who is the guy who posts the election prediction statistics?

I would like to hear his take on this "11 point lead" that TIME has published.

He has probabilities of Kerry winning in the 90% and up.

okay so how does he do all these statistics? I was a psych major years ago, and had to do some real in depth stats... what is this guys background, and where is he getting his polling from?

other than anomolous, I guarantee you TIA will let us know.

He had been predicting over 90% probability of Kerry victory and when I followed his link up stream he was some kind of physics professor or something from (I believe) Princeton. He had no access to the link through his professorial home page. OH that guy made me feel good. Now I can't find him again. I think I originally got the link off of Buzzflash

But in the past TIA has cited and used the most favorable polls to Democrats, often a single poll and completely misleading, in an attempt to boost Diebold theory, that the polls were right but we were robbed. He has never studied polls/results to determine which polls are most accurate and whether the absurdly low MOE estimates hold up.

This year he has compiled a block of polls, a huge sample and far superior method than previous. His percentages are accurate if the numbers are correct, AND the undecideds break toward the challenger in decisive number. I am skeptical about that given the extremely low number of undecideds and the illogic that any diverse group will tilt 60-70% in one direction in an otherwise 50/50 nation and election. Bush, whether we accept it or not, is viewed as the more regular and likable candidate, and that plays on the undecideds during debates, based on debate-watching gatherings at my house.

Also, tonight TIA announced he was dumping the TIME poll due to bias. That is an extreme overreaction, like the poll itself, and not something a numbers based analyst should do, if he is looking for accuracy and not one predetermined winner.

Another flaw is the polls TIA uses are not exactly synchronized, by definition. State polls, which are less frequent, that precede a national poll by days or weeks will invariably be misleading if the national numbers have changed. A 2 to 3 point national shift will automatically drag state preference in the same direction, across the country and almost without exception.