Instead of determining regime-switching, an easier job might be, given the historical data, how do we define a bunch of regimes on it... Any good ideas about how this can be done?
Try this book "Trading Regime Analysis": http://www.amazon.com/Trading-Regime-Analysis-Probability-Volatility/dp/0470987855 For more simplistic definition of trending regimes, look at bull market vs bear market comparisons. I can't find the link now but it's something like, when the xDMA is above the yDMA there's a z percent chance the market's in a bull phase (trending up).