Having issues with an outsider non-establishment candidate as a major party nominee more than anything IMO
Having issues with an outsider non-establishment candidate as a major party nominee more than anything IMO
I think silvers model is having trouble because of the uncertainty of third party voters and late stage "undecided", it's roughly 20% of likely voters giving a huge variable to any simulation run. The other elections where he based his model had pretty easy binary 0 or 1. It would be interesting and maybe he has done it but go back and use historical polls with his model for when Perot ran.
I don't know exactly how Silver's simulation works, but these models are pretty easy to construct and simulate. Something like this is going on: Each candidate starts out with 0 EC votes and each state would be treated as a simple, independent Bernoulli trial, a coin flip with a weighted probability of outcomes. If the Bernoulli trial in state i returns a 1, add the number of EC votes in state i to Clinton's previous EC total, if it returns a 0 add the EC votes to trump. So with something like that, you can run the Bernoulli trials thousands of times and sum thousands of "independent" EC vote totals. Adding 2 additional candidates (Johnson and Stein) has similar model mechanics except your Bernoulli trial is no longer a binary outcome, I'd model it as a 0,1,2, or 3 outcome with differing probabilities of returning each outcome and use if statements to add the EC votes to correct candidate. This type of simulation approach incorporates all sorts of uncertainty into the predictions. The tough part, and where Silver offers the most transparency of any site out there, is figuring out what the probability of getting a 1 or a 0 in each state. Silver offers extensive explanation of what polls he uses, and how he accounts for the quality of each poll. Anyway, the probability of winning is simply the proportion of the thousands of replicates where each candidate had >= to 270 EC votes.
What is really the craziest thing about those output distributions is that both candidates have a non-zero probability of getting nearly 0 EC votes. It is highly unlikely, so don't count on it, but what it does say to me is that Silver's model is having a really hard time converging on a likely outcome.
Based on previous debate results, Silver is predicting a 2-4 point bump for Clinton after last night.
Based on previous debate results, Silver is predicting a 2-4 point bump for Clinton after last night.
Which means Trump actually goes up 7 points, because why not?
I didn't watch the debate (cooked dinner, went for a run, etc.) but all of the post debate commentary and analysis indicates that Trump was a mature hour and that Counton did a good job of making him look unprepared and ... What's a good synonym for stupid?... because I don't want to be mean or condescending, but stupid is the only word I can come up with.
That would make it a 4.5 point race going into early voting. 2012 race was a 4 point/332 EV race. Will be a bloody GOP Civil War if a deeply flawed HRC gets close or exceeds Obama 2012. Lots of reluctant 'Pubs will vote against HRC in a general election, but 14M 'Pubs enthusiastically voted for Trump in the primaries. Those two blocs can no longer exist in the same party and will sever an alliance that's been shaky since 1992.
This is my favorite auto correct / typo.
The reverse jinx thing you've been doing re: Trump/McCrory is interesting given your optimism on Wake bball recruiting. You need to choose a lane.