Quote Originally Posted by Dan Druff View Post
It will get really interesting if somehow Trump can win PA, because then Biden will have to sweat out having to grab Ohio, Florida, Georgia, Texas, and NC. And if somehow Trump wins Arizona, then Biden really has to sweat.
Biden can win without PA, but he becomes an underdog if he loses (he definitely can't get blown out there). However, if Biden loses PA he cannot win Ohio. It's basically impossible. If Biden wins PA by 6+, then he'll probably barely win OH. If he wins by less in PA, then he'll lose OH. However, Biden can lose a state like PA and still win a state like AZ. That is possible. This has to do with demographic correlations between the states.


You can see this here:
https://projects.fivethirtyeight.com...p/?cid=abcnews


You can see they give Biden a 50% chance in OH. But now plug in PA for Trump. Now his odds in OH go from 50% to 4%. But if you look at AZ it goes from 72% to 40%.

Also, I agree in regards to Iowa. Trump won Iowa by almost 10 points in 2016. So in a vacuum, a Trump +5 in Iowa poll is not bad. If Biden swings 5 points in the midwest that's good news for him. Also, the poll is likely an outlier. Sort of like the opposite of the Wisconsin +17 poll, which was also likely just an outlier. Good pollsters will sometimes have outliers. In fact, that can be the mark of a good pollster. What's bad is when a pollster has a poll that looks like an outlier so instead of publishing it, they bury it or they play with the numbers to get it closer to what the median is. No reputable pollster wants to publish an outlier when they know it's more likely that their poll is an outlier as opposed to a poll that is capturing a legitimate swing. Well, you can see the problem with this. It's possible their poll wasn't an outlier and now they just buried it and perhaps 3 other pollsters did the same, so we never even get a sense that there was a swing. This is called 'herding', where you try to get your poll to look more like what the average is saying so your polling company doesn't look like a fool.

So you still want to look at averages, not cherry pick your favorite polls. But you don't want pollsters to dump the polls that have the potential to be cherry picked. That would produce bad averages.