Nope, only omniscience can make predictions. I.e. you have to know exactly how universe works and at what point it is now and where it is going.
That's not statistics though, you won't need any trends, correlations, percentiles, etc - you will know it all as is.
That's exactly my point.
As is always the case with statistics. Very little data.
By "very decent form of prediction" you mean probability, obviously, so not a prediction, more like - likely continuation. And it might work for a couple of times even. The problem is there will be times it won't work, there are errors in every model.