This project is licensed under the terms of the MIT license.

A python script made to create shot charts for any NBA player season or career (from 1996 to present).

These charts are primarily an extension from the work of Dan Vatterott and Savvas Tjortjoglou. I made an attempt to combine my favorite aspects of other shot charts on the internet (primarily from: Kirk Goldsberry, Buckets (by Peter Beshai), BallR (by Todd Schneider), and Swish (by Austin Clemens)) into one image, while also adding some additional features I thought would be nice in order to better distinguish the strengths and weaknesses of a player.

To run this code, download the script and player_list and put them in a directory and change the directory path in the script to your directory. To call a single player you can run the following command: python nba_shot_charts.py --player_name "Zach Randolph". If you wish to run every player in the script, you can just run: python shot_charts.py from your directory (this will take multiple days to complete).

Two lesser noticed features of these charts are the following:

• Each chart’s name will save under the template shot_charts_Firstname_Lastname_Season_eFG.png
• The color of the Player Name and Zone Frequencies key is the corresponding color to their overall eFG% for the season (e.g., if a player’s eFG% for a year was 100%, their name and key would be dark red, while if their eFG% was 0%, then those colors would be light yellow).

If you're interested in seeing more charts, I created a twitter bot (@NBAChartBot) that posts a pseudo-random player/season shot chart as a scheduled cronjob.

If you would like to report a bug, request a copy of the data, or contact me for any other reason, my contact information is at the bottom of my personal github.io page.

I'm changing the code structure a bit. I've added scrapers and some post-processing scripts so I can store/manipulate the shot data a little bit easier in a MySQL database. I'm leaving the original chart generator (nba_shot_charts_v1.py) and chart bot script (chart_bot_v1.py), but will eventually add a v2 of both of this scripts, using the MySQL db instead of re-scraping nba.com each time. Part of the reason I'm changing the structure is to help facilitate a few longer term goals of these charts:

• Speeding up the process of generating a chart (querying the database is faster than scraping the web)
• Creating some form of an era-adjustment, to be able to more accurately compare charts from players from different seasons (think of how a player like Ray Allen's shot chart may differ if his career started in 2015, or how a player like Patrick Beverley's may have differed if his career started in 1990).
• Generating team shot charts similar to how we current generate player charts.
• Generate charts for playoffs.
• Filter shots by date (create shot charts for a player/team from date X to date Y).
• Calculating some form of a similarity score, to see which players/teams have the most similar distribution and shot making ability from some pre-defined zones.
• zEFG% baselines for guards/wings/bigs at each zone.
• attempt to create a Points above Replacement for players based on position.
• Generate G-League and WNBA charts.

The new chart_bot.py and nba_shot_charts.py are active. I've changed the color scheme/added new metrics over the old charts (the old charts are archived in archive_scripts). Using the db has been a vast improvement over previous methods, and has allowed me to create some shot/zone-based metrics I didn't have previous access to. Next up will be making team charts and similarity scores.

I'm now able to create/post charts at the team level and I also cleaned up the format of the player level charts.

Creating custom (filtering by date, playoffs, multiple players, etc.) charts is now possible. Although it is very slow (multiple minutes per chart, as opposed to a few seconds), since it has to re-calculate every metric as opposed to pulling from a database.

Slowly getting back into this after the offseason; the random chart selector has been improved for the bots, as well as capacity to tweet 280 character tweets and the option to thread tweets together.

I’ve again decided to put my to-do list from last March on hold, and have instead focused on re-building some key features of the charts including:

• Restructuring the codebase (so it’s not such a mess).
• Adding other details to tweet text, including binned text descriptions for Volume, Shot Making, and Efficiency.
• Adding player pictures to the custom_charts script.
• Then ideally I would write an article better explaining the charts/metrics/methodology that I could post to my website and make my pinned tweet, making it easier for new followers to understand the charts.

Finally, after I’ve completed this, I will return to my original to-do list, and finish the following items:

• Calculating some form of a similarity score, to see which players/teams have the most similar distribution and shot making ability from some pre-defined zones.
• zEFG% baselines for guards/wings/bigs at each zone.
• attempt to create a Points above Replacement for players based on position.
• Generate G-League and WNBA charts.

Well I didn't get much done from the to-do list in December. Oh well, so it goes sometimes. Unfortunately, I will be putting this project on hold for the summer, as I will be out of the country and won't have time to tinker with these features. Hopefully when I get back I'll have some more time to work on the features I wanted to back in December. In the meantime, I wanted to jot down a few lessons I've learned over the past year-plus working on this project:

• Start small, and focus on writing good code. On this project I dove in too deep at the beginning, and now there is a somewhat intimidating amount of ugly code that I have to sift through to change anything.
• Write better comments that explain why you chose to do something a certain way.
• The "Never Write the Same Code Twice" rule is important. I ignored this when I started the project and it's created a spidery mess in the codebase
• Functions should be robust. My early functions served a single purpose, and then if I wanted anything slightly different I would have to write a new function instead of being able to specify the data I wanted via an argument call in the function.
• Overall, this project has been great and I hope to return to it after the summer. It's greatly helped my skills as a data scientist; I've learned a ton regarding data visualization, negotiating API calls, web scraping, text parsing, and sql database management.

I gave the charts a much needed update, and wrote about it too. Additionally, I’m moving all future updates to my personal website/blog, but will leave this README file up for historical purposes and will try to post links to any updates here as well.