Statistical Analysis of Jump Shots

Throughout the spring semester, I have been working with the Women’s Basketball team and the Sports Lab to try and create a program that will provide statistical analysis of the jump shot to help develop players within their program. Google’s media pipe has been extremely useful in progress towards this goal. Their program takes photos or videos, detects the person, and overlays 32 landmarks onto their joints, as you can see in figure 1. The first statistic I wanted to measure was elbow angle. I found the angle between the shoulder-to-elbow vector and the elbow-to-wrist vector. From there I plotted the change in this angle over the jump shot. I used a similar process for statistics such as leg load (how deep you get into your shot)  and forearm alignment (how perpendicular to the ground your forearm is), and hope to continue to add more trackable measurements. From there I recorded myself shooting 40 jump shots, manually tagged each one as a miss or a make, and then ran each of them through the video pose estimation, plotting the makes in green and the misses in red. As you can see from the graph on the left, this becomes very noisy. I wrote a function to average these lines based on make/miss, leading to a much easier-to-digest graph. 

Figure 1: An example of the pose estimation keypoints on my jump shot.

In the future, I hope this program allows players to build a “profile” for their jump shot. Once enough data is collected, we will have a pretty good idea of what your joint angles look like when the shot goes in. Then you will be able to more easily pinpoint the problem if you go in a slump and overall lead to more consistent shooters. To do this I need to identify the statistics that impact makes and misses, and I have written a program that takes in the data in the graph on the right and calculates the standard deviation at every point in time. From there I can calculate the chi-square to try and tell if there is a statistically significant difference between the make and miss graph. This will hopefully provide insight into what factors make a shot go in.

About the Author:

My name is Wyatt Mowery, I’m from North Carolina and am wrapping up my freshman year at MIT. I am a member of the Men’s basketball team and am majoring in Mathematics and Computer Science. 

Contact: wmowery@mit.edu