See https://en.m.wikipedia.org/wiki/Brier_score
This is about predicting individual house and senate races, as of the day before the election, not the overall presidential race.
If it's obvious that one forecaster performed better (e.g. they correctly predict the overall winner and the other predicts the opposite) then this market resolves accordingly.
If someone else's analysis showing that one forecaster performed better, then I'll go with that as long as it seems statistically sensible. I'll count an analysis from Silver or 538 themselves.
If it's not obvious and noone else does an analysis or there are multiple conflicting analyses, then I'll calculate Brier scores myself. I'll use the predicted probabilities for the house, senate and governor elections if available. Lowest Brier score wins.
1,000
3.00If I end up doing it, I would count each state (and congressional district for Maine and Nebraska) as a separate prediction.
@jacksonpolack that said, if someone does an analysis based on earlier predictions, I'm not likely to do a whole new analysis of the day before unless it seems necessary.