If Watson can win on "Jeopardy!" can similar technology do the job of a sports coach?
If humans can't beat a computer at "Jeopardy!" why should we trust them to make the right call on fourth down in the Super Bowl?
That was the fundamental question asked by some researchers at the MIT Sloan Sports Analytics Conference in Boston on Friday and Saturday.
SPORTS TECHNOLOGIES: 10 to love and 5 to hate
MORE CONFERENCE NEWS: Can't buy shares in Facebook? Invest in LeBron James
"Jeopardy!" players Ken Jennings and Brad Rutter "are some very smart people and they can't beat a computer [IBM's Watson] when it comes to one variable," says Tarek Kamil, founder of WhatIfSports.com and an executive at InfoMotion Sports Technologies. "How is a coach supposed to keep up with a computer when you're talking about 10 players on a [basketball] court," and variables including referee tendencies, defensive tempo and players' positions on the court? "There are literally 20,000 variables that come into play" in determining the right strategy, he says.
"Fifty years from now, we're going to laugh about how we used to give coaches this much responsibility," he says.
One might make the opposite argument, that Watson could only beat human players as long as the game was contained to one task -- answering trivia questions. And Watson had to consider millions of pieces of data to answer those questions. But Moore's Law ensures the exponential growth of computers' processing power, and sports experts are taking notice of their analytics capabilities after years of dismissing stat geeks as not living in the real world.
Although Kamil gave an extreme example of computers taking over human functions in sports, he also stressed the need for human guidance. "Am I saying we're going to get rid of coaches? Absolutely not," he says. "Someone has to be in charge of discipline. Someone has to be in charge of teaching players" and acting as mentors.
But human coaches are "not optimized for strategy and in-game decisions," he says.
Statistical analysis has already shown that baseball managers have over-relied on stolen bases and sacrifices, and that players are making a mistake when they slide into first base rather than running through the base. But convincing managers and players to change their habits, even in the face of unmistakable data, is an ongoing challenge.
At least in the United States, the analytics wave was led by baseball researchers, a no-brainer given that nearly every interaction in baseball can be assigned a statistic. Stat geeks, in general, have found that traditional statistics like batting average and earned run average aren't the best predictors of future performance. And their techniques are being applied to football, basketball, hockey and soccer.
Number-crunching can help teams choose the right players in the draft or trades, adjust in-game strategy, and even predict the likelihood of injuries.
Stat nuts have had to overcome intransigence in a sports world often resistant to technology. While sports teams will pour billions into enhancing the fan experience, they often won't make basic changes that could enhance the game. Major League Baseball finally allows video reviews of home runs and foul balls, but soccer referees still can't use instant replay to determine whether the ball went into the net -- even in the World Cup.
But the 1,500 attendees of the Sports Analytics Conference, naturally, are excited about the possibilities. Even Microsoft got in on the action, sponsoring the event and presenting a session about how business analytics can be applied to sports. A.C. Milan, for example, used Microsoft analytics software to predict injuries and optimize training, with the end result that it cut player injuries by two-thirds, use of medicine by 70%, and lost practice days by 43%, according to Microsoft business intelligence director Bruno Aziza.
Sports analytics still hasn't quite gone mainstream, Aziza says. It was hard to tell in Boston, though, given the star talent attracted by the sports analytics conference, including author and journalist Malcolm Gladwell, sportswriter Bill Simmons, Dallas Mavericks owner Mark Cuban and NBA announcer Jeff Van Gundy.
Attendees included executives from many professional and major collage sports teams including the Boston Celtics, Boston Bruins, Cleveland Cavaliers, Dallas Cowboys, Denver Broncos, Denver Nuggets, Georgetown University, Houston Rockets, Indiana Pacers, Indianapolis Colts, Minnesota Timberwolves, New Orleans Hornets, New England Revolution, New England Patriots, Orlando Magic, Penn State, Portland Trail Blazers, Sacramento Kings, San Diego Padres, San Francisco 49ers, Tampa Bay Lightning and the Tampa Bay Rays. Harvard University, ESPN, Nike and Google were also well represented.
Now in its fifth year, the Sports Analytics Conference seems to have opted mostly for funny, entertaining panel discussions rather than deep-dives on statistical issues. But stat geeks attend the conference to network with each other and swap ideas. Their dedication, along with rapid growth in computer processing power, should be enough to ensure a grand future for sports analytics.
While all it takes is better math and algorithms to crunch existing statistics, the next wave could be generated by new statistics from computer chips placed inside the very tools of the trade. Kamil's InfoMotion Sports Technologies developed a basketball that includes a sensor array that takes 6,000 measurements per second.
A coach might say "that kid looks like a good ballhandler," but do we really know for sure without measuring? Kamil asks.
"We can tell you his right hand is 14% more dominant than his left hand. We can break it down to that level, which has never been done before," he says. For shooters, the chip-containing ball can measure the angle of a shot, spin and release time, and compare the player against others and his own past habits.
The accuracy of referees can be enhanced by technology as well. Statistical analysis presented at the conference showed that home field advantage in sports is largely due to referee bias caused by crowd influence. Can technology help erase that bias?
NFL Super Bowl referee Mike Carey defended the reputation of officials, saying they call the game as they see it, try to eliminate all bias and make calls the same way whether it's the first quarter or overtime. But inevitably, humans make mistakes, he noted. Someday, perhaps, NFL footballs will contain a chip that can measure the location of the ball at the exact moment a player's knee hits the ground, helping referees determine ball placement more accurately, he said.
Analytics can't solve every sports problem, however, as Houston Rockets general manager Daryl Morey well knows. In the same casual, wry voice he uses to discuss trades, free agents and a player's statistical value, Morey asked, "Can we predict when the next NBA player will commit a crime?"