He called them correctly. HOWEVER, his probability for an 0bama win was >90%. There is NO WAY to prove that it wasn’t, say, 70%.
However, his model has been fairly accurate (more accurate than most) at calling many races across the country.
And no way to prove that he wasn’t correct.