In my previous articles, I have presented some relatively simple (and necessary) alternatives to official, box score statistics.  Unfortunately, measuring individual defensive performance is anything but simple.

Indeed, Ken Pomeroy wrote the following at the start of his 1/8/07 article in ESPN Insider: “One of the most difficult things to evaluate in college basketball is individual defense.  Defense is more team-oriented than offense, and even in 2007, stats for measuring individual defense are primitive.”

Measuring individual defensive performance is definitely a challenge, whereas team defense is best measured by points allowed per possession.  (This is better than defensive field goal percentage since a team could commit many fouls during a game, including some fouls when a player is attempting a 3-pt. shot.)

Number of blocked shots per game is often used as a measure of defensive performance by a team’s big man, but of course we would also want to know how many of those blocks were due to the ball being swatted out of bounds versus being in play.

In the statistical design of experiments, one commonly used design in industry is a one-factor-at-a-time design in which one factor (say, temperature) is varied while the other factors are held constant.  Then, if we have accounted for all of the relevant factors, the change in the response variable (e.g., process yield) is due to the change in temperature.

Unfortunately, we cannot easily apply that idea to basketball because if we tried to measure a player's defensive contributions by charting every statistic that seems relevant, both when the player is in the game and when the player is out of the game, the presence-absence effect of the player is not being measured independently of the performance of the substitute, and also not independent of other factors since the other team will also be substituting, the pace of the game may change, etc.  In other words, the other factors are not being held constant.

It would be very difficult to incorporate everything that is important into a single statistic, although attempts have been made to do so in the context of the plus/minus system that has been used in hockey, major league baseball, and professional basketball, in particular.

Neither box scores nor the full play-by-play will tell the complete story regarding defense.  For example, the Georgia Tech men's coaching staff charts “deflections,” which is the sum of tipped passes, steals, charges, and offensive rebounds, both for the team and for individual players, with a goal of having at least 35 deflections per game.  (Note that more than one deflection could occur on a given possession, as a tipped pass could result in a steal.  This would count as two deflections in their system.)  In a way, an offensive rebound is a “steal,” so the additive components of deflections do essentially fit together.

Obviously a tipped pass won't show up in either a box score or in the play-by-play, however, nor will limiting a player's touches by doing a good job of denying the ball.  A player's contribution to team defense, such as when a half-court trap is employed, is also very important and cannot be easily measured.

The bottom line is that any statistical measures that are devised to measure the defensive performances of individual players will be far from perfect.  More research is needed (as the Pomeroy quote implies) because of the importance of being able to measure or at least approximate defensive contributions.  Until better measures are devised, the defensive performance of individual players can best be gauged from watching the game and from studying game tape, as well as by charting statistics such as deflections.