Caleb Everett, ID: 12332250

Overview

This is a quantitative trading strategy involving perpetual futures contracts on crypto tokens. Perpetual futures contracts are instruments that trade on the major crypto token exchanges that are similar to traditional futures contracts, but have no expiration date. In order to keep the futures contract price in line with the spot price a funding rate is employed. Long holders of the perpetual futures contracts pay the funding rate to the short holders of the perpetual futures contracts. The funding rate is determined by formula based on the spread between the perpetual futures contract price and the spot price, increasing as the spread increases and contracting as the spread narrows, and flipping such that the short holders pay the long holders if the spread becomes sufficiently negative.

This strategy is designed to capture the funding rate while limiting risk related to movements in the prices of the crypto tokens themselves to de minimis levels. It is effected by establishing a short position in the perpetual futures contract when the funding rate is attractive in order to capture the funding rate payments while establishing a corresponding short position of equivalent size in the underlying crypto token in order to hedge against price fluctuations.

Assets

Our initial execution of this strategy includes ten crypto tokens on the exchange with the largest trading volume of perpetual futures contracts.

Literature

As these are relatively new asset classes, there isn't yet a significant body of academic literature covering them. There were however, three papers of note. The first is by Robert Shiller (of real estate fame) in 1993 and is the first known mention of the perpetual futures contruct, Measuring Asset Values for Cash Settlement in Derivative Markets: Hedonic Repeated Measures Indices and Perpetual Futures. The second is more recent by Carol Alexender, et al., in 2020 in the Journal of Futures Markets, BitMEX Bitcoin Derivatives: Price Discovery, Informational Efficiency and Hedging Effectiveness. Their main contribution is to show that the crypto perpetual futures can be used as effective hedging instruments against exposure to spot crypto token positions. The last one is also by Carol Alexander, et al., Optimal Hedging with Margin Constraints and Default Aversion and its Application to Bitcoin Perpetual Futures, in 2021. The main contribution of this paper is to consider leverage and liquidation penalties in determining optimal hedging strategies. While we did not utilize classical analytical techniques to determine the relative sizes of our futures and spot positions, we did refer their analysis in the modeling leverage and liquidation penalties.

Returns

Our base case return is 0.46 on an annualized basis with a Sharpe ratio of 1.46.

Sizing

Our base case returns are based on a fixed position size for each token of $1.0 million. The average daily volume of perptual futures contract over the course of this year, is over $100 billion, according to CoinMarketCap. We have left as a future enhancement the determination of optimal sizing and portfolio composition across exchanges and tokens, but believe based on the size of the overall markets that positions of significance are available.

Risks

Below are the primary risks we have identified:

  1. Overall crypto regulatory risk
  2. Risk of exchanges being hacked and/or prices being manipulated
  3. Specific exchange risk related to derivates regulation
  4. Perpetual futures contracts are a relatively new asset class, having just begun trading in 2016
  5. Operations must be established outside the United States. None of the crypto exchanges are registered in the United States
  6. Extreme volatility of underlying crypto tokens may result in unhedged positions

Future Enhancements

This is a preliminary opportunity analyis, that we believe establishes the merit of continuing to investigate this strategy and enhance it's potential risk-return profile. Below is a list of areas for further investigation:

  1. Further diversify by trading on other exchanges
  2. Optimize allocation of capital and position sizing among exchanges and tokens based on volume and proper portfolio analytics
  3. Consider further spread-hedging opitimization based on classical hedging techniques and those presented in the literature referred to above by Alexander, et al.
  4. Further optimize entry, hold and exit rules based on empirical analyses of the duration and magnitude of spreads and funding rates

Asset Features

Overview

A perpetual futures contract is like a traditional futures contract except that it has no expiration date They were first proposed by Robert Schiller in 1992 to enable derivative markets for illiquid assets To date, they have only developed for the crypto token markets and first started trading on the BitMEX exchange in 2016. Today, they are traded on all the major crypto exchanges for all of the major crypto tokens in significant aggregate volume. The average daily dollar volume of traded perpetual futures year to date is over $100 billion as of 2021-06-02. Another indicator of the rapid pace at which the market is maturing, Bitcoin futures with fixed expiration dates also trade on the [CME](https://www.cmegroup.com/trading/equity-index/us-index/bitcoin.html), albeit at lower volumes (\~\$3bn per day year to date) than the crypto exchanges.

Below is the weekly dollar volume of each of the top ten perpetual futures contracts on the Binance exchange over the last year.

We can see that the volume has grown to be quite sizeable, albeit from a relatively small base, not that long ago. The volume is obviously impacted by the amount of leverage that is being used and in some respect, perhaps the spot volume is more relevant as it relates to position sizing for us. Below is the same chart for the weekly dollar volume of spot transactions. There still appears to be plenty of volume to establish position sizes of consequence, even on Binance alone. We estimate the Binance accounts for approximately 20% of the overall perpetual futures market, leaving additional opportunities for size on exchanges.

Regulation

Derivative trading is heavily regulated in the United States and requires being registered with the Commodity Futures Trading Commission among other requirements. None of the crypto exchanges are registered and consequently, it is not possible to trade on them from the United States. Our assumption is that this strategy would nonetheless appeal to institutional investors with existing offshore operations or that if the opportunity is deemed sufficiently attractive, such operations could be established. This is also relevant as a risk to the overall strategy as some exchanges have run afoul of regulators for making their services available in the United States, which may impact their operations in the future in a manner adverse users of their services (see What’s at Stake in the U.S. Case Against a Crypto Rebel for a current description of BitMEX's trouble).

Expiration

With no expiration date, there is no need to roll fixed expiration date contracts forward.

Pricing

Many of the futures contracts are priced in units of USD or USDT, Tether, a stablecoin (a crypto token that by contract maintains a value of essentially \$1) with the crypto token as the base currency.

Leverage

Most of the contracts offer the ability to utilize high levels of leverage. Up to 125x leverage can be used for the BTCUSDT perpetual futures contract on Binance, for example. Higher leverage levels obviously increase the risk of margin calls, which can result in forced liquidation. One of they key executional aspects of this strategy is managing the dollar balances of the spot and futures positions to avoid forced liquidation and to keep them matched.

Liquidation

Another key feature of perpetual futures contracts is that there is no central counter party in the middle of the clearing. Transactions are settled automatically directly between end market participants by contract. The benefit of this is the the risk of default is essentially non-existent, since everything happens digitally, by code. Positions are liquidated automatically before they reach a zero balance - referred to as the maintenance margin. On Binance the maintenance margin is typically 50% of the initial margin. There is some risk in light of the inherently high volatility, that positions are not able to be liquidated before they reach a negative balance. The exchanges set up insurance pools, funded by the maintenance margin on automatically liquidated positions and potentially from the liquidation of profitable, highly leveraged positions if the insurance pool is ever short of funds (auto de-leveraging). Our strategy entails relatively modest leverage (less than 10x on the perpetual futures contracts) and given our frequent rebalancing of our spot and futures balances, we are unlikely to be subject to auto de-leveraging.

To reduce volatility and the number of forced liquidation events, the liquidation price is typically based on a mark price as opposed to last price, where the mark price is based on a composite index of prices from multiple exchanges.

Funding Rate

The funding rate is the mechanism that keeps the futures price tethered to the spot price absent a set expiration date upon which the contract is settled and the spot must equal the futures price. As Schiller envisioned it, the funding rate represented the flow of value from one side to the other. The crypto exchanges use a formula based on the order book that results in a bigger funding rate if the market is more bullish and a smaller, or even negative rate in which the shorts pay the longs, when the market is more bearish. Funding on most exchanges occurs every eight hours. The rates are set in advance and are paid based on the notional value of the contract at the time of payment. The amounts are paid directly from one side to the other without the exchanges extracting any rent.

Below is chart of the funding rates for each of the top ten tokens over the last year. We can see that there are periods for each of the rates where the rate is not sufficiently attractive to establish positions that will be profitable enough to overcome the transaction costs. However, across the portolio of tokens, there don't appear to be many periods where there isn't at least one attractive rate.

Strategy

There are two versions of the same strategy that offer the potential for attractive returns at relatively low risk. Both fall under the same theme, which is to look for lower returns at disproporiontatley lower risk in the high risk, high return crypto token markets. As previously discussed, both versions of the strategy outlined herein entail capturing a combination of spread between perpetual futures contracts and their underlying assets and associated funding rate.

The first version of the strategy entails simply monitoring the funding rate and establishing a position in the perpetual futures contract to be able to receive funding rate payments and a corresponding position in the opposite direction in the underlying asset to hege most of the directional risk, which may or may not result in capturing some additional profit as the spread reverts to zero. The second version, involves the same assets, but instead looks to the spread between the futures and spot prices to establish and close positions.

The chart below shows the funding rate and spread between the perpetual futures contract and the underlying asset for BTCUSDT on a daily average basis.

As it is designed to do, the spread and the funding rate track eachother fairly consistently. This is despite the significant volatility in the underlying spot market as can be seen in the series of charts below.

The return of individual trades is a function of funding rate, spread, transaction costs and leverage. Transaction cost and leverage are actually correlated. As discussed above in the overview of the assets, exceptionally high leverage is available on the perpetual futures contracts at no cost - meaning that there is requirement to pay interest. The impact of higher leverage, though, is that the intial margin amount is lower and the change in price of the futures contract that results in a reduction in the margin amount that necessitates a rebalancing, including potentially automatic liquidation is smaller. If the market moves a lot in one direction, even if the spread doesn't change, the dollar value of each of the positions can change a lot.

Our model is based on hourly tick data so that we are evaluating whether our positions need to be rebalanced every hour. The greater the leverage, the more rebalancing transactions we need to effect and the greater the number of automatic liquidations. We have assumed a 0.3% liquidation penalty if the margin amount of an open position falss below 50%, in addition to the normal transaction costs of 0.05%. We assume that positions are rebalanced if their margin amount fall below 65% of their initial margin amount. We assume normal transaction fees only on the portion of the position being rebalanced.

We have conservatively assumed that the only leverage available is that affored by the perpetual futures contracts. For purposes of determining retuns avaiable from the strategy, we have assumed a flat capital amount of $1.0 million per token. As detailed in the sizing section, we believe significantly larger position sizes are available given the size of the overall markets. Total leverage is therefore theoretically limited to 2x, i.e., if we had 100% leverage on the futures side and all of our capital in the spot position, our notional would be 2x our capital. In practice, in light of the increased transaction costs associated with the higher leverage, we use 5x leverage on the perpetual futures, which equates to a total notional equal to 1.67x capital.

$$ \begin{align} Capital &= 1.00 \, \text{million} \\ PerpetualLeverage &= 5.0 \text{x} \\ PepetualCapital &= 0.17 \, \text{million} \\ SpotCapital &= PerpetualNotional = 0.83 \, \text{million} \\ TotalNotional &= SpotCapital + PerpetualNotional = 1.67 \, \text{million} \\ TotalLeverage &= \frac{TotalNotional}{Capital} = 1.67 \end{align} $$

We note that the literature found that leverage of 8x to 9x resulted in the optimal hedging strategy given a reasonable aversion to default parameter. We leave to future enhancements the optimization of leverage versus transaction costs and aversion to being unhedged, including potentially on an individual token basis.

The funding rate is paid every eight hours on open positions based on the value of the position in USDT at the time of the funding rate payment.

Ultimately, we run our strategy on a portfolio of the top ten tokens by volume on the Binance exchange (leaving other exchanges for future enhancements). However, below is a graphical representation of the funding rate strategy executed just on BTCUSDT. For the funding rate version of the strategy, we set the open threshold to 0.05%, which is designed to cover the round trip transaction costs, including rebalancing while the position is open, plus some liquidation fees offset in part by capturing some additional profit in the spread.

Funding Rate Strategy

There are obviously some big outliers on the spread side and also keep in mind that a significant number of the trades are related to rebalancing activity - consecutive double sized markers of the same color are rebalancing transactions. However, you can see that this strategy produced a total return of 0.1574 over the year. The Sharpe Ratio is impacted by the volatility of the hourly ticks. In the bottom chart, you can see the components of the return, with most of it coming from the funding rate payments, although there is some contribution coming from the spread. The components are set up to sum to the total profit, so start with the transaction costs and then incrmentally add in the liquidation costs, spread profit and funding rate profit, in that order. Even at the 5.0x leverage and rebalancing on an hourly basis, prices were volatile enough that we did end up suffering some liquidation penalties.

Also, note that the returns in the middle chart, include unrealized returns on an hourly basis (the basis for the Sharpe calculation). Unrealized returns on a longer interval are not nearly as volatile and as shown below are rarely in fact negative.

In practice, this strategy could be set up to monitor the price movements on a real time basis to be able to avoid most liquidation costs. Althought, as it relates to the strategy for this particular token, that wouldn't result in significantly greater profits.

Spread Strategy

As you can see in the first chart in the funding rate strategy run above, the spread on an hourly basis has much more variability than the funding rate, which is set every eight hours. As such, focusing jsut on the funding rate, we miss opportunities to augment our retuns when the spread becomes sufficiently wide. The the run below opens and closes positions based on the spread instead of the funding rate itself. The threshold to open positions is set to 0.0015, which is designed to result in a profit net of transaction, rebalancing and liquidation penalties, even if the spread comes back together in a small number of ticks.

Here, we can see that we did a little better following the spread instead of the funding rate. Our return went up by over three points, trading more frequently, resulting in an increase in transaction costs, but generating more profit from the spread and giving up some profit in the funding rate.

See the appendix for runs of the spread strategy for the next top four pairs.

Returns

To analyze returns on portfolio of tokens, we take the top 10 tokens by volume on the Binance exchange and run the same spread based strategy for each. We then calculate returns on monthly and weekly bases for each of the tokens individually and in aggregtea, including:

Our totals are based on a simple average at this point, assuming implicitly that capital is allocated evenly across each of the tokens. We leave as a future enhancement determining the optimal allocation of capital to each token, based on volume and more formal portfolio construction analytics.

What we see here is a pretty good looking set of returns:

For comparison, below is the same set of statistics on a weekly basis.

Even on a weekly basis, the returns still look good. A couple of points of note:

Sizing

Given the relatively large size of the market, even on the single exchange alone, we believe available volume will not be the limiting factor on the size of our positions in aggregate. In total, over the course of the year we analyzed, we established 2,700 positions, where each position was $1.67 million in total, $0.83 million spot and $0.83 million notional perpetual futures. However, many of those 2,700 positions were actually rebalancing transactions for only a fraction of the original notional amount. If we said that 500 of those transaction were related to the establishment of initial positions, that would be a total of of 1,000 transactions to both open and close those positions. Looking at the volume in the spot market, assuming the volume stays consistent with what it has been on a dollar basis over the year to date, that looks to be conservatively $100 billion per week, just for the top ten tokens on Binance alone. This equates to approximately $5.0 trillion per year.

What percent of $5.0 trillion could we comfortably be without fear of unduly impacting the market? Even at 1.0%, that would be total traded volume of $5.0 billion, which assuming an average of 100 transactions a year across ten tokens, would amount to a sustained invested capital base of $50.0 million. We would advocate starting with a much smaller amount of capital to validate the thesis on a go forward basis and allow for the implementation of the identified future enhancements as well as other enhancements to the risk-return profile that are likely to be identified as the initial strategy is implemented.

Risks

1. Overall crypto regulatory risk

The risk here is that the entire crypto market goes away so quickly that it is not possible to unwind open positions without suffering sifnificant impairment or being completely wiped out. Given the total amount of capital invested in the space, including increasingly from institutional investors, this seems unlikely.

2 Risk of exchanges being hacked and/or prices being manipulated

This risk seems more likely and the way to mitigate this risk is to diversify across multiple exchanges.

This is also a heightened risk, specifically related to the asset class we are trading in. Again, there is significant transaction volume in the class at this point that likely includes institutional investor, who presumably have gotten comfortable that it is possible to trade on the exchanges legally and that the exchange specific risk is tolerable. However, this is certainly an area worthy of further due diligence before investing any significant capital. At this juncture, though, our hypothesis is that it is likely we are going to be able to get comfortable with this risk upon completion of due diligence.

4. Perpetual futures contracts are a relatively new asset class, having just begun trading in 2016

Similar dynamics as the preceding risk. Additional due diligence to fully understand the contractual provisions, including setting funding rate, mark price, liquidation and automatic de-leveraging is required. Hypoethesis is that we will be able to get comfortable with the provisions upon completion of due diligence.

5. Operations must be established outside the United States. None of the crypto exchanges are registered in the United States

Additional due diligence required to understand legalality, cost and tax consequnces of establishing operations outside of the U.S.

6. Extreme volatility of underlying crypto tokens may result in unhedged positions

The risk here, as discussed in the strategy section, is that prices move so quickly that one of our positions gets automatically liquidated and before we have time to close out the other position or rebalance, the remaining position move in an unfavorable way. The way to mitigate this is first of all to monitor the value of the positions on a near real time basis. Binance, for example, has a streaming websocket api, that we can monitor to rebalance or close positions on a timely basis. However, we expect that there will be instances of extreme volatility that may result in our being naked on one side for some period of time from time to time. We have not yet performed sensitivity analysis to this risk, but the next step is to do that. Additionally, we can analyze historical order books to try to form a view on how often such events might ocurr and the potetential exposure.

Appendix

Additional Strategy Runs