I Created a CFB Ranking System

[bamawriter]

Back in 2011, during the controversy surrounding the Bama-LSU rematch, I started putting together my own version of a BCS computer ranking system. I was going to use it in a future column to detail how stupid the very concept of the computers is when it comes to determining a champion. But, before I had the chance to write that column, the playoff was announced, and it seemed sort of pointless.

Anyway, I was clearing of a hard drive and found my system. Because I'm bored, I decided to update it for the past several seasons just to see what it would spit out for the four team playoff. Thought some on here might find it interesting. Certainly could spur some debate.

Here's the idea:

Bias is completely eliminated. Team names, history, and conference are all ignored. Winning % and strength of schedule are the most important factors. While the BCS eliminated margin of victory, I think that's ridiculous, so it's back in on mine. In order to adjust for home field advantage, I took the standard Vegas rule and subtracted 3 points for a game at home, and added 3 points for a road game, creating an adjusted point differential. I also included average yardage differential to help account for the way a game actually played out on the field. FCS opponents are given a standard .25 winning % and only their games against FBS teams count toward their point differential.

Here's the formula:

Winning % x 100 (too make it a whole number)
+
Opponents' Winning % x 100
+
Average Adjusted Point Differential
+
Opponents' Average Point Differential
+
Average Yardage Differential / 100
=
SCORE


As an example, 2015 Tennessee had the following score (bowl game not included):

(.6667 x 100) + (.5385 x 100) + 12.4167 + .5753 + (52.0833 / 100) = 134.0328

Interestingly, my system from last year produced the same 4 playoff teams, and even the same semifinal matchups, but with different seeding:

1. Alabama - 176.4984
2. Clemson - 175.6221
3. Oklahoma - 174.9711
4. Mich St. - 164.4958

Michigan State got in over Ohio State by .0007.

I went back to 2008 if anyone is interested in more.

 

[VolNExile]

Question about the margin of victory bit: doesn't this penalize teams who put in the bench players and freshmen toward the end? Any thought to stopping counting beyond 21 or 24? (excluding the upper outliers)

I know it's satisfying to fans to whump someone, but I like seeing the backups come in once whumpage is well established.

How many games had a margin greater than 21 or 24?

 

[TrueOrange]

I'm curious to see what the top 10s in 2011, 2012, 2013, and 2014 would look like in your system (more the latter 3).

 

[VolNExile]

--meant to say, how's your eye?

 

[bamawriter]

Question about the margin of victory bit: doesn't this penalize teams who put in the bench players and freshmen toward the end? Any thought to stopping counting beyond 21 or 24? (excluding the upper outliers)

I know it's satisfying to fans to whump someone, but I like seeing the backups come in once whumpage is well established.

How many games had a margin greater than 21 or 24?

I considered that. But in the end I felt that doing so would put a subjective standard on the game, and I wanted to avoid that. In the end, it somewhat washes because if you help out your own average with a 42 point win, you hurt your opponents' average with a 42 point loss.

 

[bamawriter]

--meant to say, how's your eye?

It's the source of my boredom. Doctor injected some sort of crap into the growth to help shrink it a little before removing it at the end of the month. It has my vision messed up, so no driving into the office.

But he's confident the growth isn't malignant, so all-in-all I'll be fine.

 

[LouderVol]

I'm curious to see what the top 10s in 2011, 2012, 2013, and 2014 would look like in your system (more the latter 3).

ditto, and of course I want to know where TN fell. maybe get the SEC rankings.

kudos to bama on the hard work.

 

[bamawriter]

I'm curious to see what the top 10s in 2011, 2012, 2013, and 2014 would look like in your system (more the latter 3).

Here you go (with the final BCS/CFP for reference):

2011
1. LSU 13-0 (1)
2. Oklahoma St 11-1 (3)
3. Alabama 11-1 (2)
4. Boise St 11-1 (7)
5. Stanford 11-1 (4)
6. Houston 12-1 (19)
7. Oregon 11-2 (5)
8. Michigan 10-2 (13)
9. Wisconsin 11-2 (10)
10. Olahoma 9-3 (14)

2012 (this year was as similar to the BCS as my system got. I debated taking Ohio State out since they were ineligible for the postseason, but it turns out they wouldn't have made the BCSNCG or playoff anyway)
1. Notre Dame 12-0 (1)
2. Alabama 12-1 (2)
3. Florida 11-1 (3)
4. Oregon 11-1 (4)
5. Kansas St 11-1 (5)
6. Ohio St 12-0 (n/a)
7. Texas A&M 10-2 (9)
8. Stanford 11-2 (6)
9. Oklahoma 10-2 (11)
10. South Carolina 10-2 (10)

2013
1. Florida St 13-0 (1)
2. Baylor 11-1 (6)
3. Auburn 12-1 (2)
4. Alabama 11-1 (3)
5. Ohio St 12-1 (7)
6. Stanford 11-2 (5)
7. Missouri 11-2 (8)
8. Arizona St 10-3 (14)
9. Oklahoma St 10-2 (13)
10. Oregon 10-2 (10)

2014 (same top 4, but OSU jumped a lot)
1. Ohio St 12-1 (4)
2. Alabama 12-1 (1)
3. Oregon 12-1 (2)
4. Florida St 13-0 (3)
5. TCU 11-1 (6)
6. Baylor 11-1 (5)
7. Ole Miss 9-3 (9)
8. Boise St 11-2 (20)
9. Michigan St 10-2 (8)
10. Marshall 12-1 (not in top 25)

 

[LouderVol]

2014 (same top 4, but OSU jumped a lot)
1. Ohio St 12-1 (4)
2. Alabama 12-1 (1)
3. Oregon 12-1 (2)
4. Florida St 13-0 (3)
5. TCU 11-1 (6)
6. Baylor 11-1 (5)
7. Ole Miss 9-3 (9)
8. Boise St 11-2 (20)
9. Michigan St 10-2 (8)
10. Marshall 12-1 (not in top 25)

interesting about Ole Miss considering how (relatively) few 10-2 teams made the list.

 

[bamawriter]

ditto, and of course I want to know where TN fell. maybe get the SEC rankings.

SEC for all 4 years

2011
1. LSU 13-0 188.2562
3. Alabama 11-1 173.7730
14. South Carolina 10-2 148.2888
16. Arkansas 10-2 146.1907
19. Georgia 10-3 141.4765
40. Auburn 7-5 116.9293 (#25 in BCS... ?)
45. Miss St 6-6 113.4671
48. Florida 6-6 112.2024
60. Vanderbilt 6-6 108.2522
68. Tennessee 5-7 103.5082
89. Kentucky 5-7 84.8307
102. Ole Miss 2-10 67.2253

2012
2. Alabama 12-1 174.6355
3. Florida 11-1 173.8223
7. Texas A&M 10-2 160.7176
10. South Carolina 10-2 154.3490
11. Georgia 11-2 152.4381
12. LSU 10-2 152.4073
43. Vanderbilt 8-4 121.4854
47. Miss St 8-4 119.8356
57. Ole Miss 6-6 111.4547
62. Missouri 5-7 108.4747
71. Tennessee 5-7 100.8810
76. Arkansas 4-8 93.0683
94. Auburn 3-9 79.9716
104. Kentucky 2-10 69.3372

2013
3. Auburn 12-1 168.2563
4. Alabama 11-1 168.1934
7. Missouri 11-2 156.5091
14. South Carolina 10-2 149.3371
21. Georgia 8-4 141.1884
22. LSU 9-3 139.4226
23. Texas A&M 8-4 139.3518
38. Ole Miss 7-5 126.6995
49. Miss St 6-6 118.4756
58. Vanderbilt 8-4 111.6524
69. Tennessee 5-7 104.4726
79. Florida 4-8 98.4718
96. Arkansas 3-9 78.6074
106. Kentucky 2-10 65.6213

2014
2. Alabama 12-1 172.8373
7. Ole Miss 9-3 158.1338
11. Miss St 10-2 153.8700
13. Georgia 9-3 148.0006
16. Auburn 8-4 146.0563
19. Missouri 10-3 139.4307
22. LSU 8-4 136.2366
28. Arkansas 6-6 131.4989
42. Texas A&M 7-5 121.2140
46. Florida 6-5 119.2841
48. Tennessee 6-6 117.6812
61. South Carolina 6-6 109.3448
81. Kentucky 5-7 95.7120
113. Vanderbilt 3-9 65.0973

 

[bamawriter]

interesting about Ole Miss considering how (relatively) few 10-2 teams made the list.

Their opponents' winning % was .6197, and their opponents' adjusted scoring differential was 4.1138. It really bumped them up. Compare that to 10-2 Michigan State: .5011 and -2.9412.

 

[TrueOrange]

Here you go (with the final BCS/CFP for reference):

2011
1. LSU 13-0 (1)
2. Oklahoma St 11-1 (3)
3. Alabama 11-1 (2)
4. Boise St 11-1 (7)
5. Stanford 11-1 (4)
6. Houston 12-1 (19)
7. Oregon 11-2 (5)
8. Michigan 10-2 (13)
9. Wisconsin 11-2 (10)
10. Olahoma 9-3 (14)

2012 (this year was as similar to the BCS as my system got. I debated taking Ohio State out since they were ineligible for the postseason, but it turns out they wouldn't have made the BCSNCG or playoff anyway)
1. Notre Dame 12-0 (1)
2. Alabama 12-1 (2)
3. Florida 11-1 (3)
4. Oregon 11-1 (4)
5. Kansas St 11-1 (5)
6. Ohio St 12-0 (n/a)
7. Texas A&M 10-2 (9)
8. Stanford 11-2 (6)
9. Oklahoma 10-2 (11)
10. South Carolina 10-2 (10)

2013
1. Florida St 13-0 (1)
2. Baylor 11-1 (6)
3. Auburn 12-1 (2)
4. Alabama 11-1 (3)
5. Ohio St 12-1 (7)
6. Stanford 11-2 (5)
7. Missouri 11-2 (8)
8. Arizona St 10-3 (14)
9. Oklahoma St 10-2 (13)
10. Oregon 10-2 (10)

2014 (same top 4, but OSU jumped a lot)
1. Ohio St 12-1 (4)
2. Alabama 12-1 (1)
3. Oregon 12-1 (2)
4. Florida St 13-0 (3)
5. TCU 11-1 (6)
6. Baylor 11-1 (5)
7. Ole Miss 9-3 (9)
8. Boise St 11-2 (20)
9. Michigan St 10-2 (8)
10. Marshall 12-1 (not in top 25)

Did you make one for post bowl games? Just curious.

 

[bamawriter]

Did you make one for post bowl games? Just curious.

I didn't. Again, my idea was to compare the system to the BCS, which stopped after the conference title games. I could easily add the bowl games in, but I don't think you'd see major shifts in ranking since bowls tend to match up fairly even teams as far as record goes.

 

[bamawriter]

If anyone is curious as to how certain teams compared to others, i.e. how they got their scores, I can break that down for you.

 

[CoaxialOrange]

Very impressive

 

[n_huffhines]

This is cool but at a glance the one flaw i see is strength of schedule might not be very accurate (if your opponents have weak schedules or tough schedules it will skew). I don't know how else you would do it tho.

 

[bamawriter]

This is cool but at a glance the one flaw i see is strength of schedule might not be very accurate (if your opponents have weak schedules or tough schedules it will skew). I don't know how else you would do it tho.

I thought about that same issue. I initially included opponents' SOS, but that caused a problem where division / conference games essentially cancelled each other out. In a conference like the Big XII where they play a full round robin, if you include opponents' SOS, then only the three OOC games wind up significantly differentiating the teams. The round robin already throws a bit of a monkey wrench in because everyone's in conference % percentage plus the in conference opponents' winning % always equals .5000. By adding in opponents' SOS, that's just one more 50/50 factor that simply flattens out each team's score.

 

[LouderVol]

I thought about that same issue. I initially included opponents' SOS, but that caused a problem where division / conference games essentially cancelled each other out. In a conference like the Big XII where they play a full round robin, if you include opponents' SOS, then only the three OOC games wind up significantly differentiating the teams. The round robin already throws a bit of a monkey wrench in because everyone's in conference % percentage plus the in conference opponents' winning % always equals .5000. By adding in opponents' SOS, that's just one more 50/50 factor that simply flattens out each team's score.

then could you just do it for OOC?

 

[bamawriter]

then could you just do it for OOC?

I'll do it last year for the Big XII and see if it makes a significant difference.

Any suggestion on how to handle FCS teams? I went with the standard .2500 winning % because the average team that transitions from FCS to FBS has a .2500 winning % during their first three years. And I only included their games against FBS opponents in their score differential so that they are treated like a transitioning FBS team.

If I were to only include FBS opponents in an FCS team's SOS, then a team like Lamar will have an SOS of 95.25 because their only FBS opponent was Baylor.

I simply have a problem including an FCS opponent's full season when assessing an FBS team. I mean, Charleston Southern went undefeated against lower division opponents, but Alabama beat them by 50.

 

[TrueOrange]

If anyone is curious as to how certain teams compared to others, i.e. how they got their scores, I can break that down for you.

2013 Baylor. How much of that jump was just from margin of victory pushing up their ranking?

 

[bamawriter]

2013 Baylor. How much of that jump was just from margin of victory pushing up their ranking?

Pretty much all of it. Their SOS - (opponents' win % x 100) + opp. adj. score diff - was a paltry 46.6042 as compared to 59.8994 for Auburn. But Baylor's average score diff was 31.0833 and their average yardage difference was 280.5000 as compared to Auburn's 15.2308 and 81.8462.

 

[LouderVol]

I'll do it last year for the Big XII and see if it makes a significant difference.

Any suggestion on how to handle FCS teams? I went with the standard .2500 winning % because the average team that transitions from FCS to FBS has a .2500 winning % during their first three years. And I only included their games against FBS opponents in their score differential so that they are treated like a transitioning FBS team.

If I were to only include FBS opponents in an FCS team's SOS, then a team like Lamar will have an SOS of 95.25 because their only FBS opponent was Baylor.

I simply have a problem including an FCS opponent's full season when assessing an FBS team. I mean, Charleston Southern went undefeated against lower division opponents, but Alabama beat them by 50.

I would leave FCS as is/I can't think of a fix. FCS really shouldn't count towards much (so 25 makes sense), it rewards teams who don't play any at all (good thing in determining the best team) and most teams can only play 1 think so it shouldn't weight it too much.

 

[bamawriter]

Okay, I came up with a solution for the FCS SOS issue. I took the worst SOS in the country last year (Marshall at 33.6822) and subtracted half a point. That way, every FCS team winds up with 33.1822.

I added out of conference SOS to all 10 Big XII teams, and there were slight changes in their ranking.

Before:
1. Oklahoma - 174.9711
2. OK State - 145.1622
3. TCU - 143.1558
4. Baylor - 142.1281
5. WVU - 121.7578
6. Texas Tech - 117.2586
7. Kansas St - 102.2939
8. Texas - 100.2789
9. Iowa St - 83.2633
10. Kansas - 31.8594

After:
1. Oklahoma - 223.0788
2. TCU - 198.3230
3. OK State - 189.6407
4. Baylor - 186.6325
5. WVU - 168.0580
6. Texas Tech - 161.4007
7. Texas - 150.4109
8. Kansas St - 142.5151
9. Iowa St - 123.4199
10. Kansas - 76.3951

So TCU and OK State swapped places, as did Texas and KSU. TCU and OSU had the same 10-2 record, but Texas went 5-7 as compared to KSU at 6-6. KSU played an FCS team while Texas did not, and that ultimately made the difference between the two.

I'm not sure I love this math, because while TCU's opponents had a better average SOS than OK State's, TCU's opponents had a combined .2781 win % while OK State's had .3611. TCU's OOC opponents' had a better SOS because they sucked really bad, and thus helped their opponents' winning % and adjusted score diff.

But I suppose this illustrates my point. Trying to use raw math to determine who was better than whom is really dumb.

 

[malinoisvol]

Bama, sorry if I missed it, can you take last year's numbers, apply them to this year's schedules to come up with a preseason top 10?

Just for S&G's.

 

[bamawriter]

Bama, sorry if I missed it, can you take last year's numbers, apply them to this year's schedules to come up with a preseason top 10?

Just for S&G's.

I suppose I could, though it'll take a little time since there's no way for me to export the information easily. That said, I'm not sure it'll tell us much, because accounting for attrition is way more effort than I feel like putting in.