But I wondered if we could quantify this impact for college basketball players. What’s the numeric impact of going from 20% of the team’s shots to 25%? Thus this summer, I spend a little bit of time looking at Ken Pomeroy’s tempo free player stats. I hoped to compare the percentage of possessions (%Poss) to the player’s individual offensive efficiency rating (ORtg) and quantify the impact. But the results were not as clear as I hoped.
Allow me to step back for a moment. If you look across all players, there is actually a positive correlation between shot volume and player efficiency. The good players get to shoot more. Thus in any study of this type, we have to look at how players change over time. And while college basketball has fewer time observations, (players have at most four years of observations), there are a lot more college basketball players to follow. So I was hopeful that there would be enough data to find some interesting results.
In general, I found for each additional 1% of possessions taken, efficiency changed by –0.25. Thus if a player used 24% of his teams possessions instead of 20%, his ORtg would be expected to fall only 1 point. Unfortunately, the results were not very robust. By varying the sample or the functional form, I could get the result to be positive or as high as about –0.80.
Here is a graph that may show the difficulty with this. This lists players who were top 100 recruits in recent years. Each player’s %Poss is listed on the x-axis, and ORtg is listed on the y-axis. The lines track the changes for individual players over time.
As you’ll note there are some lines that do slant downward. There are cases where players shoot more and end up shooting a worse percentage. But there are also many cases where players who shoot more actually make a greater percentage of their shots.
This leads me to believe I’m making a big mistake using annual college basketball data. The problem is that college basketball players are not likely to be equivalent from year-to-year. College is a key time for player development and skill improvement. I can control for the player’s year in the program (freshman, sophomore, ect.), but improvement is not uniform across players. And because improvement is not uniform, we really end up with the same problem we initially had when we compared players. The players who develop the most in the off-season (the players that show the biggest improvement in efficiency), are the exact players who are allowed to shoot more. Thus even within players, we will often find a positive correlation between usage and efficiency.
This leads me to believe I’m making a big mistake using annual college basketball data. The problem is that college basketball players are not likely to be equivalent from year-to-year. College is a key time for player development and skill improvement. I can control for the player’s year in the program (freshman, sophomore, ect.), but improvement is not uniform across players. And because improvement is not uniform, we really end up with the same problem we initially had when we compared players. The players who develop the most in the off-season (the players that show the biggest improvement in efficiency), are the exact players who are allowed to shoot more. Thus even within players, we will often find a positive correlation between usage and efficiency.
One way to get around this may be to focus on players who were regulars in all years, or who took a lot of shots in all years. And when I focus on these samples, I do get the larger negative impact discussed above.
But there are other problems to think about when using this data. Should we even expect a uniform decrease from additional shot volume? Is the marginal shot a forced three pointer or a jumper in the lane? Is the difference between 10% of a team’s possessions and 11% really the same as the difference between 29% of a team’s possessions and 30%? Players who have large changes in shot volume may be the best way to measure the overall shape of the shot distribution path. But players who show large changes in shot volume are the exact players who developed the most.
I guess this all leads me to conclude that the NBA may be a better place to study shot volume and efficiency. Certainly after the first few years, NBA players will not have the same major swings in development. Moreover, there may be some fun quasi-experiments in shot volume, as Kevin Pelton mentions here. Kevin also kindly points to this older post which summarizes the usage and efficiency discussion at length. Here is my general take on some ideas discussed in the thread. A lot of people will see a player with a high efficiency and say that player should shoot more, but that may not always work in the offense the team is running. If you have an immobile guard who is a spot up shooter, he may make a lot of wide open threes, but he might be terrible if he was asked to take an additional shot. Similarly, a player like Chase Budinger may not be the most efficient in the country, but he drew so many double teams with his shot volume that he still made his teammates better. I guess this is my way of saying that I think coaches are rarely idiots. And if a player’s efficiency seems out of line for their shot volume, there is probably a reason.
Even though my quest to quantify the impact of usage on efficiency is not definitive, because bias from player development appears to be the biggest problem, my guess is that the average effect is at least -0.25 and probably much larger.
Blatant Plug #2
While working with Ken Pomeroy’s player data, I was able to tabulate some fun descriptive statistics. For example, what’s the distribution of offensive efficiency for freshman? What’s the distribution of offensive efficiency for freshman top 100 recruits? What’s the average change from freshman to sophomore year? And so on. And if you want to read it, you can find it in the Basketball Prospectus 2010 College Basketball Preview. I was honored that John Gasaway asked me to write the Big East preview this year.
Make sure you read John’s article on experience and team performance. You may remember I had a series of posts on this last year, culminating here. I remain skeptical of the importance of experience, (because the most talented players rarely stick around), but John’s writing on the topic is starting to convince me otherwise.
Also, possibly because I am so skeptical of experience, I was perhaps the perfect person to write the Big East preview this year. That’s because almost all of the Big East teams are young this year. If you can find someone else who more enthusiastically wrote about a young Providence team, I’ll be surprised. (Believe it or not, I included a lot of stat factoids in the article on Providence, even though only three rotation players return this year.) I highly encourage you to head over to Basketball Prospectus and download yourself a copy of the whole book.
Finally, I want to acknowledge the folks over at Big Ten Geeks who are tackling some of the exact same topics. (Seriously, I did a double take when I saw they looked at the change in efficiency from freshman to sophomore year, ect.) And I want to acknowledge Villanova by the Numbers who has slowly been unveiling another Big East preview in tempo-free style. If you are interested in my Big East preview, you'll probably be interested in those as well.