In my work on the Threes and Layups NBA Net Rating Calculator, I have focused on 14 statistics (the same 7 on offense and defense) which capture a team’s performance across all areas of the game:
- 2-Point %
- 3-Point %
- 3-Point Attempt Rate
- Free Throw Attempts Per Field Goal Attempts
- Free Throw %
- Turnover Rate
- Rebound Rate (Offensive and Defensive)
The NBA Net Rating Calculator allows you to see how much a team’s regular season performance would be expected to change if you changed one of these statistics. But this leads to a natural question: which of these statistics is most likely to change?
With most teams at about the half-way point of their season, I dug through data for the last 10 seasons (2007-2017, excluding the strike shortened 2012 season) to see how well first half of the season stats correlate with second half stats. For each team, I recorded the number of standard deviations they were above or below the mean in each of the 14 statistics, for both the first and second half of the season. I used standard deviations from the mean of the particular season (rather than raw values) because there were stylistic changes in the game over the past 10 years. A high 3-point attempt rate in 2010 would be closer to the middle of the pack in 2017. With this in mind, here are the correlations from first half to second half for the offensive and defensive statistics:
As you can see, the least stable (lowest correlation) statistics have been Opponent FT% and Opponent 3-Pt %. The fact that Opp. FT% from the first half is not highly correlated with the second half of the season stat is not at all surprising. Unless a team is fouling substantially better or worse opponent shooters, this area of the game seems to be beyond the defense’s control.
Opponent 3-Pt % displaying low stability is also not a revolutionary finding. To what extent the defense can control this stat is the source of a lot of thought and research in the NBA community, and many lean toward the view that much of the variation in Opp. 3-Pt % is due to luck (though the Celtics may be a point against this theory). Also on the defensive side, it is interesting to note that Opp. 3-Pt Attempt Rate is fairly stable, perhaps supporting the theory that good 3-point defense is largely about limiting the number of attempts.
On offense, 3-Pt Attempt Rate has a high 0.84 correlation, indicating that teams do not often change their shot selection with respect to threes drastically over the course of the season. I am personally very intrigued by the fact that 3-Pt % only has a 0.33 correlation. We tend to think of good or bad shooting teams remaining consistent at this skill, but perhaps there is more noise and ability to change. Even if you include only the last 5 seasons (when 3-Point attempt rate has been higher) as opposed to the last 10, you still see only a 0.41 correlation.
Another thing I was curious about was how many teams have made “large” leaps, be they up or down, in each of the statistics. Once again restricting our attention to the last 10 seasons, the chart below tells us the proportion of teams who have had a large, at least 2 standard deviation change (positive or negative) from the first half of the season to the second half, in each stat:
* The y-axis is the proportion of teams who have made a 2 standard deviation increase or decrease in each of the stats from the 1st half of the season to the 2nd half, over the last 10 seasons.
Finally, for fun, I include a table of how many more or less wins each team would be expected to win over the course of this season if from here on out their 3-Pt % defense was average (0.362). According to my calculations the difference is a couple of wins for some teams, which is quite large over half a season. The Wizards, Sixers, and Celtics would be the biggest losers and the Kings and Bucks would be the biggest winners if Opponent 3-Pt % reverted back to average for each team in the second half of the season.