Machine Learning with the NBA’s “Shooting Machine”

Aris Papadopoulos | Published in Towards Data Science | Sep 5, 2018

Shortly after Kobe Bryant retired a couple of years ago, Kaggle released a dataset containing 20 years worth of his shots. That’s a lot of swish and I am thinking: Basketball and data science… What more can you ask?

So here is an article on how to build a simple classification model to predict if it’s in or rim. The aim is to discuss the intuitions and the practices one can leverage end-to-end, step-by-step, from data exploration to model tuning and finally evaluation. Emphasis also on simple. A decision tree (that’s what we are building) will not win you a competition but the process of building a simple model might also be 80% of the whole work in real-world situations.

https://youtu.be/gi8hXrfGsm4

Bryant was one of the most important players in the history of NBA. He played for a full 20 years with L.A. Lakers. He is the all time top scorer of the team and he is considered by many to be the second best shooting guard in the history of NBA behind Michael Jordan. Watch ten of his most impossible shots. They are most probably included in the dataset used next.

Important Note

From this point, I am omitting the author's text of this article and am just pulling out two graphics that relate to this section of the class. The article itself is quite technical with coding instructions using Pandas, Jupyter Notebooks, and Scikit-Learn, and instructions for making visualizations with Seaborn and Graphiz. The author also utilized Tableau and Python. Feel free to check out the original article if you want to geek out with the details--it's fascinating stuff!

The original post has more graphics that are pretty straight-forwad, so feel free to check them out. For this reading, I want to focus on the scatter plots below, which were made using real NBA data.

Scatter Plots of Kobe's Attempted Baskets

For all the graphics below, blue designates a successful shot, meaning that he made the basket and got the point, and red designates an unsuccessful shot, meaning that he missed the basket and didn't get the points.

Graphic 1.

The author uses Pandas to make a scatter plot of each of Kobe's shots surrounding the net. Observe that the two plots look like mirror images, but they are two different analyses. If you put these two together, it's like seeing both sides of the basketball court.

scatter plot of Kobe's shots in red and blue

More Graphics

The author uses Seaborn to make a set of more scatter plots using the same data. I don't fully understand what they represent, since the descriptions are all about the coding, but I can infer a little bit based

a array of 20 scatter plots

In summary

Notice that scatter plots may seem very simple on the face of it, but they are an important tool for getting across complex information in an easy-to-understand manner. BUT to do that requires a lot of complex coding. That is why data science and statistics are such a huge majors. Note: data science is such a new field that it's often lumped into computer science or mathematics.