What are your variables, sample size and how are you controlling for things like strength of schedule differences?
Also what regression program are you using to come up with your results?
There are twelve CUSA teams, so the sample was 12. I used MSExcel, though they were r instead of r
2-values.
Correlation coefficient (r) between a CUSA teams winning percentage and their:
Rush Offense 0.65
Pass Offense 0.41
Total Offense 0.70
Rush Defense -0.63
Pass Defense 0.31
Total Defense -0.28"
I don't think there is a need to control for strength of schedule, but there's an easy way to fix that. I also have the 119 non-transitional teams. If you think adding Western Kentucky would make a difference I can add them, it was just less to parse since the NCAA listed them in a lower table.
correlation coefficient (r) between an FBS team's winning percentage and their:
Rush Offense: 0.40
Pass Offense: 0.23
Total OFfense: 0.51
Rush Defense: -0.62
Pass Defense: -0.03
Total Defense: -0.45
So, as the sample size grows approaches the population it tells basically the same comparative story. Of the individual categories listed, rush offense and rush defense are more closely correlated with winning percentage than pass offense and defense. I would say it is a causation as well, but that is up to interpretation. When I coached I knew we could win 90% of the games where we controlled the clock, though time of possession is a 0.22 r-value...about the same as pass offense. I don't think you'll find a magic trend in any of the data, but it makes the point that passing, while entertaining, isn't necessarily what a team must do to succeed. There are too many factors contributing to a team's game plan to point out any one thing as the key other than finding those advantageous match-ups.