Part 1: An Overview of SportVU Technology
When compared with baseball and even football, basketball has lagged in its use of analytics. Given its free-flowing nature and arguably greater reliance upon the team working as one unit, analysis on a play-by-play basis is much more difficult than in its popular American sports counterparts. Fans, media, and teams alike often cling to very simple statistics (points per game, rebounds per game, shooting percentage) when determining performance and value of players. Even a popular 'advanced' statistic in NBA evaluations, the Player Efficiency Rating (PER)1, is essentially an algebraic combination of the basic statistics. Kirk Goldsberry, a Harvard visiting scholar in the Center for Geographic Analysis and contributor to the blog, Grantland, puts it this way:
From an analytics perspective, baseball is checkers; basketball is chess. One could even argue
that assigning any individual credit in basketball is a fool's errand, and that contemporary
basketball analytics erroneously conflate the circumstances of a player's environment with
his individual ability.2
In short, the analysis of the game of basketball is a fascinating and highly complex problem.
Since high quality data is crucial to any significant analytics effort, statistical analysis of the game of basketball has truly lacked any meaningful findings. If we take two of the best players in the league right now, LeBron James and Kevin Durant, and look at their traditional statistics, what can we say? They are very good: both score a lot, rebound well, and are excellent passers, but what allows them to be so good? Beyond a basic descriptive story of what happened, traditional statistics and measures don't tell us much; only through watching the game do we understand their superior ability to a greater extent, but how do we quantify this? Basketball's lack of more rigorous measurements over the years leaves us with much information lost. However, there has been a movement in the last ten years to analyze basketball in new ways, and a breakthrough technology called SportVU will change the way we look at the sport.
The SportVU technology, in its basketball application, utilizes six cameras: three per half-court, set up in the rafters of NBA arenas. These cameras capture the movement of every player and the basketball 25 times per second, then output the X-Y-Z coordinates at every interval3. The cameras are advanced enough to detect jersey numbers, and thus can attribute movements to each player correctly. The current NBA season is the first season in which the technology is being used in every single NBA arena, providing unprecedented amounts of data on every game played during the season.
STATS, the creator of SportVU, performs their own conversions of the raw coordinate data, applying complex business rules and algorithms to transform the data into something meaningful. The NBA has made a portion of this data available on NBA.com/stats. These "data marts" allow fans, media, and teams to evaluate performance at another level; for example, one can view the distance a player has run in a game, shooting percentage in certain basketball-specific situations (on a drive, on a pull-up shot), and average time possessing the ball. While these statistics are certainly new to the game, the ability of SportVU cameras to track in three dimensions gives us more than just extra numbers to look at; we can now analyze the game spatially. Though this type of data isn't available to the public, a quick Google search yields a few videos displaying some of the uses of this new technology. The ability to look at the way a defense moves as a unit or see the difference between where a defender should ideally be and where he truly is at all times completely changes the way basketball is viewed (see this link for an example visualization of SportVU data). The simple ways in which we have evaluated performance (the aforementioned basic statistics) are now greatly enhanced by the new plethora of data that SportVU allows us to analyze.
Let's look back at the traditional box score. It shows those common statistics1 that over the years have become the standard for assessing a player's performance in a game. The statistics shown in the traditional box score provide valuable information, but they do not give us the level of insight that SportVU now provides. Want to know which player hustled the most, or which player was able to create the most distance between himself and his defender to get open? SportVU can tell you. Take the LeBron and Durant example. The box score allows a coach to see that these players are playing well; with SportVU, the coach can quantify, and thus more readily analyze, why these players excel from game-to-game. It would become clear, for instance, that a player's ability to read his defender and make the best decision with the ball every time down the floor has a significant impact on his performance. Yes, some information will escape the grasp of SportVU. It is not a cure-all; however, the ability to see these new dimensions at a new level of data capture will certainly give us a richer understanding of the game.
The technology itself is fascinating, and the example above is only one of the many potential applications of the data. To reiterate, SportVU is only scratching the surface right now. What might the future hold for other stakeholders going forward?
The coach of a basketball team is like the CEO of a business. The more quality, data-driven insights the CEO has, the better his ability to effectively manage his company. The same goes for a basketball coach and his team. As mentioned above for coaches, analyzing their players' performance, their offensive and defensive strategies, and the play of the opposing teams will be taken to another level. The key here is that the coach can see the why of his players' performance and not just the what.
Contracts may be affected. When negotiating salaries, a player's past performance is often taken into consideration. Consider a player whose agent has begun having contract conversations with new teams. A team with a defense-savvy coach might say something like, "Yes, you're a great shot blocker, but SportVU shows that you have issues defending players with more of a mid-range shot." Of course, the new data could benefit negotiations as well, but it will definitely add to the evaluation criteria that teams use to set contracts.
The NBA has already started displaying SportVU information during games for TV viewers and on its website for any interested fans. It will be interesting to observe going forward how SportVU changes the way that the fan understands the game. Big fans love keeping up with the latest news and analysis on their teams, and soon it may be commonplace to hear, "we match up with them well in the post, but I think where we will win is our ability to hustle in the fourth quarter…our average [SportVU] distance is much higher than theirs."
Referees have certain positions on the floor to which they must run in order to make the right calls. These spots are the same for all referees as set by the league, and SportVU is able to monitor how quickly the referee is able to get to these spots. Those refs that are not keeping up with the bar may have a more difficult time keeping their jobs.
Keep an eye out for another blog post on SportVU, where we will take some of the data that NBA.com provides and perform an analysis.