In a lot of states in the US, that type of approach wouldn't work well given the frequency of absentee/mail voting (wherein such ballots may arrive days after Election Day and/or may be counted separately and display different patterns than the in-person Election Day vote), provisional ballots, in-person early voting, and other non-traditional mechanisms which sometimes comprise the majority of ballots cast in a state.
I know, and I understand the difficulties involved by all this possibilities. Especially in States like Washington for example, but in Austria they have also postal absentee ballot, and they manage to have a good estimate of the final result with 700,000 ballots by mail counted the day after.
We actually do have something like that. Each network sends three reps from their Decision Decks to a giant meeting of the analysts with the exit pollsters and all the data during the day on Election Day. They then break at 5 Eastern to go back to their networks. I heard that one of the reps from NBC was expecting a Clinton win slightly larger than Obama's in 2012 when he left the meeting.
Yes, the exit poll, and I understand that an exit poll (especially an early exit poll) can be wrong. But once you have some real votes counted, you should have a quick idea of the final result (compare the swing, watching what's in, what's left, ...)
The NY Times did have a live probabilities thing this year that was slightly slow to move because it started at a 84% chance for Hillary, but it showed a certain Trump win hours before the networks called it.
