A while ago I did some analysis to further Bud Elliot's look at defensive size and defensive performance, which I found interesting, but incomplete. Unfortunately, I didn't have enough mgopoints at the time to post it, but now I do!  So here you go, if anyone still cares.

I had two problems with his analysis: First, he has an extreme selection bias, focusing on the top 20 teams, along with the ACC and SEC.  Second, he compares size to defensive rank, but we have no expectation that rank would correlate with size.  For example, if a number of teams have exactly the same size, the BIG implies GOOD hypothesis predicts that those teams would perform about the same.  In other words, we expect defensive performance to correlate with size.

Based entirely on the one linear modeling class I took last fall, I have attempted to do better.  I have barely addressed the first problem, because getting more data is hard.  However I have added Notre Dame and USC to the data on his blog.  I've addressed the second problem by replacing rank with the Defensive S&P+ score from Football Outsiders.  I don't know exactly what it means, but I gather that larger numbers are better.

Tada:


So yeah, not a clear relationship, and definitely not "Staggering!".  And, given that the data is strongly biased towards the top 20 teams,  I'd be willing to bet that this relationship would diminish further,  given the full set of NCAA starting player's weights.

There is some correlation, but is it significant?  Turns out yes, but only barely.  The F-statistic gives us a p-value of 0.014.  However, the adjusted R-squared value for the fit is very low, at 0.1374. Adjusted R-squared tells us about how much of the overall variation seen in the plot is due to variation in the independent variable (the weight of the front 7).  In this case, less than 14% of the variation is explained by our variable.  Conclusion: At least  86% of a teams defensive performance is explained by factors other than the weight of the front 7 starters  (and its probably much more than 86%).

The coefficient on the dependent variable is 0.188, meaning that for every pound your front 7 puts on, you gain about 0.2 Defensive S&P+ units.  More is better . . . but it isn't clear what units that statistic is in . . . so there you have it.

The residuals are huge.  Leaving out the largest errors and only considering the middle 50% still covers a range of 25 S&P+ units - roughly equal to the range  from (14-0) Alabama to (9-5) Clemson.  Just look at all the teams with a total size of around 1800!  They cover virtually the entire range of defensive performances!! Defensive size is just not a good predictor of success in games.

One more caveat.  While searching the web for starter's size stats, it because clear that the preseason is filled with overestimates, so the 108 pound gain over last year's line should probably be taken with a grain of salt.

Go Blue!
ebv

 

Methods.  I gathered most data from Size Matters on Defense Part 1 or Football outsiders.  Notre Dame info came from 2009 Notre Dame depth chart.  USC info came from this USC roster.

I fit the simplest of linear models, which is shown as the line on the scatterplot above (intercept -222.9).

Raw data:

Michigan        1720    94.5
Alabama 1911    149.4
Florida 1883    148.3
Maryland        1864    91.3
UNC     1855    127.6
Georgia 1846    100.1
LSU     1836    126.8
Clemson 1807    125.2
Ga Tech 1804    90.7
S. Carolina     1798    122.8
Kentucky        1796    93.5
Duke    1795    86
Ole Miss        1790    116.5
Miami   1787    113.1
Wake Forest     1786    97.7
Miss St 1785    107.6
Auburn  1781    104
NC State        1777    78.1
Arkansas        1776    103.4
Boston College  1768    112.6
Virginia        1757    110.1
Vanderbilt      1754    95.2
Va Tech 1736    122
FSU     1734    90.3
Tennessee       1726    111.7
Ohio State      1784    134
TCU     1781    145
Oklahoma        1791    147.6
Boise   1808    133.6
Iowa    1821    138.7
Ok. State       1811    116.4
Penn State      1818    127.9
Nebraska        1795    143.2
Texas   1807    131.2
Utah    1837    121.2
Notre Dame      1819    98.7
USC     1750    103.8