So, Nate Silver has an article where he talks about the "128 paths to the white house", and how generally speaking, the most likely paths are either a Trump or Harris blowout where they win all 7 swing states.
I know I have that kinda crappy and amateurish simulation where i predict all the possibilities for how the election can go, but it rarely predicts all 7 swing states going the same way. But this is because it avoids the more linear model for how polling goes. It treats each state has an independent trial not linked to the others, when in reality, we can expect a candidate to over/underperform systemically.
Here's the most recent election data from today. As we can see, Trump is winning now, with a 52% chance, to Harris's 48% chance. The margins fluctuate daily and my formal predictions happen once a week. Let's see where it goes over the rest of this week before panicking about that. I'm just posting this to make a point.
If you treat elections linearly where if a candidate overperforms in 1 state they may overperform in all of them, and that performance between states is related to national trends, then I guess the above model also predicts a blowout is the most likely scenario.
31% of outcomes should have Harris winning all 7 states. On the flip side, 36% have Trump winning all 7.
And any combination of the states is gonna be the other 33%.
That's the reality of the election. The 7 states are so close any major shifts in the national environment relative to polling will produce a blowout outcome for either candidate. This is why, even when Biden was down, I gave him like a 25% chance on average (although it dropped to 13% by the time he dropped out). If Trump overperformed, trump would win, if the candidates performed as the polls predicted, Trump would win, but if Biden overperformed by so many points, he could win. His window was narrowing significantly post debate, but for a while I estimated Biden still had a 20-30% shot at winning, and that's what my prediction numbers really mean.
This is also why I say that we shouldnt worry about anything as long as it's within the 60-40 zone. because ANY over/underperformance for EITHER candidate could flip the whole thing one direction or another. Right now, Trump is at 52%, Harris at 48%, last week Harris was at 59% and Trump at 41%. Are the two that statistically different from each other? Not really. If Harris overperformed even a point, she'd win. If Trump did, he would flip the entire result. If any candidate overperforms by 2 points, they win all 7 swing states. That's the reality of the race right now. it's that close. And Silver says up to 3-4 points is just normal polling error. For one standard deviation, that's correct. I admit, my model has a little less certainty and spread if he's assuming a 95% confidence interval as you could be off by up to 8 points and still be in the margin of error, but still. That's the reality of things. And that's why this race is anyone's game. You really need to be ahead by several points for the race to really be skewed in favor of one candidate or another. Right now, ANY randomness could throw the election either way.
And yeah, I just wanted to discuss my perspective on nate's article.
No comments:
Post a Comment