- Neil Paine
You might not have felt it at the time, but the world of basketball statistics experienced a seismic shift in the early 2000s.
For the entirety of NBA history up to then, the only stats a talent evaluator could use to measure player performance came from the box score, a fundamentally flawed collection of numbers that all but ignored defense and contained categories not modified since 1977.
But roughly 10 years ago, a statistical revolution commenced, seeking to bypass the box score entirely. Working for then-new Dallas Mavericks owner Mark Cuban, mathematicians Wayne Winston and Jeff Sagarin developed a proprietary player rating called WINVAL, widely considered the first Adjusted Plus/Minus model to be used by an actual NBA team.
The revolutionary part of Adjusted Plus/Minus is that it is generated entirely from play-by-play data, which eliminates the need for the box score's distorted system of credits and debits. If a player sets tough screens, is a great defender or does any of the other proverbial "little things" well, it's captured by Adjusted Plus/Minus. The stat caught the attention of NBA front offices because it has the potential to truly be a total player rating, and one free from bias.
Although Adjusted Plus/Minus can suffer issues when sample sizes are small or if the same group of players tend to play all of their minutes together, over the past decade statisticians have developed creative ways to minimize these effects. The most recent public example is Regularized Adjusted Plus/Minus, which incorporates weighted numbers from previous seasons and is currently the most accurate version of the statistic for predicting results.
A couple of weeks ago, Daniel Myers had discussed this in his blog. We follow suit here, continuing the discussion of quantifying defense.
I was recently provided (thank you Jeremias Engelmann) a 12-year set of Regularized Plus/Minus data that includes every regular-season NBA game from 2000-01 through 2011-12. What follows is a list of players for whom conventional perceptions might need amending. We can say with quite a bit of certainty that these "underrated" or "overrated" players are doing something that isn't showing up in the raw box score numbers, for better or for worse.
Neil Paine identifies a list of underrated and overrated players using adjusted plus-minus statistics.