Seasonalities: Bad Period for Stocks?

I just finished the implementation of another approach to finding repetitive calendar behaviour, and was quite surprised that the only short period for stocks, has just began. What are the odds of this? 🙂

The calendar period I am talking about is the 21 day period from July 27th to August 17. I have been using the S&P 500 index. Over the last 15 years the period has been negative 10 times (67% of the time), while on average a 21 day period is positive 60% of the time. If one is to use a 15 trial period as a sufficient sample, the deviation is significant.

The mean return for this period is -0.62% and the median is -0.89%. The full return data follows:

Year Gain/Loss
2003 -0.80%
2004 -1.20%
2005 -1.34%
2006 2.71%
2007 -0.89%
2008 3.22%
2009 -0.25%
2010 -1.91
2011 -8.51%
2012 2.32%
2013 -2.12%
2014 -1.18%
2015 1.68%
2016 0.72%
2017 -1.83%

The same pattern holds over a longer history. Over the last 30 years, the period has been negative 57% of the time, with mean return of -0.27% and a median return of -0.73%.

Notice, I didn’t say this was the worst period ever. It was the only one short period selected by my strategy, which tries to quantify and spread all biased periods over the year. Better periods have been excluded by other criteria. For instance, my strategy of choosing periods does not allow periods which are too close to each other. If that sounds murky, wait until I find the time to open source the code. Or think how you would have done it. 😉