AbdulAlhazred
Legend
Here's the thing, if ANY REASONABLE EFFORT would produce good probabilistic ideas of the outcomes of courses of action in the real world that are non-trivially complex, then why isn't the CIA feeding the US government with incredibly accurate advice? We know from historical record that, in fact, the advice given to national leaders by extremely well-funded, highly expert, and vastly experienced intelligence organizations is famously filled with garbage. Clearly no scientific discipline has been invented which is of much use in this endeavor! Heck, as a guy with a math degree and an avowed interest in this subject I am certainly no world-class expert, but if major advances, to the level of being at all useful, had been made,I'd probably have heard about it. We would certainly see the results!You do understand the difference between "Not unlikely" and "Not the expected result" right? If you think something has a 30% chance of occurring, is it the one you're going to predict?
Less-than-expected results happen fairly often. They're rarely predicted because they're not expected. There's even a term for it in medicine; you look for horses, not zebras.
This is rapidly coming across as you thoroughly missing my point.
And to be equally blunt, I think your premise is faulty from the start. So I guess we're done here.
I mean, your point about "things happen and many of them are not the most probable" IS actually a really good point. I think the problem is, there are such a vast array of possible outcomes of complex situations that most of the probability space is filled with very low likelihood ones. So even if a model can pick out a few of the most likely, they only represent a few % of the total probability.