It’s by just a single point, well within the margin of error, but three weeks ago Monmouth had Hillary waltzing away with Florida by nine in the four-way race. A Mason-Dixon poll taken a week after that put her up by two, and now PPP sees Trump nudging ahead. Needless to say, Florida is an all but must-win for Trump in any path to 270, especially if Pennsylvania is going to be a tougher climb for him than everyone thought.

His strategy in Florida is the traditional Republican strategy except even more so: Pile up huge amounts of white votes and then pick off just enough support from Cuban-Americans to eke out a win statewide. It’s working right now:

Much has been made about how Trump will struggle in Florida because of his weakness with minorities and that’s certainly the case. Among non-white voters he trails Clinton by 49 points, getting just 22% to her 71%. But white voters still make up the majority in the state, and with them Trump has a 27 point advantage with them at 60/33. If he’s able to win the white vote by that much, Clinton doesn’t have much chance of running away with the race in Florida. The state will be the same toss up it’s accustomed to being.

Florida might go either way this year but there are signs within the poll that are good news for Democrats’ long term prospects in the state. Among voters under 45 Clinton leads Trump 57/31, and when you extend that to voters under 65 Clinton keeps a 50/41 advantage. It’s only Trump’s 59/39 lead with seniors that keeps things in toss up territory overall, but those voters aren’t going to do Republicans much good 20 or 30 years down the line.

A 60/33 lead among whites is slightly better than Romney’s pace according to the 2012 exit polls, which had him winning that group 61/37. And even slight improvement is a big deal in Florida, where Obama won by just a single point. The newsiest bit from this poll, though, is Hillary’s poor favorable rating. Her numbers have been consistently terrible in state and national polls, but she was reliably a few points ahead of Trump in that metric during August. Not here. His numbers are ever so slightly better — and they’re conspicuously better when it comes to cross-party support.

Nineteen percent of Democrats view Trump favorably. Just four percent of Republicans give Hillary thumbs up. What does that mean in terms of presidential preference? In the four-way race, PPP sees Trump drawing 15 percent of the Democratic vote(!) versus five percent of Republicans crossing over for Hillary. Offhand I can’t remember any poll this year, national or state, showing as much support in one party for the other party’s candidate as Trump is pulling here. Even at the height of her post-convention bounce, I don’t recall Hillary pulling 15 percent of Republicans anywhere.

Another, and related, interesting result is that Trump wins 85 percent of Republicans while Hillary manages just 78 percent of Democrats (thanks, of course, mainly to that 15 percent who are backing Trump). Rarely if ever over the past month has a poll showed Trump stronger within his party than Hillary is within hers. That’s been key to the grumbling about #NeverTrumpers: Supposedly they’re the critical difference in the race because they’re denying Trump the Republican votes he needs to match Hillary’s Democratic support. In this poll, anti-Hillary Dems are a much bigger force than anti-Trump Republicans are, enough so to account for Trump’s small lead overall. By comparison, Obama won 90 percent of Democrats in 2012 per the exit polls while Romney won 92 percent of Republicans. If O had performed any worse, he might have lost the state.

Another way of looking at the favorable ratings is by racial demographics:

Hillary does better among Hispanics but not substantially, and without even cracking 50 percent. Meanwhile she’s getting destroyed among whites as Trump maintains a respectable 50/44 split. On the question of presidential preference, though, it’s among black voters that she’s underperforming most dramatically. She leads among Hispanics 58/33, which is in line with Obama’s 60/39 advantage in the 2012 exit polls. Obama, however, won blacks (who were 13 percent of the electorate) 95/4; Hillary’s winning them 79/8. Any sharp falloff within a group that big in a state this tight is a major problem for Democrats if it holds on Election Day.

As further reading on pollmania, I recommend this short but smart Harry Enten piece on why the race has tightened lately. It has to do in large part with Trump’s more disciplined campaigning, although not so much because the public is necessarily impressed with him as merely because a more “normal” race favors the GOP. Remember, all other things being equal, Republicans probably win this year. The public is restless now that one party has held the White House for eight years and the economy is meh; a generic Republican would have an advantage, especially against a candidate as weak as Hillary. Models have showed this. The problem with Trump is that he turned a campaign that was supposed to be a referendum on her into a referendum on him through his own discipline. He’s remedied that a bit by not doing anything stupid over the past few weeks, making some moderate noises on immigration and racial outreach, doing more teleprompter speeches, and so forth. That gives voters a chance to notice some of the news about the Clinton Foundation’s pay for play and the FBI releasing the notes from Clinton’s interview and to remember that, oh right, Hillary’s a terrible candidate. If Trump sounds semi-coherent at the debates and lays off any self-destructive wisecracks, that’ll normalize the race further in October.