I have published my philosophical thoughts on Calendar Effects in my CFDAlert white paper previously. This was before I was able to extend my data back a further 10 years.

If you are new to the blog and have not read my abstract:

the current Calendar Effect strategy I adopt is the most naive I can imagine. It is based on the empirical observation that odd-numbered years and Decembers reject the null hypothesis of no effect at higher than a 99% confidence level. This effect has been persistent through the three decades in which I have been able to study the STI (not so much can be said for the Sell in May effect, which died out after the 1990s). So my strategy simply gives 100% exposure to market in odd numbered years and 100% exposure in Decembers, for a maximum of 200% in odd Decembers, and gives 0% everywhere ...