You can't do it, my friends

[mmcirvin]

I wrote a whole draft of a somewhat long post expressing my irritation at the recently popular parlor game of attempting to extract daily presidential polling numbers from the Gallup and Rasmussen three-day moving averages (Brad DeLong explained why you can't do this a while back).

However, that post is obsolete, since Nate Silver himself, who I think started the fad, just wrote a long and lay-readable post about why you shouldn't believe his numbers. He throws his usual giant wad of Monte Carlo simulation at the problem, and by his own admission he still doesn't get very credible results.