Hey y’all. I had the opportunity to help a professor model the election outcome in 2016. I had a full understanding of the methodology in place for these predictions. It was completely legit, and our team still had Trump with a sub 2% chance of winning for the month leading up to the election. The key factor was that ALL of these “% chance to win” models rely on a mixture of the same publicly available polling data, so if that’s not right, you can’t even begin to create an accurate model.
My point here is that the 538 website is probably doing their best, but nobody can get accurate polling data to input into the model. Models are only as good as their input data.
Hey y’all. I had the opportunity to help a professor model the election outcome in 2016. I had a full understanding of the methodology in place for these predictions. It was completely legit, and our team still had Trump with a sub 2% chance of winning for the month leading up to the election. The key factor was that ALL of these “% chance to win” models rely on a mixture of the same publicly available polling data, so if that’s not right, you can’t even begin to create an accurate model.
My point here is that the 538 website is probably doing their best, but nobody can get accurate polling data to input into the model. Models are only as good as their input data.