An Introduction to the GoodCompass Electoral Model
Humility is not always helpful.
In preparation for the 2024 presidential election, I created my own mathematic electoral model with the hope that it would be able to successfully predict the outcome of the election in most states. My model succeeded with a 98% success rate, meaning it got 49 states right (the one mess-up was in Michigan). For the Senate races, it boasted a 94% success rate (it missed PA and Arizona). I was not confident in my predictions that year, so I did not publish what my model was saying. Now, with that record, I am finally taking the step of publishing my model’s results, here on this, my blog.
What makes your model special?
My interest in election predictions came about in a successive few elections where people have made predictions based on completely malleable indicators. Estimating turnout, reading the early vote tea leaves, following the betting market, all of these things are malleable, and if you base your prediction based on these things you’re eventually going to get burned. When I created my model, I wanted to prioritize those aspects that weren’t malleable; those aspects that always correlate in some degree to the result. These are as follows:
The national environment (best exemplified by presidential approval).
The inherent partisanship of the state being predicted (best exemplified by past elections).
The result in the last presidential election (which will be useful as long as America remains polarized).
Polls from a select group of pollsters who are either well-rated or useful for rounding out the model.
These indicators make the model extremely valuable when predicting for polarized races. The further down the ladder you get, the more unhelpful my model is. My model also is designed to give predictions for margin, not vote total.
I am now somewhat-confident in my model, and I think it will perform well in the elections this year, otherwise I would not publish it.
Will there be subscriber-only content?
No. I’m not actually that good or knowledgeable, all I’m really good at is plugging in numbers. There would be very little *exclusive content!!!* I could show subscribers.
Statement of Bias
For the sake of full disclosure, I am a Republican, and have voted Republican in every election in my life. I also run an X account called CapitolMoment in which I often share Bible commentary, make statements of support for Catholic Social Teaching, and discuss other religious and political issues. I sometimes also comment on elections, but I caution you with the fact that, like I said, I know very little.
Now, that might make you think my model is therefore biased toward conservatives. You’ll see tomorrow. But I’m also not picking and choosing numbers, there’s no personal input that goes into the prediction besides the choosing of accepted pollsters, which is done months prior to election day.
When are you publishing the 2025 predictions?
Tomorrow, November 3rd, 2025, around 7 PM.
Until then, welcome to GoodCompass.