If you didn't draw it in a log scale, you wouldn't see the patterns when they occurred in the lower levels since the size of the later movements would dominate the graph.

Basically it is a way to show that there is similar movements in the value even when the ranges they move in are completely in different scales.

But doesn't this effectively just massively under-represent the actual deviation between the price and estimate and the error bounds. If you convert back in to real numbers this estimate is like saying APPL will be between $100 and $1000. I'd be interested to know if there are any other traded products that we view on a logarithmic scale.

If APPL is currently at $90, that prediction says you'll make somewhere between a lot and a ridiculous amount of money. It's not at all precise, but the action you should take if you believe the prediction is clear.

I assume because in its short history, the bitcoin price has spanned many orders of magnitude