Originally Posted by JeffDM
Originally Posted by trumptman
If you have a bunch of polls, from various methodologies and with their own "house effects" even per Mr. Silver and you thrown them together using a methodology you used to keep open but now do not reveal at all or until after the fact then what you have is precisely some set of random, unrelated facts strung together using a good "feeling" he has.
He's called the 2008 and 2010 elections pretty well, if it's just based on his feelings, how did that happen? If his biases are toward donkey and against elephant, wouldn't his figures have been way off for the 2010 turnover?
I'll be happy to read any link you have that compares his performance to others. From what I've seen, he is no different than most other folks who use numbers and rolling averages. If you just let them roll day after day until you get to the day of the event, then of course you'll hit 100% probability but until then it is about keeping things vague and keeping the door part way open.
Even now he doesn't say Obama WILL win. That would be a prediction. He says his closed model shows an 80% probability. Tomorrow it could show 81% or 79% and continue to shift until the election. It's no different than a stock price or InTrade. It reflects the consensus, slowly shifting one way or another, day by day. He calls what he does forecasting, not predicting. If you look at the weather starting from two weeks out, and adjust your numbers every day until the actual day the weather is in question, you will look pretty decent at what you are doing but really how can someone be wrong when you are allowed to adjust statements daily or hourly?
In that regard Silver gets the best of both worlds. If it were the weather, he could hit all the various weather services who have used the data to forecast the weather and he can adjust his statements hourly up until the actual weather event. Then he can take their weather forecasts from two weeks prior and declare how good or horrible they are based on being two weeks out while he does no actual weather forecasting himself.
Now that said, if SDW or myself or anyone else state that no president has been re-elected with X numbers in Y variable within the the last 40-50 years, that is a quantitative number. He isn't say Obama's going to lose because he sucks or because his friends who he shares a beer with hate him or because his knee is aching and that means rains a'coming.
If you, or I or SDW continue to toss out some numbers and are allowed to adjust our conclusions based on them up until the election, we will probably be 100% right on the day of the election. Silver's "forecasts" are about 98-99% right but they aren't making a prediction several months out. Anything predictive is given a large leeway. (I predict the Republicans will gain between 20-50 seats in the House as an example.)