Hey folks, in the interests of full transparency:
We discovered and fixed a small bug in our primary model this morning. Specifically, our model was supposed to have accounted for home states/regions when projecting bounces, e.g. Buttigieg would get a smaller IA bounce...
We discovered and fixed a small bug in our primary model this morning. Specifically, our model was supposed to have accounted for home states/regions when projecting bounces, e.g. Buttigieg would get a smaller IA bounce...
…other things equal, since he's from the Midwest. We found this improved accuracy in estimating bounces historically.
However we had a sign reversed in our code😳, so instead of giving slightly *less* credit to a candidate from the region winning, it was giving slightly *more*.
However we had a sign reversed in our code😳, so instead of giving slightly *less* credit to a candidate from the region winning, it was giving slightly *more*.
This was modestly hurting Biden, who wasn't benefiting from home region effects as Buttigieg/Klobuchar in IA or Sanders/Warren in NH. With the fix, Biden's overall delegate majority chances improved from 41% to 45%, with the gains coming variously from the other candidates. FIN
