Very true. I've posted about this tendency in other threads. That's a good rec. Eric Zivot is smart, creative, organized, and a good communicator. BTW he offers an online course on computational finance using R, via U of Washington - http://bit.ly/kqEfKL
No offense, Mizhael, because I like your questions and your proactive approach in terms of what knowledge might be important to building a viable strategy - but this is a valid point. X3
It's not that he asks questions of others -- that's fine -- it's that he doesn't do any legwork himself, or share any of his own findings or insights. He takes but never gives.
Well, a simple value on value scatter plot of auto-correlation against volatility (same rolling frame) would give you a general answer. What you are probably going to see is that at low values of volatility autocorrelation can be positive (trending) or negative (meanreverting), while at higher vol levels it is in general negative. Even better is to plot weekly change standard deviations against daily change standard deviations - the weekly volatility will gently flatten as daily volatility increases
Yeah, I know all about the beta scatter plot functionality on the Bloomberg. Got access to EuroMTS ? Pretty please ?
There are a few things that I really don't understand: A typical low-vol price series is a straight-line. It has 0 vol. It should be the perfect setting for trend-following system but worst for mean-reverting. Maybe we should define "trend-following" and "mean-reverting" system first, precisely. What's a typical "trend-following" and "mean-reverting" system, in your mind?
I just don't understand why people universally think of using volatility here... as I pointed in my other post, a straight-line with constant slope has a vol = 0 ... but it's perfect for trend-following... I am afraid that vol is only a secondary factor there, there is a more significant factor that that's more relevant... Any more thoughts?