Quote:
Originally Posted by GGG
I like 538 but the credit they get for the last two elections is a little overblown.
There might be 5 or 6 swing states out there so over two elections you might have 12 difficult choices. So assuming a 50/50 odds random chance gives you a 3/1000 chance of winning. Now even the 6 closest states probably have a 60% favourite which increases the chances of randomly going 11/12 or better is about 1/50.
So if in 2008 there were 50 websites predicting election results you'd expect in 2016 to have a few Nate Silvers existing. The other thing is that you'd expect eventually an election to occur where 538 would go 0/6 on swing states otherwise the probiotics assigned to the favourites wouldn't be high enough. In the end if you did the analysis I don't think you could say you have statisticly significant evidence that Nate is better than you would expect from random chance.
To me the value in any of these aggregates is doing all the leg work to interpret all of the polling out there in a constant methodology to provide a baseline of information around the spin. Also there work on inferring the effect of each state based on the behaviours of surrounding states, demographics and national trends.
|
So even though the echo chamber has decided otherwise, I never said that I intrinsically distrusted his mathematical methodology, as I plainly do not understand them. Neither do you, nor anyone else in here as everyone is basically relying on his results, and anecdotes, authoritative opinions etc...
That was my point. Very few of you guys can read even though apparently you are all experts at reading polls.