Interesting.
I have done a lot of thinking and study regarding the statistical testing of chart patterns over the past couple of years. One of the first conclusions I came to is that in order to even start, one would need an empirical, rigorous method to describe a particular chart "pattern". I'm not aware that any such standardized technique exists for this. Pattern recognition is a more "slippery" subject than one might think.
It so happens that another project I have on the back burner right now is work on a paper to be submitted to the Journal of Finance (AFA) called "A Proposed Methodology for the Rigorous Classification, Identification, and Statistical Analysis of Chart Patterns in Securities Markets". The purpose of the paper will be to propose a rigorous mathematical approach to address this problem.
I HAVE found some statistically significant patterns in my work on this, but they are composed of only a very few price bars and aren't any of the "classical" patterns discussed by most chartists.
I haven't worked on this research since I started trading full time however. So much to do.... yeesh!!