Beyond Polling all available election data and makes swing state projections. The highest weight is placed on polling and voter registration data, and how that data correlates to past elections. Polsters are generally judged as good or poor based on their poll point spread projection. However, our focus is on a Polsters ability to generate a percent that correlates to the actual percent a candidate receives. For instance, if a poll showed a race as being a five point win for a candidate but the candidate wins by 2, that poll may be seen as poor. However, if they predicted candidate A as getting 48% and that was the result, the poll should be seen as high quality in regards to that actual candidate.


Adjusted Polls
Part of the projections we make, are based on adjusted polls based on past polls (by the same Polster) from the previous election cycle.


Registration
A crucial data point for predicting the correct outcome of each election is voter registration data. However, not all states publish voter registration statistics that include party registration.


Independent Vote
The independent vote can be tricky to predict since the empty space in the polling is a mix between undecideds and those who vote independent. Therefore, this undecided/independent number will always be less than the actual votes independents receive. For this reason, the max independent vote in our projection will be the current undecided/independent number based on the average of all polls.

A polling issue exacerbated by polls with only two parties (and ignoring third party candidates), is that the independent vote percentage becomes unclear. The remainder percent reflects independent votes as well as those who are undecided and those who are secretive about their intentions. Historically the remainder percent greatly overestimates the percent independent candidates will receive (even previous to the 2016 election cycle).

Nate Silver is an excellent statistician, however, even his methodology is flawed when it comes to the independent vote. For instance, on October 23, 2024 the RCP average for Pennsylvania was 47.9% Trump and 47.1% Harris, which assumes 5% will vote independent, are undecided, or are shy. Conversely, Nate Silver has the race tie at 48, which would represent a 4% remainder. This assumes only 1% of voters are either shy or undecided. Our model indicates only a 1.81% independent vote in Pennsylvania. Therefore, Silver may be missing 2% of votes in his model split between Harris and Trump.