A Home Run for Sports Analytics

In sports, the perception that the biggest, fastest and strongest players win the game is not always the case. Advancements in analytics have played a stronger role in sports and they are helping teams win more games and championships without breaking the bank.

Last week the Hynes Center in Boston played home to the annual MIT Sloan Sports Analytics Conference, a gathering of stat junkies discussing new and interesting ways to look at sports outcomes.

Professional sports analysts and students alike shared and discussed models they hope will bring new insight into the game.

The boom in stat tracking comes from humble beginnings when a few dedicated fans would create homemade statistical models. In an interview with Steven Colbert, Bill James, the inventor of sabremetrics was asked about his method for tracking baseball stats, he answered, “I counted a lot of stupid stuff” to much laughter. He went on to describe how he counted stats people typically ignored, then ask how the data applies to questions people were debating. His methodology was eventually adopted by Billy Bean, general manager of the Oakland Athletics and was the focus of the movie Moneyball.

James’ basic methodology still rings true at the conference (which he attended). Students from USC presented a paper discussing basketball rebound data. Using optical tracking, the students were able to discover the probability of what team (offensive or defensive) would likely gain possession of the ball after a rebound. For example, if a rebound is located within 2 feet of the hoop, the offensive team has a 40% chance of regaining possession. While that stat may seem obvious, they also discovered shots made between 10-22 feet have less likelihood to be rebounded as opposed to shots made at 3-point distance.

While data can help, coaches have the final say about what they want their players to do. Former Houston Rockets head coach Jeff Van Gundy in a basketball analytics panel, discussed giving stats to players “If I have a stat, I use it. Sometimes, if I need a stat and I don’t have it, I just make it up.” a solution one blogger referred to a placebo analytics. The player takes criticism better, because the coach is informing as opposed to picking on them (as long as they think he’s telling the truth). ;

Basketball was the focus of a many discussions, but the conference also covered hockey, baseball, football, soccer and fantasy sports. Two MIT researchers submitted a paper that discussed a machine-learning algorithm for baseball. It would take pitch count, game state, pitcher tendency and more to determine what the next pitch would be. Using data from 2008 and testing it in 2009, the model was able to predict a fastball as the next pitch with 70 percent accuracy.  ;

Will all of these stats and models spell the end of superstar teams? Probably not since they attract a large fan base, but effective analysis can provide a means to build successful franchises that cost far less to run. In the end, the data usually helps deliver new insights into sports, but it won’t always predict the winner. If it did, who would care to watch the game?

Related Stories

Making Sport of Data Science

Along the Great Analytics Divide

The New Era of Computing: An Interview with “Dr. Data”