If you looked at the polls-plus forecast, that reduced Clinton’s odds even more.
Silver’s model was more or less given that certain assumptions are true. If they’re not, well, that will most assuredly affect the accuracy of the model. It’s like estimating the probability that at least one coin flip in three will turn up tails, but not knowing that 1) two of the flips have already happened, and 2) they were both heads.
