Original author: HangukQuant
Original translation: Luffy, Foresight News
We first came up with the idea of using perpetual contracts for funding rate arbitrage about a year ago. Since then, we have published several articles exploring this trading method and related derivative topics, including one that discusses the value proposition and source of returns of this strategy. We have also developed a fully systematic perpetual contract arbitrage robot to put the idea into practice.
Today, we would like to talk about some of the details and extensions of implementing this trading strategy.
Superimposed expected value
It is assumed that everyone understands the funding rate differential and the strategies to profit from positive funding rates. If not, please refer to the previous article. The following discussion applies to both systematic trading and manual trading.
We start with a table like this. The leftmost column is the asset, and the horizontal columns are n exchanges. We use frX to indicate that the funding rate has been normalized to account for different time intervals. We focus on combinations with large differences (the difference between the maximum and minimum), where we are long on the exchange with the minimum value and short on the exchange with the maximum value.
Here are some extensions to note. Usually, an exchange can offer different denominated assets, such as the US dollar stablecoin USDC, USDT, USDE, etc. If we choose to arbitrage a combination that is not quoted on the same denominated asset, we are actually doing an implicit triangular arbitrage. Usually, this is a good choice. For example, you can compare the prices of the following trading pairs:
BTC/USDT, BTC/USDC, USDC/USDT
Then it is found that their valuations are biased. These valuation deviations are usually not large enough to trade profitably on the same platform, but superimposing them into cross-platform arbitrage can increase the yield. In this case, you need a price oracle to help you convert the valuations. For example, on Binance, you need price information for USDC/USDT; on Paradex, you need price information for USDC/USD, and so on.
My point is that you can stack money arbitrage, triangular arbitrage, and price arbitrage in the same trade. However, in manual trading, it is better to only choose combinations of similarly denominated assets, because humans tend to perform poorly when dealing with high-dimensional decisions.
As a side note, while you can stack multiple independent strategies related to funding rates, funding rates themselves are often a key feature of perpetual futures market making (it stacks up with other factors).
Break-even
From these combination options, you should already have a combination you want to arbitrage. For example, we are interested in going long REQ/USDT on Binance and long REQ on Hyperliquid.
You need a table with transaction fees, which vary by individual (VIP level) and exchange. Funding rate differences are a source of positive cash flow. But we still need to make actual trades to open positions.
Depending on the exchanges incentives, the sum of maker and taker fees is not symmetrical. This creates a bias in which exchange you submit your order. Additionally, order book liquidity is also asymmetrical. Depending on where you execute your trade, you may get a better price basis. The combined effect of these two factors is your cost of entry.
Generally speaking, you will be compensated for providing liquidity, so you may choose to place orders where liquidity is poor and take orders where liquidity is better.
Breakeven refers to the length of time or number of cash flow intervals required to cover entry costs (for example, department exchange funding rates are settled every 8 hours).
This is a very important data. Since you are getting the future rate difference income, it is necessary to have multiple estimation methods. I use ~ to represent the break-even point estimated by historical data, and ^ to represent the break-even point predicted by the regression model.
So far, we have identified the trade combination and where to submit the bid. So, how do we execute it?
Trade Execution
If you are using systematic trading, you can use computing power to calculate some basic data at high frequency based on real-time market data. When liquidity is abnormal, you can even capture pure price arbitrage opportunities in this way. In most cases, your target is those arbitrage opportunities that are structural and can last for at least a few minutes. Your task is to ensure that the arbitrage conditions still hold within a few seconds to a few minutes of establishing a position.
The core concept is stacked expected value (Stacked EV), which is also the source of our profit and loss. The formula is profit and loss minus cost, and the break-even point can reduce the complexity of the problem, which is especially important when we are trading manually.
These are all details, and continuous iteration can make our trading strategy more rigorous. Similar rules apply whether or not we use systematic trading. In the best case, if we have a target position and can start a market maker engine to obtain the position, then everything is ready. Generally, we will use a trading model of maker and taker. If it is automated trading, we can dynamically choose to make the maker side; otherwise, the usual rules of thumb can also come in handy.
We need to determine:
Target Position Size
Maximum order size
Minimum order size
Dynamic Order Sizing Rules
For each point, we will make some relevant points. The target position size depends on risk appetite, leverage cost and available funds.
The maximum order size depends on the liquidity of the taker exchange. Once the maker order is filled, the taker hedging operation will have a (linear or quadratic) price impact. The size cap is set to reduce the impact of the transaction on the price.
The minimum order size depends on the target position size and is the lower limit of aggressiveness for accumulating positions.
The situation between the maximum and minimum order sizes is a dynamic order. If we use less capital, we can choose the order size with the least impact on the market (only taking liquidity from the best bid and ask quotes of the taker exchange). Another rule of thumb is that if you are trading manually, you can first roughly divide the order size into several pieces.
If we want to be more aggressive, we can work backwards from choosing a breakeven threshold. It’s easier to understand from a forward explanation, so I’ll explain it from a forward explanation. Let’s say we short a position worth x dollars and it gets filled, and then we want to go long to hedge. Then the actual taker price is the notional weighted sum of the price depth, which I explained in the related article:
This affects the spread between makers and takers, which in turn affects the breakeven point. We can work backwards from the breakeven point to determine how aggressive we want to be.
To swap out an existing trade for a new combination of trades, the breakeven point must take into account the exit costs.
risk
External risks include counterparty risk, hacker risk, etc. There is nothing much to comment on these risks. They are difficult to assess and are inherent in the risk premium business. This is a feature, not a defect.
What deserves more attention are the internal risks, such as margin risk. Most of these risks are caused by underestimating market volatility and overconfidence. Here, it is not uncommon for the Sharpe ratio to be greater than 10. What is expensive is volatility.
High Sharpe ratios and expensive volatility (low capital efficiency) are the perfect combination for overtrading. This is the simplest, yet most likely, recipe for trading failure. From the collapse of Long-Term Capital Management to the subprime mortgage crisis, we are always unable to control ourselves. Human madness often exceeds the predictions of the most sophisticated models.
The main operational risk is margin risk. Regardless of what position we have established, it is a good idea to transfer margin from profitable exchanges to losing exchanges. However, this is still susceptible to market crash risk factors, because network congestion may prevent margin transfers when we need to transfer margin the most.
One way to mitigate the risk of a market crash is to beta hedge. There are several options for this. Assume that the beta values are obtained from a single-factor risk model. One approach is to choose a portfolio of trades such that the beta exposure on any one exchange is roughly neutral.
This is easier to achieve when you have a larger number of target exchanges, since there are more trade combinations to choose from that satisfy this constraint. The trade-off is that the search space shrinks.
Another approach is to construct the portfolio as usual, then beta-hedge with mainstream assets to keep beta neutral. Since the hedging is done for each exchange, the hedges of the total portfolio will also cancel each other out, keeping it delta neutral. The cost is that additional capital is required.
There are other less conventional approaches, such as trading a combination of Bitcoin and other assets with the goal of generating financial gains. The trade-off is delta risk.
If we do it right, the risk of a portfolio is largely manageable. We can still use the worst-case risk engine to handle idiosyncratic risks. A single position does have the potential to disrupt market equilibrium, especially in the cryptocurrency market.
When the margin reaches a certain threshold, we can close the position first instead of waiting for forced liquidation. The advantage of this is that we can close the position gradually, and the exchange will not be so polite.
Related arbitrage methods
Last but not least, there are other forms of funding rate arbitrage that are worth paying attention to. We specifically focus on perpetual contract arbitrage, but other forms include (from a tweet by 0x Lightcycle):
Same exchange - short perpetual contracts, long spot
Same exchange - short quarterly contracts, long spot
Same exchange - borrow/short spot, long perpetual contracts
Two exchanges - short perpetual swaps on one exchange and long perpetual swaps on another exchange
Statistical Arbitrage Factor - Short all contracts with high funding rates, long contracts with low funding rates
Dynamic Funding Rate Arbitrage
0x Lightcycle has a rough comparison of each arbitrage method, so I won’t repeat it here.
Another benefit of perpetual contract arbitrage is that there are no long or short limits, which means that its performance is less dependent on market conditions and more on the structural differences in fund flows between different trading venues. Spot and perpetual contract arbitrage is usually more profitable in bull markets and relies on leveraged longs that are price-insensitive and provide liquidity compensation.
I have two more points to add about improving arbitrage returns.
Typically, exchanges offer yield-generating collateral and margin methods. For example, you can hold USDE as trading collateral and earn yield on Bybit. Synthetic yield-generating collateral is also coming to Paradex Exchange soon.
Finally, spot and perpetual contract arbitrage usually combines spot collateral to increase yields. I think this is similar to what Resolv does. For example, you can buy spot HYPE, short perpetual contract HYPE to get funding rate benefits, and pledge spot to get collateral benefits.
