What this is. Unit-level returns from two sources: the Associated Press by way of the New York Times feed, and DecisionDesk HQ's public embed API. Every unit carries a tag showing which source its numbers come from. Where both carry a unit, the board uses whichever has counted more votes, on the reasoning that both are counting the same ballots and the higher count is further along. Units tagged both are ones where the two sources agree exactly.
Not an official count. The state certifies results. This board is assembled from two private providers whose snapshots were taken minutes apart, and they disagree in places. Do not cite the total as a result.
Est. in is votes counted against the provider's expected turnout for that unit, so it is an estimate and not a precinct count. Units under 90% are flagged; those still have vote to give. A unit can read 100% and still tick up if the turnout model ran low.
Outstanding vote counts the shortfall in partly-counted units as well as units with nothing in. The two providers often disagree sharply on it, and that disagreement is shown rather than averaged away.