Truth in Backtesting (Part 4)

In http://www.optionfanatic.com/2012/11/30/truth-in-backtesting-part-3, I finished discussing Market on Close orders and raised the possibility that execution at the closing price is doubtful if not impossible to achieve on a routine basis.  This may artificially skew backtesting results to the upside.  Today, I want to discuss the truth about end-of-day (EOD) trade triggers.

EOD backtesting typically uses daily or weekly time frames with each bar having an open, close, high, and low price.  For daily studies, the vast majority of backtesting computes trades based on the day’s close.

Consider a basic moving average crossover system where a long trade is executed or held if the close is greater than the 20-day moving average (20-MA).  I am going to ask a question that may itself seem very shocking:  is trading such a system even possible?

The trading rule can take three forms.  Trading rule #1:  buy at the close if today’s close is greater than today’s 20-MA.  This is riddled with uncertainty of the sort described in my last post.  Today’s closing price is not known until after the close.  Today’s 20-MA is not known until after the close because today’s closing price is used in the calculation.  I can project what today’s 20-MA will be depending on closing price and monitor price as the close approaches.  In some cases it might be clear whether a trade should be made.  In other cases it may be impossible to make this determination if price is near the projected 20-MA and the closing ticks of the day become decisive.  The soonest I could execute this trade with certainty would then be at the next open.

I will discuss other forms of this trading rule in my next post.