In cyberspace the past couple days, a season-long conversation about Javale McGee has flared up. On one side are those who argue that McGee makes the Wizards better (looking at +/- data) and that when he plays more, the team has a better record.

On the other side are those who argue that McGee is capable of dominating individual games because of his overwhelming athleticism, but that he fails to do so regularly for several reasons, including that:
— he’s unskilled
— he doesn’t know how to play
— he continues to suffer from lapses in concentration.

The “McGee is a good player being hampered by bad coaching adherents” point to the team’s 5-2 record when McGee plays 36 minutes or more. Sounds impressive. Especially when you look at McGee’s per minute numbers when he plays that much.

But, this argument is yanked short by Ye Olde “Chicken or Egg?” question. Is McGee playing well (and the team winning) because he’s getting more minutes, or is the team winning and McGee getting more minutes because he’s being productive? And, is this analysis an example of the hazards of arbitrary endpoints?

In other words, why is 36 minutes significant? Drop the minute total to 34 and the record goes to 6-3; to 32 and it’s 7-5; to 30 and it’s 8-9. Would 6 more minutes of McGee in those games where he got less playing time really convert some of those losses to wins? On the other hand, the team falls off a cliff (8-35) when McGee gets less than 30 minutes.

An examination of McGee’s game-by-game performance reveals that no one statistical category appears to be determining his playing time. Two measures have the highest correlations with McGee’s minutes: Game Score (a PER-like stat that measures a player’s overall statistical contributions) and +/-. Here’s a list of the stats with the strongest correlations to McGee’s playing time this season (1.0 means perfectly correlated):
1. +/- -- 0.49
2. Game Score -- 0.44
3. Fouls -- -0.34
4. sOrtg -- 0.34
5. Points -- 0.33
6. Offensive Rebounds -- 0.24

(sOrtg is a simple overall measure of offensive efficiency. The formula is 100 x (pts / (fga + .44 x fta + tov)) )

This list is suggestive. When McGee is productive (Game Score), the team’s +/- is likely to be better, and he’s likely to play more. The biggest drag on his playing time is fouls.

When I’ve done this kind of analysis on other players, I found little evidence for the argument that a coach will give a guy more playing time when he’s productive and less when he’s having a bad night. In McGee’s case, Flip Saunders seems to have a feel for how McGee is performing and is able to curtail his minutes on bad nights. Or, it could just be that McGee coincidentally performs worse per minute on nights when he plays less.

Because, the game-by-game data shows that McGee’s per minute performance declines steadily when his minutes go down. For example, his Game Score goes from 16.5 per 40 minutes in games where he plays 36 or minutes to 14.2 when he plays 30-34; 13.1 at 25-29; 10.5 at 20-24; and 5.4 at fewer than 20 minutes.

He grabs nearly 14 rebounds per 40 minutes in those high-minute games, but only 10.8 in all other games. His turnovers per minute go up when he plays less; his efficiency drops; his per minute scoring falls; his fouls skyrocket. McGee’s +/- and the team’s record follow the same trend. His per minute blocked shots higher when he plays less (the data does NOT suggest that his per minute blocks and per minute fouls are related to each other -- in other words, it doesn’t appear that he fouls more when he’s blocking more shots).

In high-minute games, the team is 5-2. In all other games, they’re 11-42. In high-minute games, McGee’s +/- is +6.4 per 40 minutes; in all other games it’s -8.5 -- a swing of 14.9 points per 40 minutes. That’s huge. A more comprehensive table showing this data is available here.

To summarize: The data suggests that when McGee plays well and stays out of foul trouble, he gets more minutes and the team plays better. It does not appear to support the argument that the team would be better if only the coaching staff would get him more playing time.