Nearly any student who asks here for advice on college will get told the same thing, and that’s to go study economics. That’s true for students on their way into social work as well as into business.
There are many other useful classes in the catalog, of course, but economics offers a great framework for understanding how imperfectly even the most rational people make the choices they do, in a hopelessly complex world where everybody else is also stumbling onto their best option.
A great example of how much fun can be had comes from a Macalester College quarterback and budding economist named Benny Goldman.
His is a story about football but not really about his playing the game. His study of economics led him to build a predictive model that just may reshape traditional decisionmaking, even in a business as hidebound as the National Football League.
His story begins at Macalester in St. Paul, hardly known as a football powerhouse. He’s from a suburb north of New York City and came here, he said, “because I wanted to get off the East Coast and I wanted to play football.”
He certainly got the opportunity to play at Macalester, a Division III school that plays football against similarly brainy little colleges like Carleton College and Grinnell College in Iowa. Never a star, he’s what a traditional coach would call a “student of the game.”
His passion led him to digest the contents of NFL playbooks, a few of which he somehow found on the Internet. His first idea to improve the game was to find a far simpler way to give instructions to the team on what formation to use and what play to run.
The NFL used complex jargon, and he thought it could be done in just one word.
Goldman described himself as indifferent to academic work when he first got to St. Paul. Then he discovered economics.
“He’d been thinking about football analytically for a long time,” said Sarah West, an economics professor and Goldman’s adviser at Macalester. “Then he came into economics and realized that the approaches that we have to thinking about decisionmaking have a good deal of application to what he was thinking about on the field. He also then took a number of math courses, which deepened his understanding of the more sophisticated approaches to statistical modeling.”
Goldman turned out to be a “hybrid,” as described by the first NFL staffer who thought Goldman might be worth listening to, the former NFL scout Dan Hatman. By that, Hatman explained, he meant here was the rare football player who also understood game theory, probability and statistical modeling.
Goldman found Hatman through the social networking site LinkedIn, one of hundreds of NFL executives, scouts and coaches Goldman tried to reach. Only Hatman responded, and before long Goldman was in the offices of the Philadelphia Eagles, presenting his one word play-calling system.
“In any other industry, nobody would’ve listened to an 18-year-old,” Goldman said. “It speaks to how competitive the NFL is.”
Nothing quite came of the system, but before long Goldman and some other Macalester players began dabbling in a statistical program that evaluated each game they played, using Excel spreadsheets. They were trying to figure out how the probability of winning the game moved up and down as the game progressed.
That’s the idea Goldman has grabbed and run with. He has since developed it into a predictive model based on a database of 700,000 observations from actual NFL games.
When Goldman first started discussing this with NFL staffers he met through Hatman, he said, “what I hadn’t realized is how little analytics there are in football. Like borderline none.”
A statistical revolution came to baseball years ago, in part because it’s mostly a two-person contest between a pitcher and a batter.
“The thing about football, as you know, is it’s an enormously complicated game,” West said. “You have all these individuals interacting out there. Even with advanced statistical modeling within football, the problem remains so complicated that even an undergraduate [like Goldman] can make substantial contributions by applying new statistical methods.”
West said she hasn’t looked closely under the hood of Goldman’s predictive model. But, she said, “this guy is totally the real deal.”
How would an NFL team use Goldman’s work?
Well, imagine that an NFL game is just getting underway, with some fans not even seated yet.
The home team has the ball on its own 20-yard line and runs the ball for 5 yards. It has two more tries to get 5 more yards and a first down or it will likely opt to kick the ball away to the other team and switch to trying to prevent the other team from scoring.
What play should be called? Nobody on the coaching staff will be thinking the outcome of the game is now on the line. The game’s just started and neither team has even scored.
In Goldman’s model, what happens on the very next play does affect the outcome of the game. His model will point to the right choice for that play. Maybe it’s only measured in a sliver of a percentage point, but Goldman will know the best option for winning the game on second-and-5 from the 25-yard line.
There’s a lot of work to do to make this kind of thinking a routine part of an NFL team’s decisionmaking, Goldman explained, including finding a way around the NFL’s bizarre ban on computers in the coaches’ boxes and on the sidelines at games.
Goldman has an NFL head coach very interested in this work, although he asked to keep the team’s identity private.
Goldman, who graduates in May, sees himself as a research economist and has applied to top-tier Ph.D. economics programs. He has already gotten one generous fellowship offer, with 11 more universities yet to be heard from.
His head says that’s a better career option than working for a team, for he knows how difficult life can be for an NFL staffer. Head coaches routinely get fired after three seasons, and the rest of the staff gets swept out behind them.
On the other hand, football dreams are hard to let go.