in from outside gallup, the university of michigan, texas, maryland. we ve really taken the hood off to look at what we did last fall to try to answer your question. four things, likely voter estimation. we had obama up by three points among registered voters, which is about where he was at the end. won by about 3.85%. moved it down to a one-point romney lead, and that was more than the average for all other models of a similar fashion out there and we think we overcorrected. so the model itself is something that we really think needs reevaluating. this is an historic way we ve calculated for decades, but we re going to use new jersey and we re going to use virginia and those races working with the university of michigan to really try to figure out how in today s changing environment you do the best job of estimating. so, frank, let me ask you this, what happened, there were times you guys put out a poll and you d have romney up by 9, 10, or 11. and you talked to the obama people and