I mean, it’s not a really simple result.

I guess in a nutshell:

• There is limited Trump-vs-Harris state polling data available. Some has to be predicted based on other polls, and there won’t be many sources to work with.

• Harris polls slightly better nationwide than Biden, but it’s not by much.

• However, the model they use, which makes use of information above-and-beyond just poll results, also predicts Biden to have a better chance of winning the Electoral College.

• However, this model is built based on assumptions that may not hold for this particular unusual situation, and the authors are not sure how well those assumptions will hold up.