Udolpho.com
 

Nate Silver is a fucking idiot… The New York Times' resident election nerd--Nate Silver, the guy who does frivolous horse race columns for readers tired of substantive issues--has been all over the blogosphere or whatever this thing is.  Tongues are wagging because he has a model that is supposed to predict the outcome of the election to an absurd precision (e.g. 83.7%).  Don't forget the decimal point, it's more scientific!

 

As Sean Davis pointed out in The Daily Caller:

 

Silver stormed onto the scene in 2008 when, according to his acolytes, he correctly predicted how 49 of 50 states would vote in the presidential election (he missed Indiana). Do not remind his disciples that of the four close states -- those with margins of 2.5% or less -- Silver only forecast three of them correctly. And definitely do not remind them that the polls in swing states correctly forecast all but two states (Indiana and North Carolina).

 

One of the things that really annoys me about Silver's lapdogs, and Silver himself, is when they throw out "49 of 50 states" in 2008.  This is maybe impressive if you just fell off a haystack, because not very many states were actually close in 2008.  McCain was a terrible candidate who ran a terrible campaign as Party successor to a terrible president after the economy had just tanked.  And given the geographic split in the country, at minimum 30 states have surefire, can't-miss outcomes--they are so strongly tilted in one direction that you can predict them four years out.  (WaPo only identified 10 swing states that were in doubt.)  Silver and his ridiculous followers massively overstate his accuracy.
 
Then there's the obviously unfalsifiable nature of Silver's ever-shifting estimate that Obama has something like an 80% chance to win based on the electoral map.  Silver produces useless numbers that don't really tell you anything--anyone can already say that it is a close race and that if state polls show Obama up slightly the weekend before the election, he's slightly likelier to win--whoops, that's if the state polls themselves are accurate.  As it turns out, they've conflicted with some of the national polls (which give a slight edge to Romney), and Silver has yet to offer a convincing reason to believe the state polls are more accurate.  (State polls led him astray in 2010, when his accuracy dipped.)
 
Silver tries to weight the polls based on what he thinks are their historical bias or reliability, which is also fairly dumb.  It's like trying to figure out what the stock market will do by examining past patterns.  The problem with doing this is that the stock market and presidential elections don't run as closed loops, and there is always going to be a certain level of risk in assuming that "predictable" factors stay predictable.  Things stay the same until they change.  That's why it's ultimately silly to claim two decimal points of accuracy in your garbage-in, garbage-out Excel model.  I mean you're just bullshitting people.
 
The bottom line is that Silver is jizzing all over his fucking stupid poll spreadsheet (in which he averages polls together, as if this is scientific) but he's not telling us anything we don't already know.  No matter which way the election turns out he's not "wrong", but then again he's not "right" either.  His input is one big null value.
 
Silver is useless when the race is close, because even a small bias in a handful of state polls could foul up his model, and he's useless when the race isn't close, because who gives a shit that this Jew's spreadsheet tells you what you already know?  My model therefore gives a 100% useless rating to Nate Silver.  Well-earned.  (Next time, Nate, do three decimal points!)
 
And 538's commenters are almost all abject morons.  I'll bet most of them have libarts degrees.

 

Follow a discussion of this at My Posting CareerSaturday, November 3, 2012 - 5:26 PM  

rss feed atom feed