The NBA reviews all "officiated events" in the last two minutes (and OT) of games that were within 5 points with 2 minutes left in the game. All calls and "notable non-calls" are reviewed under the same standards as the NBA's instant replay. These reviews include the time of each call, the call type, the committing player, the disadvantaged player, and the review decision. Specfically, the review decision states whether the call was a correct call, a correct non-call, an incorrect call, or an incorrect non-call. And, the NBA publishes these "Last Two Minute Reports" (L2Ms) online as PDFs--not exactly the friendliest format.

Nonetheless, the L2Ms are exteremely interesting! Their content can be used to examine league-wide refereeing priorities, potential biases against teams and players, and referee performance (that's just a short short-list). Even if you aren't interested sports, let alone basketball, these are questions of fairness and equality within a hugely influential organization. Inarguably, the NBA has taken a leadership position in terms of organizational transparency by putting itself, or more precisely, it's referees, under the microscope.

So, I scraped all the pdfs ¹, 1,102 in total, containing 18,244 calls at the time of this writing (Dec. 3rd,2017).

To use this data to assess referee performance I also pulled the referees for each corresponding L2M. With these two datasets in hand, I calculated the number of incorrect calls and non-calls per L2M for each referee with more than 20 L2Ms and plotted the results below.

Interactive visualizations are best rendered on desktop.

So, is Mark Lindsay the best referee in the NBA? He could be. Limiting the number of incorrect calls is a huge part of being a good ref, but it's not the only thing.

One point of interest is that Lauren Holtkamp--currently the only female referee in the NBA--scored well compared to her peers. Why's that interesting? Google her. You'll find a Reddit thread titled "Can we talk about how bad of a ref Lauren Holtkamp is?", a fair bit of other questionable content, and a 2016 player and coach survey ² from the LA Times that named her as one of the three worst refs in the NBA. The public perception of her, does not line up with the numbers.

Notes

1. To gather the data in each of the PDFs, I used pdfquery, which has a handful of useful features for scraping pdfs. For example, you can find the x and y coordinates for all instances of specific string. And conversely, you can provide a bounding box within which to pull content. Using these tools, I was able to identify rows and then scrape data following the columnar structure of the L2Ms. In combination with nba_py (a python API for stats.nba.com) I filled a few gaps (e.g., the players' teams) and grabbed the referees for each L2M.
2. The survey from the LA times had a sample size of "nearly three dozen" individuals (either players, coaches, or assistants). "Three dozen" responses is laughable. The NBA has 30 teams, they maybe recieved one response from each team (I doubt it). How'd that ever get published?

Check out the code on Github and feel free to reach out to me on social media.