Microsoft Research Connections Blog
Next at Microsoft
Social Media Collective
Windows on Theory
Posted by Steve Wiens
Microsoft researcher David Rothschild is legendary for his ability to predict the future using a unique, rigorous approach to data analysis. He correctly called the results of the 2012 U.S. presidential election in every state but one. He nailed 19 of the 24 Oscar categories this year. And he’s constantly pushing the boundaries of predictive science through experimental live polling, online prediction games, and more.Now, Rothschild tells you what to expect in 2014, breaks down his forecasting philosophy, and explains why you should trust professional gamblers more than cable-news pundits.
Q: Obviously, you’re pretty good at this. What’s your secret sauce? What do you know that the rest of us don’t?Rothschild: I use a lot of publicly available data—gambling markets, polls, social-media data, and other online data. But I’m also lucky to have access to a large treasure trove of historical data at Microsoft that I use in order to understand the correlation between different inputs and outputs, things that we care about, as well as an infrastructure that allows me to gather that data continuously and continuously update it. I really owe a lot to Microsoft and my colleagues.Q: How is your approach different from your average cable-news talking head?Rothschild: Most pundits don't have or utilize data at their disposal in order to answer these questions. They're relying on their own personal insight or their own personal ideas. And that leads you down the wrong track.Also, as a scientist, I send my work and my models out to be reviewed by other scientists, and they care about how consistently accurate I am. A pundit, on the other hand, is judged by ratings or how exciting they are, how many people want to see them or quote them. So they benefit from providing predictions that are more “out there,” because that allows them to hit a few home runs. Now, a couple home runs and a couple strikeouts made Reggie Jackson a national hero—but that doesn’t necessarily make him the best guy to have on your team, right?Q: Wait, so gamblers are more trustworthy than cable-news pundits?Rothschild: Cable-news pundits are one person, not a crowd, and there is nothing to correct their errors or object if their misaligned incentives lead them to provide inaccurate answers.That being said, I rarely report the market number directly. We know how market prices historically correlate with outcomes and know the most accurate ways to turn raw market prices into forecasts, by correcting for this relationship.Q: Does intuition or gut feeling ever play a role in your predictions?Rothschild: I don't make decisions from my gut. I make decisions based on as much information as possible, and then I make them with the least emotion as I can. It's a very scientific process that you need to repeat very systematically in order to be consistent in what you do.Most people would say that I'm insanely rational in what I do all the time, and I take that as a compliment. I'm interested in thinking about how this research can actually affect decisions people make, and I take a lot of pride in approaching decisions with a very systematic rigor, regardless of whether or not it's a lofty question that I'm posting to the general population or it's a standard decision that I make in everyday life.Q: What’s an example of how that factors into your personal decisions?Rothschild: I'm sitting with my friends last weekend, and we were contemplating going to a hockey game. So I’m studying StubHub and Ticketmaster, spending probably way too much time and effort thinking about how ticket prices are changing, how people are going to behave as game time approaches. Sometimes it's at my own cost, because I'm spending too much time analyzing and thinking about the expectations of what's going to happen.Q: Are your friends ever, like, “Can we just go to the hockey game already?”Rothschild: I got told that last weekend. [laughs]
Q: One thing that’s interesting to me about your blog is that the predictions aren’t static—they’re constantly being updated as new data is available.Rothschild: I believe very honestly that in 2014 and beyond, no one is going to be happy with a static prediction or a static indicator of any sort. I mean, we just don't accept that anymore. It used to be that people would come out the night before the election, or right before a game, and say who they think was going to win. Now, we expect to see that prediction move during the course of that game. We see that prediction start before the course of the election cycle.We don't have the patience or even the idea that something wouldn't be continuously absorbing information. The Internet continuously updates. Why would your predictions be static?Q: With that in mind, let’s take a look at a few of your current predictions for 2014, starting with the Super Bowl. Rothschild: The [sports gambling] markets have been extremely consistent in showing the Seahawks and the Broncos as just absurd favorites to come out of their conferences. Both of them have an about 25 percent chance of winning, which means that, combined, all other teams have less than 50 percent chance of winning versus either the Broncos or the Seahawks, both of whom are likely to get byes.That's where we're sitting right now, and there's obviously been movement across the season. One of the really fun things to see is we have these odds on our site that update continuously every few hours during the course of the year, and it's been really interesting to watch as people talk about this game being a big game or that game being a big game and actually being able to see whether or not it has that huge of an impact.For instance, people were concerned when Seattle played San Francisco, but it didn't really impact their likelihood of winning the Super Bowl much one way or the other. And so those are the types of things we're following, versus other times where it may have a greater impact.Q: And in the World Cup, at this point you’re calling it for Brazil?Rothschild: Yeah. I mean, Brazil has a pretty nice path out of the first round. This year it’s in Brazil, and home teams have done well historically, but we're not throwing any of that in right now. We're really thinking more about what the markets are saying, because the markets have been extremely consistent and very correct in the past.Q: What about presidential election hopefuls? How is the race looking now for you, and what could change in 2014?Rothschild: Should she run, Hillary Clinton has an extremely high likelihood of victory. The election looks very different with and without Hillary Clinton in the race.Obviously the much tighter race is on the Republican side, where it's still really up for grabs. Chris Christie is maybe the favorite, but that's still a longshot. I mean, there's no individual candidate who is anything other than a longshot. If you put Chris Christie at a little over 15 percent, as well as Mark Rubio and Jeb Bush closer to 10 percent on downward, there's still a huge amount of uncertainty.Q: Well, I’m going to go place some bets now! Just kidding. Thank you very much for your time, David.Rothschild: My pleasure.See more predictions from David Rothschild.