Its not that he’s bad, its that he’s biased....
Look at the trending lines in his prediction model... Notice something that happened on 9/1? Suddenly the average angle of the line of Trumps rise lowered... and has remained lower, even though Hillary clearly had the worst few days of the campaign from 9-9 to 9-11.. and still the average linear progression of Trumps growth has remained at the slower rate...
Even while his model has given Trump NC, NV, OH, FL and has shown that Trump is a mere small state away from 270, this has all happened in the last week or two, and he still has the trending line way down from where it was pre 9/1...
So he’s biased, but not a complete fool. Based on his map Trump is basically NH away from securing 270... he clearly has momentum flipping several states, and growing, and while he has upped Trumps odds to 43% vs high 30s a few weeks ago, its laughable that if Trump only nees 3 more EC votes, and yes, if you play his map out, that’s where it is... with the momentum he is showing, that Trump only has a less than 44% chance of the win.
He’s manipulating, not as bad as some of the other guys, but clearly he’s adjusted his play to make Hillary’s collapse not look as bad as its been...
In my model, I had Trump at 40% chance, now down to 38% based on this week's polls so far.
The big difference for me is that I do not sauce the polls, so if pollsters play with the outcomes, that will carry over into my model. Hopefully, poll averaging will smooth some of that out.
-PJ