Why Moneyball and Nate Silver work, but derivatives don't

I find Nassim Taleb's style annoying, and he often mischaracterizes the views of others.  But when I teach my mortgage backed securities course, I have my students read the Black Swan because of an important point he makes (over and over again): some things are well characterized by distribution functions with small numbers of sufficient statistics (such as mean and variance), while other things are not.

With baseball and politics, we can fully characterize the distribution of outcomes (or at least come close enough to doing so).  With financial markets, we really cannot.  What is sad is that many people disdain  the insights of statistical inference when it really works; others embrace such inference when it does not.