It would appear that Google has been showing Arizona for Democrats since several days ago, and continues to do so even now when at 90% reporting there's only a 1.2% lead between D and R -- meaning that if Republicans continue to get 60% of the remaining votes as has been the case for the late counting in Arizona, then they're clearly on a path to winning Arizona. In fact, this has been widely reported in the media, and even CNN has not called Arizona yet as of Friday, November 6, afternoon.
I did some quick math myself just to make sure the reported 60% figure does check out based on Google's data of 90% Reporting for Arizona and 1.561M votes for D and 1.520M votes for R, as of Friday, Nov 6 around 13:00 MST:
- (1.561+1.520) / 0.90 / 2 = 1.711M expected total votes each when we go from 90% reporting to 100% reporting;
- with D at 1.711 - 1.561 = 0.150 M more votes required for equilibrium at 100%;
- with R at 1.711 - 1.520 = 0.191 M more votes required for equilibrium at 100%;
- the equilibrium would then be reached if the remaining votes are 0.191 / (0.150 + 0.191) = 56.0% R to 46.0% D -- not even 60% as reported elsewhere in the media; plus there's the third-party vote, which means it can even be sufficient to get slightly less than the full 56% of the remaining 10% of ballots to get the election swung back to R (as has been the case for Arizona since 2000, or even since 1952 if we ignore the 1996 performance and the lack of ranked-choice voting).
Given that Alaska is still not called for Trump -- even though it's not considered at play, and it's been Republican since 1968, far longer than California has been continuously Democrat since only 1992 -- why did Google/AP not retract their clearly premature calling of Arizona yet? Do they not expect it to flip, even though many local sources in Arizona expect the flip to still occur, after all? Or do they not plan to retract the call until it actually happens, betting on the possibility that it might not happen?