Nate Silver of 538 / NY Times fame has been one of those number crunchers who has enjoyed a long run of success, accurately predicting a number of high-profile races by averaging polling data and providing readers with outcome probabilities that are generally spot on (or at least he hedges really well). For the last week, Silver has caused Florida Democrats to fret because he was projecting a Trump win in the Sunshine State. Huffington Post even accused him of putting his thumb on the scale in favor of Trump.

But he just moved Florida back into the column for Hillary Clinton.

Over that same period, however, Democrat strategist and communications consultant Kevin Cate has held fast, buoyed by his own data predicting a Hillary Clinton “blowout” (by Florida standards) in the Sunshine State. Indeed, his website, FloridaTurnout.com, supports the claim that Florida “leans Democrat,” and has been leaning that direction for more than a month. Cate is officially on the record saying that Clinton currently holds an early voting lead of almost 90,000 votes. And before Republicans can even point to the fact that they won election day by 1.1 percentage points in 2012, Cate is quick to throw cold water on the notion: Republicans have already “cannibalized” their 2016 Election Day turnout in order to keep pace with Democrats in early voting.

Data show that Republican early vote totals are up 1.4% over 2012, while the Democrat share is down 5%. But when Cate crunches the numbers against the daily polling averages, Clinton comes out far ahead.

“Our model has Clinton up by 93,875,” Cate announced on Twitter after early voting ended Sunday night. “Trump now needs a 4.1% polling shift of remaining votes.”

To hit those kinds of numbers, Donald Trump would have to defy all of the existing polling with a performance well outside the typical margin of error in virtually every poll taken so far.  Cate’s website, which he jointly developed with Proxeme.com’s Travis Brooks, isn’t a true analytics or polling site. The model they use doesn’t delve into demographics or complex algorthims. They simply take the RealClearPolitics.com polling average for Florida and apply it to the early voting numbers as each indvidual ballot rolls in by county.

Could the projections be wrong?

“No chance,” says Cate, with a bit more confidence than anyone else is willing to project, at least on the record. He points to the data coming out of Duval County, a traditional Republican stronghold, that shows Democrats winning there by 2,000 votes. “Trump needed to dominate Duval early and by mail. He’s not. He may win the county, but not by enough.”

What about white Democrats in the Florida Panhandle? Might they shake things up by voting for Trump?

Cate says the numbers just aren’t there to support that hypothetical scenario.

And there is no evidence in polling to support it, either. Republican strategists seem at a loss when asked to point out potential bright spots that might portend a Trump upset in Florida. Trump’s campaign is pointing to the fact that Republicans have already cast 140,000 more votes at this point in the election than they cast for Mitt Romney in the 2012 cycle. As previously noted, however, much of those totals will cost Republicans on Election Day, because fewer of them will have a vote left to cast. Still, they are encouraged by the early voting performance against Democrats this cycle. The race will ultimately boil down to whose voters show up on tomorrow.

So what would it take for a Trump upset? For starters, five of the last seven Florida polls would have to be wrong. But, maybe the gang at Trafalgar Group have it all figured out. Their latest polling shows Trump up +4 in Florida, and taking a decisive share of key demographics for Clinton in other states, too.

One thing is certain: at some point on Tuesday night, we’ll finally know who’s right and who is full of hot air. That is the beauty of Election Day in America – it’s a crystallized moment of truth in a world filled with guesses, propaganda and outright lies. And it can’t get here fast enough.