Interactive graphic plotting each shot on the court
Flipping through the options, there’s a couple observations one can make. There’s the potential to identify or verify how play style changes to accommodate differing players on opposing teams. For instance, Kobe took about the same number of shots against the Pacers and Pistons, but the Pacers graph shows a much denser cluster of layups. If I knew anything about those teams, I might be able to guess why. :)
This is super interesting. Baseball is a game played by statistics, as most people know, but no one ever really talks about the possibilities for other sports… or at least, I never hear about them.
Basketball’s covered a bit at the data-journalism sites that care about sports, like fivethirtyeight.com. For example, this article.
Much of the really interesting data wasn’t available until the past few years though. It’s too tedious to collect by hand, so only started becoming available after the NBA installed a system that uses a camera array to track players and shots. Even then not much was done publicly with the data until people discovered around 2011 that a bunch of it was available for scraping, and started doing the modern “data journalism” type visualization and analysis.
Some of the especially useful data, like whether shots were early/late in the shot clock, how close coverage was on the shot, etc., is even newer, only collected since 2014.
As of the beginning of this season, the nba npm package was the best I found for retrieving statistics from the NBA. The stats.nba.com site uses Angular; you can see a full list of endpoints if you search for “var routes” in http://stats.nba.com/scripts/custom.min.js. There’re some helpful starting points in Python out there as well (see @alphacc’s comment below).
Hockey is also going through a statistical revolution right now. I think baseball came first because it was a lot easier to record most of the raw data. But the other sports certainly seem to be catching up.
In 2008 on a contract I spent lots of time pouring through sports data. Outside of B*Ball there were plenty of stats for other team sports such as NFL, Hockey, Cricket, Football(soccer).
Really the question is how deep do you actually need to go? There is the obsession over having the most granular stats ever, but honestly they are not really needed. For example with NFL - do you really need to know the exact position of every player on the field at each second, or the exact vector of the throw? Probably not. I look at this chart for Kobe and I think to myself, that its just lots of noise!
I look at this chart for Kobe and I think to myself, that its just lots of noise!
Maybe… But, let’s not look at a single player, but all shots against a single team. As @tedu points out, Kobe, against the Pacers, shot more lay-ups. Is this common for all players against the Pacers? And if so, should the Pacers change their defense to accommodate? If the field goal percentage is higher there, than maybe?
Disclaimer: I haven’t followed sports since 1997, nor am I a statistician.
FiveThirtyEight has a pretty interesting post on this also from last fall http://fivethirtyeight.com/features/an-ode-to-kobe-bryant-in-two-charts/
A colleague found this, I though I would share :
I didn’t know who this was so I assumed this was going to be shots fired by some LA gangster.
I like the gap around the outside where he purposely moves out or shoots early an extra few centimentres to get outside the 3 point line.