Seeking Relationships In The Big Ten Passing Game

Submitted by LSAClassOf2000 on December 13th, 2012 at 10:21 AM



The object of this experiment was to look for highly correlated data within some of the regular season passing statistics. Data from all 144 regular season games within the Big Ten conference was collected from box scores and team sites. What is being sought is to confirm essentially some of the relationships that people notice with the so-called “eye test”, plus potentially find ones that are not so obvious.


A matrix of statistics was created using six variables – passing attempt, pass completions, total passing yards, yards per attempt, yards per completion and completion percentage. For purposes of this diary, interception data was not included, but is part of a future planned diary along this vein of discussion.

So, in all (n=144 for each variable), there were 864 individual values which were either reported or calculated from existing data. The assumption made was that 144 trials would be sufficient to show some interrelationships inside the data.

Fifteen potential relationships were examined and their r-value calculated.


The six most correlated variables are as follows –

Completions / Attempts - r=0.88

Yards Per Attempt / Yards Per Completion – r=0.82

Completions / Yards – r=0.77

Attempts / Yards – r=0.65

Yards Per Attempt / Comp. % - r=0.57

Yards / Yards Per Attempt – r=0.49

Two relationships showed a relatively meager positive correlation; Completions / Completion % (r=0.37) and Yards / Yards Per Completion (r=0.36)

A few relationships even showed very slight negative correlation; Attempts / Yards Per Attempt (r=-0.29), Attempts / Yards Per Completion (r=-0.29) and Completions / Yards Per Completion (r=-0.26)


Below is the table of summary conference statistics (including a thumb of the matrix I used to calculate "r") and a cat photo -


MEAN: 18 31 207 7.0 11.8 59.1%
MEDIAN: 17 30 192 6.6 11.3 59.3%
MODE: 16 30 171 8.6 12.7 60.0%
STD. DEV. 7 10 81 2.3 3.2 10.5%
VARIANCE: 46 108 6567 5.4 10.4  




December 14th, 2012 at 8:35 AM ^

...does this lovely data actually tell us lay people? Perhaps adding a section "Conclusions" would be valuable for us non-maths geeks who are interested in math-geek findings.

Also, was the objective to roll around in the data to try and find correlations or were there specific hypotheses you wanted to test?


December 14th, 2012 at 9:06 AM ^

...because forgetful (and, right now, cold-stricken) me did not realize until this morning that the conclusions, which I typed separately, were not in here. My fail really, but I choose to blame pseudoephedrine.

Long story short, the idea was to see, within the context of a season, if the relationships we can infer in the passing game exist statistically, and they do. More attempts correlate strongly to more completions, yards per attempt relate strongly to yards per completion (i.e., the greater one is, the greater the other tends to be typically), and more completions tend to mean more yards. Those obviously aren't new, but it is merely interesting to see them worked out this way.

Another one that shouldn't be new, although I am intrigued at the result, is the merely mild correlation between yards and yards per attempt, but when a team just isn't connecting on certainly plays or has a receiving corps that tended to disappear, like Iowa most games this year, that isn't shocking either.