Ken Goldstein ’87 is involved in some very hush-hush business. All spring, he has ridden Amtrak’s Acela from the nation’s capital to Midtown Manhattan on Tuesdays, as well as some Saturdays, to an undisclosed location known as the quarantine room.

As soon as he gets there at noon, his cellphone is confiscated; bathroom breaks are supervised. Not even an inkling of the information he sees can leak out.

What exactly does Goldstein do?

When this professor of politics isn’t teaching about campaigns or talking about political ads, he’s crunching numbers and calling winners for the Election Night Decision Desk of ABC News. It’s the political-junkie equivalent, he says, of playing in the Super Bowl.

In fact, Goldstein, 51, has worked for a network news station behind the scenes in every national election since he was a fresh-faced recent grad with a degree in political science. These days, calling elections is a chance to put academic theory to practice. For Goldstein, a presidential year, especially this one with its long-running nomination process and conventional-wisdom-be-damned campaigns, is a thrill ride.

“It’s real time, ‘Can we call this now?’” he says. “It’s a rush. It’s history in the making. … That’s pretty fun.”

Goldstein grew up surrounded by politics. His father is a recently retired professor of government at Smith College; his mother, a former deputy mayor of Northampton, Mass., where Goldstein grew up; and his paternal grandfather was a New York state attorney general.

But Goldstein’s focus in college was baseball, though he says he was no standout. High-energy and trim, with dark hair, he has kept both his athlete’s physique (he played in an adult baseball league until recently) and his Haverford friendships with teammates Jon Trohn ’87 and Chris Siedem ’87, and lifelong friend John Robinson ’87.

Out of college, Goldstein snagged a researcher spot with CBS News. “This was before everything was online,” he says. “You would travel around to the states and county clerk’s offices and literally get out the dusty books to do past-voting-results research. That was enormously fun.”

Goldstein nixed law-school plans, joined CBS News’ Nightwatch (hosted by Charlie Rose) as a researcher/producer, and soon went to grad school at the University of Michigan, completing his Ph.D. in political science in 1996. There, he met his future wife, Amanda, a onetime TV reporter; the couple, who live in Bethesda, Md., have two teenagers. And it was as a grad student that he also fell upon a treasure trove of data.

Once, while co-teaching a politics class in D.C., his flight was canceled and he killed time by connecting with a friend whose office was next door to a new company, Campaign Media Analysis Group (CMAG), which tracked political advertising. After meeting the company’s head, he immediately realized the value of its data to academics and asked for the numbers—for free. He got them. That bit of serendipity—and networking chutzpah—morphed into the game-changing, Pew-funded Wisconsin Advertising Project, which issued timely reports based on CMAG information and provided scholars a first-of-its-kind database. Goldstein started the project in 2000 after joining the faculty of the University of Wisconsin–Madison.

Paul Freedman is an associate chair of politics at the University of Virginia who also studies ads and works on ABC’s Decision Desk, and he lauds Goldstein’s vision: “For the first time, we were able to bring real-world, comprehensive tracking data of political ads into the academic world. It revolutionized the way we study campaign ads. The most cutting-edge work in the last 10 years has used this data.” (The project has since moved to Wesleyan University, where one of Goldstein’s former graduate students is an assistant professor.)

In 2011, Goldstein took a leave to dip into the world outside academia, leading the group that gave him his first tracking data, now known as Kantar Media CMAG. “I had a lot of theory,” he says, “but this was the learning process.”

But Goldstein missed teaching, and in 2013 he was hired to direct the University of San Francisco’s USF in D.C. program, which gives students an opportunity to study and work in the nation’s capital through a consortium of four universities. It was his chance to be in the thick of politics.

“I’m fortunate,” he says, “that the stuff I research, the stuff I think about, the stuff I teach, and the stuff I sometimes blabber about on the media are all the same thing.”

In the quarantine room, Goldstein and ABC’s executive director of elections, Dan Merkle, typically huddle over data from early exit polling until 5 p.m. When the lockdown is lifted, they head to ABC’s studios to debrief other analysts who gather for multistate races. Two more waves of data arrive and get vetted. The marching orders are clear: “Our first instruction is to get it right,” Goldstein says, which can mean studying precinct-level results till the wee morning hours. No one wants a repeat of the 2000 race, when networks declared Al Gore the winner, then George Bush, only to find that the results were not clear, and would not be for weeks.

“Living through something like that makes you even more careful and cautious,” Goldstein says. “Look, it’s a lifetime of preparation. I understand the models and what to look for. I do my homework on the states. What are the absentee-ballot laws? When are they counted?”

As Goldstein talks, he fidgets, throws out the occasional expletive for emphasis, digresses to an anecdote, and jokes about his Haverford GPA (3.0 “with generous rounding”), the stale brownies on election nights, and the addiction to black coffee he’s had ever since the 2004 presidential election.

But when he works the Decision Desk, he’s known for his equanimity and laser focus. “It’s very serious, no chitchat, no joking around,” ABC’s Merkle says. “It’s kind of like solving a puzzle with teamwork. … Ken is a very quick thinker. He’s decisive on election night and not sitting there thinking, ‘Should we do this or that?’”

Once the race’s winner is called, the breaking news is delivered into the earpiece of election-night anchor George Stephanopoulos, who immediately tells his audience of millions.

“And it’s out,” Goldstein says. “Which is terrifying-slash-thrilling.

“Primaries are hard in general, because the models are built for general elections and turnout can be unpredictable,” he says. “And this season is tough. You do it and do it. Tuesday, Saturday, Tuesday, Saturday.”

On top of the longer-than-usual nomination run, 2016 has had its, um, peculiarities, which have challenged decades of campaign wisdom on any number of fronts.

One example is advertising dollars. Of the $3.8 billion spent in 2012 on television ads for all aces, $1.3 billion went to the presidential race. This cycle, the overall TV ad pot is expected to rise to $4.4 billion, but the presidential portion will likely be smaller, in part because of Donald Trump’s “focus on and complete dominance of the free media,” Goldstein says. As of May, the presumptive Republican nominee spent a mere $20 million on ads—compared to, say, Jeb Bush, who spent $80 million (including the money spent by PACs), and everyone knows how that turned out. “Trump is clearly using a different playbook,” Goldstein says.

Meanwhile, down-ballot Republicans might spend more on ads to set themselves apart from Trump, Goldstein notes. In other words, ads still matter, and especially in close races.

“You have these absolutist claims when it comes to the effect of advertising,” he says. “It’s either advertising completely determines the election. … Or this doesn’t matter at all. The fact is, it matters at the margins.”

Remember that word: margins. It is the main theme of the recently published Inside Campaigns: Elections Through the Eyes of Political Professionals, which Goldstein co-authored. (His other books include Interest Groups, Lobbying, and Participation in America, which is based on his Ph.D. thesis, and the co-authored textbook Understanding American Politics and Government.)

Many political scientists argue that fundamentals—the economy, the incumbent’s approval ratings, and the partisanship of likely voters—ultimately decide wins and losses, not the hundreds of daily decisions campaigns make.

“Fundamentals can get you within field-goal range,” Goldstein argues in Inside Campaigns. “Then it’s the margins.”

The right ad, for instance, can affect who comes out to vote, or which candidate swing voters support. Consider the attack ad that showed John Kerry windsurfing. It labeled him as both insincere and an East Coast elitist and helped sink his candidacy. Then there’s the recent Indiana primary. Bernie Sanders was heavily on air and Hillary Clinton was not, and he won an upset.

Goldstein’s insights have made him a sought-after nonpartisan voice on the subject of TV political advertising. He frequently appears on television and radio and in print.

“We love to look at the data of the election,” says Tom Johnson, executive producer of With All Due Respect, the Bloomberg News show where Goldstein is a regular contributor. “That’s a real sweet spot for Ken. By actually digging into the numbers, you can tell a lot about the campaigns. You can reverse engineer a bit. … He can put it in terms that are very relatable and understandable, a rare quality.”

The primaries may be over, but Election Night looms, and it may hold some surprises, Goldstein says. “The models work best when the present looks like the past,” he says. “So when you have years that are asterisks, that are unique, that can stress the models.” Sounds like a long night of figuring out what data to trust—but if anyone is up to the task, it’s numbers guy Ken Goldstein.