>>12249623>what the fuck is going on here?people are braindead idiots and don't understand stats or how to interpret it
probability is a mathematical "best guess" when we have incomplete information (uncertainty) using a model or some sort (all models are wrong, but some are useful, etc).
Further, people have this idiotic idea that "greater than 50% means yes, less than 50% means no" and toss probability completely out the window, instead opting for black/white "yes/no" answers and then getting angry when their black/white interpretation is wrong
Any model that puts one candidate at >98% is most likely too biased. Still, 98% is not 100%. A model can predict a candidate to win with 98% certainty, and that candidate can lose, and the model will have still been correct. Because 98% is not 100%. Not even close.
98% means you expect that the other candidate would win 2/100 times. Lottery odds are way, way worse than that, yet people still win the lottery.
Using a more conservative mode, 538 gave Clinton a 3/4 chance to win, roughly, near the end of the election (if I remember correctly). Would you play Russian roulette with a 4-chamber revolver for $100? Why not, its 75% to win! Would you be utterly surprised if you shoot yourself first try? I wouldn't be. It's a decent chance to still happen.
Don't do this: interpret a model as "right" if a candidate wins when they were projected to have >50% chance of winning and "wrong" if they lost instead. That's just your monkey brain making assumptions (aka turning it binary by assuming that >50% or >80% or any number is a safe cutoff to say "Candidate A will win if the model is correct"). Again, a model can suggest a 98% chance of victory for Biden, and Trump still win, and the model be completely valid.
