Maybe my thinking is too philosophically deterministic, but is it fair to think of probability for some of these major events not as: there’s an 80% this happens and a 20% it doesn’t, but rather: based on our best information, we’re 80% confident this will happen, and therefore there’s a 20% we’re wrong. I mean -- by election day 2016, the die was set and there was no longer a 70% chance of a Clinton win actually happening -- we just didn’t have all the information necessary to know that and based on the info we *did* have there was reason to believe with 70% conifdence Clinton would win.
(Perhaps what I’m saying is painfully obvious, or I’m out to lunch, but something rubs me the wrong way in saying there’s some percentage chance of something happening or not at a moment in time, where it’s too late for an actual course change for a different result, we just don’t know the result yet at that moment)
That's a reasonable perspective that hits on some really fundamental issues of philosophy and physics. A good way to think of it is this: If we could re-wind the universe to that morning in 2016 -- that is, return everything to EXACTLY the way it was -- and let the film roll again, do we again get President Trump that night? In a deterministic understanding of reality, the answer is yes. In this view, expressing a probability is only, as you say, a reflection of your imperfect understanding of reality. If we understood reality perfectly, every forecast would be binary: It will happen or it won't.
The contrary view is one that I (and many people far more informed and smarter than I am) hold. If we were to replay that day, it may or may not end with a President Trump. In fact, if we were to replay that day, say, 1000 times, a percentage of those outcomes would end in President Trump while a percentage would produce President Clinton. The split would represent the real probability of each outcome. In this view, the goal of forecasting should be to get as close to that split as possible.
Pursuing this debate is above my pay grade, however.
Maybe my thinking is too philosophically deterministic, but is it fair to think of probability for some of these major events not as: there’s an 80% this happens and a 20% it doesn’t, but rather: based on our best information, we’re 80% confident this will happen, and therefore there’s a 20% we’re wrong. I mean -- by election day 2016, the die was set and there was no longer a 70% chance of a Clinton win actually happening -- we just didn’t have all the information necessary to know that and based on the info we *did* have there was reason to believe with 70% conifdence Clinton would win.
(Perhaps what I’m saying is painfully obvious, or I’m out to lunch, but something rubs me the wrong way in saying there’s some percentage chance of something happening or not at a moment in time, where it’s too late for an actual course change for a different result, we just don’t know the result yet at that moment)
That's a reasonable perspective that hits on some really fundamental issues of philosophy and physics. A good way to think of it is this: If we could re-wind the universe to that morning in 2016 -- that is, return everything to EXACTLY the way it was -- and let the film roll again, do we again get President Trump that night? In a deterministic understanding of reality, the answer is yes. In this view, expressing a probability is only, as you say, a reflection of your imperfect understanding of reality. If we understood reality perfectly, every forecast would be binary: It will happen or it won't.
The contrary view is one that I (and many people far more informed and smarter than I am) hold. If we were to replay that day, it may or may not end with a President Trump. In fact, if we were to replay that day, say, 1000 times, a percentage of those outcomes would end in President Trump while a percentage would produce President Clinton. The split would represent the real probability of each outcome. In this view, the goal of forecasting should be to get as close to that split as possible.
Pursuing this debate is above my pay grade, however.
I once played a game of DnD where all I needed was a 2 or higher to succeed, and I rolled a 1.
A superb read - and a timely reminder of the relevance of Superforecasting. Take the time to read this slowly and with care.
Love how you've transformed this fuzzy theoretical point that usually causes eyes to glaze over, into a concrete relatable *feeling*. Works wonders.
I'm tempted to turn this into a visual version, if I may.
Also love Marc's point, thanks for clarifying both views of probability can coexist in the wild.