Even if you are a believer in crypto as an asset class, building a fund to buy tokens on a fundamentals basis is really hard. It’s hard to underwrite the ultimate demand of the token, its true price/value and with so many market manipulators, even if you can strike a fair value to a crypto, it’s unlikely the rest of the market will agree in the near term.
As the market plunged in 2018, and a lot of the early crypto investors who managed the “successful hedge funds of 2017” got wiped out, people started looking at each other, trying to figure out how to build market-neutral strategies. Especially as many, who still believed in the market, became less confident in the timing of another bull run.
Many of these people tried launching funds that did exchange arb, others focused on market making, and some of the established funds, who already had a captive capital base, tried pursuing quant approaches.
But exchange arb is hard, and everyone is either doing it or already looking at it, so finding significant spreads is near impossible. And market-making became ultra competitive, with a lot of high frequency shops entering the space early with infrastructure that is hard to replicate.
So as an alternative, many people (including established crypto funds), spent a bunch of time looking at quant strategies to decide if they were a viable way to find returns in an increasingly confusing market.
The unrelenting volatility that saw Bitcoin rise 1300% in 2017 and fall 73% (so far) in 2018 seemingly lends itself well to quant trading. But building out a quant strategy within the crypto ecosystem is harder than you’d think – in large part, because the asset class is so new and immature. A lot of the things that make the market so attractive for a quantitative approach, also hurt it. Unreliable data sources and centralized order-books, increased counterparty credit risk, limited liquidity, and trade slippage all serve as barriers to entry and many funds in the space have not been able to get around these issues.
Below we list a number of the challenges we have identified that prove quant trading, while still an attractive strategy, is more nuanced than people might think:
Data & Centralized Order-books
Quant trading is models driven, and building robust models in traditional asset classes is difficult enough. But building models for cryptocurrency markets is even more difficult – in large part, because the data is so bad.
Historical price/order book data is sparse, and without reliable/accurate data you can’t build functioning quant models, period. While some alternatively rely on a host of 3rd party data vendors, it is hard to validate where they received their data cache/know how accurate it is. To know, you’d have to quiz them on how exactly where they were able to procure data from past years if they didn’t have websocket or API connectivity with various exchanges – and even then, it’s hard to know who is being truthful.
Once you have “clean” data, how do you get comfortable with it? Unlike traditional markets, crypto doesn’t benefit from a centralized order book that every market participant has access to. Instead, there are dozens of exchanges scattered across the globe. To complicate matters further, new exchanges sprout up all the time. How do you ensure that you have websocket/API’s that are directly connected to the most relevant exchanges? The answer is to hire engineering talent that is able to stay on top of the latest developments in the space, and ingest priority data accordingly. And this is expensive.
Let’s say you have that engineer and she has procured every data point that you could want on price, volume, and order book. Now what? This brings us to the second issue…data scrubbing. Prior to 2018, there were large enough price discrepancies across exchanges that arbitrageurs were able to buy bitcoin at one price on Exchange A and sell bitcoin on Exchange B simultaneously, and lock in a risk-free profit. While that no longer happens with regularity, you still have to understand that price disparity across exchanges will certainly affect you back test results. Model A might return 20% on Exchange 1, but only 12% on Exchange 2. Believe it or not, this is the rule, not the exception in crypto.
Counterparty Credit Risk
So now you have an approved model and are hunting for the right place to trade it. How do you choose an exchange? Counterparty credit risk is as pertinent in crypto as it was in the credit markets before Lehman Brothers went bankrupt. In a sense, you are essentially trading bilateral swaps every time you don’t take physical delivery of the coin within your own cold storage solution. To overcome this issue, you would ideally trade with an exchange that has insurance on its cold-storage holdings. But read the fine print. Being insured is not the same as insuring all client funds. One must ask the exchange what percentage of client assets are actually insured. Hackers aren’t targeting a small sliver of exchange assets, they want the whole pie. Not only that – but it’s worth reading the insurance policies if possible. The first time we looked at the insurance products out there, and looked into the actual details of what was covered, the claims process, etc. we were shocked by how false the insurance wrappers often felt.
You should also examine where your exchange of choice is located. What is the regulatory framework that they operate within? Have they ever frozen funds? How do they deliver profits….in bitcoin or fiat? The delivery of bitcoin means you have to add additional transaction costs to your bask tests as you pay bid/offer to convert that profit from crypto to fiat. Are you comfortable with the exchange margin requirements? In traditional asset classes, margin requirements are designed to give you leverage to a particular asset. You only put a fraction of the total value of the position you are trading. Margins are quite stable, and you are given notification far in advance of any change to those requirements. For example, the front month Crude Oil contract requires 6% Initial Margin for retail traders. That gives you a 16.67 leverage ratio. No such luck in crypto. You are forced to put up 40% or more Initial Margin for longs and up to 125% margin for shorts. You read that correctly – 125% Initial Margin for shorts. This means you have negative leverage! To compound this pain is the short notice exchanges have been known to give clients when hiking their margin requirements. Knowing your exchange’s track record is crucial to avoid this pitfall.
Alpha Hurdle: Slippage and Exchange Fees
Once you are comfortable with an exchange, you’ll ideally repeat the process a few times until you have a few counterparts to trade with. Now comes the interesting part…slippage. Slippage is the difference between your desired entry level and where you actually get filled. When buying the S&P500 future, slippage is only a few basis points. In crypto, slippage can be as large as a few percent if you don’t have a plan in place for executing trades. This is true even in bitcoin, despite being the largest, deepest crypto market. You need a plan for what venues and instruments you plan on trading and what to do under various slippage conditions. This slippage will naturally cause a basis between back-tests and reality.
Another key factor you must incorporate into your back test are exchange fees. This process is actually quite straight forward and a little research will indicate what fees your exchange charges and how they have deviated over time.
Beta vs Alpha
Perhaps the biggest stumbling block of all for traders are the quant strategies themselves. Many are fooled by terrific results in 2017. The reality is that anyone made money in 2017 if they bought crypto before mid-December. The true test of a manager’s skill is whether they made money in 2018. It’s no different for systems. Many managers are fooled into thinking they’ve created a brilliant quant model when all they did was benefit from the exponential move of 2017. Right place, right time and nothing more. A solid model makes money both long AND short, including in 2017. Anything else is simply beta disguised as alpha. Beta is easy, alpha is truly hard.
There will be winners who apply quant strategies to crypto. But being a traditionally fundamental fund, hiring an engineer, and trying to pick up easy nickels via repeatable arb strategies probably isn’t going to get you there.
The combination of:
- The significant cost of hiring the right talent and building the right infrastructure
- The poor data available
- The high alpha hurdle on each trade due to slippage and fees
- And the likely speed of model decay
Means it takes real investment by whichever GPs try to pursue it.
2019 will likely bring more institutionalization into the space. While prices dropped dramatically, the flood of high quality talent coming into the space didn’t slow down as much as we had expected. And a lot of the traditional long-short hedge funds also got smacked pretty hard, and has to let a lot of their talent go.
So the stakes of launching a successful fund going forward will continue to get higher. Just like launching a VC fund in 2018 is harder than launching one in 1998 was. And the investment in talent and infrastructure to get these strategies right will be higher than we think most people are ready for.