This article has been updated.

Just when you thought that the seemingly endless Senate race in Alabama was over, the candidate who was long expected to win it has announced that it isn’t.

After Republican Roy Moore’s campaign chairman took to the lectern to assure the candidate’s supporters that declarations of victory for Democrat Doug Jones were premature, Moore himself stepped up to do the same.

“When the vote is this close . . . it’s not over,” Moore said. Why? Well, if a race is within half a percentage point after all the votes are tallied, an automatic recount is triggered that could conceivably flip the result. And with a narrow Jones lead and military ballots still needing to be counted, Moore assured the crowd that some miracle still might happen.

Alabama Secretary of State John Merrill was less sanguine when talking to CNN’s Jake Tapper a short while later. Merill walked through the various reasons that a recount was unlikely and, even if it happened, that the result was unlikely to change. A walk-through that we will now re-create.

Let’s start with the numbers, as of midnight on Dec. 13. Jones has 671,151 votes and Moore 650,436. That’s 49.9 percent of the vote for Jones and 48.4 percent for Moore — because 22,819 voters wrote in someone else.

Quick math shows us that 49.9 percent minus 48.4 percent is 1.5 percentage points, three times the margin that would be required to trigger an automatic recount. The vote was close, as Moore said, but when the vote is that close, 1.5 percentage points close, the state actually thinks it is over. Merrill noted that either candidate could call for a recount regardless — but that candidate would then have to foot the bill.

So let’s assume that Moore doesn’t want to incur that cost. Is it feasible that somehow Moore could pull within 0.5 percent of Jones?

Moore’s camp mentioned those military ballots. To Tapper, Merrill said it would be “highly unusual and highly unlikely for that number of ballots” — the 20,715 margin between Jones and Moore — “to be outstanding” from service members. It might, however, be enough to draw Moore within 0.5 points, he said.

We can be confident, though, that it wouldn’t. As of May 2016, there were only about 8,700 people from Alabama serving in the armed forces. Assuming that’s held fairly steady, even 100 percent turnout from that group would account for only about 42 percent of the gap that Moore needs. Even if all of those servicemembers voted and all of them voted for Moore, Jones would still have a lead of 0.9 percentage points. Not enough.

Tapper then asked about the write-in votes. Might those somehow affect the results?

“They might, Jake,” Merrill replied, “but they might not even be impactful.”

That seems likely. If all of those ballots were thrown out, the result would be a bigger lead for Jones, not a smaller one. If you have five apples, one friend has four and a third has one, you have 50 percent of the apples. If the third friend eats his, you’ve got five of the nine remaining apples — or 56 percent of those remaining. Perhaps some of those write-in votes are for Moore, in which case they’d presumably be added to his total. But that seems unlikely.

What if Moore just paid for a recall out of his pocket? Could that change the result? Merrill was skeptical.

“I would find that highly unlikely to occur, Jake,” he said when asked. He mentioned a recent race, when a gubernatorial candidate won by 160 votes to make a runoff. “When they had a recount that was paid for by the third-place finisher,” he said, “the actual total didn’t change more than three or four votes when all of the ballots were run back through the machine after that was requested and paid for by the unsuccessful candidate.” That’s because of the system the state uses, he explained, which, being mostly electronic, meant that votes were counted the same way the first time as they would be the second.

Update: What’s more, election law expert Rick Hasen argues that Moore can’t pay for a recount anyway since he sought a federal office.

In situations like this, when a candidate is favored to win after months of campaigning but comes up short, it’s hard to tell if a refusal to concede is motivated by sincere concern over the result or simply a dogged refusal to accept reality. In 2014, Chris McDaniel narrowly lost a Senate election to Sen. Thad Cochran (R-Miss.) and, according to the Clarion-Ledger’s Sam Hall, simply never conceded. Perhaps that’s Moore’s eventual fate: Always being the guy who almost won but never admitted it.

The evidence at hand makes one thing seem pretty clear, though: He’s not likely to suddenly become the guy who actually did win.