Data analysis is only as useful as the data you collect. That’s my core belief when it comes to the “field” of data analysis, or data science, or statistics or whatever it’s called now.
The quality of publicly available basketball analysis has always frustrated me, because the underlying data doesn’t wholly capture the necessary data to draw solid conclusions. For example, who is the best passer? Use assists. But assists only capture passes of made shots in the field. What about passes on missed shots, or on fouls? And when we compare across teams, this potentially biases toward players whose teammates may be better scorers.
We currently have numerous “advanced” stats which do fancy weightings of box stats to estimate the things we care about. These work well as an first pass analysis, but what if we want to go further?
I believe the answer is not to throw more statistical modeling and/or machine learning techniques at the problem (sorry, RAPM), but to collect more, different, better datapoints.
Turns out, collecting data (“scorekeeping”) in basketball is pretty difficult. To make it less difficult, I started building a tool which I call the Game Logger.
Vorped Game Logger
The Game Logger tracks both play by play and location data, and derives a boxscore based on that data. It differs from existing scorekeeping tools in a couple ways:
- Logs the location of any event (fouls, rebounds, etc.), not just shots
- Track per-possession stats, not just field goal percentages
- Be flexible enough to define your own play by play events – you’re not limited to points/rebounds/assists/blocks/steals
If you’ve ever attempted scorekeeping a game, you know a basketball game moves very fast, sometimes faster than you can log data. The Game Logger tries to minimize this issue with a few features:
- Text-based and keyboard-based – no hunting and clicking buttons to log the right play and player
- Enter multiple events at one time
- Easy to fix/update past events
Using a text input lets you spend more time watching the game instead of your screen. I found click-based systems problematic when scoring a game in-person. It also allows you to track as many events types as you wish, now that you’re no longer constrained by your device’s screen space.
You can try it out here: http://vorped.com/1-nba/2014-2015/game_logger. Click Learn More on the front page for a short tutorial on how to score the basics of a game.
You may be wondering, why is this useful at all? After all, Sportvu tracks all location data (literally), and Vantage Sports tracks almost every event in an NBA game, why would anyone use this?
To which I answer, why not track *any* basketball game? Why not track your college or high school basketball team’s games? You can. Many people ask me how they can create their own shot charts, and now, you can use the Game Logger to do so.
In fact, Game Logger was designed to be used live at a game. The actual Game Logger application is written as a Chrome App, which you install through a Google Chrome browser, which allows you to log games without being on the internet.
As with all things, there are many, many bugs and missing features, so please feel free to give me your feedback here or on Twitter.