BoardWatch · analysis
Vote predictor accuracy scorecard
A prediction is only worth something if it was made before the vote. This page scores every projection recorded in advance, and nothing else.
No scored predictions yet
This scorecard begins with model v0.2, published Aug 24, 2026. It will show a score once the board has decided a matter that was projected here beforehand.
Why it does not simply score the past. The model’s rates come from the board’s existing record. Testing it against those same decisions would be marking its own homework — it has already seen the answers, and any accuracy figure produced that way would flatter it. The previous version of this page reported such a figure. This one will not.
How accuracy is measured
The headline figure is the Brier score: the average squared difference between the projected probability and what actually happened. Zero is perfect. Predicting 50% every time scores 0.25, so anything worse than that is worse than a coin toss.
It is shown beside a baseline — what you would score by ignoring the model entirely and always predicting the board’s historical approval rate. This board approves the large majority of what reaches it, so a high hit rate proves very little on its own. Beating the baseline is the only result that means the model is contributing anything.
Corrections
If something here is wrong, tell us at editor@monumentindependent.com. See also our editorial policy and corrections.