Monday, October 14, 2019

Optimizing Trading Portfolios and Shifting to Return on Risk Metric

I started exploring testing portfolios of trading strategies. For this, I decided that in looking at return on capital where capital is the face value of a futures contract doesn't make much sense. It seems to make more sense to look at the return on money at risk. It makes the most sense to measure that in dollars as a share of a total risk budget. This then leads me to making the primary measure of return also to be in dollars. Here is an optimized portfolio of Bitcoin, oil, and palladium:


The portfolio has a total risk budget of USD 5,000. This is then allocated across the three markets with the resulting profit curves. The analysis assumes that you can trade fractional futures contracts. Oil and palladium help diversify Bitcoin and increase the information ratio. Using a zero benchmark and daily returns on risk the portfolio IR is 2.96 rising from 2.17 for Bitcoin alone.

Oil hasn't gone anywhere in the last year, but did well in 2018 when Bitcoin struggled. Going forward, I will test whether each new market I look at improves the portfolio IR or not.

This approach then also leads to computing the prices for continuous futures contracts additively rather than multiplicatively – so that differences in dollars are preserved -– and to focusing on a constant risk budget in dollars. It also allowed me to simplify the back-testing program quite a bit. In the following chart the blue line is the daily profit curve for trading one contract of Bitcoin futures:


The red curve is based on completed trades only. The green curve has a constant dollar risk equal to the average of the single contract. To be more realistic I have rounded the number of contracts to the nearest whole number, which could be zero. This keeps the strategy out of the market in late 2017 and early 2018, when the single contract strategy had a big drawdown. The constant risk strategy has a higher return and a smaller maximum drawdown than trading a single contract. So, this is the strategy I am adopting for Bitcoin going forward. This morning, we switched from one contract long to one contract short...

No comments: