Friday, September 04, 2009

By the Numbers, Michigan State and Virginia Tech Suck

Thanks to the Comcastic wonder of having hundreds of cable channels, I had the pleasure of watching the January 1, 1986 Cotton Bowl the other night. It was a trip down memory lane for me, as it pitted the Bo Jackoson-led Auburn team for which I rooted passionately (it's a long story) against Texas A&M. While watching Auburn's feeble passing game hand the game to the Aggies, I remembered devouring Allen Barra's Football by the Numbers the following summer. Barra argued in the book that Jackson was by far the best player in college football in 1985 and he was penalized by some Heisman voters for playing on a team that couldn't throw the ball.

Football by the Numbers had a real effect on me. For instance, Barra described in the book his NEWS Rating, which was a deceptively simple way to measure quarterbacks. Take a quarterback's passing yards, add ten yards for each touchdown, subtract 45 yards for each interception, and then divide by attempts. Voila, a nice way to evaluate quarterbacks. I liked Barra's focus on yards per play as a measure. As an 11-year old, Barra's approach made more sense to me than the usual drivel that football analysts threw out to rank quarterbacks.

In honor of Barra, I decided to take a similar approach to ranking college football teams before the season. What if we just ranked teams by looking at their per-play yardage margin (yards gained per play on offense minus yards per play allowed on defense) and returning starters? I wish that I would have thought about this earlier in the summer so I could go back to previous years to see if this method would produce better results than the more subjective preseason rankings. I also wish that I had a name for these rankings. Anyway, looking at yards per play on offense, yards per play on defense, and returning starters, removing the decimal from the per-play margin, and then creating a crude sum of apples and oranges, here is the list of BCS Conference teams that had non-negative per play margins in 2008:
  1. Florida - 2.6 yards per play margin / 18 returning starters / 44
  2. USC - 3.0 / 12 / 42
  3. Georgia - 1.7 / 15 / 32
  4. Oklahoma - 1.7 / 14 / 31
  5. Ole Miss - 1.5 / 16 / 31
  6. Penn State - 2.1 / 9 / 30
  7. Cal - 1.5 / 1.5 / 30
  8. Georgia Tech - 1.2 / 18 / 30
  9. Iowa - 1.4 / 1.4 / 28
  10. Oklahoma State - 1.4 / 13 / 27
  11. Texas - 1.2 / 15 / 27
  12. Oregon - 1.7 / 9 / 26
  13. Texas Tech - 1.5 / 11 / 26
  14. Missouri - 1.6 / 9 / 25
  15. Alabama - 1.2 / 12 / 24
  16. Ohio State - 1.0 / 12 / 22
  17. West Virginia - 1.0 / 12 / 22
  18. Clemson - .7 / 15 / 22
  19. Baylor - .6 / 16 / 22
  20. Rutgers - .8 / 13 / 21
  21. Florida State - .7 / 13 / 20
  22. Nebraska - .7 / 13 / 20
  23. Wisconsin - .8 / 11 / 19
  24. Cincinnati - .9 / 9 / 18
  25. UConn - .6 / 12 / 18
  26. North Carolina - .4 / 14 / 18
  27. Oregon State - .7 / 10 / 17
  28. Boston College - .4 / 13 / 17
  29. Miami - .2 / 15 / 17
  30. Notre Dame - .2 / 15 / 17
  31. Pitt - .2 / 15 / 17
  32. Tennessee - .4 / 12 / 16
  33. Arizona State - .1 / 13 / 14
  34. Northwestern - .1 / 13 / 14
  35. LSU - 0 / 14 / 14
  36. Maryland - .3 / 9 / 12
  37. South Carolina - 0 / 11 / 11

Some thoughts:

  • I'll acknowledge that I wouldn't necessarily use these rankings without some subjective component. For instance, I'd obviously have LSU higher than 35th because the raw numbers do not take into account either the Gotterdamerung of a quarterback situation that they sorted out towards the end of the season or the major upgrade that they have at defensive coordinator. That said, if I would have gone through this exercise before doing my rankings, I would not have had LSU in the top ten.

  • I'll also acknowledge that there should be some modification for strength of schedule. I'm mulling over a good way to integrate that factor into the rankings. I wish there were some way to take into account the quality of the players replacing departed

  • Some notable teams that do not appear in these rankings because they had a negative per-play margin in 2008: Arkansas, Michigan State, Minnesota, N.C. State, Wake Forest, and Orange Bowl Champions Virginia Tech. Interestingly enough, Virginia Tech and Auburn have the exact same characteristics: they were outgained by .4 yards per play in 2008 and they return 15 starters. One team fired its coach and is unranked; the other is a preseason top ten team. Oh, and Alabama beat Auburn 36-0.

  • I was surprised that Texas was only fifth in the Big XII in yards per play margin last year. They strike me as being overvalued going into 2009.

  • The numbers love Ole Miss.

  • If these rankings turn out to have merit, this will be a terrific season in the State of Georgia. I'll freely concede that Georgia Tech is a much better prospect by the numbers than I gave them credit for being when I did my rankings. We'll see if my subjective Spidey sense about them has any value.

4 comments:

Anonymous said...

I think you should call the method Luba

Jesse said...

Nice. I like it as it's a fesh way to look at things. Just wondering, but what made you choose to go with the sum versus say simply multiplying the YPM by the RS to get a YPM/RS?

I ran the numbers, and while generally things are the same or basic flipping of consecutive teams in the original rankings, there were a couple of movers. Tenn went from 32 to 29, Maryland went from 36 to 33, and Penn St went from 6 to 9. So really, in the end I'm not sure it matters much. If anything it would only give you a definition for the final number, that being Yards Per Play Margin Per Returning Starter. Yeah, scratch that, haha.

Another thing that might be worthy of looking at is the amount of offensive and defensive production returning. Thinking through this process it seems as though it might net more accurate numbers considering that it would actually remove production from those who are no longer returning, whereas these numbers are comparing totals against returning starters. It would also require a little more work to compile for 40 teams or more since you would need to pull every players stats.

Jesse said...

Haha, how about not Per Returning Starter. I have no idea why I said that, but I kept getting interrupted while writing that comment so I guess I wasn't thinking straight. So it really is no different than getting the sum of the two.

I'd still like to pursue the % of production returning, but I haven't thought of a good way to automate the process across 40+ teams. I guess I will sit down this weekend and see if I can't create a plug-n-play spreadsheet. I'm definitely open to any suggestions or sites that might help my efforts.

Anonymous said...

Interesting post. the numbers might not be far off on LSU; they looked 35th-ish against Washington on Saturday night.