RCP rehabilitates ARG, pisses off 538

[Mr. Morden]

I agree that it makes no sense to purge ARG from the database, and then reinstate it once it starts showing results that you like.  However, I'd like to speak out in favor of subjectivity with regard to choosing which pollsters to include.  Nate Silver implies that RCP is doing something illegitimate by disregarding certain pollsters based on fuzzy criteria that aren't spelled out anywhere.  But is that necessarily a bad thing?

Recall how most of us on this forum considered ARG a joke poll *even before the voting started in this year's primaries*.  This was not because of any past record of inaccuracy.  It was because ARG's polling for the primaries was so wildly off from every other poll in existence.  It was only once the voting started in the primaries that we had definitive proof that ARG was a joke, but we pretty much knew beforehand.  ARG's final polls before an election have always been bad, but their polls taken weeks/months before the election are typically much worse, and this is not reflected in anyone's "pollster ratings" because one typically only considers the accuracy of the final poll taken before the election.

Additionally, ARG's polling seemed to magically improve after Super Tuesday.  I doubt this is because they suddenly became much better pollsters.  More likely, they just started copying more credible polls.  But there's no way to prove this, and no way to score that in any kind of objective analysis of their accuracy.  So I think excluding them from a poll average for no other reason than that your gut tells you that they're faking it (and not because of any objective criteria) is a defensible position.  But yeah, once you do that, stick with the decision.  Don't renege on it for no apparent reason.

[Mr. Morden]

no defense of ARG or Zogby is legitimate.

Wait, I'm confused.  Was that a response to my post?  Because I wasn't defending ARG.  I was doing the opposite.  I was explaining why it's legitimate to exclude ARG while keeping other pollsters that rank worse in "pollster accuracy", because all of the "pollster accuracy" scorecards miss certain clues that tell you that ARG is fraudulent.

[Mr. Morden]

no defense of ARG or Zogby is legitimate.

Wait, I'm confused.  Was that a response to my post?  Because I wasn't defending ARG.  I was doing the opposite.  I was explaining why it's legitimate to exclude ARG while keeping other pollsters that rank worse in "pollster accuracy", because all of the "pollster accuracy" scorecards miss certain clues that tell you that ARG is fraudulent.

You were defending RCP's defense of ARG.

Not really, no.  I was defending the general concept of using somewhat subjective criteria to decide which pollsters to include in averages.  The specific example I was using to explain this was a defense of RCP's original decision to exclude ARG.  Nate Silver thinks ARG shouldn't be excluded because there are other pollsters who are "worse".  I questioned whether ARG's badness can really be captured in the criteria that Silver would use.  That's roughly the gist of what I said.

[Mr. Morden]

ARG's methodology is actually quite easy to understand.  They use this formula.