npr had a person who mentioned even if you take the official margin of error into account there is still a lot of sources of uncertainty. better to just fuck polls.

You need to consider both statistical and systematic errors.

10,000 responses is enough to give you a statistical error of 1%, but the systematic error on 2016 was around 2%, and you can’t beat that down with more responses.

Also, the red team has a structural advantage of a few percent.

All of this means that the blue team needs to be up by at least 5% in national polls before they can be considered to be ahead.

I don’t think that’s correct. Even when the margin of error is high relative to the difference between the politicians, the poll still provides some non-zero amount of information about which candidate is ahead. If the results are 49% vs 51% with a margin of error of ±5%, you should still bet on the candidate with 51% if you have to bet.

Now my pet peeve is when the media presents national polls as if they are the right polls to look at in order to predict which candidate will win. They should look at just the swing states!