It's on the aggregate of predictions that you can see the accuracy. If my weather app tells me hundred times there is 10% change of rain but it rains 50% of the time after 100 days, it's completely useless.

Same for FiveThirtyEight. They have been massively wrong on multiple counts.