An operating theory of political primaries goes something like this: There are groups of voters, bound by demographic or ideology, who tend to agree on their choice of candidate in a crowded field. Candidates, then, try to secure support from these groups to use as a base, then try to peel off voters elsewhere to add up to the required amount of support. Politics is like a computer: Our eyes are glued to the ridiculous nonsense that shows up on our screen, but what's really going on is just addition.

So far in 2016, though, that pick-a-lane strategy doesn't really seem to be working. Donald Trump consistently leads with evangelical voters, despite not exhibiting a strong history of religious sentiment. (That's what's known in the writing business as an "understatement.")

You can see some examples of how candidates benefit from support from a particular demographic group. Take Mike Huckabee. The graph below shows how much support Huckabee has gotten in Fox News polling since May. The dark line shows overall support, and the colored lines show support from certain demographics. He has fairly consistently received more support from evangelical voters (purple line) than other groups. That faded in the most recent poll — a poll in which he did the worst.

In most cases, the demographic groups cluster pretty closely around the overall vote total, though. Here's Carly Fiorina. Even as she has climbed in the polls, she's not seeing much variance in support from different groups.

Two important caveats at this point. First, there are higher margins of error in demographic support measurements. Second, related to the first, variability between groups correlates to overall support. That is, the more overall support a candidate gets, the more likely different groups will diverge from that level of support.

It's where the divergence happens that things get interesting. Rand Paul, for example, has seen more support from college graduates and men for most of the past few months.

Marco Rubio has also done better with college graduates, and he saw more variance among demographic groups earlier in the campaign.

Ted Cruz has been hoping to appeal to the evangelical base and has done better with them since May than with everyone overall. But that support has fluctuated as his overall support has fluctuated, just like Huckabee.

Ben Carson has similarly done better with evangelicals and conservative voters.

And Jeb Bush has done worse.

Again: Support from those groups flows up and down with overall support. Sometimes, the up-and-down ebb influences overall support. Often, we can assume, it follows it.

Which brings us back to our first example, which we will now visualize.

Trump does about as well with evangelicals as with everyone. He does much worse with college graduates and much better with non-college graduates, but that's a larger split in part because he has so much more support. But all that support from conservative Republicans and evangelicals! Even college graduates have increasingly preferred Trump since June, though he trails Ben Carson with them at this point.

What's happening here is one of three things. Perhaps, as with state-level support, candidate choice in 2015-2016 is nationalized and universal in a way that it isn't usually. Or maybe looking at it this way doesn't accurately reflect a real base of support as we might be seeing with Huckabee.

But I'd put my money on the third option: The lock-down-a-lane strategy of winning elections was always presented in a more simple fashion than reflected reality. It, too, is one of the pretty things on our screens while Donald Trump is adding up support in the processor.