Yeah, but those three are what people remember, maybe especially so in Minnesota. Nevertheless, as Florida descends yet again into Recount Madness in three statewide races, the Miami Herald tries to set expectations with this analysis today of recounts and their likelihood of changing Election Night outcomes. Forget Al Franken, if you can — if recounts were baseball hitters, they’d be outmatched by Mario Mendoza:

But a recount that reverses an initial margin of more than a few hundred votes would be unprecedented in the recent history of American elections. According to an analysis by the nonpartisan group FairVote, which advocates for electoral reforms that make it easier to vote, out of 4,687 statewide elections between 2000 and 2016, just 26 went to a recount. Of those 26, just three recounts wound up changing the initial result of the race: The 2004 Washington governor’s race, the 2006 Vermont state auditor’s race and the 2008 Minnesota U.S. Senate race. The average swing in those three elections after the recounts? About 311 votes.

In Minnesota, the swing was nearly double that at more than 500 votes. There were a number of reasons why that recount succeeded. For one thing, the scale was so small that normally meaningless computation and ballot errors became much more significant. For another, as I reported in 2009 and re-ran here at Hot Air two years ago, Republicans got caught flat-footed by the ferocious but legal strategy employed by Franken’s legal team. I linked it earlier but I’ll excerpt a bit from it here:

The Franken team had a clear mission. Since they were behind in votes, they needed to either find new Franken votes or discredit Coleman votes in order to close the gap. Learning from the Gregoire campaign, they trained their volunteers to understand the limits of Minnesota law and to aggressively challenge ballots.

Coleman’s team, on the other hand, had a much more delicate mission, as they explained to their volunteers. Each of the precinct workers interviewed for this story had the same description of the instructions given by the campaign: Do not get overly aggressive in challenging ballots. They did not want to be seen as the campaign that “disenfranchised” Minnesota voters, as successful ballot challenges do by removing ballots from the count.

One Coleman volunteer explained the instruction as an explicit message from Team Coleman that “we don’t expect to be in the business of suppressing Franken votes, and we’re not trying to fi nd new Coleman votes. … Don’t go out of your way to make what we think will be frivolous challenges.”

This instruction came specifically about overvotes—instances where a voter filled in two or more bubbles on the same race, which would have led the counting machines to reject the ballot for that race. Franken’s team latched onto the overvotes and tried to argue on as many as possible that the intent of the voter was to support Franken. Coleman’s team knew from the beginning that Franken’s volunteers would use this strategy, another Coleman recount worker said, but believed that “we should not engage them like that.”

But according to Minnesota law, as the state discovered during the process, the question of voter intent on overvotes is a legitimate area of challenge in a recount. In this case, it appears that both campaigns understood the parameters of action, but Coleman’s team simply didn’t want to avail itself of the entire range of action allowed by Minnesota statutes. They trained their recount volunteers to engage only on the most obvious cases and to refrain especially from giving the appearance that the Republicans wanted to invalidate ballots on a massive scale. Predictably, this led to many missed opportunities for Coleman challenges.

The new Florida recounts differ greatly from Minnesota’s experience as one of only three statewide recounts in the past two decades to change an election outcome. For one, Republicans have mounted an aggressive legal strategy from the get-go; Rick Scott has not only called the cops, so to speak, he’s filed a flurry of lawsuits to keep Broward and Palm Beach counties from playing with the ballots. Democrats won’t get a head start on ballot issues the way they did in both Minnesota and Washington.

More to the point, though, the gaps are far wider here than in other states. Leads of 12,500 (Senate race) and 32,600 (gubernatorial race) are too large by an order of magnitude to be the result of simple computational error. That kind of error would have to be a systemic defect in the state’s election system rather than ballot counting errors. And thus far, no one’s found any.

Not that they haven’t looked, of course. Bill Nelson’s legal team (which includes Minnesota recount veteran Marc Elias) claims they’ve found one in the ballot design, where the Senate race was located below the voter instructions. However, there are three big problems with that argument. First, more than 95% of voters managed to find the contest on the ballot anyway. Second, the elections commission that created the Broward County ballot is controlled by Democrats, and was approved by Democratic elections supervisor Brenda Snipes.

Third — even if one calculates the projected undervote, Nelson still comes up thousands of votes short:

But there are a number of factors working against Elias’ theory. For one thing, even if every one of those 25,000 undervotes are discovered in a recount and given to Nelson at the 69 percent rate he won votes in Broward, the senator would still be about 4,800 votes shy of Scott. For another, voting machines simply don’t make mistakes of this scale very often, experts say.

“I don’t know why the tabulators would miss that race and pick up all the others,” Barry Burden, a professor of political science at the University of Wisconsin at Madison said.

They wouldn’t, and they didn’t. Bear in mind that 25,000 comes to about 3.5% of all ballots cast in Broward for governor, which means 96.5% of voters managed to find the Senate contest on the ballot. Coulda woulda shoulda might make for an interesting philosophical exercise and a basis for determining Snipes’ competence, but a legal argument for challenging election results it ain’t.

There’s a reason why recounts have a .113 batting average in statewide races, and why the average swing is far under a thousand votes. Ballot counting technology is incredibly accurate, even the old punch-card system used by Florida until after the 2000 election debacle. Optical-scan technology is even more accurate and reliable. There is no precedent at all for reversing these kinds of margins through recounts or even challenges … that won’t start this year, either.