Florida’s early vote numbers are in:

One number looks good for Trump, and that’s the raw white vote of 900,000 people. But consider that if Floridian whites split 60-40 for Trump, this would only net him 180,000 votes. A 90-10 split for Clinton among blacks would net her 56,000, and an 80-20 split among Latinos would net Clinton 272,000.

President Obama won the state in 2012 by only 72,000 votes, so that 87% increase in the Latino vote has to be concerning for Trump.

According to Joshua Green’s reporting for Bloomberg, the Trump team has long identified the Miami “designated market area” as the most critical one for him in the nation, but nearly 30,000 county residents registered to vote in October, and “through Saturday…(of the) 707,844 county residents (who) had already voted: 44 percent were Democrats, 30 percent Republican, and 25 percent had “no party affiliation.” That last group, unaffiliated voters, skews younger and more racially diverse than the electorate as a whole. “The demographic mix of early voters also looks highly favorable to Clinton: 58 percent Hispanic, 17 percent African-American and 20 percent white.”

But the late registrants, Smith says, give the clearest indication that sentiment in Miami-Dade is running strongly against Trump. Of the 29,657 voters who registered last month, 41 percent are Democrats, 44 percent are unaffiliated, and only 12.5 percent are Republicans. “That’s nuts,” said Smith. “These are the barometers that indicate the hostility toward the GOP candidate.”

Campaign organizers like to bank as much of their vote before election day as possible. For one thing, fewer outstanding votes means fewer people you need to contact and drag to the polls. But you’d ideally like to get the unlikely voters to show up early because likely voters generally get themselves to the ballot booth sooner or later. If all you’re doing in the early voting period is “cannibalizing” you’re election day vote, that’s only modestly helpful. It avoids losing a few votes here and there to car problems and family emergencies, but it otherwise doesn’t much improve your candidates’s prospects. The thing is, a lot of the people who have voted in Miami-Dade County have either never voted before or have only voted sporadically.

“Of the 707,844 voters in Miami-Dade, 201,000 did not vote in 2012—and 127,000 of them are Hispanic,” he says. “Hispanic voters are over-performing their share of the electorate.”

Despite early concerns, it appears that black turnout in Florida will exceed the raw numbers from the last presidential election and, from the Clinton campaign’s point of view, getting 127,000 Latinos (in Miami-Dade alone) who didn’t vote in 2012 to the polls already is a big boost to morale.

But it gets better. In the Miami “designated market area,” which includes Broward and Monroe Counties, “351,000 (of the early voters) did not vote in 2012—and 47 percent are Hispanic.”

These are very big numbers for a state that was decided by only 72,000 votes four years ago.

In the Orange County belt around Orlando, (where there Latino community is more Puerto Rican and less Cuban) the numbers look similarly promising for the Clinton team.

Democratic optimism in Florida needs to be balanced against concerns about black turnout in North Carolina (it’s way down due in large part to Republican efforts to suppress it with “surgical precision“) and Cuyahoga County (Cleveland), Ohio. Early voting in Iowa also looks bad for Democrats as their advantage is far below where it should be. But Trump cannot become president without Florida so, on balance, the news out of the Sunshine State is looking potentially decisive.

As a word of caution, Nate Silver gives Clinton only a 52% chance of winning Florida in his Now-Cast, which assumes that the election will be held today instead of tomorrow. On the other hand, the folks at Huffington Post give Clinton a 89% chance of carrying the state.