clock menu more-arrow no yes mobile

Filed under:

The Virginia Cavaliers' Individual Defensive Points Per Possession

This post ranks the Hoos by their defensive performance through the first nine games of the season, describing and using the metric of Individual Defensive Points Per Possession.


Basketball fans universally love long three pointers, between-the-legs passes,1 and thunderous dunks. For Tony Bennett, though, it doesn't get much better than a forced shot clock violation or an effective hedge on a ball screen. Bennett's Pack-Line defense has become the cornerstone of the improving Hoos. Offense is nice, but the Hoos continued success hinges more on stifling opponents than out-shooting them.2 We recently took a look at individual offensive performance and efficiency. But on a team that relies so heavily on defense, who are the Hoos best defensive players?

Steals, blocks, and defensive rebounds are the obvious defensive statistics. These metrics only tell part of the defensive story, especially for the Pack-Line that is designed to force difficult outside shots and abhors the risk inherent in attempting a steal or block.3 The main tenet of the Pack-Line is team defensive shape. An off-ball defender, by preventing lane penetration and forcing a pass, could have made the best defensive play of the possession.

Last season I developed4 a metric that attempts to better capture all elements of an individual defensive performance. Those who read my old blog or religiously followed Stat Geek Idol5 will be familiar with Individual Defensive Points Per Possession ("IDPPP").6 I'll explain the metric first, before presenting an individual defensive ranking of the Hoos through the first nine games of the season.

[Explanation of IDPPP]

Plus-minus is a deceptively comprehensive and somewhat controversial7 statistic that simply measures the difference in points scored and points surrendered while a given player ("Player X") is on the floor. This statistic captures such difficult concepts as Player X's effect on the team as a whole, but also includes extraneous information of individual teammate performances. Theoretically, in the long run, as Player X is on the floor with various combinations of teammates, the effect of teammate performances will diminish.

To turn plus-minus into a defensive metric, I thought: why not just delete the offensive portion? Track the total points surrendered while Player X is on the floor.8 Better yet, standardize this metric by dividing Player X's number of points surrendered by his number of defensive possessions, giving Individual Defensive Points Per Possession.

Now for the qualifier. The esteemed Ken Pomeroy9 wrote an article seriously undermining the widespread usage of plus-minus due to the excessive randomness contributed by its comprehensive nature. In other words, we love plus-minus' ability to capture player contributions not measured by basic statistics, but in doing so the metric includes so much that it is only meaningful in large data sets. Despite his arguments, I think IDPPP can still provide a useful measurement of Pack-Line prowess.

At the defensive end of the floor, based on Pomeroy's analysis, we can estimate a 2.5 point error per player per game in a 20 game sample.10 The Hoos play approximately 60 possessions per game. So when this analysis hits 20 games, the IDPPPs will have an expected error of slightly more than .04 PPP. I can live with that.

For this post, is nine games a large enough sample size to draw concrete conclusions about the precise hierarchy of Pack-Line defenders? Probably not. Are the outputted defensive rankings completely unrepresentative of defensive ability, and would we be better off randomly picking Hoos names out of a hat? Probably not. IDPPP through game nine doesn't tell us everything, but it does tell us something, and it will improve as the season wears on.

To measure IDPPP, I went through the play-by-play logs and tracked the players on-court, the possessions, and the points surrendered. An example from the George Mason game:


That chart creates the following individual breakdown, giving IDPPP.


Now that we're familiar with the metric, we can move on to the rankings of the current season.

[Hoos IDPPP through Game 9]

Compiling the defensive possessions and points from all nine games gives the following IDPPP ranking:


Bennett has been attributed with various versions of the quote "if you don't play defense, you don't play;" all aforementioned qualifiers duly noted, the top five players in the IDPPP rankings are the recent starting lineup. Further considering that many of the lower ranked players saw a higher percentage of the possessions against bad teams, the bunching of starters at the top emphasizes the stout Hoos defense.

The buzz surrounding Teven Jones as the second best defender on the team also seems warranted. We can't crown him first at this point, but he's at least in the top half to top third of the team. Atkins has received a lot of credit for locking down some marquee post players in the early going. IDPPP also supports his cause. Three freshmen and the one walk-on player who have seen minutes this year11 occupy the bottom of the rankings, and no freshman outside of Jones appears in the top five, reflecting the storied learning curve12 of the Pack-Line.

It will be interesting to see how the rankings change as the data set grows and the error decreases. Look for the freshmen to move up with more possessions against the remaining four lowly out-of-conference opponents, beginning with Mississippi Valley State on Saturday.


1 I'm looking at you, always-flashy Paul Jesperson

2 see Virginia 46 - 38 Tennessee

3 missing either can easily take the defending player grossly out of his precisely planned Pack-Line position (precisely planned Pack-Line position, precisely planned Pack-Line position)

4 I say "I developed" because I haven't seen it anywhere else. The chances that I actually came up with it, though, are very slim.

5 the religious followers from the limited pool of people who have even heard of it. And no, I didn't make it up.

6 and for their own health and interest level, should probably skip to the second section. Some of the first section is, ahem, recycled. #green

7 controversial in the "causing heated nerd arguments" sense. More on this later in the section.

8 effectively, "minus"

9 who is somehow both respected and a Tech graduate

10 Pomeroy calculates an average plus-minus error of 4.8 points per game for a 20 game sample. Since the offensive and defensive parts of plus-minus are just reflections of each other, they should have the same error over time. Since both contribute equally to plus-minus, the error on each end of the floor should be half of the overall plus-minus error, or roughly 2.5 points.

11 I ignored Thomas Rodgers and his six minutes of playing time. Sorry Thomas.

12 hmm, post idea?