Statistical Oddity in 538 Senate Likelihood Prediction


What strikes me as weird is that the likelihood of the Republicans ending up with 51 seats is less than both the likelihood of 49 or 50 seats and the likelihood 52, 53, or 54 seats. This is unusual.

The only thing I can think of is that it’s being driven by some binary variable (or correlated variables) which flip(s) several seats at a time. So that if X happens the Republicans gain a whole bunch of seats, and X either happens or doesn’t happen with no intermediate possibilities.

But I’d appreciate other insight from statistical-minded folks.

[Note to moderators - and others: this is not intended as a political thread about the likely outcome of the election or the validity of 538’s predictions. It’s solely about the statistical aspects of this prediction, as described.]

Nate Silver explains.

OK, that answers the question.