"Completely Inaccurate!" is not an expression that will win a lot of trust from me. This piece is most interesting when it avoids such generalizations in favor of somewhat more nuanced analysis. But you can't win points for being subtle and then turn around and punchline me with: "Long run... don't bet on them." Either give me some stats on how much more or less often they've been accurate than polls in a particular "long run" or, better yet, tell me why the data is inconclusive on that and under what conditions you would consider it conclusive.