Most of life is inherently uncertain. How can we predict these events — and should we even bother trying?
As humans, we naturally think of chance in a normal distribution. This is probably because, out of all probability distributions, normal is by fair the easiest to calculate given brainpower.
Of course, chance never takes on such a simple concept. It is rife with heteroskedasticity, fat tails, black swans… all sorts of events which allude our natural heuristics.
So, we built models. We drove forecasts with all these things in mind — building bigger, more complex probability distributions, with more variables and a higher success rate. To some extent, these models are valuable — for example, robustness factors can help us better understand data, and make logical conclusions based on correlation or even causation. However, this doesn’t predict. Just because we know X follows Y doesn’t mean we know that X will follow Y tomorrow, or will instead follow Z.
A great example of this is with financial crisis. Obviously, we cannot predict financial crisis — otherwise, it would not happen! Instead, risk managers look at their models and see that smooth sailing (X) follows current conditions (Y), though the reality is that current conditions lead to economic collapse (Z).
I have spoken about this before, and my opinion stays the same: I think forecasting is possible, but we are a long, long ways off. At this point, we can only pretend to know the complexities of chance. It is, like emotion, an abstract concept, fully protected in a fog of war until we mine closer to its deposit. And so, why should we pretend that we know about the uncertainty principle?
A good risk manager does not build a model, then listen to that model to tell whether they’re safe. A good risk manager simply plans for the worst. If you have a plan for the worst uncertainty, then you are safe. You can do all the cost-benefit analysis and probability distributions you want, but nothing is going to protect you from a naturally unforecastable event. The best way to protect against an unforecastable event, is to assume that it will happen. Because some day, it will.
Of course, the question then is how we get to perfect forecasting. I mentioned why we should pretend to know about the uncertainty principle — in reality, that’s a pretty farce question. After all, we have to pay all those forecasters somehow! If we one day decide to say we won’t forecast because it’s irresponsible, all those people down below will look up, scratch their heads, and say “So you are telling us you just made all that shit up?”. Some might even demand to have their forecasts back, even if they are faulty. Either way, people won’t be very happy.
So perhaps there are some less consequential places to test and build up forecasts in the meantime. While financial institutions certainly shouldn’t try it, perhaps sports betters could. Perhaps an individual trader could set a (small!) portion of their portfolio to try out some new forecasting techniques. Or an amateur meteorologist sets up a model to see if it will rain tomorrow. Obviously, if we’re wanting to break the uncertainty principle, we’re going to have to develop the research somehow. It’s just best not to stress too much about it.
