Kevin, you did a pretty decent job of explaining. You even came to the correct conclusion:
"The answer to Nate Silver isn't to go back to traditional political punditry... but it's to develop ever more statistical models in the hopes of finding a better way... to forecast."
But you HAD TO throw this in:
"Circling back to Silver, if Romney pulls out a victory... it'll be a data point for Silver to add into his model to try to 'get it right' next time."
WRONG, WRONG, WRONG, WRONG, WRONG, WRONG! It will do NOTHING to change statistical modeling, it will simply show that a 22% probability (actually 27% at the time of my post) is INFINITELY greater than 0% probability.
IT'S A MODEL!