Defending Defense

It’s time to talk a pet peeve of mine. When we talk basketball a topic that comes up over and over is defense. It’s used as an explanation of why basketball stats are behind other sports. It’s the biggest complaint used as a “flaw” in Wins Produced. Anyway I want to explain why I think the state of defense in basketball stats is not as bad as people think.

Defensive stats are kept in basketball!

Alright, so the first major point that comes up is that defensive stats aren’t kept. Defensive rebounds have defense in the name for pete’s sake! Steals, turnovers and personal fouls and blocks are also defensive stats. There are a few more that could be kept on the standard box score (tips and charges taken) but even those can be extrapolated by the play by play. The point is there are quite a few defensive stats kept. It’s not some nebulous area we can’t examine. There are some aspects of defense outside of individual box scores. Guess what? That’s our next point!

The impact of team defense is not that large

The Wins Produced metric actually looks at defense not directly captured by the box score. This includes

  • Three Pointers made by the opponent
  • Two Pointers made by the opponent
  • Opponent Turnovers that weren’t a result of a steal.
  • Team Turnovers – these are turnovers not directly attributed to a given player (e.g. shot clock violation, or in other words, turnovers that are not steals)
  • Team Rebounds — these are rebounds not directly attributed to a given player that resulted in a change of possession.
  • Team blocks – total blocks by the team (this used to be in Mate48 in the classic WP)

Now the thing is when we look at how much of an impact this has it is not a lot. For most teams the impact last season was less than 5 wins. What is also critical is that by accounting for team defense Wins Produced makes sure the individual stats are still accurately represented. That means the value of a made shot or rebound has accounted for any defensive factors not directly attributed to an individual. In regards to how we handle team defense? That’s our next point!

Defense is a team activity

While it is fun to picture players like Bruce Bowen or Kobe Bryant single handedly locking down the opponents that is not the case. Unlike offense where one player can basically take every shot (Kobe) or make sure they touch every play (Paul or Nash), on defense a player is actually limited. If the opponents choose to pass to another player then the defender is limited in what they can do. That’s why defense is largely a team activity and we notice that defense year to year seems to be based on the system. This also makes it hard to pin down individual blame. Is the reason the opponent got an easy shot because the player is a bad defender or that another player missed a rotation?

The Wins Produced formula treats defense as a team activity and distributes the credit based on minutes played. While it is possible that some players may be better or worse than this number indicates it’s important to note that this number is not that big to begin with. In the average case it’s less than 0.010 WP48 for the whole team and in the worst/best case it’s around 0.025 WP48 for the whole team. This of course brings up the next point!

What people want may be really hard and not that useful.

Ok when people say defensive stats aren’t kept they’re wrong. However, the stats they want revolve around the following question: How can I measure how much of an influence player A has on player B’s shooting?

To figure this out would basically require the following equation.

  • Every time a player takes a shot figure out what percentage of the time it should go in.
  • Figure out how much an impact the defending player had on that.

This requires that we guess what bad shots (Kobe double teamed) can still go in and good shots can still miss and then factoring in the situation. A problem with this is that match-ups in the NBA aren’t that frequent. How many times does Arron Afflalo cover Kobe taking a mid-range jumper during the season? The answer is: not enough to be significant. Any attempts down this path will likely result in noisy and inconsistent data. While this hasn’t stopped people with metrics like Adjusted Plus Minus, the truth is that this method isn’t a very good way to attempt to model the NBA.

Defense isn’t what is holding people back

Another peeve I have with defense is that it is treated as a major flaw that until solved will keep basketball stats from being relevant. That is simple not the case. Wins Produced for instance explains 95% of wins and Adjusted Production per minute is consistent year to yer (essentially we don’t know how many minutes or what position players will play but we know if players will be good)

Here are examples of some hard problems that people bring up that Wins Produced can’t fully answer:

  • The Draft – How do we find the best talent?
  • Coaching – We don’t know if coaches will play the right rotations.
  • Potential – Young players improve but big breakouts are not easy to spot.
  • When players fall apart – We know old players get bad. How do we spot which ones sour quickest?

These are all really hard problems that can mean a ton to teams. The thing is finding out if a player impacts team defense twice as much as an average player is simply small potatoes to finding out if your superstar is about to fall off a cliff or your 22 year old up for a contract extension will have a breakout year or stay mediocre.

Closing thoughts

My personal opinion is that people have their set of beliefs as to what makes players good and stories they use on them. Patrick did a great post on the illusion of validity about this. When they are confronted with stats they can give credit to the stats and possibly change their view. Their other option is to delve as far away from the stats as possible and use that as an explanation for why their player is still good and why the stats are wrong. Defense is a very common argument that comes up in this venue. The truth, though, is that by focusing on a problem and explaining it is noisy and tricky does not answer a bigger question. Namely, is the information gathered there tremendously useful?

I don’t discount figuring out defense more would be helpful.  But I suspect that the truth is that measuring all a player does on defense would not be as helpful as people think, and the focus on defense ends up being a very convenient argument to bring up to dispute the stone cold stats.


Comments are closed.