[GGG]

Originally Posted by Iowa_Flames_Fan

I do. Firstly the 538 numbers you cite above are projections of chance of winning, not vote share. And anyone who plays poker knows there is a WORLD of difference between a 30% chance and an 8% chance. For what it’s worth 538 gave Trump a higher chance of winning than just about anybody else (and that’s true in 2020 too).

I’ve said this before but people really need to stop saying the polls were “wrong” in 2016. With the exception of a measurable polling “error” in Wisconsin, the aggregate polling averages actually provided a very accurate picture of what happened. The “error” was pundits not catching the fact that there were very few high quality polls in certain swing states (most notably Michigan and Wisconsin) and assuming that voter preferences in those states would follow the national popular vote. As we know, that isn’t what happened.

Pollsters didn’t get 2016 wrong. Poll WATCHERS did. There are important lessons to be learned from 2016 but “don’t trust polls” just... isn’t one of them.

Totally agree, this idea that polls were wrong in 2016 isn’t founded in evidence.

In general polls for 2016 were highly accurate. Within 2% of the final results. This is more accurate than average.

I think the big issue was the Princeton poll aggregator saying 98% and then pissing al over the 538 model publicly. This created the perception of certainty. But even the aggregators that has high odds of Hillary winning like the New York Times (85%) had warnings over how uncertain an 85% chance is.

The Upshot’s elections model suggests that Hillary Clinton is favored to win the presidency, based on the latest state and national polls. A victory by Mr. Trump remains possible: Mrs. Clinton’s chance of losing is about the same as the probability that an N.F.L. kicker misses a 37-yard field goal.

So people’s perception of what 85% or 70% or even 98% (1 election in the history of the US) means is really the issue. Though given that lotteries are successful we shouldn’t be surprised.