I devoted an entire piece in Monday's WaPo to making the case that Senate Minority Leader Mitch McConnell (Ky.) is in more trouble in his fight for another term than conventional wisdom in Washington dictates. And yet, our "Election Lab" model gives the Kentucky Republican a 97 percent chance of winning his race against Democrat Alison Lundergan Grimes. So, what gives?

I reached out to "Election Lab" co-creator -- and Monkey Cage blog author John Sides -- for some explanation of the discrepancy. Here's what he told me:

To me, the key thing happening now is this: the polls are moving in the direction of the forecast. That is, the earlier polls suggested a tied race, but have shifted somewhat in McConnell's favor.  (Pollster gives him a 93 percent probability of a lead.)  This fits with a key finding from political science, which is that the polls tend to move in the direction of a forecast based on the underlying "structure" of a race.  In this case, the structure -- a well-funded Republican incumbent running for reelection in a red state in a midterm election -- favors McConnell.

Remember that, as Sides explained when we unveiled the model in May, the "Election Lab" model is built less on individual candidate qualities than on broader structural pillars of the race. So, the national political environment or the relative financial strength (or weakness) of the candidates matter more to the model than whether McConnell or Grimes is, say, a "good" candidate.  ("Good" is, of course, in the eye of the beholder and therefore extremely hard to quantify across a whole nation's worth of races.)

And, as Sides, notes, recent polling in the race -- post McConnell primary -- does suggest the contest is moving, albeit it slightly, in the incumbent's favor perhaps as Republicans split by the intraparty fight have begun to line up behind him.

KY Senate

Here's what the "Election Lab" model -- and a similar one from 538 that gives McConnell an 88 percent chance of winning -- see in this race: A very well funded incumbent who has won a number of past races running against a relatively inexperienced challenger (in terms of time in elected office) in a state that favors Republicans in an election year that seems likely to favor Republicans. "The key point is for Grimes to win she needs to maintain support among constituencies that seem unlikely to favor her," noted Sides.

Democrats -- and some nervous Republicans -- don't disagree about the underlying structure of the race favoring McConnell. But, they also look at McConnell's numbers -- particularly his job approval numbers -- and wonder how someone with those numbers can win no matter the landscape on which the race is being contested.  These people would argue that candidates and the campaigns they run matter far more than our model or the one built at 538 accounts for.

My belief, as I wrote in the piece, is not that McConnell will lose. Instead, it's that the idea that McConnell is a stone-cold lock to win -- a sort of mantra that is repeated by the D.C. chattering class -- isn't entirely right. Our "Election Lab" model says I am wrong.