“How was I able to predict the outcome of the TCJ polls 2 polls in a row? They aren’t taking polls at all, maybe, just feeding us numbers that we like?”
How do we know PPP or Hays research or FDU aren’t doing it? All you have to do is use a turnout model that achieves your desired result and just plug the numbers in. Monmouth, FDU and even Rasmussen are way off because the turnout model is deeply flawed.
Right. But those people are actually taking polls. We can look closely at what real pollsters are doing, and pick apart their work, and get our own answers. FDUs most recent poll was crazy. I looked at it, plugged in my own turnout model, and got Christine down by 12. You can do that when there’s actual data to work with.
I started with an assumption, which I think is correct, that unknown bloggers do not do 5 huge polls twice a week, and that they’re just giving us information that we will like and we will tell other people about. They want to give us the “best” polls. They took the results from 5 races that we cared about and gave the Republican 3 points. TCP started us with -8. I then noted that, and was curious about what they’d do next. Ah, add one to Christine. Perfect answer for us. We’re gaining a little. We’re happy, he’s getting hits. I presume that he’ll continue doing the exact same thing. And he did. I predicted he’d say 6 the next “poll” and he did, and then 5, and then he did. I don’t believe for a second that any poll was done.
Now, don’t get me wrong, his guess will be closer than FDUs most recent guess. But I don’t believe that any poll was taken by TCJ.