Defensing A Team's Top Option

UCONN'S DEFEAT OF DOOK IN THE NCAA title game showed a couple things. First, it showed that Dook Sux (a required statement from any North Carolina grad). Second, it demonstrated that Dook could be beat by shutting down their All-American center, Elton Brand. The second part is the main focus of this article because the first part is too obvious to be worthy of an article.

It was reported in Sports Illustrated that Connecticut's strategy for beating Duke in the 1999 NCAA title game consisted of 2 principal things:

1. Double-team Brand when he got the ball, not allowing him a lot of scoring opportunities, and
2. Run a lot, on offense and on defense

How would UConn know to use these strategies? The story is that they watched tape for hours before the game and saw something. I believe that to a significant degree because Jim Calhoun is a good coach and because I would have believed the strategies a priori, especially the first one. But coaches are also often wrong, coming up with strategies that don't work. Is there any way to use numbers to check whether such strategies work?

Specifically, is it possible to determine whether Duke is best shut down when limiting Brand's possessions or points? Uh, it's not possible for me because I don't have Duke's boxscores for 1999; if I did, I would definitely look. I don't think I'd find anything conclusive, however, because I took a similar look at NBA stars....

...Does Stopping an NBA Star Stop His Team?...

The basic question a team has to answer in prepping a defense for an opponent is whether to stop the main weapon or to stop the supporting cast: "Stop Jordan or stop the Jordanaires?" as they used to say. But how can this be phrased so that it is testable? What is "stopping Jordan"? (I know, you can only hope to contain Jordan.) What is stopping the Jordanaires? How does one measure success/failure of a strategy?

For the purpose of preventing a ranting rambling article that talks around the subject, I'll do the scientific thing and come up with a hypothesis:

This is testable. We can look at the top scorers of the NBA teams and see whether their team's offensive ratings are worse when they use a relatively small percentage of their team's possessions. If so, our hypothesis is supported (not necessarily proven). If not, that doesn't disprove the hypothesis -- something your sixth grade science teacher should have blasted into your head along with the definition of a hypothesis; it means that you might want to rephrase your hypothesis.

The reason I mention this is because the following table provides some specific results that don't all support the hypothesis. Below is a table showing a list of top scorers in 1997-98, how their teams performed on average, and how their teams performed in the 10 games where the players had the smallest percentage of their teams' possessions*. The grayed greater than signs (>) indicate players whose teams performed worse when they didn't use many possessions. The significant differences -- the ones that can be distinguished from random chance -- are highlighted in yellow**.

Allen Iverson was the most important player to shut down, according to this. Reducing his possessions worsens his team's offense more than than reducing any other player's touches. Given Iverson's performance in '99, people will start to recognize this and an equilibrium -- which I'll talk about later -- will set up.

Both Keith Van Horn and Clyde Drexler showed similar, if less significant, trends. Van Horn is like Iverson in that he's young and teams are still trying to understand his skill. His name being on this list indicates that opponents probably should try to shut him down more, which they did indeed do in '99 as his efficiency numbers did go down. Drexler's presence on this list is probably due to his playing with both Charles Barkley and Hakeem Olajuwon; teams generally considered Barkley and Olajuwon the big threats, but Drexler was doing more damage than they realized.

The other player with a significant difference was Shawn Kemp, whose Cavaliers actually played worse as a team when he used a lot of possessions (shown above with a less than sign: <). Weird, unexplainable, and possibly just a fluke.

The other players don't appear to have significant effects. Several teams actually show better offensive ratings when their stars don't use a lot of possessions . There are some competing factors here causing this effect:

1. A player may have a small percentage of his team's possessions because his team gets a big lead early on (has good offense) and he can coast.
2. This competes with cases where, like Duke, the team needs a player to score for the offense to succeed.

...Are Teams Successful When Their Star Uses A Lot of Possessions?...

If we now look at how teams performed when their stars had a lot of possessions, we have the following table. You'll notice right away that there are no highlighted rows...

When there is a < sign, that means that the team performed better when the player used a lot of possessions. Only 13 of 29 cases met this condition and there were no cases that were statistically significant at 90% -- meaning that every difference could have been just luck.

Again, there are competing factors. Is a player using a lot of possessions because they are playing well? Or are they using a lot of possessions because there team is doing poorly and they have to throw up a lot of shots, even if they're not playing well.

For years, there was the argument about whether you should shut down Jordan and let his teammates beat you. This table and the one previous indicate that Chicago figured out how to beat you no matter what you chose.

...General Correlations...

Finally, let's simply look at whether there is a general trend where, with certain players, the more possessions they use, the better their team's offense (a positive correlation). Or there may be some players where their team performs better the fewer possessions they use (a negative correlation). Here, we simply correlate a team offensive rating with the percentage of possessions a player uses and determine whether that correlation is statistically significant:

Again, Drexler, Iverson, and Van Horn show significant positive influence on their teams by using a lot of possessions. There are now a couple others with significance greater than 90%, two of whom -- Reggie Miller and Karl Malone -- "cause" their teams to be worse when using more possessions.

The problem is that there is no "cause" here. Cause and effect are impossible to determine from these statistics. It seems likely that Miller and Malone end up taking more shots when their teams need them to -- in a close game and/or a game where their team offense has not played well. It is their role.

...Equilibrium...

Obviously, very few players had numbers reflecting the hypothesis. This implies that the hypothesis needs revision (don't say it's wrong).

The strongest exceptions to this equilibrium are, not coincidentally, teams with young stars leading the scoring: Philly with Iverson and New Jersey with Van Horn. These exceptions have to go away with time; you shouldn't see the pattern in the 2000 season (it's already gone for Van Horn in '99, but only lessened for Iverson). Because defenses will have figured out how to play them, and their teams will have had to find a way to handle that defense.

The reason for this, I believe, is that most teams are at a type of "competitive equilibrium". If an opponent is going to shut down the main threat, then it knows how to get other teammates involved to maintain the overall team offense. We debate whether to shut down a main threat or his teammates precisely because the answer is not clear. It depends. Teams find their way of handling the different strategies.

TV commentators will describe this in different ways. Sometimes they will preach that a star has to get a lot of shots. Sometimes they will preach that a star has to get their teammates involved. Sometimes, they will insist that a star should be taking the shots only when his teammates are struggling. Bill Walton will say whatever floats through his drug-damaged head at the moment. But the point is that it's a dynamic equilibrium, with teams and stars constantly reacting to the defense they face, with the opponents of those teams and stars playing defense the way they feel best. If a dominant strategy for playing any team or individual existed, everyone would do it enough so that the statistics wouldn't show it. Right now, teams haven't figured that out with Iverson and Van Horn, but they will.

...Conclusions...

It may be tempting to say that the hypothesis is completely wrong, but that would be foolish. If every team were to start giving Steve Smith or Shaquille O'Neal or Michael Finley or Mitch Richmond or Reggie Miller or Karl Malone a few more chances per game, their respective team offenses would almost assuredly go up. But teams won't do that. Competition drives teams to look for this "equilibrium", where the statistics cannot see the effects.

My personal opinion is that UConn coach Jim Calhoun got a little lucky in his strategy. Sure it worked once -- and at a time when it really counted -- but would it work consistently? Probably not. Duke would devise a strategy to counter it. Just as the correlation we see with Allen Iverson will die away, so would the effectiveness of UConn's strategy.

Fortunately for UConn and Tar Heel fans, it only took one game...

Final Notes

• The fact that Iverson has one of the strongest correlations between his possessions and team offensive rating implies that Coach Larry Brown's decision to move him to shooting guard was a good one. No, he didn't shoot too much even as a point guard.
• You also really want to keep Van Horn out of the lineup. When he's in, the percentage of team possessions isn't all that important, but when he was out, the Nets performed much worse offensively.
• My thought is that negative correlations for established players is probably normal. They may not be statistically significant, but it seems like a common trend that fits in with the current offensive strategy of always going to your best player in close games.