Introduction
Given the name of the blog, its fair to say that I am a huge NBA fan. I have been watching the NBA for over 2 decades and saw first hand the rise of analytics in basketball. Statistical modeling has greatly impacted basketball strategy from drafting players all the way choosing what play to run. However, there is still a long way to go as there are many problems with the current state of basketball analytics. They often times only paint half the picture, and people just run with the result, rather than taking a moment to reflect on the short comings of these stats.
Hot Hand Theory
The best example I can think of where statisticians blindly followed the data and jumped to a conclusion is the hot-hand theory. The hot-hand theory is when a person experiences a successful outcome and then has a greater chance of success in future attempts. In basketball, it means there is positive auto-correlation with shot-making percentages. I am more likely to hit the next shot if I have already made a few, whether this is do to confidence, focus, or some other unexplainable reason.
For years, statisticians and media members quoted a paper by Gilovich, Vallone, and Tversky which claimed that the hot hand did not exist. By looking at sequences of shooting, the authors concluded that the various 'hot hand' streaks were nothing more than random sequences of chance.
Players and coaches were shocked at this conclusion. From a person who has been playing basketball at various levels since I was 5, I could not accept this conclusion either. I was doing mental gymnasts trying to reconcile what I was feeling and the data. This was a very controversial debate with people who played basketball on one side arguing against those who analyzed the data. Most of the basketball community had accepted the paper as gospel and the hot hand was renamed the hot hand fallacy.
Finally, in 2014 the NBA got tracking data and the hot hand fallacy was debunked. A paper using tracking data was able to control for shot difficulty and disprove the hot hand fallacy. Players who are 'hot' shoot their next shot from further away and face a tougher defense, lowering the expected value of the shot. The conclusion is that players who are outperforming will continue to do so, conditional on the difficulty of their present shot.
The Blog
Similarly to the hot hand fallacy, our current statistics are have flaws when trying to describe the game, which leads to people often drawing wrong conclusions from the data. When people analyze a game, they often look at the box score which will often have around 15 categories, ranging from minutes to FG to points. I wish to tackle basketball analysis from first principles and examine the pros and cons of each box score statistic. I will do deep dives on new statistics that are important in describing the outcome of a game and show how these new metrics can help create optimal game play.