Sports data analytics isn't always a slam dunk

Emotional X-factors

But things didn't go as well as the analysts had expected. After a Western Conference finals appearance in the 2014-2015 season, the team exited in the first round of the 2016 playoffs and Howard subsequently left Houston and signed with the Atlanta Hawks. So what went wrong? Reports indicate that Howard was upset about his role in the offense and didn't like the coach the team brought in for the upcoming 2016-2017 season. Issues of human personality often affect in-game performances. In the case of Howard, who has had a history of disunity with teammates, his personality traits might have been foreseen, though impossible to predict with data.

There's another problem with sports data analytics, and it's not so much that outcomes are unpredictable in an absolute sense. An article published in 2014 in the journal EPJ Data Science found that, in fact, the timing of scoring events and the outcome of games are remarkably predictable in American football -- both collegiate and professional -- and ice hockey. Basketball games tend to be less predictable, though they still adhere loosely to a discernable pattern.

No, the problem is that knowing likely outcomes matters little to in-game decisions. A coach may know with a high degree of certainty that teams facing a scoring deficit similar to the one his team is facing go on to lose games 65% of the time. But does that mean it's time to pack up and go home? Of course not, the coach will try to optimize every play to help his team be among the 35% that go on to win.

Just because you know the general likelihood of something happening doesn't mean you know precisely what will happen in a specific case. Regardless of whether we're talking about football or basketball, you try to score when you possess the ball and keep the other team from scoring when they have the ball. In most cases, the decisions don't change substantially just because you know your chances of winning or losing.

A game of inches and gigabytes

Overall, I still think it's a good thing that sports teams are becoming more data-driven. Sports data analytics can show baseball teams where player investments tend to pay off according to a player's position. It can help basketball teams understand the right mix of two- and three-point shots to optimize scoring. It can help football teams determine the amount of practice and training a player can handle before his risk of injury increases.

But today's vogue of proclaiming that everything in sports is predictable ignores substantial problems for which analytics poorly redresses. It also drives an unnecessary wedge between the people who have adopted analytics into their understanding of sports and those who haven't. There's a lot of talk these days about people who "believe" in analytics.

Statistical methods that have been around for centuries aren't something in which individuals can choose to believe or disbelieve. But the believers have turned it into a partisan issue. Instead of pushing an ever-expanding one-sided agenda, proponents of analytics in sports should acknowledge that data is useful in some areas and not in others.