A slightly revised version of my previous idea (linked below). It goes to 2.8k :-( :-/ :-| :-)
Basically the same as my last idea, except I thought I would throw the 2016 channel in there. Strangely, the lower trendline of the channel matched my "low mean" almost exactly. Hmm.
I zoomed in on the 2014 decline from top and it's interesting to look at how the price was bouncing off the death ray initially as resistance, and then as support,.. but ultimately followed it all the way down, until the market was satisfied.
Peter Brandt tweeted today that most of the time parabolic advances end with an 80% loss when the parabola is violated, so he puts the low at under 4k somewhere.