More fun with numbers

Especially after reading this post of Kyle’s, I’ve done some more thinking about Sunday Morning Quarterback’s analysis of the correlation of certain statistical performance categories and wins/losses.

It gnaws at me a little that SMQ’s chart ranking all of the D-1 teams that finished the last season with at least 10 wins shows Georgia and Hawaii to be almost functionally equivalent (although it should be noted that he places more distance between the two in his chart of ranked teams) and that Kyle has riffed off of it as to Ohio State and BYU in the premise of his post. I watched all four of those teams play at least twice last year, and my (admittedly subjective) perspective is that Georgia was a far better team than was Hawaii and that OSU was a better team than BYU.

All of which led me to consider tweaking SMQ’s yeoman’s work in assembling his numbers. And did, hence this post.

Where I started was with his chart of 10+ win teams. I noticed that while he compared the rankings of those schools in nine categories, those weren’t the same nine categories that he found most closely correlated with wins and losses. When I substituted the two stats that had been omitted (3rd down conversions for offense and defense), here’s the chart that I got:

School Rush D P/E D Total D 3d Dn D T/O Mar P/E O Total O 3d Dn O Pass D
                   
Kansas 8 9 12 31 1 7 8 9 49
LSU 12 3 3 14 2 37 26 29 9
W. Va. 18 28 7 8 9 11 15 36 14
S. Cal. 4 6 2 28 41 36 29 8 15
Ohio St. 3 4 1 17 76 12 62 14 1
Okla. 17 43 26 5 25 1 19 13 59
Boise St. 35 66 25 3 47 6 12 17 26
Hawaii 41 21 34 10 93 3 3 6 37
BYU 9 18 10 21 93 28 25 32 32
Va. Tech 5 5 4 85 14 53 100 4 31
Georgia 16 36 14 24 18 61 74 23 36
Cincy 19 34 50 66 6 8 16 26 89
Ariz. St. 21 15 30 68 38 17 56 11 61
Missouri 25 45 59 2 11 13 5 82 96
Bos. Coll. 2 23 19 44 30 59 33 47 88
Texas 6 70 52 11 47 30 13 63 109
UCF 69 27 49 71 54 67 45 55 69
Tenn. 73 66 70 56 27 29 54 72 73
Tulsa 108 94 108 6 92 4 1 62 108

Just to save you the trouble, here’s how the teams ranked on that raw data after averaging.

  1. Kansas 14.89
  2. LSU 15
  3. West Virginia 16.22
  4. S. Cal. 18.78
  5. Ohio State 21.11
  6. Oklahoma 23.11
  7. Boise State 26.33
  8. Hawaii 27.56
  9. BYU 29.78
  10. Virginia Tech 33.44
  11. Georgia 33.56
  12. Cincinnati 34.89
  13. Arizona State 35.22
  14. Missouri 37.56
  15. Boston College 38.33
  16. Texas 44.56
  17. UCF 56.22
  18. Tennessee 57.78
  19. Tulsa 64.78

I didn’t stop there, though. I thought that to get a more representative sampling that the averages should be weighted on the basis of how closely each statistical category correlated to wins/losses. In other words, SMQ’s work shows that rushing defense is almost twice as significant an indicator as passing defense. To me, it only makes sense to give more weight to a school finishing first in rushing defense than in passing defense.

The way I went about doing this was to award points on an inverse scale to finish (a team finishing first in a category received 119 points; one finishing last received 1 point) and then multiplied the points a team received by the factor that SMQ assigned to that statistical category. Thus for example, Kansas finished eighth nationally in rushing defense. The Jayhawks received 112 points for their finish; those points were multiplied by .437 (SMQ’s factor) and that product (48.94) was the figure that was used to average Kansas’ statistical performance.

With that, here’s the order of the schools:

  1. Kansas
  2. LSU
  3. West Virginia
  4. S. Cal.
  5. Ohio State
  6. Oklahoma
  7. Boise State
  8. Hawaii
  9. BYU
  10. Virginia Tech
  11. Georgia
  12. Cincinnati
  13. Arizona State
  14. Boston College
  15. Missouri
  16. Texas
  17. UCF
  18. Tennessee
  19. Tulsa

The only change that resulted from all that was to flip Boston College and Missouri in the order.

The final step I took was to factor in strength of schedule. I did this by multiplying each school’s weighted average by Sagarin’s SOS rating (not ranking). I then ranked the schools by the number of points generated. Here’s the end result (with the number in parenthesis being SMQ’s ranking of the averages):

  1. LSU (2) 2860
  2. West Virginia (4) 2716
  3. Southern Cal (3) 2691
  4. Kansas (1) 2562
  5. Oklahoma (6) 2541
  6. Ohio State (7) 2514
  7. Georgia (14) 2342
  8. Arizona State (11) 2322
  9. Virginia Tech (15) 2315
  10. BYU (8) 2290
  11. Missouri (9) 2276
  12. Boston College (13) 2208
  13. Cincinnati (10) 2115
  14. Texas (16) 2016
  15. Boise State (5) 2005
  16. Hawaii (12) 1906
  17. Tennessee (18) 1700
  18. UCF (17) 1489
  19. Tulsa (19) 1218

Those numbers seem more in line with what I saw last season. In another post, I’ll try to do the same with the teams ranked in the AP, to see how this matches up with the results that SMQ obtained.

A few observations:

  • Everytime I play with team stats, LSU winds up on top. Again, over the course of the year, the Tigers look like the best in the country.
  • West Virginia was underrated.
  • Ohio State may have been a little overrated, but not to the extent that so many would have you believe after the BCS title game.
  • BYU looks like it should have been the school to crash the BCS party last season, not Hawaii.
About these ads

2 Comments

Filed under College Football, Stats Geek!

2 responses to “More fun with numbers

  1. James

    Even with SOS factored in, this type of analysis is fatally flawed. When a team only has to get “up” for one or two games per year, that team will perform better week to week. After a team from a strong conference plays rivals 4 out of 5 weeks and then plays a weaker foe, they don’t perform well. Just look at the difference in performances by UF and UGA against Troy. UGA was the better team, but the score against Troy didn’t bear that out as they had just come out of a string of SEC games.

    Additionally, UGA has a history of not running up the score on outmanned opponents, while some schools try to humiliate their opponents, seeking the adulation of articles like this.

  2. James, while I think “fatally flawed” may be an overstatement, that’s an interesting point you raise about scheduling. I’ll see if I can explore that in a post in the future.

    I don’t think the numbers bear your second premise out, though. Georgia finished 34th nationally in scoring offense and 18th in scoring defense. Steele tracks those same numbers, but takes out all overtime points scored and allowed and also meaningless points scored or allowed in the final minutes of blowout games. If your argument holds water, the Dawgs’ standings should improve; in reality, Georgia drops to 38th in points scored and 25th in points yielded under Steele’s analysis.