After watching the Pacers hang with the Heat through seven games on the strength of Roy Hibbert and David West pushing the smaller Miami defenders around – and having significant concerns about Nerlens Noel and his 206 lbs. coming to Cleveland with the number one pick – I want to look at the impact of weight on team performance.
I will weight the player weights by minutes played to put an average size on the lineups being rolled out by each team throughout the season and limit myself to the past five seasons so I don’t have to steal too much data from www.basketball-reference.com
Promisingly for Nerlens Noel’s prospects, the weight of the average lineup appears to have almost no correlation with winning percentage (0.05), ORtg (0.08) or DRtg (0.002). Statistics related to playing inside, as you would expect for a heavier team, show bulkier, more substantial correlations: turnover percentage (0.36), ORB% (0.25) and FT/FGA (0.23). The
The distribution of teams is about as close to the textbook definition of random data as you can get. In fact, here is the distribution of random normal data recentered to the mean/standard deviation of the NBA data.
On the other easy-to-visualize metric of team physical appearance – height – there is a similar lack of relationship with the key metrics. Correlation with winning percentage (0.01), ORtg (0.11) and DRtg (0.09) is in the vicinity of perfectly unrelated – the mild improvement in offensive performance of taller teams is offset by an equally mild decrease in defensive performance (lower DRtg is better). Height has a hefty correlation with ORB% (0.28) similar to weight’s correlation with that metric, but the TOV% (0.11) and FT/FGA (0.12) correlations are much slimmer than those for weight.
In the current NBA Finals matchup between the Spurs and the Heat, the season-average lineups for the two teams have the Spurs outweighing the Heat 217.5 to 215.9 pounds (about 0.4 standard deviations) while San Antonio's height advantage is 78.76 with Miami averaging 78.43 inches (a difference of 0.75 standard deviations).
All in all this analysis doesn’t deliver the big insights I was hoping for, but does allow me to unload a small number of weight-related puns. For that I am grateful.