Given past history, it is always in the back of my mind that there could be some kind of Republican hanky-panky with the election results. My question is, could the forecasting models I live and die by these days (538, Princeton, etc) be used to identify possible voter fraud? Remember that Nate's daily number of EVs (276, 299, etc.) is actually an averaged result of thousand of discrete simulation runs. These runs in turn are based upon certain relationships between the voting trends of different states (I think). Something like if Obama is up by 3 points in Colorado, it is not likely that he would down by 14 points in New Mexico.
The point being, the election outcome on November 6 will actually be one of Nate's runs. Somebody should be able to go back into the model, identify the run and determine what the probability of that outcome is of occurring. If it's down in the general mush where the majority of the outcomes occur so be up. But suppose it is a low probability anomalous event (0.01% chance or whatever your trigger point is). Can that alone be used to justify an investigation of potential irregularities? Any thoughts on this out there?