Understanding the art of prognostication is not that dissimilar to understanding weaving. Few things ever occur out of the blue - they just hadn't emerged out of the background noise just yet, and as such when they do appear, you can usually see how the threads extend backwards, and add these things into your model.
I've been thinking about models a lot lately. The human brain is a remarkably powerful modelling tool, probably the best such tool on the planet. It keeps track of those things that it can perceive, adds in probabilities for those things it can't see, and uses this to insure that you can get through the day without walking into the side of a building or stepping in front of a car.
Thus, what most people perceive as "reality" is a model that they keep inside their heads, that's constantly changing, and that is for the most part heuristic - it doesn't try to ascribe reasons to why things happen (reasoning is a considerably higher function that came along fairly late in the game) but works by keeping this floating stream of metaphors going - this is like that, so this will behave like that ... until it doesn't. This fact then gets added into the metaphor, that doesn't-ness is a very distinct possibility for "this".
A set of prognostications, then, is a model, one that tries to incorporate past known history with the interaction with other threads. These interactions are where things get exciting. It's relatively easy to follow a thread into the future as long as it doesn't interact with other threads, but the moment that it does, you have to make a guess, based upon what is known, about something that is currently unknown.
This is where prognostication becomes analysis - trying to think through how the interaction of two or more threads will shape each (or whether in fact they will merge and become a broader thread). It's doable ... in most cases if you mix two known things for which analogs exist that are also known then you can have a pretty good idea about what comes next. But not always. This is why prognostication is still more art than science.
One final aspect about prognostication, before launching into my predictions for 2009. There is sometimes a delicate line that you have to walk between being an observer and being a participant. The observer, having no stake in the final outcome, is usually able to interpret those threads and interactions with comparatively little bias, while the participant has a definite bias that colors his or her perceptions, and that can consequently amplify trends in a given model that aren't that important or mask trends that are.
Unfortunately,in order to understand a system, you also have to participate in it. Walking this line between observer and participant is challenging, and the consequences of it should be obvious if you look at what happened on Wall Street, where analysts were paid significant sums of money in order to not only observe the system but also to game it. They grew to see what they wanted to see, not what they needed to see, and in the end this lack of vision brought The Street down.
This is a long report, and one that hits a number of different topic areas. Both in order to make it easier to digest this and for me to enter it into the O'Reilly CMS, I have broken the Analysis 2009 story into a number of separate articles. In order to see the whole listing, click on the Analysis 2009 link. You can also download the article asa PDF File, here.