcoin on every single exchange you might want to sell one on.
Clay: I think that something that you might have given a [00:52:30] lot of thought to is the creation of indices. in a highly regulated financial environment like we have in the United States, we take these indexes for granted. A company sets forth a methodology and maybe some committee meets quarterly and decides if they want to change it. But we can more or less trust it. If you’re creating an index in crypto that is not directly collateralized [00:53:00] by the things in the index, it seems like it presents a whole lot of problems. Like, whose exchange’s data are you believing? If you have a basket that represents all […]coins, do you include stable coins? Especially if the stable coins are collateralized by USD, should it really count?
There’s just a lot of things that you have to trust if you’re going to let someone create an index that is not [00:53:30] directly collateralized by the assets being held. Of course, that’s the name of the game in derivatives often is that a lot of these things are not directly collateralized. What are your thoughts on the proper way to create crypto indexes that are not directly collateralized by the underlying assets?
Sam: The first thing I’d say is it’s important to understand the use case for it. As you said, there’s a big difference about whether it’s collateralized, and it’s sort of the difference between a future and an ETF. [00:54:00] You can imagine an ETF or a mutual fund or something like that where it’s this thing that owns a lot of things but actually physically holds them. That’s the whole point of it is that you’re buying into this fund that goes and buys a lot of things for you. You expect to be able to actually turn it into those underlying things, that it is just sort of a basket of things.
If you want to, you can always think of a stable coin as an ETF on the dollar, like what is tether? It’s a token that says it holds $1 [00:54:30] per token, and you can redeem it and get your dollar out and you can send your dollar in to create it. It’s sort of a boring ETF, but it’s an ETF and it’s a pretty useful one. But when you’re looking at stable coins, the whole point of them, part of the point is to have a token that’s worth about $1, but part of the point is to have a deliverable so that you can actually physically turn your token into a real dollar in your bank account.
When you look at index futures, on the other hand, it’s sort of the opposite. It’s explicitly saying this is a futures contract. Nothing is getting delivered here. It’s all cash settle. [00:55:00] So you’re not doing this because you need to settle some OTC trade in Grosso coins, so I’m buying the […]coin index future to get the Grosso coin. That’s not happening. The whole point is you’re betting on price action.
Then, what becomes important is not the deliverability exactly of the index components. It’s whether their price action is what people are expecting. What you end up asking yourself then is, “What do people expect?” There’s this thing that’s listed on FTX, which is called mid. It’s the midcap [00:55:30] index future. What is that? When we tell people we have a midcap index future, what are they expecting? We want to deliver the product that’s a product they want and expect.
You get to questions like, “How bout stable coins? How bout a midcap stable coin? Should that be in mid?” And the answer is you can do it either way. It works either way. What’s the right decision? By far, the most important thing is that you don’t lie about what’s in it. By far, the most important thing is that you publish exactly what it has. “This is [00:56:00] exactly what this future will expire to.” You just go compute it and it’s got some stuff in it.
So, that’s just by far the most important thing is that you’re really, really, really clear about exactly what it is so that everyone is on equal footing when it comes to pricing it. I think the right answer is no, it shouldn’t have a stable coin. It’s not because it would destroy the index. It’s because it would not represent what people actually want. No one’s like, “Oh yeah, […]coin index. I bet that’s going to be 30% US dollars because of all the small cap [00:56:30] stable coins.” That’s exactly what people don’t want. People want things that look nothing like a dollar, that are tokens with nothing behind them but speculation. That’s what you’re betting on when you’re betting on small market cap cryptocurrencies.
So, the key thing is first of all, be transparent about what’s in it. Second of all, build the product people actually want. Think about what the use case is, and think about does this represent the use case? And so, we drafted some criterion for the alt-coin index future. We thought about it for a second, and we were like, “Oh God. [00:57:00] I guess we need more criteria because this thing’s going to be 25% tether.” And obviously, that’s not what people want.
So, yeah, we exclude all stable coins from the index futures, not because it would destroy them but because it’d be lame. That’s not the part people want to be trading. Then, you get to questions like, “Well, exactly what data feeds do you use?” And that does actually become quite important. This is when you get back to deliverability a little bit. There’s nothing deliverable about an index future but the thing that proves that it’s [00:57:30] constructed fairly is that if you wanted to, you could deliver the exact thing that it expires to.
If you could look at every single index component, every market of every token in the index, you could buy that token on that exchange, that market of that exchange at the price the index says it’s trading at, and then withdraw it to your wallet. If you can do that, then that should directly be worth what the token is worth. [00:58:00] If it’s not, if that market’s trading at 50% the price of the token, then you can fix that. You can just go buy it and sell it somewhere else and do an arbitrage.
In the end, one of the key things about your index is that it could theoretically be delivered. If it could theoretically be delivered, then it is in some sense it’s directly a fair representation of the products you’re claiming it is. The second thing beyond that is how liquid is it? Theoretically, this could be delivered but let’s be honest. When the […]coin September [00:58:30] expiration comes, I don’t expect traders to go out and be like, “All right, we’re hedging these 400 markets right now. Have we successfully bought our three Grosso coins on the […] Grosso coins tether market?” People aren’t actually going to do it. It’d be a nightmare.
In addition to theoretically, you could do it, you want to do your best to choose as liquid of markets as you can. Now, obviously for the bitcoin index, this isn’t that hard. There’s still some tricky things, but it could be harder. Coin-based bitcoin USD, yeah, that’s basically a bitcoin. The honest truth [00:59:00] is if your index just expires to that, it’s kind of okay. You’re going to be within like a bip or two of the right expiration.
Clay: Hey, this is Clay cutting in to quickly define what a bip is for those of you who are uninitiated. Bip stands for Basis point. A basis point is one hundredth of a percent. Basis points are used as a unit of measurement in contexts where percentage differences of less than 1% are discussed. For example, a difference of 0.10 percentage points is equivalent to a change of 10 basis points. And, an increase of 100 basis [00:59:30] points means a rise by 1 percentage point. Okay, back to the show.
Sam: With […]coins, you’re not going to be close. Just look at this and be like, “[…] a token that’s in the index and it trades on four markets, and every single market is at least 20% wide.” Even if there’s not an arbitrage, their centers are just 10% away from each other because they’re really illiquid and wide, and all of a sudden it matters which markets you choose. It’s not really hedgeable yet, theoretically it’s hedgeable if you want to cross [01:00:00] the 20% spread. Now you’re saying, “Oh God, am I marketing to mid, to last, to bid, to offer? Like I don’t even know.” All these numbers are really different because this is a super illiquid market.
Now you start getting bogged down in the weeds, and the thing that saves you here is strength in numbers. The thing that saves you is as long as you make generally reasonable decisions, the […]coin has like 100 tokens in it, and it’s got like three markets per token. So, it’s expiring to an average of 300 different markets, which means that [01:00:30] if a single one of those markets is off by 5% from the fair price, what’s the actual impact on the expiration price of the future? It’s less that a bip.
That’s the thing where you have to make sure that you’re making reasonable decisions, that you’re not making systematically biased decisions, that you’re not choosing markets that are so bad that they’re just all egregious and the index is just expiring to a random number. As long as you can generally make reasonable decisions if your index is big enough, they’re going to [01:01:00] wash out.
Clay: I think you make a really good point about thinking about what most people would expect. Often, it’s not about the letter or what’s technically true or not true. It’s about what would the average investor who hasn’t read every page of the docs come to expect?
We ran into this with our market data API. We were publishing all-time highs for various coins and we were getting a lot of flack over something that we thought was really cool. We were publishing the true all-time [01:01:30] high for each coin. We thought it was super cool because we were really transparent. We told people, “Not only is this the data, the all-time high but this is the trading pair and the exchange where the all-time high happened.” People were coming to us and saying, “That’s not the right all-time high. These four other websites say that this is the all-time high and I don’t remember ever being able to buy it at that price.” It was like, “No, this is a true outlier.” So, even among the all-time high, people didn’t want the all-time high. They wanted something that they could have bought at on exchanges [01:02:00] that they use.
Sam: Exactly. Like the highest price that it trades for any sustained period of time on a moderately liquid exchange or something like that.
Clay: Yeah. That’s not Korean or someplace where they sort of…
Clay: I think maybe the next component of this is pricing data. Okay, so you would exclude stable coins because people aren’t betting on USD. It’s the opposite of that. But what about pricing feeds? Whose version of reality do you use? [01:02:30] We have customers that create indexes and they’re like, “We’d like to use these five exchanges that don’t prohibit us from creating indexes based on their data,” but it’s still highly reputable and no one, no person would look at this and say, ‘This is … that an index is created from these prices.'”
I think the other component is outlier detection. Do you exclude things that are blatantly outliers? Do you include or exclude flash crashes or things that are outside of maybe [01:03:00] an interquartile range and stuff like that? What are your thoughts on using exchange data, what exchanges you use, and then outliers?
Sam: So I think I’ll start with just the general pump, then I’ll go back to front with those. The first thing is that this is something Alameda Research has to think about it a lot because we’re trading these coins. If we make bad decisions about this, we’re wrong about basically what an asset’s worth. To give you a good example of this, [01:03:30], which is earlier today, the story, and I’m just repeating what other people have told me. The background behind this is, not saying I know this to be true but this seems to be the consensus, something something AWS Tokyo forgot to turn on an air conditioner something something some computers overheated. That part of the story is maybe a little speculative but that is what has been told, though.
Anyway, Amazon Web Services Tokyo servers had issues. There are [01:04:00] a lot of things in those servers. One of those things is Binance. Actually, FTX is another one of those things. Something something Binance something. Again, it’s not clear exactly how this chain of causation happened. But it’s not shocking that it did. The AWS data issues caused Binance to briefly spit out nonsense market data. As far as they know, there weren’t any crazy trades on Binance. But it did briefly publish trades that didn’t happen [01:04:30] because it was getting bad data from its servers at crazy prices, including some Bitcoins trading at like 30 cents or something.
This then had a bunch of downstream ramifications. One of them is that crypto Twitter had a field day. But maybe more impactfully, there are a lot of people who are pricing points around Binance. Maybe around multiple things but Binance is obviously an important part of that. If you’re pricing around Binance and Binance’s feed publishes that Bitcoin’s worth 30 cents, what do you do? [01:05:00] Well, it kind of depends on exactly what your algorithm says to do in that case. But for some people, the answer’s that they’re excited to sell Bitcoin for 50 cents because it’s arbitrage by a factor of two and a half. So, Binance didn’t have any crazy trades. But the garbage market data published by Binance because of the AWS data center issues caused some market makers to sell coins at 10% of their fair value on other exchanges, thinking they’re doing an arbitrage.
Now, all of sudden on [01:05:30] unrelated exchanges, you get actual crazy prints that happen and each exchange has a different response to this. Some freeze withdrawals, some take down their servers, others let everything go. Some are thinking, “Oh God, do we have to bust these trades?” There is a lot of mayhem caused by exactly this earlier today. Again, I’ll get to what our actual policy is, but I think the approach to this is God didn’t come down from his mountain and say something like, “And the fifth commandment is this is how thou shalt handle outliers in cryptocurrency market data.” [01:06:00] There’s a lot of parameters that you have to choose and it’s kind of arbitrary. There really just isn’t a consistent convention here.
So, rather than figuring out the perfect answer, what this comes down to is two things basically. First of all, do you choose a reasonable answer? And second of all, do you clearly communicate ahead of time what answer you’re going to choose? As long as you tell people, “This is exactly what we’re going to do,” that at least helps because now everyone’s on equal footing about what the product is. They see these data issues and they’re like, “Whether or not [01:06:30] the index is doing something reasonable, at least I know what it’s doing. I know what it’s going to do and I can trade accordingly.” So, those are the two commandments. There is make reasonable decisions and communicate clearly about what decisions you’re going to make.
What do we do on FTX? The first thing is we do have some outlier handling. For each coin, we have some set of markets that represent the price feed for that coin. By default, what we do is we just average those prices, but we cap each contributor at, I think it’s [01:07:00] something like 35 basis points away from the median of those data points. So, we just take wherever the middle value is and we don’t let anything deviate from that by more than 35 bips.
Is that perfect? No, obviously not. Thirty-five bips off a piece of that is an arbitrary number, but it does reasonable things, like if one index goes to zero and there are foreign marketers here and there are four things in, then that’ll move this index down seven bips or something like that, which is okay. [01:07:30] It’s a reasonable thing to do. […] perfect but it gets the job done. So that’s what we do within each market.
Then there’s a question of how do we choose these markets in the first place? What exchanges do we use? And the answer is basically, we try and go for the exchange that maximize liquidity and withdrawability. What I mean by that is first of all, obviously, order book liquidity matters a lot here.If there’s a super liquid order book, then that’s probably a true representation of what the coin is worth. [01:08:00] Because if it weren’t, you could make a lot of money trading against it.
If your book’s super thin, maybe it’s off by 50% but it’s 80% wise, there’s nothing you can do. The other key thing again is even if it’s the most liquid order book in the world, if you can’t withdraw the coin from that exchange, there’s not actually anything that stops it from trading at half the fair price, because what is that coin? It’s not actually exactly the same thing as the coin you’re saying it is. You can’t turn it into that coin on any other exchange or in your wallet. It’s sort of one exchange is dubiously ever expiring future [01:08:30] on that coin.
Those are the things we’re going for the most is exchanges that have liquid order books and that reliably process withdrawals. We also have to balance out with reality. Reality is that when Bitcoin is your coin, you’ve got good choices. When you’re looking at Grosso coin, you got to take what you’re given.
We drew the line at how real of an exchange where we cannot allow. The place we drew the line, basically, stemmed from [01:09:00] our experience trading on all these exchanges. In a sense of these exchanges, almost all of their order books are at least mostly real and together, they make up the bulk of the real spot volume. You’re not going to get insane order books here. They’re going to be, in general, pretty good.
That doesn’t mean they’re perfect. There’s going to be wash trading on some of these. When you get down to the lower coins, you kind of have no choice. But even in the markets that have wash trading, they’re going to be real moderately liquid markets that you could do real trading on as well.
So, [01:09:30] basically what it looks like is what I’ll call the regulated exchanges, which is basically the US and Japanese exchanges, although Japanese exchanges basically don’t factor into this though because it’s mostly the JPY pairings. So US exchanges and then the six or so most liquid Asian exchanges. What are you looking at there? Basically, you’re looking at Binance will be okay X. Bitfinex, I don’t know if you want to call it Asian but it’s real. Then, BitMax, [01:10:00] Gate.io, KuCoin. Those are the exchanges where most order books are basically real to the extent that there is liquidity in a coin. You’re highly likely to find it on one of those exchanges.
Clay: Let’s say we’re doing the […]coin index. There’s no stable coins. Maybe you’ve got some limit on the cap and then the constraint then is the exchanges. You’re not explicitly naming particular coins. You’re saying every token trading [01:10:30] on these exchanges below this market cap that isn’t a stable coin is included. There might be listings and delistings and when that happens, coins will be added and removed. Is it roughly correct?
Sam: That’s basically right. I think our metric, if I remember correctly, is that the coin has to be listed on at least two of those exchanges. We said at least two just because it’s kind of awkward if a coin is only on one exchange and then that exchange goes down for maintenance.
Clay: Then, in terms of outliers, assuming [01:11:00] semi-efficient markets, most of these prices should be roughly in a line. If one falls out of the pack but the rest of the pack is basically in sync, then there’s a good chance that it really is an outlier and I think most reasonable. But if they all flash crash together, I think, yeah, again, what’s reasonable, most people would assume, “Yeah, that really should go down.”
Sam: Right. Maybe it’s kind of stupid that’s going down, but it’s hard to argue that you shouldn’t use that lower price if every exchange in the world has that lower price now. [01:11:30] Maybe you think it’s good to buy it there, but I think the index basically has to use that lower price. That’s what it’s trading at now.
Well, that concludes part 1 of this series on crypto dertivites with Sam Bankman-Fried from FTX. I hope you enjoyed it. Stay tuned until next week for part 2. Before you go, however, [01:12:00] I want to mention that we’ve started producing episodes at a much higher rate around here and we now have room for a few more sponsors. If you like the work we do and would like to support this show, then a sponsorship might be a good fit for you.
I can say from our own experience that Flippening sponsorships work. Each and every time we put out an episode of this podcast, we mention our own API. To date, every single one of those advertisements has resulted in at least one customer. In fact, we would do these shows even if nobody else sponsored because of the business it brings to us. Over 80% of paying customers mention [01:12:30] that they heard of us through our podcast. If you’re interested in sponsoring the show, please hit us up at email@example.com.
All right, that wraps up things for this week. Stay tuned for next week’s episode. Until then, take care.
That’s it for this week. To sign-up for our free crypto investing newsletter, listen to other episodes, or get the show notes from this episode, please visit flippening.com. I also invite you to check out the startup that funds this podcast, Nomics [01:13:00] at nomics.com. Finally, if you get value from the show, the biggest thing you can do to help us out is to leave a five-star review with some comments and feedback on iTunes, Stitcher, or wherever you listen to podcast. Thanks for listening, and see you next week.
Part 2 Transcript
Clay: Welcome to Flippening, the first and original podcast for full time, professional, and institutional crypto investors. I’m your host, Clay Collins. Each week, we discuss the cryptocurrency economy, new investment strategies for maximizing returns, and stories from the frontlines of financial disruption. Go to flippening.com to join our newsletter for cryptocurrency investors and find out just why this podcast is called Flippening.
Clay Collins is the CEO of Nomics. All opinions expressed by Clay and podcast guests are solely their own opinion and do [00:00:30] not reflect the opinion of Nomics or any other company. This podcast is for informational and entertainment purposes only and should not be relied upon as the basis for investment decisions.
Welcome to the second installment of this deep dive on crypto derivatives trading. I’m joined by Sam Bankman-Fried. As I mentioned in last week’s episode, Sam is CEO of FTX, which is a crypto derivatives exchange and may be one of the fastest-growing derivatives platforms to date. Launched in April of [00:01:00] 2019, FTX does around $100 million in daily trading volume at the time of this recording, and has launched several features and products including leaderboards, spot orders, altcoin index futures, and much more.
Sam is also CEO of Alameda Research, a quantitative trading firm that provides liquidity to crypto markets worldwide. Alameda manages over $100 million in digital assets and trades $600 million to $1.5 billion per day across thousands of products. They are responsible for approximately [00:01:30] 5% of all legitimate crypto exchange volume. That’s up there with the high end of what a high-frequency trader would do in a US equities market.
This deep dive is broken up into 7 chapters. Chapter 1 through 3 were covered in the previous episode. In Chapter 1, we discuss quantitative trading. In Chapter 2, we explore the relationship between Alameda Research and FTX. In Chapter 3, we dove into the current state of crypto derivatives trading.
In today’s episode, we conclude the conversation by covering chapters 4 [00:02:00] through 7. In Chapter 4, we cover the similarities and differences between leveraged volume and spot market volume. In Chapter 5, we discuss the current state of FTX. In Chapter 6, we delve into growth hacking and brand differentiation in the crypto derivatives exchange space. Finally, in Chapter 7, we close our conversation by exploring what the future might look like
We’ll get right to this episode in just a second, but before we get started, I’d like to pause for a moment to tell you that this episode is brought to you by the Nomics API and CSV Data Export Service. If you need an Enterprise-Grade Crypto [00:02:30] Market Data API For Your Fund, Smart Contract, or App, or if you need historical CSV dumps of trading data from top exchanges or even obscure ones, then consider trying out the Nomics API or our historical data export service.
Our API enables programmatic access to clean, normalized, and gapless primary source trade data across a number of cryptocurrency exchanges. Instead of having to integrate with multiple exchange APIs of varying quality, you can get everything through one screaming fast fire hose. And if you’d like to order historical cryptocurrency market data [00:03:00] as CSV exports from top exchanges, email us at firstname.lastname@example.org.
Okay, back to our regularly scheduled program. Here’s part 2 of my conversation with Sam Bankman-Fried from FTX and Alameda research. Enjoy.
Let’s transition to chapter 4, which is about the similarities and differences between [00:03:30] leveraged volume and spot market volume.
I won’t say who. I had someone from one of the top exchanges that is a customer of ours send me a screenshot of some data and say, “Do you think this volume is real?” and it was a while ago. At first, I was like, “No, I don’t think that’s real.” I talked to a few market makers, then I came back to them and I was like, “Yeah, I think there’s a chance that that is real.” Then now talking to you, I don’t doubt those volumes were real. It’s really impressive what [00:04:00] you guys are doing.
Let’s talk about volume for a second. Do you think that leveraged volume is equal to spot market volume? Because on the one hand, it does seem to me like, someone’s getting hosed if this goes in the wrong direction and someone’s making money to the exact tune of the leveraged amount if it goes well. Do you think it’s apples to apples or is it apples to oranges? What do you think about volumes?
Sam: The first order, I think it’s apples to apples. I mean, there’s obviously some differences here, but I think that the similarity’s [00:04:30] worth the differences. Fundamentally, buying a Bitcoin future and buying a Bitcoin are just pretty similar. There are some differences, like one is deliverable and one’s not. So, if you’re looking at a particular product where it’s all about the delivery, then you see interesting divergences between them.
A classic example of this is staking coin. If the whole point of a coin is you can stake it for rewards then the physical delivery is what matters and maybe the fair price of a future is half the price of the coin because [00:05:00] you can’t stake the future. There are specific cases where futures and physical diverge, but by and large, in almost all cases, they’re pretty similar. And the volume, a dollar in one is similar to a dollar in the other.
Fundamentally, if you’ve got a one month out future on Bitcoins, you’ve got a Bitcoin. If there’s a dollar that trades in the future, you can sell there. You can buy Bitcoin and do an arbitrage. Yeah, it eats up a month of capital. It’s not exactly the same thing but it basically is, like it’s getting turned into a Bitcoin in a month. So, I think that [00:05:30] the first-order answer is that yes, a dollar of the leveraged volume and a dollar of spot margin volume and a dollar of spot volume are all about equally important.
Clay: Let’s kick off chapter 5 which is about the current state of FTX.
Touching on FTX and what’s going on here, where’s your home base? Where most of your folks are operating out of?
Sam: It’s decentralized is what we like to say. I’ll give you a little bit less […] of an answer than that. Part of the real answer is that it is somewhat [00:06:00] spread out. We’re incorporated in Antigua, we have servers in Tokyo, we’ve got employees in four different countries, we’ve got subsidiaries in a lot of places. There really are things all over the place here and that sort of represents the pretty global nature of crypto where, if you want to service customers from all these places, you’re going to end up somewhat decentralized.
I can tell you that I’m taking this call right now from Hong Kong. We’ve got a bunch of people out here. We also have smaller offices in a bunch of other places [00:06:30] as well and some of us are flying around a fair bit between them. Part of this is to have local presences to service clients in different jurisdictions. Part of that is to build up regulatory infrastructure in different jurisdictions. There are a lot of places where if you want a bank account in that country, you need to have boots on the ground, you need to have an entity. Part of this is to maintain flexibility.
Traditionally, this is basically just meant like corporate flexibility and jurisdictional flexibility so that [00:07:00] we can pivot if we need to. But at this point, speaking to someone in a city with giant weekly protests right now, we’re appreciating regional flexibility more and more. So, it is a pretty global operation but a lot of it’s happening in Asia, the bulk of it’s happening in Asia and a decent chunk of that in Hong Kong.
Clay: It doesn’t surprise me that so much exchange activity is Asia-based. It seems like a lot of US companies that have become comforted [00:07:30] by the US financial infrastructure and just do the templated, “Hey, we’re going to start a Delaware C Corp. We’re going to create a bank account at Silicon Valley bank. We’re going to do everything roughly the normal way.” They start hitting limits, slowly but surely.
Whereas, if you grew up in a place where the government can shut you down or where there isn’t a lot of security or certainty, you become accustomed to being multi-jurisdictional. [00:08:00] You’re comfortable being on a plane, setting up things in different places. You understand the international banking system a lot better because you have to when you’ve got hedges on hedges on hedges so you can make payroll. Your funds are pretty seizure-resistant or at least the majority of them are. You spread your risk around.
Do you think that’s why Asia has done so well is because people from that region seem to be willing and able to navigate the global [00:08:30] market for jurisdictional market and financial banking products in a bunch of different jurisdictions? Why are you in Hong Kong and why is there so much activity there?
Sam: I think there are a few things going on. And I think that it’s part of it is that we talk to really, really local exchanges from any jurisdiction. They try and maximize the hell out of their market but they’re really crap at getting other markets because it’s just not what they’re built for. This is true in the United States, but it’s not the only place. There is US exchanges, [00:09:00] Japanese exchanges, Korean exchanges, local Chinese exchanges that basically only have customers from their home country. Part of this is that, I would say Asia but even more so global Asian cities. Places like Hong Kong, like Singapore are sort of the melting pots of Asia in some sense, natural places for a pretty global business to use as a touchstone. I think that, that’s a piece of it.
I think that there are other pieces there, too. [00:09:30] One of them is that I think you end up looking at governments that take favorable looks at emerging industries. Say what you will about the US regulatory regime. In general, you contrast it with some other places, it tends to move slowly at the beginning. It tends to take the position of, “Look, we don’t need to be the first country in crypto. We’re going to wait till we have a regulatory structure that we trust, and then try [00:10:00] and be huge in things eventually. Once we do, develop the most comprehensive regulatory infrastructure in the world.” But with crypto right now, particularly in margin trading, it’s basically taking the position like, “We haven’t figured out how to regulate that yet. So you can’t really do it here.”
Whereas, you contrast that with a lot of places in Asia, which take much more of the approach like, “Go for it. We love to see new emerging markets and technologies. We’ll accept some messiness and we’ll regulate things as we need to.” So, I think that’s another angle of this [00:10:30] that you see. I say that without casting judgment on what the better system is. It really depends on what your goals are.
The last thing I’ll say is, if you look at use cases for crypto, obviously, some of this is TBD. Some of this is like, “It could be huge in the future for all these reasons. We haven’t figured out why yet. We’ll get around to that, we’re working on it.” But if you’ll be like, “Right now, what does crypto do?” There are two obvious use cases and it feels like store value in and means of transfer thing, [00:11:00] like if you want something to throw all your money in and you don’t have any better, or if you want to move some money around.
The key thing here is that this is going to be a lot more useful in countries that have really bad alternatives. The United States has spent an enormous amount of time and money and labor building up infrastructure to provide these services. Bank transfers are terrible at paying for groceries but it’s kind of okay because there’s this entire credit card sector that takes care of that.
Say what you will about the dollar, [00:11:30] it’s generally considered the most safe, stable asset in the world. The calling point of Bitcoin in America is in some sense is substantially weaker because America is already relatively good at the things that Bitcoin is currently used for.
On the other hand, you look at more emerging markets, places where, first of all, just the financial ecosystem isn’t that developed. So, the whole banking credit card industry is more scattershot. Actually, just the fact that you can pay for things [00:12:00] easily with it is valuable, combined crucially with a less mature currency.
If you don’t have the US dollar as an option, you don’t have the Euro, you don’t have the Yen, and you’re thinking, where am I going to store my wealth? A lot of people, instead of thinking it was this incredibly speculative crypto thing or the safe fiat thing, they’re like, “Well, I’ve got two unsafe options. Which do I want to choose?” and a lot of people choose crypto. So I think that’s one piece of it is just countries where the currency is less developed, less stable, less mature, or countries [00:12:30] or regions where moving money around is both important and difficult.
I think it’s easy in the west to forget about this because you almost never have to move money around and when you do, it just sort of happens via some magical process you never have to look at. You set up AutoPay and you have no idea how that works or what that means but it just does it. Other than that, you just never move money around except for credit cards.
But then you get to a lot of other places. There’s a lot of remittances being sent across borders between banking systems that don’t [00:13:00] talk to each other. There’s countries that restrict their currencies, you can’t move it out of the country. There are people who are worried about government seizing their bank account. And now all of a sudden, the appeal of a decentralized currency that you can buy with anything and move anywhere in a minute starts to become a lot more obvious.
Clay: You’re located in Hong Kong but you’re decentralized. You’ve got options if […] hits the fan there. How many employees do you have, specifically for FTX? [00:13:30]
Sam: It’s about 20.
Clay: That’s incredible. It sounds like a really efficient team. What can you share about users and revenue?
One is that we’ve tried to make a platform [00:14:00] with a ton of different features for a ton of different people and a really global one with growing communities in lots of different countries. On the other side, it also has the most liquid order books in crypto. So, if you are a person or a firm that wants to trade huge size, it’s a great place to get that off.
Those are two different types of users that comprise it, with the first being the bulk of the user base in log-ins and things like that, and the second being the bulk of the daily trading volume.
Clay: Given how liquid your [00:14:30] books are, I imagine that even though you have 10,000 users, you probably have a really fat, fat tail.
Sam: That’s right.
Clay: On one side of things.
Sam: Yeah, and just to give a sense of that from Alameda’s perspective, there are exchanges through Alameda is 30% of the volume. There are exchanges where Alameda trades $100 million a day. The tail really is fat and it’s all about unlocking the really key pieces to [00:15:00] being able to do huge size, which are smooth functionality, easy to get money in and out, and a ton of liquidity.
Clay: In terms of revenue, can you share any revenue numbers?
Sam: Absolutely, I can give some. This is actually just basically fully transparent. You can just calculate this. You can estimate it obviously because we have public volume numbers and we have a public fee schedule. Now, you can’t figure out exactly what it is because there’s fees here, there’s FTT discounts. There’s all this stuff going on that means you might be off by a factor of two, [00:15:30] but you can ballpark it.
Whatever plausible deniability we had is lost because we also do a weekly buy and burn of a third of our revenue. That’s a public number. Obviously, it goes up and down over time, like some days are just crypto trades more than other days. Our fraction of crypto is of sort of overall crypto volume is more constant than our actual raw crypto volume is, because just like higher volatility days lead to more activity everywhere.
I can say that right now, FTX is trading something like $200 million a day, is the right ballpark. We had one day [00:16:00] where I think we got up to $500 million or $600 million. On a really quiet day maybe we’ll be $70 million. But $200 million is a pretty typical day. $200 million times whatever, three, four […], that’ll get you the ballpark. Then you can look at our buy in brands and it’s going to show you the same thing. Each week we’re buying and burning right now something like $50,000–$100,000 worth of FTT and that’s a third of revenue. So, you can sort of back it [00:16:30] out from those numbers there.
Obviously we’re happy about that revenue, that’s real money, but that’s not the upside. I think that if things never changed from where they are now we’d be like, that as a cool project, but we’re kind of sad it didn’t get bigger. Part of that obviously is they are fixed expenses, but the bigger part is just whatever the expenses are, more is always better. We do think that there’s the opportunity to 10X. We think that because the top exchanges are doing 10 times as much as FTX is.
If you had said, [00:17:00] “In 12 months, how would I determine if FTX was a success?” Obviously it’s not binary, but my rough answer would be like, “It was one of the top four exchanges by trading volume on average, trading somewhere between $1 billion and $10 billion a day depending on market environment.”
Clay: In terms of revenue stream, you’ve got the FTT token. That’s probably just unrealized income or whatever. I don’t know if that really counts as income strictly, but you’ve got that. You’ve got your indices. Are those just [00:17:30] products that are available on your platform and you just collect trading fees, or is there other revenue connected to the indices?
Sam: Not from the indexes. We’re not selling the computation and I guess anyone can use them if they want. The data is public and we’re just collecting trading fees on the futures on them, on the platform. The biggest thing there is FTT, which as you said, is mostly unrealized. You ask how much money have we made and the answer is like either [00:18:00] X or like 300 times X, depending on whether you count that. That’s something we’ve been debating internally, is it used to be that this stuff was simple. We’re a trading firm, we had a daily P&L. If you add that up and now we have these businesses with valuations, we have all these tokens, we’ve got all these entities with different tax treatment. It becomes hard and more and more nuanced.
So basically, there’s exchange fees, there’s FTT, there is leverage tokens as well is another piece of this. Leverage tokens have a [00:18:30] three bip a day management fee. Right now, that’s not making much of the outstanding market cap of leverage shown is not very large, but we’re obviously hoping that that grows a lot. Maybe there will be more things in the future, but those are sort of the pillars in terms of revenue.
Clay: So you’ve got trading on Alameda, you’ve got the OTC desk, you’ve got FTT, you’ve got the leverage tokens with a management fee. That’s basically it, right?
Sam: Yeah, that’s right.
Clay: Let’s transition to chapter 6 which is about brand differentiation in the crypto [00:19:00] derivatives exchange space.
No one’s going all Apple on branding. I don’t see a lot of differentiation on really product in this space, at least at a high level. They’re all kind of pretty much accomplishing the same things. It seems like the ones who are winning at this stage, maybe this just due to the infancy of the space, but the people who are winning are winning on liquidity, availability of trading products, and then just pure bread and butter execution. Do you think [00:19:30] there’s anything else that you guys are differentiating on right now other than those three things?
Sam: I think that there is sort of this really long tail of things, which it’s hard to categorize. I could give you a bunch of random examples, but they’re really random and have nothing to do with each other. I think maybe the thing that ties them together is something like, just trying really […] hard. Taking just 50 different areas of the business and not half-assing it and being like, “We’re going to do the best we can here. We’re going to try and build the best product we can, do as much as we can [00:20:00] to get users and to get volume, not settle for ‘pretty good’ in any area, keep working as hard as we can continuously, keep taking feedback, keep improving every aspect of it,” independent of how successful or not it has been so far because you can always get better, independent of how good you are. There’s always better.
Just to give some random examples of that, we’ll often talk with other exchanges and be like, “Here’s some advice. I think you should change this seven to a four,” [00:20:30] and they’re like, “Yeah, you’re probably right.” We’re like, “Cool. I assume it’s a number in a database somewhere, right?” and they’re like, “Yeah, it is.” We’re like, “Cool. So, do you agree you should do it?” and they’re like, “Ah, yeah. I think we probably should.” We’re like, “So, do you think it’ll happen?” and they’re like, “Yeah. I don’t know. Maybe in a year or two.”
We’ll sit there scratching our heads like, “What just happened? You agreed you should probably do it, and that it would take a day of 2 minutes plus another 10 minutes to make sure they didn’t make a typo maybe,” if you want to be generous there, “tack on another 10 minutes of code review. I mean, we can keep [00:21:00] tacking on things like this, we’re not going to get above a couple of hours of time, and I think this might take your revenue up by like 10% if you did this. And you’re not going to do it for two years because you’ve got other things on your mind.” And we’re just like, “What the […]?” We’re like, “What’s going on there?”
There’s just things like this all over the place. Just to give you some random examples of this, fiddling with the UI. We get a ton of UI feedback and we’re just constantly making changes. We should make the font smaller here so we can fit more in. Or like, “Yeah this is [00:21:30] too many rows, lets show fewer rows in this table so we can show more other stuff on the page.” Or like, “Man, people seem to be confused by this button. Let’s change what it says. Put some tooltip on it so it’s more clear.”
There’s things like what products do we list? We’ve already listed more types of products than any other exchanging crypto, basically, and we’ve got another six that we want to roll out over the next two months. We’re getting bottlenecked. We got perpetuals, we got quarterlies, we have spot markets, we have leverage tokens, we have index futures, we have OTC, we want to add [00:22:00] options, hash rate futures, Bitcoin dominance futures, prediction markets.
Eventually, we run into a limited number of things we can add, but there’s so many things that would obviously be cool for this platform. We understand your rate limit and how much you can do, but you’ve been around for like two and a half […] years. Why haven’t you started doing this?
We talked to the developers. It’s not like we’re so overwhelmed with this. Yeah, I don’t know what I do all day. The important stuff was done two years ago and now we have these giant projects that’s like, [00:22:30] “I don’t really know what they’re doing,” and I’m like, “What the hell? There’s so many things you could do be doing.” On the other side we’re like, “How about getting customers?” Some exchanges are really good about this. Some exchanges just really […] try like hell to get customers, while other exchanges are just like, “Yeah, we reached out to a bunch of people and we’ll see what happens.”
We’re sort of like, “Look, we’re not saying it’s easy to do. Again, you’ve been around for three years. You’ve had a lot of time, and somehow we’ve already tried more things than you have. Have you tried, oh boy, like you haven’t looked into [00:23:00] WeChat yet?” Two weeks in we realized that an important thing to keep in mind is that WeChat, which is one of the most popular Chinese messaging apps, is super popular in crypto, but it has a limit of 500 people per group. So if you want to have a huge group, you can either do a bunch of WeChat groups or you can use QueQ, which is maybe like a less user friendly service, but doesn’t have a limit.
We’re like, “You haven’t even heard of these platforms. That’s how all of China communicates.” We’ve opened up both platforms because of this 500 limit, [00:23:30] and we’re like, “Boy, we’ve already tried to create communities in eight countries. I’m not saying it’s been enormously successful, but at least we’re trying and we’re making progress on some of them. We’ll continue to make more.” There is just so many things here.
We have just documents upon documents upon documents of ideas that we think would obviously make the platform better, and we’re churning through them as fast as we can. And it helps. I mean, we get a lot of feedback of people being like, “What the hell? I asked for a feature and it rolled out the next week. That’s not how things work. [00:24:00] I was surprised you even responded to the fact that I wanted it, let alone implementing it.” We’re sort of sitting here thinking like, “Yeah, we’re working really hard, but I honestly don’t see why other people aren’t doing this.”
Clay: Personally, I think it’s because a lot of the people who were early adopters in this space or who have the appetite for risk in this space, just don’t have a lot of operating business experience. They haven’t managed businesses at scale. So, anytime they get any kind of growth that they realize that their org chart really isn’t that tight or [00:24:30] that they don’t have systems in place to scale, or they’re not willing to make tough decisions around firing people, or what have you and they just can’t execute at that level.
That was one of the things that sort of surprised me. There’s not a lot of trading firms that start exchanges, in the same way that there aren’t a lot of ecommerce businesses that start Amazon web services. It’s not traditionally what’s done, and yet, you guys and gals were able to build this capacity.
Did you hire an entirely new team? Do you have a CTO that has shared [00:25:00] responsibility around both businesses? How segregated and focused are the teams working on Alameda versus FTX?
Sam: Basically, what I’d say is split us up into three different departments. There’s the trading team, which is the Alameda team, always has been, probably always will be, and basically doesn’t do FTX stuff other than run market-making, liquidity-fighting bots on FTX, but just with the same interface that they do on other platforms. [00:25:30] Again, the one caveat that obviously they’re incentivized to really go all out on fighting liquidity on FTX. That’s one approach, which is basically fully Alameda.
Then you have the developers, which are split between the two. As you alluded to, we have a CTO, Gary, who is a co-founder, a college roommate of mine who, two years ago, basically wrote the first draft of Alameda’s trading systems.
Clay: Hey! I wanted to pause for a second [00:26:00] to let you know that this episode of the Flippening podcast is brought to you by the Nomics API and CSV Data Export Service. And as a sponsor and producer of this podcast, I wanted to give you an announcement that I’m doing a webinar every week day, on crypto data and how it works. You should join me. The webinar is called Crypto Market Data 101: Fake Volume, Exchange Spam, and How The Seedy Market Data Underworld Actually Works.
On the webinar we discuss (1) how exchanges use exchange volume spamming and ticker stuffing to spam CoinMarketCap and other aggregators, (2) what everyone is getting wrong about transparency [00:26:30] and fake volume, (3) why most price aggregators are displaying bad data, (4) the three types of pricing data and why everyone is using the wrong one, and finally, the two transitions you must make in order to move from inaccurate crypto data to good crypto data. And much, much more. To join me on a webinar, go to nomicswebinar.com. Okay, back to the show.
Sam: When we started FTX, he pivoted mostly to the FTX side and took the lead on [00:27:00] building out the exchange product. Both businesses do require devs, so that team is split between the two, with most people having defined role in one side or the other, but we have some in both. That’s a thing where our hiring criteria are pretty similar for the two. It’s not like we needed to hire a totally different type of person. We could pivot half the dev team, or two thirds really, to FTX. Now. we’re trying pretty desperately to hire more developers who’d want something like three more, [00:27:30] and the plan being I will put two of them on the FTX side, one on the Alameda side, just according to whatever people are most excited by.
Then you get to serve the ops and BD side, which is almost entirely FTX. Some of that was rotating ops and BD from Alameda, but a lot of that is new hiring. That’s the biggest place where these are people who weren’t working for us before, largely because it’s just a lot less important for Alameda, especially on the business development side. Almeda doesn’t really have many customers. It mostly [00:28:00] just deals with exchanges and it’s a huge effort managing relationships with all these exchanges, but at least it’s 30 relationships instead of 10,000.
Especially for things like customer support, KYC and compliance, account reps, community management, things like that, for FTX we’ve hired four or five new people solely for that in the last few months, and are likely to hire another tronch soon. Also something worth noting is that on the trading and dev side, we almost exclusively hire people who didn’t come from crypto. [00:28:30] We hire people who have a trading background for traders or a developer background for the developers. On the BD side, we do primarily hire from crypto, because knowledge of the industry is super valuable there. A lot of those are people that we’ve had relationships with because of our dealings in the industry.
Clay: That makes a lot of sense that when you’re doing these-domain specific hires or these skills specific hires, like you’re hiring a developer. It doesn’t matter if they’re cool with the crypto scene. It’s like, “Can you do your job?” [00:29:00] Same with trading. It doesn’t matter to some extent what the asset is.
Sam: We take the extreme. You’re making a matching engine and you’re like, “All right, that ticker should be BTC instead of Apple or whatever.” The matching engine looks similar in both cases. Whereas the BD side, a lot of this is familiarity with where are large but unpredictable pockets of crypto interest.
Clay: At the headline level, when I think about popular exchanges, I think people largely turn to Coinbase or maybe Gemini [00:29:30] for fiat on ramps, at least in the United States. For Binance, I think the headline was that they just had a lot of coins and a lot of liquidity. They executed very well and have a good product. For probably your largest competitor, BitMax, I think the headline was just that there’s leverage.
Sam: Yup. They were the first players in the game when there was no one else offering the leverage they were.
Clay: And that’s not a knock on them. I’m just talking about the high level, “What’s the headline?” It’s like, “Well, you can do this.” [00:30:00] What do you think is the headline for FTX?
Sam: I don’t think so. I think it really is more a collection of four headlines or something. I think it’s a lot of liquidity on the products that you want that no one else has. I think it’s a system that’s continuously getting better, is going to become the product people want in a lot of small ways put together, the product that’s going to break new ground in the crypto derivatives industry, and launch products no one else even thought about launching before. [00:30:30] Sort of combining all of those four together. Maybe one last thing I’ll say on the risk controls and liquidation side, something that just does it right the first time, doesn’t have markets randomly crashing, blowing out, causing clawbacks because of mismanaged liquidations, and just combining all those together to create a way better trading experience on a lot of fronts.
Clay: Let’s transition to chapter 7 which is about the future.
As someone who came from a quantitative and HFT background we’ve got ice data services [00:31:00] selling these totally antiquated like, “Hey, you can co-locate with us, we’ll charge you $1 billion for this colocation service.” it feels like the modern version of that now is like, “Oh, you’re in AWS East. I’ll be there, too. Let me spin that server up and it’s there.”
Do you think that is the future of co-location in this space? What do you see happening with data? From my perspective, it feels like when you have a monopoly on trading, like if you’re [00:31:30] the New York stock exchange or NASDAQ, the bulk of the trades sort of happen on those platforms, so then you can seek rent on the data because it’s almost all happening there. But when you’ve got a pretty nascent space like crypto, in a lot of ways if your product is these trading pairs, then it behooves you to distribute data about what’s happening on those trading pairs and distribute data like, “Are there new trading pairs coming online? What’s the price? Can we get listed on all the aggregators?” It feels like [00:32:00] your data is a distribution tool for your product, which is the trading pairs.
Where do you see data going in the future? Do you see it going the route of super expensive co-location and rent seeking, or do you think it is going to end up being fairly free and open, or do you see exchanges monetizing it eventually?
Sam: My best guess is eventually it will go down the same route that everything else went down. There will be a huge HFT industry, but that we’re not close to that. One thing you [00:32:30] can think about is how much advantage did you get from that? From getting that last bit? The answer is, you shave a 10th of a millisecond off of your round trip time. How valuable is that? When everything else about the infrastructure is basically solved, which it sort of is in US finance. The pricing isn’t, but the infrastructure, it’s kind of been the same for a while and it kind of just works. It does what it does and everyone’s [00:33:00] staring at exactly the same stuff. I’m sure a 10th of a millisecond faster […] matters. That’s the trade for some set of firms.
In crypto, you talk about a 10th of a millisecond and then you’re like, “Okay, well the average exchange has latency of a half second.” So that’s one thing. This is another key thing, is let’s say that this 10th of a millisecond gets you, I don’t know, how much does Bitcoin moving in 10th of a millisecond? Or how much can you tell it’s going to move? Maybe you can squeeze a bip out of that.
If you can squeeze a bip out of it, but exchange fees are five bips, [00:33:30] it’s not really clear there’s a trade there. Maybe there is, but that really puts a damper on it. It becomes a lot less about everyone staring at it until it becomes good, a lot more about there’s this giant chasm of fees, and there is a lot more edge in even just getting lower fees is worth way more than that 10th of a millisecond.
Then you look at Tether, which is an interesting thing. What’s a Tether worth? Is it worth exactly a dollar? Is it worth zero? Is it worth 70 cents? Is it worth $99.90? What do you think Tether’s trading at right now, if you had to guess?
Clay: [00:34:00] In the aggregate, it’s probably a little bit more than a dollar. And if you’re only looking at fiat trading pairs, it’s probably a little bit less than a dollar. But I don’t know.
Sam: It’s pretty close. I think it’s a little bit more than a dollar everywhere, but it is lower against fiat pairs. It’s about 22 bips over a dollar right now. All of a sudden, you’re trying to get this one bip out of this 10th of a millisecond, and now you’re like, “Wait, but it’s trading against Tether, which is trading in the market 22 basis points of a dollar, but we can’t create it for a dollar? Maybe I can.” And now all of a sudden you have is 22 basis point [00:34:30] effect maybe, which just totally washes out that 10th of a millisecond advantage.
I think what it’s saying is basically, the amount of edge that you get out of that last level of latency, is just really small compared to the lowest hanging fruit right now. The industry has a really long way to go before it gets efficient enough and standardized enough that, that last fraction of a bip, that last fraction of a millisecond, is really the biggest thing to fight over. Once it is, people are going to fight over it I think, [00:35:00] but I don’t think we’re close to that yet.
A lot of exchanges try to make big waves by saying, “We’re setting up a data center in Aurora, Chicago right next to CME and it’s going to be this big thing,” and then they end up just not even bothering to do it because no one wanted it. Like no one cared. They went to the HFT for […].
Clay: That’s not a specific example at all.
Sam: Right. Oh, obviously not, but this happens a lot of places. They make this big waves about the these co-los, and it’s sort of [00:35:30] a nothing burger.
Clay: Well, that concludes part two of this deep dive on crypto derivatives trading with Sam Bankman-Fried from FTX and Alameda Research. I hope you enjoyed it. Before you go, I wanted to mention that since we’ve started producing episodes at a much higher rate, we now have room for a few more sponsors. If you like the work we do and would like to support [00:36:00] this show, then a sponsorship might be a good fit for you.
I can say from our own experience that Flippening sponsorships work. Each and every time we put out an episode of this podcast, we mention our own API, and to date, every single one of those advertisements has resulted in at least one customer. In fact, we would do these shows even if nobody else sponsored because of the business it brings to us. Over 80% of paying customers mention that they heard of us through our podcast. If you’re interested in sponsoring the show, please hit us up at email@example.com.
All right, that wraps up things for this week. Stay tuned for next [00:36:30] week’s episode. Until then, take care.
That’s it for this week. To sign-up for our free crypto investing newsletter, listen to other episodes, or get the show notes from this episode, please visit flippening.com. I also invite you to check out the startup that funds this podcast, Nomics at nomics.com. Finally, if you got value from the show, the biggest thing you can do to help us out is to leave a five-star review with some comments and feedback on iTunes, Stitcher, or [00:37:00] wherever you listen to podcast. Thanks for listening, and see you next week.