Stuart Rothenberg treats politics like sports, and his race ratings are the batting averages everyone wants to know.

Rothenberg still remembers being in graduate school and memorizing stats on American politicians as if they were baseball players.

Three decades later, his Rothenberg Political Report is one of the premier predictors of electoral politics. When Rothenberg moves a race as leaning one way or the other, it can make headlines.

We asked him to explain how he crafts those ratings:

How did you become the ratings guy?
I've been doing this about 30 years but I intended to be a college professor specializing in U.S. foreign policy or international politics.

Like a lot of people in graduate school at the time, I had some background in American politics and I had a roommate for a couple years who always had a book, "The Almanac of American Politics," sitting in the living room.

I used to thumb through that and we would watch sporting events and treat the political candidates and members of Congress as though they were athletes.

So I came to know some of the facts and figures. But I very much stumbled into this. I knew somebody who was writing infrequently for Roll Call and they had to give up their position.

That person referred me and I started writing for Roll Call. I had hired somebody who left me and went to ABC and then CNN, and I started getting requests for soundbites and to do appearances on CNN.

It took off dramatically in the early '90s.

How has the ratings landscape changed in the time that you've been doing this?
Right now everybody and their brother has ratings. But back then, people didn't pay a lot of attention to the House. They figured it was just structurally impossible for the Republicans to win.

The reality of it is that my business is better when there is a lot of uncertainty and a big election cycle. So ratings became more important, I think, in the early '90s when the house really kicked in as a body in play. Obviously since '94 we have had a series of really strange dramatic elections.

Who is your audience?
We have a lot of interest groups, political action committees, trade associations, labor unions, as I say political party types and some contributors.

Then there is a small but all-important universe of political junkies out there: people who are just obsessed with politics but also individual campaigns.

It's not a huge universe. This is not tens of thousands of people but people who care passionately. Many of them follow elections and need to know who's going to win to do their jobs well.

Would you count yourself among the political junkies?
I used to be. Time has taken a little bit of a toll on my enthusiasm.

After 20 or 30 years, you hear every two years campaign politicians making the exact same arguments and the exact opposite argument than they made two years earlier when they were in a different position. I think you probably reach a level of fatigue and cynicism that from time-to-time you worry is about to drown you.

But I still get the biggest kick out of it as a sporting event. I still like sports and I still like the sport of politics.

So how do you do your ratings?
The survey data is very important in guiding us as to what is happening. We extrapolate from one district to another district, one race to another race, and we see kind of how different groups are moving. So there's a lot of that.

Some of it is we've been doing this a long time, so we bring our experiences and races we've seen.

Some of it is psychology – interviewing the candidates and trying to put yourself in the mindset of the voters – and some of it is just the heavy reporting function – making lots of calls, trying to get lots of information and data – so that when we make predictions or assessments or projections or whatever you want to call them, they're based on something hard.

We don't just intuit these things. We don't just say, "Okay, well in 2008 this happened so it's going to happen this year." We want to look at the numbers.

What do you look for when interviewing a candidate?
We sit usually for about an hour. We get their backgrounds we get their position on issues. We get who they've hired, how much money they've raised, and we try to push them on some issues or some themes to see how they respond to that. So [we look at] how they perform.

We get a whole range. You get people who have never run for office before or it's their first run. Six months from now they're going to be very good, but right now they're green. We understand that.

We get some who are very poised. They've done this forever. They run for office every 2 years. We try to evaluate them differently based on their experiences and how the voters will see them and over time.

Do you have access to insider information?
We talk a lot to pollsters, media consultants, and general consultants. I've been doing this a long time so you just meet with consultants who have been doing races for the last 25 years. So you've built a relationship.

This is a straight reporting function except that we do it all on background. There's no gotcha journalism in this.

We just want to get the handicapping right. We want to call the races correctly. And so, again, we talk to the consultants and try to learn stuff about the campaigns and the candidates.

In that regard, we hunt for and at least from time to time are successful in getting our hands on otherwise private survey data that we use.

How early can you call most elections?
The first half of an election cycle is all about scenarios. You go to one committee or one campaign and they have a scenario about why they're going to win or why they're going to do better than historically expected.

It's all scenario. We can come up with scenarios about anything here at the Rothenberg political report, and we have heard scenarios about everything.

That's why I'm always a bit more tentative in my ratings until we hit this last few months period.

Is there an art to what you do, or it just a matter of culling data?
To us that's half the equation. There's a lot of art.

We look at those numbers too, but we think people are interested – and, frankly we're interested – in the individual candidates. At end of cycle, I want to say, "I met this candidate who won and here's what they will be in Washington. Here's why you should look at this person and not this person."

You get a flavor for the cycle.

How do the ratings you come up with affect the money flow?
Apparently we are all part of this conventional wisdom that develops as to who are the top-tier candidates and who are the long shots. I gather that affects some money flow, but you'd have to talk to the party committees about that.

I think they do send around what we say as of a way to create momentum for candidates, and momentum creates fundraising.

But we don't do that. They may use our stuff but I don't care where they send their money. As long as they send me their subscriptions.

What insight do you think race ratings provide?
We labor over them because they have our name on it and we want to get it right. That's one of the services we provide to our subscribers and to other people. But I would think somebody would be really getting only 10 percent of the value of what the Rothenberg Political Report hopes to offer if they just looked at the ratings.

We do ratings but with the newsletter then, but we'll try to explain why we put the ratings. And ratings change. The rating today may not be same rating three weeks from now. But we always like to explain what looking for, what we've seen. But if circumstances change, then we'll change the rating.

We put a lot of time into writing. We write a lot. To do ratings, all you have to do is put a name on a list.

Then why do the ratings?
Doing the ratings I think does help us think about all the factors.

We go over the polling, talk about the candidates, dissect the race, and talk about everything that goes into rating a race. The ratings do help us think about the candidates and the races.

Sometimes, early in cycle, we'll say we're not sure. So we'll make a best guess.

But hopefully over election cycle as see more of the candidates, as we get more of the polling, we're able to make better distinctions between the races.

One of the dangers about doing ratings of individual races is you are investing yourself in that race in a particular way.

We've just had to move Nevada away from "leaning Republican" to "toss-up" because there's enough survey data there, there's enough chatter about Sharron Angle, even though eight months ago I absolutely, absolutely never thought Reid would win short of the Republicans self-destructing.

When you rate a race, you have to be willing to say we were wrong.

Will the ratings change a lot between now and November?
We like to let the party in power make its case to voters.

This is what we did in 2006. Republicans were in trouble, but let's see if they can spend enough money, they change their trajectory, they localize enough races, and the squeeze out victory. Now, we came to a conclusion pretty early that they weren't going to do that.

But that's why put emphasis on September: to see whether the Democrats can change things.

If Democrats don't change the trajectory of this election cycle and if they don't change the fundamental dynamics – what people are seeing, how they're evaluating things, and the choice that they are going to make – I would expect our target of net change to continue to grow with the Republicans benefiting.

Ambreen Ali writes for Congress.org.

This interview has been edited and condensed.

Editor's Note: You can test your skills against Rothenberg's with CQ Politics' new "Rate the Races" feature.