<< Back to Home
2016 Presidential Forecasts
Take a deep dive into the 2016 Presidential Forecasts.
Forecasts
1 ∘ signifies qunatitative model. • signifies qualitative ratings.
2 The lower the brier score the better. It is calculated by the averaging (prob - outcome)^2. For example, a 70% race goes the other way, (.70-0)^2 = .49. Brier score is more complete compared to just win %, as win % is binary, and brier takes into account the given probability. Presidential Brier and call % are weighted by electoral votes.
3 Experts who use tossups will be 50% correct on all tossups.
4 Midpoint Derivative Polynomial Test (MDPT) is new function to quantify a forecaster's model calibration. The way it works is it takes a third degree polynomial of the forecasted probability and the outcome. From there the derivative is taken, and the derivative is taken taken at the midpoint. This tells us if the forecaster is overconfident or underconfident. If is derivative is less than 1, the score is 1 / derivative. A score of 1 is perfect calibration, and the lower the better. *The more races in the forecast, the more accurate the score.*
In-Depth Look at MDPT
MDPT is a way at looking calibration and how well a model or ratings proejctions match the actual outcome.
Consensus Forecast
Take a look at the average forecast for each race, and what % of forecasters got it right.
|
Ratings |
% Correct |
Ratings
See how each forecaster rated each race. Hover over a forecaster's column to see their ratings for each race.
Created by: Jack Kersting
Download Data
