Breaking News Emails
For most fantasy football fans, Kansas City Chiefs running back Jamaal Charles looks like a surefire number one pick. While he regularly bulldozes defenders, there is one obstacle he can’t break free from — Boris Chen’s algorithm.
Chen, a data scientist at the New York Times, is part of a new breed of fantasy football obsessives that do more than parse injury reports and stay glued to the TV on Sundays. Many are former software engineers and financial analysts who use complicated algorithms to score big in a market where millions of Americans spend billions of dollars trying to build the perfect team. For some, it's not just a hobby — it's the path to a six-figure salary. Others just really, really love data.
“I actually have only been playing fantasy football for four years,” Chen told NBC News. “I hated football before that.”
The glut of statistics produced in each game has motivated 37 million other Americans to play fantasy football, according to the Fantasy Sports Trade Association. (That could be bad news for bosses. The hobby has cost American companies $13.4 billion in lost work output, according to consulting firm Challenger, Gray & Christmas.)
Many fans pick their favorite players or stars with gaudy highlights. Chen, however, sticks to the numbers, which is why he recently moved Charles down and the Philadelphia Eagles' LeSean McCoy up as his top pick.
More Than a Feeling
Like a Jedi, a true fantasy sports master can’t give in to their feelings.
“You want to be as emotionless as possible if you are going to take this data-centric approach to fantasy football,” Chen told NBC News.
That doesn’t exactly conjure up images of high-fives, foam fingers and all of the other stuff that makes watching football fun. Don’t algorithms make the game kind of, well, boring?
“Personally, not for me, because I think winning is pretty fun,” he said. “And being objective helps you win.”
While the image of running back Charles scoring an absurd five touchdowns in a single game against the Oakland Raiders last year might still be fresh in people's minds, Chen says fantasy players should pay more attention to the fact that his right tackle, Donald Stephenson, was just suspended for a month for violating the NFL’s performance enhancing drugs policy.
Unlike Chen, Nik Bonaddio was a football fan before he started playing fantasy football. Bonaddio, who holds a computer science degree from Carnegie Mellon University, used money he won on "Who Wants to Be a Millionaire" to start the sports analytics company numberFire.
"All sports fans are amateur statisticians," he joked to NBC News. He just happened to be better at manipulating those stats than most people. He admits, however, that his personal history with football can be more of a curse than a blessing.
"I'm a Steelers fan, so I know I'm biased towards starting a Steeler," he said. "An algorithm does not have any of those biases."
Bonaddio claims that people who use his algorithm — an evolving formula similar to the regression models used by financial firms — win 30 percent more often than they do when they pick on their own.
Where these algorithms are really helpful are the later rounds of a fantasy draft. It's pretty clear who the elite players are in the first round. It's the second round and beyond where people really start to let their emotions and biases choose their teams for them, Chen said.
His solution involves something called the Gaussian mixture model. It's complicated, but essentially it uses a clustering algorithm that sorts massive amounts of data with the goal of creating tiers — groups of players that have the same value in a fantasy football league.
The idea is that instead of picking a wide receiver because you remember a spectacular catch he made, you would just defer to the tiers, picking any remaining tier two players over those in tier three or worse, while always diversifying across different positions.
Diversifying picks across tiers sounds like something you would hear on Wall Street, not a sports bar. In reality, playing fantasy football isn't that different from what many people in the finance world do.
Cory Albertson is currently in business school at Notre Dame. He also makes thousands of dollars each day playing in fantasy leagues. In a profile in the Wall Street Journal, he likened what he does with "securities trading" where the "athletes are the commodities."
He isn't the only one. Peter Jennings used to trade stocks. Now he makes a living in Colorado playing in different fantasy sports leagues. Some of these leagues have big payouts: the first prize in this year's FanDuel Fantasy Football Championship is a cool $2 million.
"It’s becoming, for a subset of people, a full time job," Bonaddio said. "There is definitely more money around fantasy sports than there ever was before."
Even without hitting the jackpot, top players with a good algorithm can make more than $100,000 a year. To make that kind of cash, you need to have your fingers in a lot of different pots, like basketball, hockey, baseball and the other kind of football — known in the United States as soccer.
"Our algorithm could really be applied to any kind of team sport," Sarvapali Ramchurn, a computer science professor at the University of Southampton in the U.K., told NBC News.
He and his team developed a virtual team manager that placed in the top 1 percent of players in the English Fantasy Football League. Eventually, he hopes the predictive modeling technique he developed could be used beyond sports, like coordinating large numbers of emergency responders during a disaster.
For now, however, its main focus is soccer, where it's good at winning, but not perfect.
"There are subtleties that we can't capture in the algorithm that humans are better at grasping," he said, like a team's motivation while safely atop the standings or the effects of an off-field scandal. "The algorithm is better than most players, but it's not better than the best players."