Third-party candidates usually start off as media darlings, and end up as afterthoughts. Gary Johnson was supposed to be different, though, and not just because of his status as a two-term Republican governor. With both major parties nominating unlikeable candidates, Johnson and the Libertarians had an opportunity to argue for a way out of the Republican-Democratic duopoly.

According to the Washington Post’s analysis of the latest CNN/ORC polling, the opportunity has passed:

One big change is that Gary Johnson, the Libertarian candidate, saw his support drop by half. He can console himself with the fact that the change was within the margin of error, but that change is both explained within the poll — and also predictable.

Polling that includes Johnson has shown him drifting lower in support over the past month or so.

This comports with what often happens to third-party candidates as the election nears. Polling in the mid- to late summer often over-represents how much support third-party candidates will get. Once it comes time to actually vote, a lot of that support moves back to the major-party candidates.

In the case of Johnson, the drift appears to have been a function of men walking away from his candidacy. Clinton and Trump each gained 2 percentage points between the two CNN-ORC polls. Johnson lost 4 percent. Johnson lost 7 percentage points from men, but none from women. Those men split between Clinton and Trump roughly evenly.

Give Johnson this much credit: usually it takes a shorter period of time for voters to abandon independents and return home to their parties. This dynamic comes from a predictable cost-benefit analysis that understands human nature. People want their votes to count, and want to have at least a chance of contributing to a win. Only two candidates have that potential in 49 states, and neither of them are named Johnson. Voters are much more likely to engage in protest support than in protest voting, which is why most independent candidates draw so few votes.

One possible exception to this might be Utah, but that doesn’t involve Johnson either. Evan McMullin’s support has put him in competitive position against both Hillary Clinton and Donald Trump, two candidates who have gained almost no traction at all in that state. McMullin’s tied for second in the RCP aggregation, and has a four-point lead in Emerson’s latest report. As long as it looks like McMullin has a realistic chance to win, his support will stick with him — and those voters will get the benefit of both a protest action and an opportunity to be part of a win.

Otherwise, the CNN/ORC poll doesn’t show much change in the race since the first debate. Hillary still leads by five points, this time 49/44. In a two-way choice, it’s still 51/45, the same as it was three weeks ago. Her positive support has gained six points to 66%, while Trump’s has only gained one to 60%. The overall gender gap gives Hillary a ten-point edge, though, and she’s still leading among white college graduates by double digits — a demo that Republicans have controlled for a very long time. Despite Trump’s attempts to appeal to working-class Americans, Hillary has a 14-point lead among those who earn less than $50,000 a year.

The five-point lead is almost identical to the current RCP averages for two-way and four-way polling, so this is at least the most representative of current polling. It may not be as dramatic as ABC’s 12-point differential nor have anywhere near that series’ dynamism, but overall it indicates that Trump’s behind, and he didn’t do anything over the last three weeks to catch up. Not even Johnson’s very predictable fade benefited him, and with two weeks to go, Trump’s running out of both time and opportunities.