A Follow-up on "A Volatiltiy Trading Strategy That is (Almost) Too Good to be True"

This is a follow-up on the previous post. As I mentioned, backtest shows that the VIX term structure based strategy has faltered since late 2010. The strategy has seemingly lost its magic between late 2010 and late 2012, as reflected in the fluctuating backtest balance. What is interesting is perhaps that starting from late 2012, "doing the opposite" seems to be a working strategy:

Recall that in the original strategy, we hold VXX when recent realized vol is greater than VIX; hold XIV when recent realized vol is less than VIX. So the opposite strategy would be to hold XIV when recent realized vol is greater than VIX; hold VXX when recent realized vol is less than VIX. Of course, looking at backtest results in this manner is bordering data-snooping, but it does get us thinking: what kind of regime change has occurred in the volatility market since the end of 2012? When would it flip back again?