Each time we revisit this number, it's amazing. Fifteen days after the end of the 2016 election, and California still has 1.8 million ballots to count, more than were cast in 23 other states and D.C.

In other words, the final results of the election are still in flux, even if the final winner isn't. Hillary Clinton now has a national lead of more than 2 million votes, thanks to the fact that these ballots are creeping in from America's largest state — a state that preferred her over Donald Trump by a wide margin.

In fact, as it stands, California's margin of support for Clinton is the second-highest percentage that a Democrat has received in the state since the emergence of the Republican Party. The only Democrat to get a higher percentage of the vote in the state was FDR in 1936.

So that's a large margin of victory for Clinton in a large state with a large number of ballots left to count (including 750,000-plus in Los Angeles alone).

California's results on election night came in after it was pretty clear what the eventual outcome was likely to be. With Donald Trump's wins in Florida and the Midwest — and given that he wasn't going to win the West Coast — it was obvious that no matter how the vote in California ended up, it wouldn't really matter. Trump would be the next president.

What that meant, though, is that the picture of the election we got wasn't clear. Certainly polling suggested that Clinton would win Wisconsin and Pennsylvania, which she didn't: The polls were wrong. National polls, though, suggested that Clinton would win more votes nationally — and she did.

The final tallies from RealClearPolitics' polling average gave Clinton a projected 3.3-point win. As of now, her national vote win will be about 1.5 points, a miss of 1.8 points. Wrong — but not egregiously so.

Why was it wrong? Two reasons. First, Donald Trump outperformed his final polling average by more than Clinton outperformed hers. Second, Gary Johnson and Jill Stein got less support than polls suggested — a trend that we noted the weekend prior.

If we compare that result to the past three cycles, it's not too far out of line. In 2004, the polling average was off by 0.9 points from the final result. In 2008, it was off by 0.3.

But in 2012, in a cycle where Mitt Romney's support was broadly overestimated as pollsters misjudged who was likely to come out and vote, the polling average was off by 3.1 points.

That's clearly worse than this year, but it mattered less because the polls still anticipated that President Obama would win reelection. Missing the percentage but nailing the result is more acceptable to the public than missing the result but getting the percentage right — especially when “the result” is not the same as “won the most votes.”

Note, too, that these numbers are going to change more. Remember those 1.8 million ballots in California! If all of those ballots are determined to be valid (and they probably won't be) and if the percentages that the two main candidates are currently earning in each county hold, Clinton will gain another 600,000 votes in her advantage over Trump, making the overall national gap between them wider and the national polling averages closer to the end result.

Polling missed the mark in 2016, as we've noted. But national polls predicted Clinton would earn more votes by a comfortable margin, and she did. And if California ever figures out how to count, we'll soon know by exactly how much.