Our group recently started testing our algorithms against specific regimes for the AAPL stock which are inclusive of high/low volatility uptrend/lateral/downtrend periods. Previously, I had been backtesting my algorithm over arbitrary dates for different ETFs and my results were very positive compared to our project sponsor Chester’s algorithm. I had let this get to my head a bit and thought that I had created a very profitable algorithm that would perform well under any circumstances.

I was quickly proven wrong when backtesting over the new AAPL regime dates as the total net profit generated by my algorithm over these dates was 83% worse than Chester’s algorithm and 74% worse than just investing in AAPL. This just proves the importance of regime testing and I plan on spending time over the next week creating additional regimes for ETFs to get an even better idea of how my algorithms are performing.

I did manage to improve the profitability of my algorithm by making some changes and testing alternative versions, and one of the alternative versions actually did outperform both AAPL and Chester’s algorithm, but I will need to test this version over different regimes to determine if it really is an improvement over Chester’s.