SPY SMA Optimization – Revisited

Today, I wanted to revisit the Simple Moving Average cross for trading SPY since it is still a major holding of my portfolio. Originally, I was optimizing the cross by maximizing the Sharpe ratio across the entire 10 plus-year history from 2007 onward. However, I thought it would give a better idea if I had broken down the time period so that the Financial Crisis was in one window and the rest afterwards. As well, rather than using the Sharpe ratio, I’m going to optimize using the Sortino ratio because upside return deviation is actually a good thing and can be ignored. Doing this would let me see whether 25-220 is actually independent of the market “regime”.

I chose to split my time windows to be 01/01/2005 to 01/01/2011 and 01/01/2011 to 01/01/2017. The following table gives the optimal long windows for the specified short windows given the time period you’re in. It’s noticeable that the short window is optimal between 25 to 35 and that the long window is optimal between 210 to 220 regardless of the time period. This leads me to believe that my current cross is a decent enough indication of whether we are in a major correction or bull run. I don’t intend for this cross to allow me to time the markets perfectly, just enough for me to avoid the majority of a major market drawdown.

One key note is that the correction in the summer of 2015 was almost unavoidable by most of these crosses, except for a severely over-fit model of 40-45.

Table of Sortino Ratios for Different Time Periods

Sortino Heat Maps Across Different Time Windows