As the markets seem to move more frequently between “risk-on” and “risk-off,” we decided to look at two fairly volatile instruments to see if any trading opportunities exist. We went to different asset classes as well – selecting Tesla and Bitcoin (see our volatility graphs below).
First, we looked to see what kind of correlation relationships they had. For this, we wanted to see the longer-term relationship rather than the day-to-day one which we would use for trading. Using the “Watchlist” function in Excalibur Pro, we plugged in Tesla (sorry for the pun) and Bitcoin and selected the rolling correlation calculation of a 5-day average in a one-month window. We had a strong sense that these two would be highly correlated – and they were… highly NEGATIVELY correlated. That’s a study for another day. The key here is that they were highly correlated; it doesn’t matter really if negative or positive for our pair-trading purposes. (see correlation graph below).
We ran stationarity tests to see which pair-trading approach would work best – it failed Cointegration and Distance. But not Correlation. In our back-test, we played around with different triggers to see which would be most profitable. The winner: Entering trades at 1 standard deviation of the correlation relationship and exiting the trade at 0.75 standard deviation.
The returns were eye-popping (see graph below). As you can see, for the 5 trades that were entered, the actual returns ranged from 8.35% to 38%.
The least profitable on an annualized basis were the trades that exited on 12/16 at 156% and the trade that exited on 5/5 at 1604%. Four of the trades involved going long Bitcoin and shorting Tesla while the other involved shorting Bitcoin and going long Tesla.
Comments