This article is from the archive of our partner .

Within minutes of polls closing on Tuesday, media outlets are likely to declare winners in at least two of the day's three big contests. But in politics, as we've seen (infrequently) in the past, anything can happen. Here's what it would take for each of the races everyone is watching to turn into an upset.

The key thing to watch in any race is turnout — how many people get to the polls. It seems obvious of course: if no one votes, no one can win. But there are patterns to how people vote that differ only marginally from state to state. Off-year elections see more high-frequency voters (by definition), people who vote in every election. Those voters tend to be in Republican-leaning demographics — older, wealthier — so Republicans tend to do better in those races. When you're looking at what happens on Tuesday, last year's presidential election — featuring massive media attention and the nation's first black president — isn't a good guide for predicting who will vote. Better to look at the 2009 elections in each place. Even that, though, comes with lots of caveats.

Let's start by looking at Virginia.

Virginia

Democratic change between 2009 and 2012: 140.8 percent increase
Republican change: 56.6 percent increase

A quick note on the graphs above, which we've generated for each race. At left, the change in vote total between the 2009 and 2012 elections. In each, you'll see that more people voted in 2012; in each, you'll see that Democratic turnout increased more than Republican turnout between the two elections. At right, the percentage of the vote won by each party in 2009 and 2012 and what polling suggests about the races in 2013.

Expected winner: Terry McAuliffe, Democrat

How he could lose: "The only poll that counts is the one on Election Day," McAuliffe's Republican opponent, Ken Cuccinelli, likes to say. And that's true, if not particularly insightful.

What Cuccinelli and his supporters should be keeping an eye on is that thick bar of yellow in the bottom right of the graphs above. Eight percent of the electorate supports the libertarian candidate — but 5 percent hadn't yet made up their minds. In a race in which the Democrat leads by 6 points, that's not insignificant.

There are caveats: McAuliffe is the second choice of even those libertarian voters; his approval, though negative, is substantially higher than Cuccinelli's. There are various theories for how undecided voters break on Election Day, but they will not break heavily for Cuccinelli.

But if turnout is down, with more Democratic voters staying home than expected, the race is close enough that something dramatic could happen. When McAuliffe warned over the weekend that he could lose, he was doing two things. First, he was reflecting the universal concern of all candidates that they've misjudged something badly and they will, in fact, lose. But, second, he was warning against complacency. McAuliffe already has an uphill battle, demanding Democrats turn out heavily in an off-year election. If Democrats stay home because it's raining (it's not expected to), that's one thing. If they stay home because they think McAuliffe is going to walk away with the race, that's another thing altogether — and could, possibly, be a serious problem.

What won't make any difference: Any last minute ads or mistakes. The campaigns are now solidly focused on turnout operations (or should be); dragging people to the polls, almost literally. So even if Cuccinelli were to come out and reveal that he has been endorsed by Jesus Christ himself, it wouldn't actually matter. Too late. (Though he might win races in the future.) (In the South, probably.)


New Jersey

Democratic change between 2009 and 2012: 95.5 percent increase
Republican change: 25.9 percent increase

Expected winner: Chris Christie, Republican

How he could lose: In one sense, Christie should be more worried about complacency among his base than should McAuliffe. After all, his lead is much larger, well beyond what any undecided voters could make up.

But Christie could still lose. His opponent, Barbara Buono, hasn't engendered robust support from the Democratic establishment. But she has one hope: that the polling is flawed. Polling relies on the same sort of estimates of turnout that we touch on in our graphs, trying to guess how many people and how many people of each demographic type are going to get to the polls. This was the heart of the "unskewed polling" trend last year, the assumption that the turnout model (as it's known) underrepresented Romney voters and were, therefore, incorrect.

It is possible that the models used for Christie are, in fact, flawed. That the Democratic electorate that handily supported Obama will return to the polls this year, or, perhaps, that more women will come out to support Buono. These things happen. In 2012, the unskewed poll trend was stupid largely because the preponderence of polls used similar models, developed from parallel assessments of voting patterns. There have been fewer polls in New Jersey, largely from the same pollster, Quinnipiac University. What if that pollster's model is wrong? (In 2010, the last off-year election, Quinnipiac "performed strongly," according to 538's Nate Silver.)

Buono's other hope may be that Christie supporters, figuring that the race is a done deal, stay home, and that the labor unions backing Buono (like the teachers) can make up some of that difference by knocking on doors. Unfortunately for Buono, field efforts like that usually only make a difference of a couple of percentage points.


New York City

Democratic change between 2009 and 2012: 273.1 percent increase
Republican change: 25.4 percent decrease

Expected winner: Bill de Blasio, Democrat

How he could lose: De Blasio's base is heavily Democratic voters in Brooklyn and Queens — areas of the city that a 2012 study found had lower turnout than the rest of the city. And, as one voter discovered on Tuesday morning, polling places in some areas of Brooklyn are already having problems.

So how does de Blasio's Republican opponent, Joe Lhota, win? Shortly after 1 p.m. on Tuesday, a massive power outage sweeps across Brooklyn, spreading across Queens and into the lower Bronx. There's turmoil in the streets as businesses close and traffic clogs the streets. De Blasio, in a private car sounding its horn as it navigates the sidewalks, heads into Manhattan to demand that the election be delayed. Voting continues apace in Manhattan's Upper East Side and — completely unobstructed — in heavily-Republican Staten Island. By 5 p.m., as de Blasio voters are getting off work and heading home, they find polling places shuttered, near-anarchy from the East River to the well-maintained lawns of Long Island.

By now, de Blasio has made it to Lower Manhattan (having comandeered a small boat) and is confronting Mayor Michael Bloomberg. But Bloomberg, whose first primary race, on Sept. 11, 2001, was delayed only in the face of a much more dramatic crisis, refuses to reconsider the election. (Mostly because Bloomberg doesn't want de Blasio to win.) Bloomberg summons police commissioner Ray Kelly to subdue an irate de Blasio. When polls close at 9 p.m., despite broad outcry from across the Brooklyn Bridge, the vote tally comes in — Joe Lhota has won by three votes in record low turn-out.

That's how de Blasio can be upset.


Polling data via Quinnipiac. Vote history via USElectionAtlas and Vote NYC.

This article is from the archive of our partner The Wire.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.