With the Blazers continued road woes and the importance of scheduling in the Western Conference playoff race, I thought I'd take do a little analysis of home court advantage in the NBA this season.

Measuring Home Court Advantage

The simplest way to measure home court advantage is to compare a team's home court and away records. Because schedules in the NBA are unbalanced, however, it is possible that some teams have played more difficult schedules at home than away. In addition, as John Hollinger and others point out, margin of victory is a better predictor of future record than wins and losses. Since most of us are interested in projecting how the rest of the season and playoffs will unfold, it makes sense to measured home court advantage by margin of victory rather than wins and losses.

To create a measure of home court advantage for each team, I started by downloading a log of all of the 2009 regular season games from basketball-reference.com (through Saturday March 14th). Then, if the game went into overtime, I set the margin of victory to zero. Then I used standard statistical/econometric methods to control for strength of schedule and create an estimate of each team's home court advantage.  (For those interested in the technical details, I created the measure by regressing margin of victory on a set of dummy variables for each team, a dummy variable indicating which team was home and set of team*home court interaction terms. The team dummy variable is the team "Fixed Effect", it accounts for each teams average margin of victory. Each team's home court advantage is the coefficient on the home court variable plus the team*home court interaction. The team*home court interaction is described as the "Extra" Advantage). The results are below.

Results

Home Advantage is the difference between a team's average home margin of victory and average road margin of victory, controlling for their opponents' average margin of victory and their opponents' home v. away advantage. Note that, by construction, a team's away disadvantage is just their home advantage multiplied by -1. The average home-court advantage in the NBA this season is 7.87 (if overtime games count as ties).

"Extra" Advantage is the difference between each team's home court advantage and average home court advantage in the NBA. As with the raw numbers of home court advantage, a high number can be taken to indicate that a team plays well at home or poorly on the road. Similarly, low numbers can indicate that a team plays well on the road or poorly at home; it's impossible to distinguish between those characterizations of what the numbers mean.

Discussion

The results conform, more or less to my expectations, with a few surprises. Portland has a "larger" home court advantage than the average team, but Utah and Golden State have even larger home court advantages (or road disadvantages). It's interesting that thee top four teams in terms of "extra" advantage are also relatively young, while three of the teams with the smallest home advantages are also young. It is amazing to see that Minnesota has actually played worse at home. I think the depressing faces that Phil Jackson referred to must be in Minneapolis-St. Paul, rather Portland.

Obviously teams like Utah and Portland would like to be able to play as well on the road as they do at home. However, given that the advantage of playing at home seems to be connected to how loud fans are... I'm not sure it's possible to just flip a switch and make the team play as if 20,000 people are enthusiastically supporting their every move.  Perhaps some improvement can come with experience, but I would not expect any big changes this season. Thinking about this a bit reaffirms to me the importance of the Blazers getting home court advantage in the playoffs.

A few limitations of the method that I used: it assumes that each team's "strength" and home court advantage do not vary systematically throughout the season. Each team plays poorly some games and well in other games, but any variance other than home court advantage and opponent strength is (random) noise. One way to improve the analysis above would be to include "month" indicators for each team to account for streaks that occur during the season. There is no reason that "hot streaks" would care about the calendar, however, so this would be a pretty dirty fix. Another approach is to include lags and leads of margin of victory. I played around with this a bit, but saw little evidence of auto correlation. A more promising approach would be to include variables that indicate when key players have been out due to injury (or even playing while injured). If anyone knows of a nice data set on player injuries, I'd be happy to give that a shot. Indeed, I've already looked into this for the Blazers and the results suggest that having key players out matters quite a bit, in some cases. More on this later.