Calgarypuck Forums - The Unofficial Calgary Flames Fan Community - View Single Post

octothorp · 08-04-2016, 11:55 AM

I'm not sure which 538 model you're looking at (they have several), but if you're looking at something like this page, keep in mind it's not showing you percentage of votes in polls, it's showing you percentage likelihood that they'll win based on current polling. (In other words, a candidate might have 55% of the support in a state in a polling aggregate, but there might be a 75% chance that a candidate with 55% of support at the beginning of August wins that state).

http://projects.fivethirtyeight.com/...-forecast/#now

Also, 538 has three different forecasts: polls-plus, polls only, and nowcast (which one of the 538 staff recently described as a squirrel on mountain dew trying to predict the election). The now-cast basically takes the polling data as it stands now and uses that to predict what would happen if the election were tomorrow. Polls and polls-plus both take into account that we're still a long ways out and the numbers can move a lot; the difference is that polls-plus takes into account things like anticipated convention-bounces, as well as economic data and such. Right now polls-only looks a little better than polls-plus for Clinton, because polls-plus is essentially saying "yeah, but this is right after a convention, so we'd expect Clinton's numbers to be high."

In the polls-plus model, you'd need to include every state where Clinton currently has 65% or greater odds to get her to 270.

08-04-2016, 11:55 AM	#44
octothorp Franchise Player Join Date: Oct 2002 Location: not lurking Exp:	I'm not sure which 538 model you're looking at (they have several), but if you're looking at something like this page, keep in mind it's not showing you percentage of votes in polls, it's showing you percentage likelihood that they'll win based on current polling. (In other words, a candidate might have 55% of the support in a state in a polling aggregate, but there might be a 75% chance that a candidate with 55% of support at the beginning of August wins that state). http://projects.fivethirtyeight.com/...-forecast/#now Also, 538 has three different forecasts: polls-plus, polls only, and nowcast (which one of the 538 staff recently described as a squirrel on mountain dew trying to predict the election). The now-cast basically takes the polling data as it stands now and uses that to predict what would happen if the election were tomorrow. Polls and polls-plus both take into account that we're still a long ways out and the numbers can move a lot; the difference is that polls-plus takes into account things like anticipated convention-bounces, as well as economic data and such. Right now polls-only looks a little better than polls-plus for Clinton, because polls-plus is essentially saying "yeah, but this is right after a convention, so we'd expect Clinton's numbers to be high." In the polls-plus model, you'd need to include every state where Clinton currently has 65% or greater odds to get her to 270.