This article is part of our Nerd Alert series.
Back in the final week of February, I took a look at how often NBA players were getting their shots off at the rim. Because there is undoubtedly a little bit of luck in shooting a basketball - these mechanical acts are only so repeatable by human minds and bodies - it's better to look at something the player does control: where he takes the shot. Or, at least, so the theory goes.
It's tough enough to predict what will open over a full season. Limit your scope to one-half of a season and allow a few random acts to play a much bigger role and the task becomes even tougher. But that is essentially what this analysis attempted to do.
Overall, the data did a decent job. Although I didn't exactly present the data in this form, the suggestion was that we can project field goal percentage for the rest of the season given last season's field goal percentage and the change in the player's at-rim shot rate. The formula the data gives is as follows:
Expected FG percent = [Last Year's FG percent] + 0.237*[Difference In At-Rim Rate] – 0.008
And it turns out this does a reasonable job of predicting field goal percentage, with a coefficient of determination of .575 – that is, 57.5 percent of the variation in second-half field goal percentage is explained by these two variables.
So I wasn't selling you snake oil, at least. But here's the real question: is this really an
Back in the final week of February, I took a look at how often NBA players were getting their shots off at the rim. Because there is undoubtedly a little bit of luck in shooting a basketball - these mechanical acts are only so repeatable by human minds and bodies - it's better to look at something the player does control: where he takes the shot. Or, at least, so the theory goes.
It's tough enough to predict what will open over a full season. Limit your scope to one-half of a season and allow a few random acts to play a much bigger role and the task becomes even tougher. But that is essentially what this analysis attempted to do.
Overall, the data did a decent job. Although I didn't exactly present the data in this form, the suggestion was that we can project field goal percentage for the rest of the season given last season's field goal percentage and the change in the player's at-rim shot rate. The formula the data gives is as follows:
Expected FG percent = [Last Year's FG percent] + 0.237*[Difference In At-Rim Rate] – 0.008
And it turns out this does a reasonable job of predicting field goal percentage, with a coefficient of determination of .575 – that is, 57.5 percent of the variation in second-half field goal percentage is explained by these two variables.
So I wasn't selling you snake oil, at least. But here's the real question: is this really an improvement over the basics? Is it better than shooting percentage at midseason? What about last year's shooting percentage?
Unfortunately, the answer is no:
Both this season's FG percent and last season's FG percent have nearly the exact same amount of predictive power as using last year's FG percent with the shot location data as well.
What does this mean? Most likely, it means we can't just assume these stylistic changes in players are real after 25 or 30 games. Just because a guy gets to the rim early in the season doesn't mean teams won't adjust and employ strategies to keep him away. Maybe a player will come back and change his role – see Rudy Gay in Memphis with Zach Randolph returning, for instance. Still, 40 of the 82 players this sample is examining shot within two percent of their expected field goal percentage.
And that's why this data is just another tool, not a be-all end-all. It takes close examination to figure out which players have really changed their style and which haven't, and we need to keep an eye out for changes within the season as well. I still feel this data can be very useful, but it appears it needs to be taken on a case-by-case basis, and not used as an absolute.
