Tail event: Difference between revisions

Now here is the thing. When we calculate probabilities — when we roll dice — we are in situations of ''known risk''.  We don’t know exactly how this particular event will turn out, but we know the range of possible outcomes and the calculated probability of each. No matter how deft our technique, no matter how surprising or extraordinary their trajectory, the dice must come to rest showing one of six equally-sized faces uppermost. No two throws are the same. The trajectory is chaotic, but all this intractable uncertainty is wiped out when the dice come to rest. None of it matters. All that matters is the average. By that we can set our watches.  

Every fair die has the same characteristics. It is not just that on some some dice the probability is more like ⅐, on others now like ⅕ but, on average, dice work out at all about ⅙. Every die must, within tolerance, behave exactly the same way. Therefore probabilities are a valid means of predicting behaviour.

When we roll dice to ''determine'' an outcome we do not build a statistical model that predicts a ⅙ probability: we build the dice to yield that outcome. A die is part of what [[Nancy Cartwright]] would call a “[[nomological machine]]”:<ref>This is a ''terrible'', typically ''academic'' label. No doubt it is etymologically accurate, but it is forbidding to a lay reader. Academics, like lawyers, tend to do this while they train and occupy the junior rungs as a self-credentialising device. By the time they sit on the higher rungs, they don’t know any different way of writing.Cartwright is a brilliant thinker, but her writing is dense and hyper-academic. </ref> a carefully designed, constrained, hermetically-sealed [[simple system]], designed to generate a specific theoretical outcome. If over time our dice don’t yield the ⅙ outcome we want, we don’t conclude the ⅙ outcome is wrong: ''we throw out the'' ''dice''.