That's a preface to the broader point. There's always a chunk of the electorate that never comes around to making up its mind on the candidates in an election. There are a lot of reasons for expressed uncertainty in polling, including genuine indecision (the vast majority) or an unwillingness to talk to pollsters (some). But you rarely close out a poll at the end of the campaign and find that everyone has a stated opinion. What you do find is that the number of people who haven't made up their minds tends to be lower the closer you are to Election Day, as people learn more about candidates and decide.
This year, though, there's an interesting phenomenon. Hillary Clinton and Donald Trump — the two most unpopular major-party candidates in history, as you've heard over and over again — are getting lower support at this point in the cycle than President Obama or Mitt Romney did four years ago. That's cumulative; the number of people saying that they plan to vote for one of the two major-party candidates is lower now than it was in 2012 in all of the swing states (according to the RealClearPolitics averages of polls).
If that's a function of people not having made up their minds, it could suggest a big wave breaking for one candidate or the other in the past few days, bigger than we saw in 2012. (We've already seen a link between fewer undecided voters and increased Trump support.)
But that's not really what's happening.
The answer, quite simply, is the third-party candidates. Gary Johnson's down a lot in polling, but he still draws a decent amount of support. Jill Stein, too, draws more than zero.
If you look at the number of voters who weren't planning to support the Democratic or Republican candidates for president in 2012 and 2016, over the past two months, the latter figure has been higher each day. But separate out those Johnson and Stein voters, and the number of undecided voters is about the same this year as four years ago — and often less.
As of the most recent RCP averages in each state, here's where the number of voters who either don't plan to support the Democrat or Republican or who aren't picking any of the four candidates stands vs. 2012.
In Colorado, there are more people who are uncommitted to any candidate at this point this year than were uncommitted to Romney or Obama at this point four years ago. In New Hampshire, at the other end of the spectrum, minds are more made up than they were then.
There are a lot of ways in which these numbers are squishy, including that poll averages are contingent on recent polling. It's also the case that we've seen some signs that softening support for Johnson has been a boon for Trump. But the polls suggest that there is not, at this point, a giant pool of voters waiting to go one way or the other. At least, no more than there was in 2012.
Assuming, of course, that all the poll respondents aren't lying?