Exploring the effect of size on NFL combine drills: Offensive (skill) positions

Just as we did with defensive positions, let's explore the relationship between size and speed within the combine drills by positions, from smallest/fastest to largest/slowest.  Again, we transformed all timed events using the rate (total distance/time taken) to ensure larger numbers are better in all drills.  The data is was combine participants between 1999 and 2015.

Let's start in the backfield (quarterbacks excluded, as the possible subject of a future post):

There doesn't appear to be much happening with respect to height, though there appear to be some strong relationships with weight.  Again, we'll examine the partial correlation coefficients (adjusted for position differences between RB/FB) along with the p-values for the test for significant, non-zero partial correlation between these drills and weight/height:

Pearson Partial Correlation Coefficients, N = 301 Prob > |r| under H0: Partial Rho=0
	dash_rate	shuttle_rate	cone_rate	VertLeap	BroadJump	BenchPress
Weight	-0.24607 <.0001	-0.32696 <.0001	-0.27348 <.0001	-0.06997 0.2269	-0.11010 0.0568	0.17279 0.0027
Height	-0.06765 0.2427	-0.14032 0.0150	-0.06441 0.2661	0.10126 0.0799	0.10990 0.0573	-0.06758 0.2433

Indeed, all the rates appear strongly inversely proportional to weight.  We see a similar trend in bench press that was observed in the secondary:  larger weight (likely in muscle) corresponds to better bench press reps.  Also, we see a similar trend in broad jump that was observed in the linebackers with a small weight penalty and slight height advantage.  Shuttle rates are somewhat inversely proportional to height, but this may be because taller individuals tend to weigh more.

Moving onto the receivers, we see some similar trends to that observed in the backs:

Pearson Partial Correlation Coefficients, N = 361 Prob > |r| under H0: Partial Rho=0
	dash_rate	shuttle_rate	cone_rate	VertLeap	BroadJump	BenchPress
Weight	-0.26015 <.0001	-0.20277 0.0001	-0.25864 <.0001	-0.08922 0.0910	0.00048 0.9927	0.21709 <.0001
Height	-0.11617 0.0275	-0.17604 0.0008	-0.10181 0.0536	-0.04737 0.3702	0.11930 0.0236	-0.00177 0.9732

Again, the rates are strongly inversely correlated to the weight and somewhat positively correlated with height, again likely due to the tendency for height and weight to co-correlate.  Again, we see being taller is advantageous for broad jump, and we also see that weight (muscle?) is an advantage for bench press.

Finally, let's look at the offensive linemen:

The regression lines are so similar that it may not be necessary to adjust for outside/inside position, but it couldn't hurt.  In all likelihood, the partial correlation coefficients below are identical to the standard correlation coefficients if we just pooled all the offensive linemen together.  

Pearson Partial Correlation Coefficients, N = 361 Prob > |r| under H0: Partial Rho=0
	dash_rate	shuttle_rate	cone_rate	VertLeap	BroadJump	BenchPress
Weight	-0.26015 <.0001	-0.20277 0.0001	-0.25864 <.0001	-0.08922 0.0910	0.00048 0.9927	0.21709 <.0001
Height	-0.11617 0.0275	-0.17604 0.0008	-0.10181 0.0536	-0.04737 0.3702	0.11930 0.0236	-0.00177 0.9732

Much of the same here:  rates are inversely proportion to both weight and height.  Being taller is an advantage for the broad jump, and being heavier correlates with better bench scores.

In summary, it may be more meaningful to compare (weight x rate) for the timed drills.  Furthermore, we might also stratify by height before comparing broad jumps of players or use some sort of linear transformation.  Finally, bench to weight ratio may transformed to serve as a surrogate for lean muscle mass (and it's a little ridiculous that the combine doesn't track this already).