Rating Perimeter Defenders; Tony Allen Still King

By Steve Shea (@SteveShea33)

February 1, 2017

It has always been more difficult to quantify a player’s defense than his offense.  Box scores offer little beyond blocks and steals.  While many players play 60-70 defensive possessions a game, most only accumulate 1 or 2 steals or blocks.  What is happening in the other approximately 95% of possessions?

Each season, NBA offenses are becoming more perimeter-oriented.  Unquestionably, the 3-point shot is a huge component of today’s game.  In this modern landscape, it is all the more important that analytics be able to assess perimeter defense.

But how?  There are two approaches one can take.  One approach is a top-down metric.  Top-down metrics quantify individual performance by looking only at full lineup production, such as plus-minus or adjusted plus-minus. The motivation for using a top-down approach for defense is that defense is a team endeavor.  Four players can be playing great d, but if one player gets caught ball-watching and loses his man on a backdoor cut, it can lead to an easy layup.

Unfortunately, even when trying to adjust for quality of teammates, top-down metrics for the individual can be heavily pulled by who the player plays with.  The perimeter defender that has the benefit of help defense from a great rim protector tends to look better.

Alternatively, one could use a bottom-up approach, which is to use statistics based on individual actions.  In this case, that would be steals, fouls or something of that sort.  Historically, there have been very few such bottom-up stats that pertain to perimeter d, leaving a purely bottom-up approach impossible.

Fortunately, with recent improvements in data gathering (including spatial-tracking data gathering), a bottom-up approach becomes more feasible, and this is the approach we’ll take here.  No, we will not use defended FG% as presented by NBA.com.  This metric looks at opponents’ FG% when each defender is the closest.  Krishna Narsu recently explained why this is troublesome, especially on 3-point shots.

Instead, we’re going to use new metrics from NBA.com that get at some of the nuanced actions on defense, including deflections.

In this new metric, called Perimeter Defense Rating (PDR), we’re approximating a player’s production as a perimeter defender.  When measuring production, it’s always important to consider context.  This metric cannot account for important components of a player’s context on defense.  Perhaps the most important factor is the defensive scheme.  Certain schemes put players in better positions to succeed.

PDR should not be viewed as purely a measure of a player’s intrinsic skill set as a defender.  It is largely influenced by a player’s skill set, but there are also extrinsic variables.  It is possible that a player’s PDR is enhanced or diminished by the system he plays in.

Below are Perimeter Defense Ratings (PDR) for NBA players as of January 31, 2017.  We only included players that played at least 800 minutes and excluded players that are typically classified as centers.