Forget the ‘have-a-beer-with’ question. Think nuclear codes.

The quote is of ambiguous provenance, but brilliant nonetheless: “If you torture the numbers long enough, they’ll confess to anything.”

Too often, we believe something not because the evidence supports it; too often we pay attention only to evidence that allows us to believe (or continue believing) what we prefer to believe.

We’ve seen this a hundred times during the campaign. After 10 polls in a row from six different pollsters show Donald Trump trailing Hillary Clinton, one comes out showing him tied or ahead. Having never mentioned any of the previous polls, Trump announces the “good” one to one of his rallies and the place goes wild. Maybe it’s natural or normal, but it’s a case of this selective perception/confirmation bias in action.

I haven’t written much about polls this year. That’s on purpose. I decided it wasn’t my function to say who was winning, nor who will win, in part because I don’t know, but also because it’s the wrong question. At its core, the job of the voter is not figure out who will be president, but to decide who should be president, based on which candidate you believe will do the better job, and then to vote for that candidate. And the job of a public scribbler like me is to pass along information and analysis that might reasonably make that decision a slightly more thoughtful or informed one.

I know this terribly schoolmarm-ish of me. Of course, the “horse-race” poll is only a small subset of things polls tell you that won’t really help you make up your minds how to spend your vote. There’s a jillion (or do I mean two jillion?) poll questions, which aren’t exactly horse-race questions but are used in stories to nonetheless advance the horse-race analysis. Whom are voters more likely to vote for, the one they trust more on the economy or the one they trust more on foreign policy.

Well, at least those are valuable considerations in deciding whom you should trust with the job. But my absolute least favorite are polls that delve into the “likability” stuff. Can I just say that a) I feel I have little valid basis for liking or disliking far-away strangers whom I know mostly from watching them on TV, striving (and often failing) to project likability; and b) my hunch that I would like one of them more as a friend carries very little overlap with which one would be a better chief executive and commander in chief.

Likability (and a few other things) got us George W. Bush instead of Al Gore in 2000. When I wrote about famous debate “gaffes” recently, I didn’t include (but could have) Gore sighing audibly and rolling his eyes obnoxiously at the falsehoods Bush was spewing to justify some of his policy positions. In almost all cases, Gore was correct (but obnoxious) in attempting to signal by sighs that Bush’s facts were off. But who would you rather have running the Executive Branch, a smarty pants, or a regular guy with whom you would much rather have a beer?

Oy, the have-a-beer-with question. Drives me nuts. It strikes me as almost encouraging voters to attach more weight to the least relevant factors.

Actually, I don’t remember the beer question being used before the Bush-Gore race. Bush, as it happens, was a recovering alcoholic, so good luck having a beer with him. But even if you could, how do those beerable qualities translate into the guy you would also rather entrust with the nuclear codes? But, as best we could understand, preference for Bush as a beer buddy had much to do with why he won (if we won).

Trump, by the way, apparently out-beer-withs Clinton (by 45-37, with 17 percent unsure or, I hope, refusing to answer the stupid beer question.) And Trump, like Bush, claims to be a nondrinker. But the poll that included that question, perhaps wondering whether the beer question had a tinge of gender bias in it, added a newer question about which candidate you would rather have to your home for dinner. On that one, Trump bested Clinton by a statistically insignificant single percentage point. So there’s that.

Now then, for those of us who know we will, in all likelihood, neither have a beer with, nor bring home for dinner, either of the candidates, how about that nuclear codes question?