Welcome to this two-part deep dive on high-frequency cryptoasset trading. Our guest is Ed Tolson, founder of kbit. Kbit is a high-frequency trading firm that’s responsible for about half of a percent of the daily total crypto exchange volume. For example, in May of 2019, kbit did just shy of $1 billion of total volume.
This two-part deep dive is broken up into five chapters:
- Chapter One: What high-frequency trading is
- Chapter Two: What high-frequency traders actually do as well as the categories of high-frequency trading that exist in cryptoland
- Chapter Three: Data, fake volume, and toxic activity on exchanges
- Chapter Four: Kbit’s business operations
- Chapter Five: What the future of high-frequency trading might look like
In part 1 (episode 0049), we explore chapters one and two. Part 2 (episode 0050) covers chapters three through five.
Topics Discussed In These Episodes
- Ed’s background in crypto and finance
- The role that artificial intelligence plays in high-frequency trading
- What Ed was doing before founding kbit
- The definition of high-frequency trading
- Who kbit is competing against
- Whether front running is possible in crypto
- Why traditional high-frequency traders have not moved to crypto
- Categories of high-frequency trading in crypto
- What a high-frequency trader does
- The concepts and mechanics involved in a trade that all beginners need to know
- When to cancel orders
- Where changes in the market show up first
- What Ed sees happening in the crypto space that others might not be aware of
- How exchanges outsource wash trading
- What “classic crypto wash trading” looks like
- Types of data that are most useful to high-frequency traders
- When it’s necessary to manually execute a trade
- What business development looks like between market makers and exchanges
- The two biggest pain points that market makers have with exchanges
- Maker-taker fee schedules
- How latency impacts high-frequency cryptoasset traders
- How kbit plans to enter the OTC space
- Where Ed goes for industry information
- Cryptoinvestor Weekly Newsletter
- Clay Collins
- Ed Tolson
- Ed on Twitter
- Kbit on LinkedIn
“In my mind, we’re competing with other high frequency, market making liquidity providing firms. We’re competing based on our technology, our engineering, our models to provide better liquidity, better quotes, more narrow spreads.”
“You can make markets on a single pair. Whereas capturing on or capitalizing on fragmented liquidity is more roping in different pairs on the same exchange or the same pair or different pairs on different exchanges.”
“If you put a bid for Bitcoin slightly below the current price of Bitcoin, what you’re really doing is writing an option, or giving the ability for somebody to sell Bitcoin to the whole market.”
“We’re in the business of automating, so we’re a four-person company trading this kind of volume 24/7, 365. We can’t do things manually.”
“But the business of high-frequency trading, it requires us to be on exchange at all times. And if there’s not a need for that value, then we take it out. We move it to a safer location.”
“What we can do is use our capabilities, our fee structure, our forecasting on exchanges to provide liquidity to exchanges that want to build an agency OTC platform, so we’re expanding in that, today.”
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 [00:00:30] do 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.
Clay: Today I’m joined by Ed Tolson, founder of kbit. kbit is a high frequency trading firm that’s responsible for roughly half of a percent of total exchange volume daily. For example, in May of [00:01:00] 2019, kbit did just shy of $1 billion in total volume. One fun fact about kbit is that they turn over their entire portfolio between 20 and 40 times per day.
This is a two-part deep dive in high frequency trading is broken up into five chapters. In Chapter 1, we explore what high frequency trading is. In Chapter 2, we discuss what high frequency traders actually do and dive into the categories of high frequency [00:01:30] trading that exist in crypto land. In Chapter 3, we talk about data, fake volume, and toxic activity on exchanges. In Chapter 4, we look at kbit’s business operations. Finally, in Chapter 5, we peer into the distance to see what the future of high frequency trading might look like. Today’s episode contains Chapters 1 and 2.
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 Nexo. Here’s a word from them. [00:02:00] Nexo is the world’s largest and most trusted crypto lender, offering automated instant crypto credit lines, which allow you to use your crypto, for example, Bitcoin, Ether, or XRP, as collateral to get cash in over 45 fiat currencies and stable coins without selling your crypto assets.
Nexo also offers interest earning accounts yielding up to 6.5% per year for stable coins and euros. Interest is paid out daily and you can add or withdraw funds at any time. [00:02:30] Nexo is also a strategic partner of exchanges, OTC desks, traditional and crypto funds helping them earn interest on idle stable coins and fiat. The company’s growing portfolio of structured institutional products includes fully collateralized continuously rebalanced swap agreements allowing counterparties to effectively manage their balance sheets. I love this product. This is something that I personally endorse if you are a believer in Bitcoin or Ethereum or whatever you believe in, then it [00:03:00] does not behoove you to sell those assets but sometimes, you need some cash. A good way to get that cash is to give Nexo your favorite asset as collateral and borrow cash against it. That means that you are not triggering a taxable event. I’m not a tax accountant, so this is not tax advice, but this is, personally, what I have experienced. I’m a huge fan of this way of going about things.
Bitcoin is the best performing [00:03:30] asset of the last 10 years and there are a lot of people that wish they have never sold their Bitcoin, but instead done what Nexo provides and borrowed against that Bitcoin, taken the cash, waited for Bitcoin to do whatever it’s going to do, has historically gone up quite a bit, and they can pay off their loans with a small fraction, a small fraction of the Bitcoin held as collateral by Nexo, for example. Anyway, thanks to Nexo for making this program possible [00:04:00] and for everything they’re doing to make this space more efficient.
In addition to Nexo, this episode is also brought to you by Nomics, which is a company that produces this podcast, and like I said before, we fund it. I wanted to give you an announcement that I’m doing a webinar every weekday, and also on some weekends. You should join me for this webinar. The webinar is entitled, Crypto Market Data 101: Fake Volume, Exchange Spam, and How the Seedy Market Data Underworld Actually Works. [00:04:30] On the webinar, we discuss how exchanges use exchange, volume spamming, and ticker stuffing to spam coin market cap and other aggregators, we talk about what everyone is getting wrong about transparency and fake volume, we discuss why most price aggregators are displaying bad data, and the three types of pricing data and why everyone is using the wrong one. Also, on this webinar, we discuss the two transitions that you must make in order to move from inaccurate crypto data to good crypto data and much more. [00:05:00] To join me on a webinar, go to nomicswebinar.com. Again, that’s nomicswebinar.com.
Okay, back to our regularly scheduled program. Here’s part one of my conversation with Ed Tolson from kbit. Enjoy.
Clay: Before we get into kbit and [00:05:30] high frequency trading, I’d love to hear a little bit about your background and how you got into crypto. What’s your story in finance up to the point that you started kbit? It’s not that the average person that decides they want to start a high frequency trading firm in crypto.
Ed: I think of myself as a computer scientist first. My parents were both computer entrepreneurs, [00:06:00] starting from before I was born. I grew up in a house with a lot of computers, I was always on the computers, and I started coding when I was 10 years old, believe it or not, which seems crazy now. By the time I got to high school, I knew several programming languages, and I was the guy programming the Texas Instrument Graphing Calculator. Maybe I’m dating myself, but if you can remember those things, you know programming them in high school to play games and things like that.
I’ve always been kind of a computer guy first, computer scientist first, [00:06:30] and I sort of stumbled into the financial industry while I was studying artificial intelligence at MIT. I kind of bumped into somebody who wanted to “apply” artificial intelligence to day trading, and that’s how I sort of stumbled first into high frequency trading. And then spent some time in some large hedge funds, and that’s kind of how I spent my career is this intersection between computer science and investment and trading.
Clay: Do you have any formal education or background in finance or is that [00:07:00] primarily self-taught through the use cases that you applied artificial intelligence to?
Ed: I didn’t take any finance or econ courses at school. I would’ve if I could go back now. I learned kind of everything on the job. I got some certifications along the way, but yeah, mostly it was kind of on the job training and learning from others and also some experimentation and trial and error.
Clay: Does artificial intelligence [00:07:30] currently play a significant role in the work that you do as a high frequency trader, or are there other roles that you primarily deploy?
Ed: It sort of depends on the definition of artificial intelligence. If we’re talking about fitting or machine learning, there’s a component of that. But really the business of high frequency trading is probably most systematic trading. I tell people it’s a lot of boring work. It’s a lot of pipes and plumbing [00:08:00] and gears and mechanics. It sounds really sexy, and we trade a lot of volume. It sounds super exciting, and we build trading models. And that part is a lot of fun. But a really high percentage of the work and the skillset is sound engineering, robust thinking through error cases and use cases and things like that.
Clay: It looks like you started kbit in 2017. Prior to that, you were director of [00:08:30] trading technology at Citadel. What was the impetus for you taking the leap? Where were you at when you decided to found kbit?
Ed: It was fortuitous and sort of unintentional. I was on a garden leave period and was basically looking for a side project, something to do to keep my skills fresh and keep myself interested and engaged and in the markets. I thought crypto would be a fun place to do that. I really started basically [00:09:00] messing around, and I was applying technologies because I wanted to learn them, not because I thought they were going to be the best fit for the job, and things like that. But the timing was really great. I was in position for the 2017 craziness that happened, and it exploded into an actual business instead of a side project right in front of my eyes. That wasn’t really my intent; that wasn’t my intent at all when I started it. But it quickly became a real business.Clay: In the early days or the first days of [00:09:30] kbit, was it just you in your house or in an office somewhere with a computer, or did you have a co-founder? Did you raise venture capital, or is this just purely like you did something, seemed to work, so you’re like, “Let’s recruit more resources to scale this.”
Ed: Yeah. It was just me for a long time, and with hindsight I probably under invested in that way. Like I said, I started as a side project. I put a small amount of capital into it and got up on a few exchanges. [00:10:00] Before I knew it, I was doing material volume and material revenue. It just kind of reinvesting that and continuing to deploy and expand. I was working solo for the first year and a half or so until late 2018 or second half of 2018 when I started to add to the team and hire a few people.
Clay: Let’s kick off chapter one of this deep dive, which is an exploration of high frequency trading, [00:10:30] liquidity, market making, whatever you want to call it. There seems to be many names for the universe of things or the constellation of things that you do. How do you describe what you do to folks who ask?
Ed: Yeah. It’s tricky. Lately at least I prefer the term high frequency trading. I think that’s the most applicable. It’s the best analog from the non-crypto, traditional financial markets. [00:11:00] We run a very high turnover, high sharp strategy, which means high risk adjusted return, high trading volume every day, and capacity limitations, meaning I can’t just put 10 times the money in and get 10 times the dollar amount out. The profits or the scale is really generated through engineering and research and not just through putting more capital in. In all those ways, it’s very much like a traditional high frequency business from the stock or other markets.
Clay: Did you run [00:11:30] high frequency trading systems at Citadel? Was that a large part of what you were doing there? Were you essentially taking the skillset you had at Citadel and mapping it onto crypto, or were you doing sort of a lot of general work around trading systems and then you decided you wanted to double down on high frequency trading?
Ed: At Citadel I was a part of the asset management team, which is the hedge fund business. It was more about doing huge, very huge trades and how to get the best execution for that. [00:12:00] It’s very analogous, there’s a lot of overlap there, but it wasn’t the same business by any means.
Clay: Got it. That was sort of around strategizing around executing large block trades without burning through order books and increasing prices or decreasing prices when you’re selling, etc. Is that accurate?
Ed: Yeah. It’s about how to get the best execution on large trades.
Clay: Okay. Got it. Although it wasn’t [00:12:30] 100% mapping on of that skillset to this one, it was primarily around trade execution and that skillset is what mapped, is what you took to your current business.
Ed: Yeah, that’s right. Yeah. It’s about trade execution. It’s about how and when to post liquidity versus take liquidity to sort of optimize the microstructure of the market. A lot of overlap, but not by any means the same thing.
Clay: I know this is sometimes a difficult question when you live in a world [00:13:00] to actually like define the term that’s the umbrella under which you perhaps operate. But how do you define high frequency trading?
Ed: Yeah. It’s a term that means a lot to different people. For us, if you’re turning over the portfolio, we turn over the portfolio 20, 30, 40 times a day, that’s high frequency. Probably the best way to explain it is anything that meets that definition, that kind of turnover is probably high frequency. And then there are different types of high frequency [00:13:30] strategies. The kind of most basic are traditional of which would be on-exchange or on-lit venue market making, which is to say posting bids and offers and trying to profit from the bid offer spread, which is the difference between which people can buy or sell the asset.
Clay: You mentioned that you are turning over the entire portfolio 20 to 30 times per day. During a previous conversation, you mentioned that you’re executing a trade [00:14:00] every 1.5 to 2.0 seconds. Another thing that you said that was pretty interesting was the percentage of overall trading volume that you estimate that you’re responsible for. Could you share that with us?
Ed: Yeah, sure. So, in the crypto spot markets, which is the non-derivative markets, we believe we’re north of a half a percent of the bona fide, non-wash trading volume. To put a specific number to that, last month, which was [00:14:30] May 2019, we did just shy of $1 billion of total traded volume.
Clay: Wow. Congratulations…
Clay: …on building a business of that size and magnitude. When you think about high frequency trading as a business, do you have a customer? I could imagine, in some cases, a customer could be an exchange or a trader, maybe it’s a token project, [00:15:00] maybe it varies on a case by case basis or maybe you’re operating like any other trader and the primary goal is not to serve the needs of a customer, it’s generally just to generate a return on the capital that you have under management. How do you think about who the customer is, or do you not think about that at all?
Ed: The core business is, as you mentioned, it’s prop trading. We trade the companies on capital, and the intent is to produce a very high risk adjusted [00:15:30] return on that capital. And so, in that sense, there is no customer; the company is its own customer. But because of the scale that we have across the crypto markets, because of the kind of technology that we have to accomplish that, it does open up some other kind of second order opportunities. But those really aren’t our primary focus. And so, you’ve kind of alluded to some of those. We can partner with exchanges to kind of provide liquidity on pairs or markets that don’t naturally have that much liquidity [00:16:00] and find some kind of a customer kind of relationship there, receive some kind of compensation for that. There’s also the possibility of partnering with tokens that are listing on an exchange and have certain liquidity needs. We, to date, haven’t done that and haven’t really had a serious conversation about it. But that is a business out there as well.
Clay: Maybe a better version of this question would be to ask, who are you competing against and who do you think wins and who do you think loses as a result of you [00:16:30] operating in the world? A lot of people that are in the space are perhaps familiar with Flash Boys, which made the term high frequency trading kind of a dirty term. From my understanding, it’s a little bit different in crypto. How is the ecosystem affected by you operating within it?
Ed: In my mind, we’re competing with other high frequency, market making liquidity providing firms. We’re competing based on our technology, our engineering, our models to provide [00:17:00] better liquidity, better quotes, more narrow spreads. In my mind, we’re competing against other firms doing similar things to us, and from a retail customer, institutional customer that wants to go onto an exchange and find liquidity, all of our participation and our competition is making that market more liquid for them so that when they go on an exchange, they see a tighter spread, they see more volume quoted and can get an execution with less slippage. The strategies described in [00:17:30] Flash Boys that you alluded to, I would describe them as predatory strategies, unethical strategies. I don’t think they’re a large percentage even of the traditional financial markets. Certainly, we don’t do any strategies like that.
Clay: Is front running even possible in crypto?
Ed: We’re certainly not doing it, and we don’t know how to do it. I assume it’s not possible. Yeah. I guess the way it would be possible, you’d have to be an exchange or have a partnership with an exchange. [00:18:00] If an exchange was trading against their own customers, that’s generally considered not ideal, not a good practice, and that’s one of the reasons why is that they could front run and the customer would have very little recourse or maybe even not know it’s happening.
Clay: In order for you to front run, you’d need a very deep partnership with an exchange. It could be a DAX with an open source code. I imagine that if that kind of thing were possible and they were okay letting it happen on their [00:18:30] platform, they’d probably want to take the profits themselves.
Ed: For sure. I could tell you just as a specific like for us, the vast majority of the orders that we place on exchanges are liquidity making orders. They’re orders that are not intended to be filled immediately and cannot be filled immediately, and they’re there to provide liquidity. Obviously, we intend to profit off of them by capturing the spread or some other means. We can sort of prove we’re not front running because we’re not running. We’re not hitting an order most of the times. When we place [00:19:00] an order, we’re not getting an immediate execution; we’re posting an order to provide liquidity.
Clay: In other words, the orders you post are net new orders. They’re not created in real time in response to a specific order that exists on the order book. Is that right?
Ed: Yes. The way front running would look is you go to buy Bitcoin on the exchange, and before your order gets processed, somebody else buys it in front of you. And they do that by lifting [00:19:30] the offers, that is taking the offers that are currently available, and then offering it back to you. Let’s say you placed a market order; you’re going to end up buying at a higher price than you would have. The front runner is by necessity placing liquidity taking orders. He’s lifting the existing offers, and then posting liquidity, placing order that you’re going to lift.
Clay: When you think about what you do and your business and your strategy, is there a pretty tight corollary in the traditional financial markets? [00:20:00] Is this similar to maybe Forex market making? What do you think is the closest analog?
Ed: I think the crypto markets have some pretty unique characteristics, but it’s a lot of things ported over or applied from the traditional markets. There’s market making in every market. I’ve never been really involved in the FX markets, but that’s probably the best analog. I can’t speak to it with firsthand knowledge, but in terms of activities, strategies that take place in the [00:20:30] stock markets, which is where I have more experience, I think there’s a lot of analogs there.
Clay: When you look at folks operating in the traditional stock market doing high frequency trading or market making or whatever you want to call it, what are you jealous of in terms of what they have? And vice versa, what do you think they would be jealous of you about if they fully understood your business model?
Ed: What they have is a much larger market. A couple orders of [00:21:00] magnitude, right? And so there’s a much bigger opportunity set. They also have 100 years of market data history to work with instead of a few years with very different regimes in each year. They have mature and meaningful risk models so you can divide the equity space up into risk factors, different industries, and things like that. That sort of existing crypto, but it’s not at all very mature. [00:21:30] In terms of what crypto has, that the traditional stock markets don’t, it’s kind of naturally an inefficient market. Bitcoin trades on exchanges globally with different regulatory regimes, different customer races, different time zones. It’s just naturally a fragmented market, and that creates inefficiencies that we can capitalize on and help to rectify.
Clay: They would be jealous of just [00:22:00] having those inefficiencies to capitalize on.
Ed: The inefficiencies have been wrung out of the U.S. stock market, so it’s spread tiny. You have to beat the speed of light between New York and Chicago to have any kind of an edge. The amount of capital that’s been invested, technology that’s been invested. It’s a hyper competitive and hyper saturated market. To the benefit of retail investors where spreads are penny instead of when I started in 2000, they were a quarter or an eighth of a dollar. It’s a [00:22:30] hyper saturated, hyper invested market. As compared to the crypto market, it’s still quite immature in a lot of ways.
Clay: Why do you think more high frequency traders from the traditional world have not moved to crypto? I’m somewhat deep in the space. I personally can count on one hand the number of crypto high frequency traders. There doesn’t seem to be a lot of them or at least they’re not trying to advertise very broadly what they do. You certainly don’t. Is it because there’s challenges around [00:23:00] data and infrastructure and models? What do you think are the largest barriers to entry for someone starting now? What are the pain points? What are the things that would cause someone, when entering this space maybe from the traditional high-frequency trading world, to just kind of give up, pack their bags, and go home?
Ed: I don’t think the premise is quite right because there are several large high frequency firms, mostly based out of Chicago. They are in the market and involved. And then there are firms like mine. There are lots of firms like mine that are smaller [00:23:30] and have a niche or looking to fill a spot in the market. I think there are plenty. The reason there aren’t more, and larger names, is only because of the size of the market. The market’s tiny at 280 billion market cap and call it like maybe daily trading volume, and the support market, call it like 2-5 billion a day. It’s just tiny. The opportunity set is not big enough for some large firms. It’s just a rounding error.
Clay: Let’s move to chapter two, which I’m [00:24:00] creatively naming, What the Hell Do You Actually Do? Obviously, there’s a lot of volume. You’re turning over the portfolio quite a bit. But what does a high frequency trader do? I guess maybe before I ask you what you specifically do, are there categories of high frequency trading that exists within crypto? Would it be possible to say, “Here are the four types of high frequency trading someone could do—there’s arbitrage, there’s this, there’s that, [00:24:30] there’s sort of this defined set, and we do one of them.” Does that exist or not?
Ed: Yeah, I think so. Probably the first thing that existed for the first several years in crypto was what I’ll call pure arbitrage, and that is buying a crypto on exchange A, sending it to exchange B, and selling it for a higher price on exchange B. And then cashing out or moving that money in a circle, or however you do it. I think that opportunity set is [00:25:00] probably completely gone at this point. It probably existed through maybe 2017 at the most.
Clay: Hey! I want to pause for a second to let you know that this episode of the Flippening podcast is brought to you by Nexo. Here’s a word from them. Nexo is the world’s largest and most trusted crypto lender offering automated instant crypto credit lines, which allow you to use your crypto for example, Bitcoin, Ether, or XRP, as collateral to get cash in over 45 fiat currencies and stable coins [00:25:30] without selling your crypto assets. If you believe these assets are going to go up, borrow money against them, do not sell them. Nexo also offers interest earning accounts yielding up to 6.5% per year for stable coins and euros. Interest is paid out daily and you can add or withdraw funds at any time.
Nexo is also a strategic partner of exchanges, OTC desks, traditional and crypto funds helping them earn interest on idle stable coins and fiat. The company’s growing portfolio of structured institutional products includes [00:26:00] fully collateralized, continuously rebalanced swap agreements allowing counterparties to effectively manage their balance sheets. Check them out at nexo.io. Okay. Back to the show. Nexo is a fantastic partner to huddlers of various crypto assets. If you’re huddling assets and you want to earn interest against them, Nexo is perfect for you. If you are huddling assets and you do need some fiat cash to do things like put down a down payment on a house, then Nexo [00:26:30], is a fantastic for you. Okay, back to this interview.
Ed: I think the other form of providing liquidity is the sort of block trading—or OTC if you prefer that terminology—offering, and so there are several really large well known firms in the crypto space that do that and probably lots of smaller ones as well and some exchanges are building offerings in that space. That’s a kind of separate business from what we do. [00:27:00] It’s more about taking a really big block, if you want to buy $10 million of Bitcoin, and then they’ll unwind that, or they’ll try to hold it and offset it against another client. And then maybe within the kind of on exchange space, the easiest, simplest thing to talk about is like market making on-exchange market making, and we’ll talk more about that as that’s a core part of our business. And then there’s also kind of just without pure arbitrage, kind of like capitalizing on fragmented liquidity, so finding opportunities without moving coins to be able to [00:27:30] capitalize on different pricing across different exchanges.
Clay: Which of these do you fall into? Is it mostly on-exchange or is it the last two, the kind of on exchange…?
Ed: The last two is our business, and the core business is market making, and we’ve kind of added to our repertoire of bridging liquidity between exchanges.
Clay: When you talk about capitalizing on fragmented liquidity, is that [00:28:00] intraexchange trading or is that maybe fragmented liquidity on an exchange where you can get there in terms of filling the order, but you have to take a path through some of the trading pairs in order to provide sufficient order book depth?
Ed: Yeah. I think both those exist. There’s the same kind of pair, the same asset, trading on different exchanges, obviously with different order books. There’s going to be some inefficiencies and opportunities there. And then there’s the fact that giving crypto can trade against [00:28:30] two or three different “currencies” on the same exchange, so there can be opportunities that arise out of that, inefficiencies that arise out of that.
Clay: It sounds like the on-exchange activity is your bread and butter, and this capitalizing on fragmented liquidity is an emerging thing at kbit.
Ed: All that is on exchange. The core strategy is like market making a pair. That’s like a very simple, like when do you post bids, when do you post offers, [00:29:00] it doesn’t require trading on other exchanges or trading on other pairs. You can make markets on a single pair. Whereas capturing on or capitalizing on fragmented liquidity is more roping in different pairs on the same exchange or the same pair or different pairs on different exchanges.
Clay: Let’s dive into the first one, on-exchange market making for a given trading pair. Let’s say you have no experience in market making. You don’t even know how to [00:29:30] even start thinking about what this is or what someone in your business does. How would you unpack this for them? What’s a strategy, maybe basic, that someone might employ if they wanted to start doing this?
Ed: Conceptually, if you log onto an exchange and you want to buy or sell Bitcoin, you’re going to get different prices. As a regular retail trader or investor, [00:30:00] you’re going to probably use a market order. And so, if you use a market buy versus a market sell, you’re going to end up with a different price. When you go to buy, you’re going to pay a higher price; when you go to sell, you’re going to get less money than if you were to buy. That’s the fundamental issue is that you want to buy or sell at a given point and time, but there’s a, what we call the bid offer spread there. What a market maker is doing is trying to, “Well, wouldn’t it be great if I could be the one to buy at the lower price and sell at the higher price? That sounds like a great business.”
Clay: [00:30:30] Hey, this is Clay cutting in to define what a market maker is. According to the SEC’s definition, a market maker is a firm that stands ready to buy and sell a particular asset on a regular and continuous basis at a publicly quoted price. Market makers create opportunities and advantages for investors by creating avenues for market liquidity, in eliminating delays in cases of orders, and in ensuring that spreads are well-stabilized. Okay, back to the show.
Ed: That’s fundamentally what a market maker [00:31:00] is trying to do is to be there for sellers when they need liquidity; be there for buyers when they need liquidity. Those will be at different times. If things go well, you might be able to capture that spread.
Clay: Would you say it’s accurate that the majority of the market making that you do for a given trading pair is sort of vis-à-vis market orders? Is that correct?
Ed: We don’t really know that. We’re placing the liquidity [00:31:30] making order. Probably the people hitting us are placing market orders, but if they’re like us, if they are a systematic participant, they might be for their own protection putting a limit price on that order just so that if something goes wrong there’s a flash crash or a flash pop right as their order is going in they don’t get filled on a crazy price. For instance, anytime I mention that the vast majority of our orders are liquidity-making orders, and we intend for them to be executed immediately. They won’t be able to be executed immediately, they won’t find a match. But when we do place an order [00:32:00] that we intend to find a match, we still put a limit price on it so it’s not a market order. We don’t really know when someone’s hitting our making order if they have a limit price or not. It’s not information that’s provided to you.
Clay: The matching engine is agnostic to the order type, and it’s probably not data that you’d be given, although I’m sure you’d love to have it.
Ed: Yeah, it would probably be an interesting data point.
Clay: Let’s say you were teaching a class at a college called [00:32:30] Market Making 101 or High-Frequency Trading 101. Let’s say it’s a computer science class. You’re tasked with providing this class kind of a bare-bones 101 rudimentary market making project. What would you task them to do, and what strategy would you encourage them to deploy? How might you come to the conclusion that you should [00:33:00] sell at a given price and buy at another price?
Ed: Let’s work through a single trade and we’ll try to keep it simple and illustrate the concepts and mechanics of what’s going on here. Let’s say that we are trading on the exchange for Bitcoin and we want to make that market Bitcoin versus dollars. What we’ll do is we’ll place both a bid, that is we’ll be willing to buy. We’ll place an offer, we’re willing to sell, and we’ll put a price difference. Maybe that price difference will be a dollar or something. [00:33:30] Our hope is that before the price of Bitcoin moves by too much both of our orders will get filled. If our buy order gets filled, our sell order gets filled, we’re back to our original position, but we made $1 because that was the arbitrate price I said, that we’ll put between our bid and our offer. That’s how the trade goes if it works out well for us in this kind of simplistic example.
But, unfortunately, if before both of those orders get filled if the price of Bitcoin kind of drifts up or drifts down, or moves up or moves down rapidly, we’re going to have a bad outcome, [00:34:00] and that bad outcome is just one of our orders is going to get filled. If the price of Bitcoin moves down just our buy order is going to get filled, and before we know it, we’re under water and we’re going to have to eventually sell at a loss. Those are the two scenarios if we kind of keep it simple and pretend that it’s binary and not continuous, but we’re going to place a bid and an offer.
We’ve got to figure out at what price level and at what price spread to place them. What are going to be the inputs to that decision? You can see right off the bat [00:34:30] that the more activity there is on this Bitcoin versus dollar pair on this exchange, the more people are trading and placing market orders, the more likely we’re going to get filled on both of our orders very quickly. If tons of people are placing market orders all the time, then our bid and offer are going to get filled right away. We have very little risk that the price of Bitcoin is going to move one way or the other before we make our dollar, so that would work in our favor. All things being equal that would mean maybe we don’t need to have a spread of a [00:35:00] full dollar. Maybe it could be something smaller.
On the other hand, the volatility of the asset is not our friend. If the price is likely at any given moment to kind of swing up or down wildly then we’re that much more likely to have just one of our orders get filled before that price movement happens. All things being equal, we can see that higher volatility we’re going to want to choose a higher spread. Those are some of the really basic inputs that you could kind of put together and start to build a model for what is the [00:35:30] probability of different outcomes happening and what is going to be my profit or loss in that scenario?
Another way to conceptualize it—that helps me a lot—is to think about options. Because if you put a bid for Bitcoin slightly below the current price of Bitcoin, what you’re really doing is writing an option or giving the ability for somebody to sell Bitcoin to the whole market. If the price goes down for sure your order’s going to get filled, so somebody’s going to take you up on that option. But if it doesn’t move, [00:36:00] hopefully, you’re compensated by somebody filling your order and you’re going to get, basically, a profit that we can think up there is the difference between the bid price you put and what the current fair price is, of Bitcoin. That’s another way of kind of conceptualizing why we’re getting compensated, or what we’re compensated for, is that kind of the time risk of an option, where an option starts to then be called the theta value of the option.
Clay: So, in most cases are you placing orders that sit on the book until they get filled, [00:36:30] or are you seeing existing orders on the order book and deciding that you’re going to pick those off?
Clay: In most cases, are you placing orders that sit on the book until they get filled, or are you seeing existing orders on the order book and deciding that you’re going to pick those off?
Ed: The vast majority of our orders, especially if we’re not in a current position, are the quitting-making orders. We’re putting orders out there. We’re going to let them sit there, and we did some calculation involving some of the inputs, all the inputs I just described plus some other inputs, in terms of what price we want to put that order at. Then, as the market changes either it will get executed, in which case we have a new situation to react to, [00:37:00] or things will change so much that we say, “Okay, well this order doesn’t make sense for us anymore. We need to change the price on it or cancel it altogether.”
Clay: Is it advantageous to have at your disposal multiple complex order types at an exchange, or are most of the order types fairly simple that you place?
Ed: We don’t use those today. We just place orders; we just place regular limit orders. From what I’ve seen in the equity markets, [00:37:30] primarily these order types are useful for highly saturated markets, basically when you want to kind of jump the queue. You want to say, “Okay, if the price moves up one tick, I want to be the first person on the bid at the next price level.” It’s that kind of thing that only really makes sense in a highly saturated market. Most exchanges don’t have orders like that. The only complex orders I think I’ve seen in the crypto markets are hidden orders or iceberg orders. I don’t think we make use of either of those today.
Clay: Let’s say [00:38:00] a competitor of yours were listening to this episode of the podcast, what are the types of things that they’d hear you say, or that you could say, that would be completely obvious to them? What are the types of information, or the categories of information, that you would deem to be proprietary and not want to be shared with? Is it your methodology for deciding [00:38:30] how to price things? What do you hold close to the vest, and what is common knowledge in this space?
Ed: I think the trade secrets are exactly that model that we just kind of talked about at a high level, which is like what are all the inputs and then how do you bake all those eggs into a cake or whatnot? Like, how do you stir those inputs up into a decision? That’s really the key of it. I see people post on Reddit all the time like, “Oh, I tried to make a market maker on this or that exchange, and I just keep losing money. [00:39:00] What’s going on here? I’m always posting on the liquidity-making bids and offers but somehow I’m losing money.” It turns out that if you just put a bid up, if the market goes down, you’re guaranteed to get executed. If the market goes up, you probably won’t get executed. The key is in making that decision, “At what price do I want to put the order? What is the fair market value right now of Bitcoin?” That type of thing. “What kind of signals am I getting from the order book, from other exchanges across the planet?” That kind of thing is the [00:39:30] secret sauce.
Clay: When do you decide to cancel orders? If there are moments of just crazy volatility, do you cancel? Or if you suspect that there will be, do you cancel orders? Under what circumstances do you just like halt the operations on an exchange?
Ed: Only for protectionary measures. Obviously, if there is some kind of an outage on the exchange. If we don’t think the market data, we’re getting is good then there’s absolutely nothing we can do, obviously, or it wouldn’t be wise to try and do anything. [00:40:00] But generally, for two and a half years from when I started this, we’ve basically never been 100% out of the market. We’ve been out on some exchange because they’re having an outage or having an outage of our activity to them, but we’re always trading 24/7 365 days a year. But we have protections in place mostly around protecting us from things going horribly wrong. If we see things happening on a particular exchange, or more likely on a particular pair where we just keep taking losses, or the balances [00:40:30] on the account aren’t what we thought they were supposed to be, that type of thing. We very, very quickly back off and say, “Well, we can’t do anything here. We need a human being to get involved.”
Clay: I imagine you probably have things in place for if the volatility is crazy, like you’ve got like a flash crash bot or something, it just buys up a bunch of stuff if it gets crazy low.
Ed: That’s right. We don’t back off, and I didn’t specifically answer that. We don’t leave the market because of high volatility. We’re in the market during periods of high volatility. [00:41:00] It can, as you mentioned, often be quite profitable, and it’s an input to the model, right? If the volatility is going up then by all means we should be quoting wider spreads. If there’s a serious flash crash where obviously, we’re willing to buy, but we’re not willing to sell. But other than that, we’re in the market in periods of high volatility.
Clay: Do you shift from one model to another when the circumstances change, or is it one kind of omnibus model that takes everything into [00:41:30] consideration?
Ed: No, that would be a changing. We would consider that a change of parameters or you have different parameters flowing into the model. It would be the same code with different calculations give you different results.
Clay: Are there any questions that I didn’t ask that I should have on this chapter of, What the Hell Do You Actually Do?
Ed: The other kind of general category we can talk about is inputs into that model. We covered volatility. We covered the volume [00:42:00] on the exchange. Maybe it goes without saying, but we’re using global crypto markets as, let’s say, an input into our market making on every pair. When we’re making markets on a regulated North American Exchange we are by all means reviewing and using in real time what’s happening on order books on regulated exchanges in Asia, for example, and kind of all that is flowing in as input. In that sense, [00:42:30] it’s a globally integrated market and we’re using the global market as inputs into our model.
Clay: It would strike me that there are things that are happening globally, and there’s probably leading indicators and trailing indicators. If your model has a good sense of what the leading indicators are that would probably be beneficial to you. I’m sure you have a good sense of what those are over time.
Ed: Yeah, exactly. That’s right.
Clay: Related to that, where do changes, in your opinion or experience, where do [00:43:00] changes usually show up first? Do you notice that they show up in Asia first and then they come to US, or do they originate in the US? When something is about to happen, I guess in the general sense, where do you see that showing up first? What markets are most sensitive to what is eventually going to happen with the price?
Ed: Yeah, the markets are integrated, and they’re driven by volume. They’re driven by the exchanges that have the most volume, [00:43:30] the most legitimate volume, because that’s where things are happening. An exchange that has $20 million in daily volume is not driving the market, so it’s following the exchange that does $1 billion a day in volume.
Clay: Well, that makes sense.
Ed: There is, as you alluded to, there is a region kind of time zone aspect to it. Well, first off, obviously during Asian hours, the Asian exchanges have a bigger influence. But also, there’s this kind of like waking up phenomenon [00:44:00] that you see in the markets where kind of North America wakes up and if something happened in there overnight they may react to it so there may be a big move one way or the other, and then Asia wakes up 12 hours later, whenever it is. There are often large moves at those kind of like morning hours in key markets.
Clay: Putting your models to the side and super proprietary computer science algorithmic knowledge, what do you think you know about this business that other people might not? [00:44:30] Do you think you have a sense of what the toxic exchanges are, who’s fudging data? How do you see this world? What do you think you see in the crypto space that other people might not?
Ed: We’re experts in execution, exchange activity, order book, market microstructure. We have relationships with almost all of the exchanges that have material volume, and we have a sense of what’s going on there, a pretty good sense of what’s going on there. [00:45:00] We have a pretty good feel, or figure, for where wash trading is happening, what percentage of volume is wash trading on various exchanges. We’re experts in that kind of what’s really happening on exchanges and where is volume. There was the Bitwise report that came out and reported like, “These are the 10 exchanges that have volume for Bitcoin.” I was like, “Well, I think you missed one. Other than that, this looks pretty good.”
Clay: Was that one liquid?
Ed: It was liquid. I think I realized that they were [00:45:30] excluding JPY, but they also missed—joking aside—they also missed a couple top 40 exchanges that I won’t name, but that have material real volume but some inflated volume, but they have much higher volume than the bottom five on their list.
Clay: Well, that wraps up part one of our conversation with Ed Tolson [00:46:00] from kbit. 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 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 them [00:46:30] because of the business it brings to us. Over 80% of paying customers of the Nomics API and CSV data export service mentioned that they heard of us through our podcast. If you’re interested in sponsoring the show, please hit us up at firstname.lastname@example.org. Anyway, that’s it. Stay tuned for part two of this deep dive on high frequency trading with Ed Tolson which comes out next week. Until then, take care.
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 crypto currency 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 [00:00:30] do 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 and final installment of this deep dive on high frequency trading in cryptoland. I’m joined today by Ed Tolson, founder of HFT firm kbit. As I mentioned in last week’s episode, Kbit is responsible for [00:01:00] over one half of a percent of all trading volumes that happen on open exchanges. In May of 2019, kbit did just shy of $1 billion in total volume. One fun fact about kbit is that they turn over their entire portfolio between 20 and 40 times per day.
This two-part deep dive is broken up into five chapters. Chapters one and two were covered in the previous episode. In Chapter one, we explored the nature and character of high frequency trading. In Chapter two, [00:01:30] we discussed what high frequency traders actually do and dove into the categories of high frequency trading that exist in cryptoland.
In today’s episode, we conclude the conversation by covering chapters three through five. In Chapter three, we talk about data, fake volume, and toxic activity on exchanges. In Chapter four, we look at kbit’s business operations. Finally, in Chapter five, we peer into the distance to see what the future of high frequency trading might look like.
We’ll get right to this episode in [00:02:00] 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 of trading data from top exchanges, or even obscure ones, then consider trying out the Nomics API or our historical data export services. We’ve got the best trade data available, according to my research.
Our API enables programmatic access to clean, [00:02:30] 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 found that you or your developer have to spend too much time cleaning up and maintaining datasets, instead of identifying opportunities, or if you’re tired of interpolated data and want raw primary source trades delivered simply and consistently with top-notch support and SLAs, then check us out at NomicsAPI.com. [00:03:00] Or if you’d like to order historical cryptocurrency market data as CSV exports from top exchanges, email us at email@example.com.
Okay, back to our regularly scheduled program. Here’s part two of my conversation with Ed Tolson from kbit. Enjoy.
Let’s move to chapter three [00:03:30] on data, fake volume, and toxic activity. You mentioned the Bitwise report and some of the things that they got right and perhaps some of the things that they missed. I think it’s always worth mentioning with that report that they limited their scope analysis to, I believe, the top 81 exchanges on coin market cap, and they were only looking at exchanges that have continuous order books and that had like either Bitcoin to Tether, or Bitcoin to USD pairs.
Is it your experience that [00:04:00] wash trading and fake volume and toxic activity that it’s generally the case that if this happens on one or more markets on an exchange that it’s happening on all the other trading pairs, or does it seem to be segmented by trading pair?
Ed: If it’s happening in high volume on one pair, then all other pairs that have high volume it’s happening on that. It’s not being applied to some pairs that don’t show much volume, but it’s being applied roughly in proportion to all the pairs that are showing volume [00:04:30] or showing high volume.
Clay: In other words, they’ve got some method in place for creating fake volume and in your experience, it seems to be like they’re applying that across all their trading pairs and not just one.
Ed: Over some of them. They’ve got a method or a technique, or they’ve outsourced it, and it’s being applied to in pairs. It could be 1 pair, it could be 10 pairs, but whichever pairs it’s applied to those are the top on their coin market cap pitch.
Clay: You mentioned outsourcing it. [00:05:00] I’ve definitely heard of organizations or firms that provide volume as a service. That seems like it’s a little bit different than what you’re doing, because there are business models to charge people for creating this kind of volume. Have you encountered activities that look like it’s volume as a service? What is it like to trade against that volume?
Ed: Okay, so yeah, two questions there. We have been approached to provide–I don’t think anyone used that terminology, but to wash rate on exchange to drive their volume up, right [00:05:30]. They didn’t word it exactly like that. I think they said something like, “We could give you zero fees maker and taker if you hit certain volumetrics per month, and those volumetrics are like really high, right, so there’s only one way you’re going to hit that. If you have zero maker and zero taker, and they’re measuring you by your volume, you’re fully incentivised to wash trade with yourself. Certainly exchanges, I don’t know if they’re all outsourcing it, but maybe that’s a cleaner model for them, right, because if push came to shove [00:06:00] they’re not even doing it themselves. They’re just giving a customer a good fee deal.
Clay: An exchange could do that, or a token could do that to get their markets to the top of the Binance page, on a website like ours, or a coin market cap.
Ed: Those to me are two slightly different things. It comes back to your second question there which is, “What is it like to trade against wash-traded volume?” The answer is you can’t trade against wash-traded volume, that’s what it’s there for.
The classic crypto wash trading, it’s sad but there is a classic crypto wash trading [00:06:30]. The classic crypto wash trading is hitting your own order right in the middle of the spread. A pair has a really high spread, if you’re watching or recording, as we are, you may for one millisecond see the liquidity-making order pop up right in the middle of the spread and then immediately gets filled, and it gets filled to the exact size. There’s no residual on the other order. That is noneconomic trading. However, it’s structured that is not a real economic trade, and no fees are being paid, or if they are they’re being rebated. [00:07:00] We can assume that.
Versus let’s say a token made the poor judgment of trying to inflate their volume on an exchange after listing, or whatnot, and hired somebody to do that, or did that themselves. In that scenario, as I understand it, they would still be paying fees. There might not be an advantage for them to trade against themselves, because they might be paying higher taker fees than maker fees, so they might be placing a lot of making orders. They might be doing a lot of economic activity with other parties [00:07:30]. I say it’s economic because it’s a real trade if they’re not on the other side of it. They’re doing something kind of like what we’re doing, except they’re measuring their performance in a different way. Just because their net buys and net sells at the end of the day equal in size doesn’t mean it is economic trading, because they’re different people on different sides of those trades. So those, I think, are different. You can trade against one but not against the other.
Clay: It seems like that’s probably a pretty good opportunity. The first example we talked about wash trading where the order comes in right between the bid and the ask [00:08:00], and it’s 100% of the order. You can’t trade against that, but you could probably clean house against the second category. Is that correct or not correct, not correct at all?
Ed: Because we trade on exchange, we never really know who’s on the other side of a trade, and they may have all sorts of different ways that they’re measuring their performance. Part of what they’re trying to achieve may be to boost the price. They may be buying for a while and then trying to sell. They may be buying while intentionally trying to have an impact on the price, and they’re trying to sell while not trying to have an impact [00:08:30]. It’s hard to say exactly what the impact would be to us. I don’t think it’s as simple as they just lift the offer and then turn around and hit the bid. That would probably be the least effective way for them to do it.
Clay: We talked about wash trading. We talked about volume as a service on specific trading pairs that a specific token project might want to incentivize. What other types of toxic activity do you encounter?
Ed: For sure [00:09:00] wash trading is out of control. Probably there’s price manipulation, not of the majors, not of Bitcoin and Ethereum. I think it would be impossible and prohibitively expensive to move the price of any of those. But there are probably some price manipulation, or attempted price manipulation, going on in some of the kind of lower liquidity cryptos. I think that’s it.
Clay: I come from the data background. We’re a data company. We’re very interested in data in this space. What kind of data is most valuable to you, is it like streaming [00:09:30] order book data? Is it streaming trade data? What’s useful, and just what isn’t useful at all?
Ed: The primary feeds we use are the real-time order books from global exchanges and the real-time trade feed from global exchanges. Those are the primary inputs for us. The way we massage those and make decisions and make calculations off those, I think, is more of the secret sauce, but that’s kind of the bread and butter marketplace information of what the market’s doing, what the order books look like. [00:10:00]
Clay: Do you have any shout outs to good exchanges? Are there a few exchanges that you’ve just had a great time working with and you appreciate the integrity of their business, and you would with confidence encourage your mother or father to trade on these exchanges? Are there any outstanding exchanges that you’ve found to be particularly and impressively sort of forthright? Or would you rather not answer this question?
Clay: When you do need to manually execute a trade, what are the circumstances under which you might do that? Is it where there is like a large order and you really want that order and [00:11:30] you just don’t trust the computer to do it or. When does that happen? The execution EPI is down? It’s probably not worth it, though, for you given how frequently you trade. When would you do that?
Ed: Yeah, we do tens of thousands of trades a day, 99.999% automated. We’re in the business of automating, so we’re a four-person company trading this kind of volume 24/7, 365. We can’t do things manually. That’s kind of all we’ll do. We try to automate everything, but there are times [00:12:00] when there’s a certain activity that only happens once a month. There’s a certain thing that needs to happen once a month and so it’s just not worth automating in terms of all the other priorities we have going on. It’s really just little one off situations that it hasn’t been worth automating yet but it’s somewhere on our list. Like once a month I have to log into this exchange and they give us some kind of a rebate and some coin and I sell it to Bitcoin or something. That’s not really a good example because that would be automated but some kind of one-off situation. [00:12:30]
Clay: Can you provide an example of a one-off, or is that you’d rather not?
Ed: No, I’m just trying to think of what it is.
Clay: I know some exchanges have like daily auctions. Are those the types of things where maybe there’s a monthly auction? I really don’t know what that could be.
Ed: We automate our participation in auctions. Anything that’s going to be daily for sure we’ve automated it. It’s usually not a trade, it’s something like I need to move a coin [00:13:00], or something like that and we haven’t automated the coin movement from that exchange, because we’re obviously really sensitive about coin movements and the automation around there, because of the serious attack back there. It’s usually more about movements than trades.
Clay: Let’s kick off chapter four which is about operations. You mentioned you’re a four person company. What functions are represented there?
Ed: Yeah, we’re four computer scientists, and that’s probably just because that’s my bias and that’s how I think about the world, and I think we can automate everything, right? What can’t we automate? When I see a task that’s manual and we’re going to do it everyday, I say, “Well, we should automate that task.” We haven’t yet figured out how to automate business development, so I guess we need to work on that [00:13:30]. But, we’re four computer scientists, and the kind of roles of operations, or relationship management stuff we handle between the four of us and we just kind of split it up. Maybe one day we’ll be at the size where it will make sense to have real roles for that, or maybe we’re there already but it’s just not my mindset. Yeah, we’re four computer scientists today automating things.
Clay: It’s not like you’ve got a biz dev person, you’re got an accountant, and you’ve got some security/cold storage guru or something like that. They’re all computer scientists [00:14:00]. I imagine you share several roles together, just cross functional team.
Ed: Yeah, I kind of do the everything else, accounting and that kind of thing. We have somebody on the team who is really good at biz dev so maybe he wears that hat 10% of the time, or something. He’s kind of just taken to it. It’s worked out.
Clay: Let’s talk about biz dev a little bit. You mentioned that you have relationships with some of these exchanges. How does that usually work out? Do they notice your activity and it flags [00:14:30] something in their system and someone reaches out to you and next thing you know you’re having a steak dinner with a CEO of an exchange? How does this look like? What does biz dev look like between exchanges and market makers?
Ed: What you’re describing is the traditional market, not the cryptomarket. What happens in the cryptomarket is we go completely unnoticed and completely neglected. I’m exaggerating. Generally there’s kind of let’s call it three categories. We could talk about the progression, as well, because exchanges are improving and they’ve improved vastly [00:15:00] over the past call it 18 months in this area.
There are some exchanges that we do a ton of volume on. We might do 1%, 2%, 3%, we might do 5% of their volume every day, and I’ve never heard from them. I get like the automated emails with their marketing information or whatever. I think 18 months ago that was probably 100% of the exchanges. The reasons are because these exchanges went from nothing to millions of customers, maybe more, in a couple months. Most of their [00:15:30] customers are retail customers. They’re often very focused on that. I think they’re focused on number of unique website visits and accounts and things like that and maybe don’t understand the role, yet, that liquidity providers, large customers play in making the exchange work.
The second category which is a growing category and probably the bulk of exchanges these days are exchanges that give us a representative. You reach out one day and say, “Hey. It looks like you’re a high-volume customer. [00:16:00] I’m going to be your representative here at this exchange. If you need help with anything, please let me know.” And this is, obviously, better than the first and well-intentioned. It works well for some things, right?
If we have a support issue, a basic this problem happened, it absolutely works great for that. They chase down the support team and make sure it gets looked at in the first day or whatever. But what I think it doesn’t work as well for, the more of a challenge, is the more strategic interaction with the exchange. [00:16:30] We have ideas about how you can improve your API, improve your fee structure or your merchant making program. You send those through to the rep. Maybe, if the rep connects us with the right strategic people internally, it might work well. But more often, they’re the point of communication. You get this game of telephone. Usually, I hear back nothing or very little and nothing happens. Nothing comes of it.
[00:17:00] A lot of exchanges fall into that category these days. It’s better than that first category, but I think it’s an area of development. These exchanges, like I said, they’ve gone through rapid growth, so it’s natural that there’s going to be growth pains. It’s natural that there’s going to be a misunderstanding of your customers and how to get the most from your customers.
The third category would be ones where either myself or somebody on the team, we have a direct dialogue with decision makers. [00:17:30] They’re are a couple exchanges where I got the CEO on Telegram, or I’ve got the CTO or the president or the strategy guy, and we communicate directly. That can obviously be really powerful.
Obviously, I’m biased. But given the volume that we do under these exchanges, the visibility we have across what’s going on in different exchanges and what’s going on in market structure, fee schedules and things like that, I think we can add a lot to that. [00:18:00] It’s not just about us getting something out of that type of relationship. But I think we have a lot to offer the exchange as well.
Clay: Yes, first category, don’t do anything at all. Second is an account manager. But really, they’re not going to put things on the product roadmap that are going to benefit you, necessarily. Or, if they do, there’s a long road there. The third is maybe you’re hooked up with someone at the C-level or someone who actually has influence over the product roadmap.
The exchange there is that… What is it? That you can provide [00:18:30] them a lot of insight into how to make their platform more enticing for market makers? Or just maybe more friendly? Just the market maker UX is better, whatever that should be. But do you also, sometimes, share data with them that, maybe then, they might not be able to arrive to on their own about things happening on their exchange that they should be aware of? Or what? How do you provide value to them?
Ed: The C-Suite can’t be in communication with every customer, right? But if you’ve got really large customers doing 1%, 2% of your volume, you probably want to reach out to them and see [00:19:00] how you can make that relationship more productive. And yes, absolutely, because we are active 24/7, crypto market participants across tons of exchanges, I think we bring a certain perspective to that, and because all of our backgrounds on the team. We’re all, basically, computer scientists from large hedge funds and high-frequency shops.
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 NomicsAPI and CSV Data Export Service. [00:19:30] And as a sponsor and producer of this podcast, I wanted to give you an announcement that I’m doing a webinar most week days, on crypto data and how it works. I think you should join me.
The webinar that I’m doing is called Crypto Market Data 101: Fake Volume, Exchange Spam, and How The Seedy Market Data Underworld Actually Works. On the webinar, we discuss, one, how exchanges use exchange volume spamming and ticker stuffing to spam CoinMarketCap and other aggregators. [00:20:00] Two, what everyone is getting wrong about “transparency” and fake volume. Three, why most price aggregators are displaying bad data. Four, 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 so much more.
To join me on this webinar, go to nomicswebinar.com Again, that’s nomicswebinar.com. If you enjoy the kind of content that we make available on this podcast [00:20:30], then you will also enjoy this webinar. Again, just go to nomicswebinar.com to register. Okay, back to the show.
Ed: I think that I’m biased. I don’t mean to pat myself on the back but I think we have a lot to offer in terms of our perspective on how you can improve your product which may or may not benefit us, but both in terms of your API, your UI, your user experience, your fee structure for not just for us but for retail customers, like how can you make it more attractive? [00:21:00] We can be a really valuable resource in terms of the color that we have on the market.
Clay: What are the kinds of asks that you often have of exchanges? If you could wave a magic wand and say, “Every single exchange is going to do the following things that are going to address my largest pain points,” would it be around maybe having maker taker fee structures? Is it that their execution or their data API would add certain things? What’s your wish list [00:21:30] for the exchange space, in general, as a market maker?
Ed: Two things come to mind on there from different aspects. The first is it sounds simple, but one of the most painful things is just not being in the loop on the strategic changes that are going on in the exchange, like you’re going to launch this or that product or get into this or that market. I can be there to help you right at the day you launch it. I can be there, providing liquidity, if I know about it. But too often, I hear about it on Twitter or Reddit [00:22:00] or whatnot. It’s more just like keep me in the loop. And I would suggest that, if I’m in the loop, I may have some useful insight or feedback into it. And the second more specific, less strategic is a lot of the exchanges, the APIs stink, you know?
Clay: That’s why we’re in business.
Ed: Exactly. Exactly, yes. You’re just dealing with the market data side. God forbid you are placing orders and trying to see what your order status is. The number of exchanges where the strategy that we have to employ [00:22:30] is that we pull for active orders, so we have 38 active orders. And then, that’s probably a small number. We have to call this rest call which is like a web call to say, “Oh. What are my active orders?” And we come back, and we see 36 of them. Okay, two are missing. Did they get executed or canceled?
I have to make another call. Did this one get executed or canceled? Oh, it looks like it got one execution. And then, it get canceled. That’s crazy. We shouldn’t be doing that. Sorry to dwell in on that little pain point. But that would be my [00:23:00], I think, really tangible, non-strategic request for exchanges. “Here’s how I think you can improve your API.” And then, probably have some suggestions on, “You don’t have to rewrite the whole thing, here. Here are two changes you can do that’ll make it much more effective for your high-volume traders. And it’s better for you. You don’t want me polling for those 38 orders over and over again. That’s so much traffic for you.”
Clay: In that instance, you would want to, with one call, get the status on every single order that you’ve placed, [00:23:30] defined by some parameters? And you wouldn’t have to go back and ping and check and ping and check and maybe make another call to find out what’s happened. Is that accurate?
Ed: Yes. The ideal is that you’re going to push many changes, right?
Ed: I’m not going to ask you, “Did anything change? Did anything change?” You’re going to push me changes, so that’s the ideal. But that’s a big, architectural overhaul. I hope all the exchanges are moving towards that, thinking about that, and know that they should do that. That’s the preferred way, a web socket is a standard [00:24:00] in crypto markets. You may have heard of the fixed protocol from traditional markets. I’m sure you have. But it’s a way that the exchange can push changes to make this order just got an execution. Here’s the current state. Here’s the price on that execution, fee on that execution.
If that’s a big undertaking, you’re working on that, that’s on your strategic list, it’s going to come out in three quarters, in the meanwhile, we could make a REST endpoint that’s, “What are the changes either since last time or in the last 10 seconds?” Or whatever it is. I give you a timestamp, you tell me what’s happened since then so you can just tell me. It’s kind of the same thing. [00:24:30] But instead of you pushing those changes to me, you say, “Okay, in one call, here are all the changes that happened since the last time we spoke.”
Clay: Do the larger exchanges, for the most part, have maker taker rebate schemes or something where a percentage of the fee that the taker pays is given to the maker? Or is there anything like that happening at scale right now in the space? Or not at all?
Ed: Maker taker fee schedules are, more or less, the standard.
Hey this is Clay cutting in to shed some light on maker-taker [00:25:00] fee schedules. The maker and taker model is a way to differentiate fees between trade orders that provide liquidity–”market orders”– and trade orders that take away liquidity–”taker orders”. Traditional maker taker fee schedules, the maker and taker orders are charged different fees. Often, the maker is earning money, or at least paying less money. The taker is paying more in fees.
Okay, back to the show. [00:25:30]
Ed: Getting a rebate, so that’s when you’re actually getting a negative fee. You’re getting some money back, or whatever, is pretty rare. There’s only a handful of exchanges that have that as a part of their fee schedule. In discussion, not on the public fee schedules, it’s not very many exchanges, I think, that do that.
Clay: Also in this category of operations, what about servers? Are you co-located with these exchanges and specific data centers? Can you pay extra for high-bandwidth [00:26:00] connections? Does that kind of thing exist in the crypto space at all? Or is it more like oh, you’re in AWS, US East, so we’ll put our servers there, too? How do you go about getting better data or better access to data?
Ed: You’re asking about latency. And most of the crypto exchanges are hosted in the cloud. There are a handful that have physical data centers and then offer co-location as a [00:26:30] service for a fee. The crypto markets are not as saturated, not nearly as saturated as the traditional financial markets because of the things we discussed earlier. Because of the size, the opportunities that’s not there. People haven’t been investing tens or hundreds of millions of dollars in the infrastructure to do high-frequency trading on crypto for a decade or more.
Latency, at least in our experience, is not a big profit driver. It’s much more your model, your data. Now, that being said, I’m not going to trade an Asian exchange [00:27:00] from North America. We cloud co-locate, which is what you were describing. We’ll try to run either in the same cloud location that the exchange is in or in something nearby, certainly, the same continent. That’s how we deal with that.
Clay: What can you share about operations as it relates to cold storage and strategies for moving assets on and off exchanges? How do you approach those types of problems? And [00:27:30] what can you share about your activity there? I mean, it sounds like maybe an alternative is you just keep it all one exchanges 100% of the time, and you’ve got, I guess, your turnover the portfolio 20 to 30 times a day.
Ed: Yes. Not a lot of room there to move things in and out of cold storage, by any means. Yes. We trade, actively, on exchanges. And this is a business. It’s not if I want to huddle Bitcoin, which I do. Separately, that’s outside of the business and then cold storage [00:28:00]. But the business of high-frequency trading, it requires us to be on exchange at all times. And if there’s not a need for that value, then we take it out. We move it to a safer location. When I say that, I mean we take it into the bank.
Clay: It seems like your business model is somewhat or particularly vulnerable to exchange hacks. I imagine you allocate capital in a manner that’s somewhat proportional, more or less, to the volume on these exchanges [00:28:30]. If Binance or Liquid or one of these large exchanges gets hacked, you’re probably out a good chunk of your portfolio. What measures can you take to secure your business in the face of those types of an event? Is there insurance you can buy, what do you do?
Ed: We try to be very capital efficient, so we try to keep as small of capital as we can in our exchange accounts and use leverage, use margin-ing capabilities and things like that to reduce [00:28:59] our risk exposure. But it is an operating risk. We have to be able to deliver a return on capital that is high enough to exceed the long term expected value of what we’re going to lose in incidents like that.
Clay: You need a business that generates so much profit that, if there were this kind of incident, it just wouldn’t take you out of business.
Ed: It has to be high return on capital. And one of the things exchanges do to help us is to provide margin [00:29:30], to provide a probably higher level of margin than other customers could do. And that helps us reduce that risk footprint.
Clay: You estimate, of the legit volume, about a half a percent on the spot exchanges excluding derivatives, contracts, and things of that nature. Do you see, at some point, your business model extending to derivatives platforms, options contracts, et cetera? Or does high-frequency trading just not work [00:30:00] in those environments?
Ed: No. It works great. It’s one of our strategic initiatives. When I started this project, I just happened to start in the spot markets. I think there was more opportunities there at the beginning, but there’s absolutely a huge opportunity set in derivative markets today. We are actively working on expanding our business there.
Clay: What about decentralized exchanges which are often ERC-20 based? It seems like that could be a headache [00:30:30] for your business model. First off, a lot of these tokens that are trading on decentralized exchanges are just really obscure. And you’ve got to get Wrapped Ether, I don’t know. You’ve got to trade from your own wallet. What challenges do decentralized exchanges present? And have you overcome them? Or is that something for the future?
Ed: No. We’re monitoring them. I think today the opportunity set is very small. The operational complexity, the risk footprint, they’re both very high. [00:31:00] For today, we’re just in monitor mode. And we don’t have any active plans.
Clay: Let’s transition to chapter five which is on the future of your business model. What are some of the short, medium, and long-term changes that you foresee occurring in your space and with regards to the high-frequency trading business model?
Ed: I think, as the crypto market matures, as crypto market expands, as the market cap goes up, as the legitimate daily trading volume goes up, [00:31:30] there will be more players than there are now. And there will be the players that are already here are going to continue to invest. The profit margins are going to continue to compress over the next five, ten years, whatever it is. Our plan is that we’ve been here early, we’ve got a nice footprint. We’ve got a sustainable business, and we’re going to continue to invest, continue to research, continue to build and expand on what we’re doing.
We’re going to expand in the derivative markets. We’re going to, hopefully, continue to grow [00:32:00] our footprint in the spot market. Today, we’re doing north of a half of a percent. A really dominant high-frequency business can do 5%.
Ed: We’ve still got the potential for 10X growth in the spot market. We could double our … We could ramp all of that business over into the derivatives market, which is, today, pretty close to zero. Our footprint’s pretty close to zero, so there’s tons of room for expansion. The options markets are beginning to show real liquidity, so we’re going to get involved in those over the next few quarters. There are more regulated [00:32:30] derivatives markets coming online, so we hope to be involved in those as well.
We’re here. We have an existing presence and infrastructure. And we’re going to look to support it and expand it and build on it as the market continues to develop.
Clay: How much of what you do, do you think is specific to crypto exchanges and how they operate? I guess what I’m getting at is, let’s say hypothetically, one day, all these tokens are traded on traditional exchanges [00:33:00], the New York Stock Exchange, NASDAQ, et cetera. Do you think you’d participate in those markets, or not?
Ed: Yes, I think we would. I think the crypto markets behave pretty differently both from a tech standpoint and from a market structure standpoint. You can’t just pour it in equity high-frequency model over to crypto and plug it in. It’s not going to work. So, no. If every crypto just ceased existing on the crypto markets and was only the NASDAQ, then they would probably just trade like stocks [00:33:30] and they’d probably work. Those kind of models would work. But I don’t think that’s what’s going to happen. At the very least, they’re going to continue to have a footprint in these kind of current crypto markets.
Ed: I think what we have there is the ability to bridge the two, the kind of unregulated and the regulated markets. And with our footprint in the unregulated markets, I think we have a leg up or something to build on top of as the crypto regulated market matures and expands.
Clay: As you said, these markets are a little bit different. They’re trading 24/7 [00:34:00], and you’ve got quite a bit of history, so you’ve iterated on these models for a while. You also mentioned that these markets tend to be pretty well integrated. You’ve got an eye on what’s happening everywhere. I’m sure that will allow you to perform pretty well if the NASDAQ or the NYSE or whatever did it.
What about OTC? Do you ever think you’ll get into the OTC space? It seems like you guys can provide an interesting service where maybe [00:34:30] it’s low touch in terms of customer service, but you guys could provide, in a pretty automated way, fairly instantaneous, like down to the millisecond, price quotes. No one would even have to pick up the phone, they just get a quote from you and they’d be off to the races. Is that something you’ve contemplated? Or is that not compatible with four computer scientists writing code?
Ed: That’s right. That’s the key issue. It is on our roadmap. [00:35:00] In fact, we dabble in it today. But I think what’s not on the roadmap is building out a sales team and getting MSB licenses and the things you would have to do to open the door and say we are here as an OTC shop. What we can do is use our capabilities, our fee structure, our forecasting on exchanges to provide liquidity to exchanges that want to build an agency OTC platform, so we’re expanding [00:35:30] in that, today.
Clay: When you look at the derivatives space, do you think you’d be more drawn to exchanges like Bitmax or Deribit? Or would you be more inclined to trade on CBO and the CME Groups products? Or is it really all of the above?
Ed: Yes. We have plans to trade on all of those in the near future, in the next quarter or two.
Clay: When you consider what’s coming out in the way of institutional [00:36:00] products and institutional infrastructure in the space, exchanges like Bakkt, et cetera, is there anything that those exchanges can or, to your knowledge, might make available to you that would just make the possibility of trading on them so much more seductive? Or does it really just come down to the boring basics like liquidity, solid APIs, things like that?
Ed: Yes. For exchanges that are launching, I think we can be a partner in terms of providing liquidity. And what we would [00:36:30] need to make that work is, yes, a sound API, an attractive fee schedule, and that’s really the key to it. In terms of not just from a market making perspective, but in terms of what I think the market would like to see is regulated physical futures products. I think that’s really interesting and could really help make the market more efficient and get better price discovery. I’d like to see it regulated, maybe physically delivered options, this kind of bread and butter [00:37:00] derivatives that crypto doesn’t have yet. It’s crazy that there is no physically delivered derivatives market yet.
Clay: Yes. I think LedgerX has it, and Bakkt is planning on having it. But LedgerX is pretty small in the scheme of things, right now. They’re probably the smallest derivatives platform. But it’s funny that phrase, physically settled even though it’s Bitcoin. But yes, like versus cash settled, right?
Ed: Right, yes. LedgerX is a good example. Yes, if you buy a call option, [00:37:30] you should get a Bitcoin at the end. My understanding, LedgerX is I think if you’re writing options, I think you have to fully collater-ize them. There should be a better solution than that.
Clay: Where do you go for your information? Do you listen to podcasts? Do you go to events? Is there some kind of industry thing for market makers? Do you have a set of peers that you swap ideas with? What point of connection do you have to the industry other than your computers trading with their API?
Ed: [00:38:00] Yes. I mean, we have our closest dialogues and relationships are with some of the exchanges. We trade, I think it’s almost 300 crypto assets. As I mentioned, we’re doing tens of thousands of trades a day. It’s neither feasible nor advisable for those trades to be reviewed by human beings. Oftentimes, usually, more often than not, for the vast majority of the times, we have no idea what’s going on on price, what’s driving price. It’s funny [00:38:30] because people will come to us and say, “Hey, BSV just gapped up 100%. Do you know, is there news? Or what’s going on?” And our response is not even that we don’t know. It’s that, “Oh? It did? Well, that’s interesting. We didn’t know that.”
Ed: Our systems knew that, and it’s acting accordingly. But we can’t respond. It’s not scalable for us to respond, manually, every time a crypto moves 50% because that happens every day.
Clay: When it comes to inputs to your models [00:39:00], is it mostly just market data? Or do you also incorporate sentiment data, signals from Twitter, Reddit, Telegram? Is that stuff just bogus for the most part? Launching data? How broadly do you look when it comes for data that informs your models?
Ed: Yes, we don’t use that. The way I view that is traditional markets are so saturated, STAT-R or quant trading in equities, [00:39:30] in US equities in particular, so saturated. You’re looking for tiny sources of alpha or information anywhere, right? If you put enough machine learning on Twitter, you might get this tiny piece of positive alpha or positive information value. But there’s so much more opportunity set in crypto than that. It’s not a saturated market. We’re not. We don’t have the people or the time or need to try to squeeze a tiny bit of alpha out of Twitter. Yes, we’re just using market information. [00:40:00]
Clay: Just the bread and butter stuff provided by exchanges? Trades and orders, right? Streaming through whatever is the fastest way you can get them? And that’s about it, right? Is that correct?
Ed: Yes. That’s right. We have the global order book, if you will, for all the 300 crypto assets that we’re trading. And we’ve got it in real time. We’ve got the history of it, and we’re crunching those numbers 24/7.
Clay: What about technical indicators? Do you look at anything like the exponential moving average or whatever, whatnot? [00:40:30] Do you look at any of that stuff?
Ed: I advise people not to use that stuff because it’s complete garbage. I’ve looked at these in the equity markets. And to my belief, the only effect there is that enough people believe that they might make the effect happen. But no, we don’t apply that in crypto.
Clay: Okay. That’s interesting. I thought you might say something to that effect. What are some of these technical indicators? [00:41:00] Castic fast or the ultimate oscillator or the Williams percent range or any of this?
Ed: Some of those like RSA and stuff, those started as math projects. They did real research on it and stuff, but those are less troublesome. I think the most troublesome are the patterns like the head and shoulders. I don’t even know the name, the teacup handle or whatever it is. Sometimes, see a pattern here. No. Please, don’t [00:41:30] trade on that.
Clay: Well, that concludes our two-part deep dive on high frequency trading with Ed Tolson from kbit. I hope you enjoyed it. Before you go, I want 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 this show, then a sponsorship might be a good [00:42:00] 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. And 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 firstname.lastname@example.org. [00:42:30]
Alright, that wraps up things for this week. Stay tuned for next week’s episode. Until then, take care.