Future Basketball StatisticsPosted: March 12, 2013
I have no doubt that NBA teams probably have sophisticated (and proprietary) databases and technically-savvy analysts looking at all kinds of basketball data to assess strengths and weaknesses of teams and individual players.
But as an outsider looking in, I was thinking about what areas of data analysis in basketball seem most promising for the future, given that we are now in a “big data” world.
First, basketball is really hard to analyze because there are countless variables interacting with each other all the time. In baseball, you can analyze a pitcher and a batter and ignore everyone else, basically, without losing too much information. In basketball, there are always people active on the court beyond just the shooter. There are help defenders, there’s the guy that fed the shooter the pass, there’s the guy that fed the guy who fed the guy the pass, there’s the guy that set the screen, and there’s the guy on the opposing team in the paint who may be deterring a drive that was never attempted. So it seems like focusing on problems of interacting variables is something that should be given more thought than it has been given thus far.
Second, basketball is spatial, so it would be interesting to do more in this area. Already TV networks visually show shot selections for various players, but I don’t get the sense that it has gotten much deeper than that.
Third, as I alluded to in a previous post, people look at averages but not so much at the variability of players’ performances. Getting some measure of how consistent players are seems important in correctly assessing value.
Finally, analyzing injury impacts on players in a statistically-rigorous fashion would be interesting, and probably really helpful for teams who are considering signing players who are just coming off injuries. Andrew Bynum is one example of this. A database of ACL-injuries, pre and post-injury, and whether things improved over time, or whether you could predict the quality of a comeback based on the first 10 games, for example, seem like interesting questions to try and answer.
Some questions can be answered as more and better data are collected, but other questions are simply hard to quantify and answer with data. Quality of coaching seems like one area in particular where this holds true.