≡ Menu

Week Three College Football Ratings (2014)

Mariota and the Ducks look as good as any team in the country

Mariota and the Ducks look as good as any team in the country.

Regular readers know that I publish weekly college football ratings using the Simple Rating System. The catch is that the SRS isn’t a viable option in the first few weeks of the season; until we have more interaction among the top teams, we can’t really generate computer ratings.  Frankly, running an SRS program today would be pretty useless.

Consider that a team like Arizona State has played Weber State, New Mexico, and Colorado. Auburn has played Arkansas (the Razorbacks are not very good) and San Jose State. Oklahoma has played Louisiana Tech, Tulsa, and Tennessee (the Vols are not very good). So what can we do?

One thing we could do is to use the concept of Elo Ratings. But calculating Elo ratings in this context is no simple task, and there’s a good chance my buddy Neil is going to do that, anyway, so I thought I would try simpler process. I’ll give a high-level overview of the process here, then present the rankings, and then provide all the nuts and bolts for those interested at the bottom of the post.

First, we need to start with pre-season ratings. Any set of preseason ratings would work, but I used Brian Fremeau’s FEI ratings. Next, I converted those ratings into SRS scores.1 That gave me a set of preseason set of SRS ratings. From there, I discarded all games that looked like a mismatch prior to the game, assuming the better team won.  For all other games, I modified each team’s rating following the result of that game, with 85% of the new rating coming from the old rating, and 15% coming from that single game.

Below are the ratings through three weeks. As always thanks to Dr. Peter R. Wolfe for providing the weekly game logs.
Post-Week 3 SRS Ratings

Wk 3 RkTeamWk 3 SRSRecConfConf RkPres RkPres SRSDiff
3Florida St61.72-0ACC1362.4-0.7
8Texas A&M55.13-0SEC418505.1
11Notre Dame52.93-0Ind12049.13.9
12Michigan St52.91-1B101953.8-1
13Brigham Young52.53-0Ind23146.75.8
14Oklahoma St52.52-1B1221451.80.7
15South Carolina52.22-1SEC7656.8-4.6
17Arizona St51.23-0P1241551.20
18Southern Cal50.52-1P1251152.5-2
19Kansas St50.22-0B1231750.20
23Ohio State48.82-1B1031352-3.2
25Oregon St47.12-0P1272647.10
30Mississippi St45.43-0SEC93645.40
31Virginia Tech45.32-1ACC43545.6-0.4
33West Virginia44.92-1B1265138.46.5
36Boise St43.72-1MWC12747-3.3
37Central Florida43.50-2Amer12447.8-4.3
41North Carolina41.92-0ACC53843.6-1.7
44Georgia Tech413-0ACC643410
45Penn State40.63-0B1074838.91.7
46Miami FL40.42-1ACC74441-0.6
49Northern Illinois40.23-0MAC15038.71.4
54Texas Tech37.32-1B1284242.3-5
55East Carolina372-1Amer46433.13.9
57North Carolina St35.93-0ACC1273314.9
62Texas-San Antonio35.41-2CUSA17828.76.6
64Iowa St351-2B1296333.51.5
67Washington St34.91-2P12105237.7-2.8
70Boston College34.22-1ACC137430.63.5
72San Diego St32.81-1MWC38027.94.9
73Louisiana Tech32.42-1CUSA310322.510
75Bowling Green30.82-1MAC35834.9-4.1
77Colorado St29.62-1MWC48426.43.1
78Utah St29.32-1MWC56233.7-4.5
79Western Kentucky28.71-2CUSA48925.63.1
82Florida Atlantic28.11-2CUSA59923.24.8
85South Alabama27.41-1Sun187261.4
86North Texas271-2CUSA68825.81.2
87Fresno St260-3MWC66732.8-6.8
88Arkansas St261-2Sun29424.51.6
90Middle Tennessee St25.22-1CUSA79723.51.7
93San José St24.71-1MWC79324.70
95South Florida23.71-2Amer107728.8-5.1
96Central Michigan23.32-1MAC49025.3-2
97Ball St22.91-2MAC58327.3-4.4
100Wake Forest21.61-2ACC149823.3-1.7
104Air Force21.22-1MWC911018.92.3
106Kent St20.90-3MAC89524.2-3.3
108Western Michigan20.61-1MAC910719.41.2
113Ohio U.19.41-2MAC1011417.42
115Florida Int'l18.51-2CUSA1111218.6-0.1
116Southern Miss17.21-2CUSA1211118.7-1.5
117Texas St-San Marcos15.41-1Sun511815.40
118Old Dominion152-1CUSA13121141
120New Mexico13.60-2MWC1212014.8-1.2
121Appalachian St12.11-1Sun712312.10
122Georgia Southern11.31-2Sun812411.30
123Georgia St111-2Sun912213.1-2.1
125Miami OH10.20-3MAC1112510.20
126New Mexico St9.52-1Sun111283.85.7
128Eastern Michigan7.11-2MAC131268.6-1.5
  • It’s important to understand some of the flaws in this methodology.  If two teams have very different ratings, and they play a close game, that might provide more insight into the teams’ strengths than this methodology assumes.  That’s most clear with Georgia Southern.  Right now, GSU is playing more like a top-60 team than a bottom-feeder.  If the Eagles can keep this up, they will look pretty good when we run the first iteration of real SRS ratings.
  • Baylor was not given a high pre-season rating, but so far, the Bears are dominating terrible teams just like they did a year ago.  This methodology gives them no credit for crushing SMU, Northwestern State, or Buffalo — which, I think, is a pretty reasonable way to handle things — but it’s worth monitoring.
  • Memphis (7-point loss in the Rose Bowl against UCLA) and Hawaii (close losses to Washington and Oregon State) are two other teams that might be underrating by eliminating games where one team appeared far superior to the other, but escaped with only a close win.
  • Vanderbilt already dropped because of the loss to Temple, but discarded games like a 41-3 loss to Ole Miss and barely escaping at home against UMASS means the Commodores are even worse than this system thinks.
  • Texas Tech beat FCS Central Arkansas and UTEP by a combined 11 points in the first two games, but both scores were discarded.  After getting blown out by Arkansas in week 3, the Red Raiders dropped in the rankings, but maybe not far enough.
  • Washington (close wins against Hawaii and Eastern Washington), Iowa (+12 in points differential in home games against FCS Northern Iowa and a not-very-good Ball State), Oregon State (only +23 against Portland State and Hawaii), and UCLA (+15 against Virginia and Memphis) may all be overrated, too. On the other hand, Michigan State beat Western Michigan by 13 and USF by only 15 to start the 2013 season, before winning the Rose Bowl.  As a result, there’s at least some logic in not downgrading good teams for playing down to the level of competition early in the year.
  • There are 19 teams whose ratings have not budged at all, as a result of either three discarded games or two discarded games and a bye. Leading the way here are Alabama, Oklahoma, Auburn, Arizona State, Kansas State, Baylor, Washington, and Oregon State. While that might seem like a lot, how much more do we know about those teams than we did before the season?
  • BYU, Missouri, Texas A&M, and Notre Dame are the teams whose stocks have risen the most in the first three weeks, at least among top 25 teams.

Methodology Explained

Let’s get into the finer details of the system, which is mostly for the two or three of you who are curious, and for future me when I review this post.

Week 1

There are 128 teams in the Football Bowl Subdivision. Six of those teams did not play in week one.2 Of the remaining 122 teams, 46 played an opponent that was at least 15 points worse in the SRS — it may not have been a cupcake, but a game that we would have expected the better team to win. It turns out, teams went 45-1 in these situations, with Texas-San Antonio beating Houston 27-7.

For the other 45 teams, the margins of victory varied wildly. Texas State beat Arkansas-Pine Bluff 65-0. Memphis beat Austin Peay 63-0. Oregon beat South Dakota 62-13. These games tell us very little that we did not already know. Texas Tech squeaked by Central Arkansas, 42-35. South Florida beat Western Carolina 36-31, and North Carolina State edged out Georgia Southern, 24-23. These games probably do tell us quite a bit more about these teams, but I decided to simply not make any adjustments in those games. So Rule #1 is that if Team A plays Team B, Team A has a rating at least 15 points better than Team B, and Team A wins, neither team’s rating will change following the game.

Of those 45 games, 16 came against FBS schools, which means 16 other schools qualify as “Team Bs” here and don’t have their ratings move, either. Add in the six teams that were off, and 67 of the 128 teams will see no changes in their ratings after week one.

For all other games (which would include the Houston-UTSA game), we calculate a single-game SRS rating in the usual process: three points are awarded for home field, and the margin of victory is added to the opponent’s SRS rating. Then, we get to rule #2: For each team, their SRS rating moves by 15% based on the result of the game. Why 15%? That number just felt right to me; it was not the result of exhausting research, but it seemed like it struck the right balance.

Let’s use the Temple/Vanderbilt game as an example. Prior to week 1, Vanderbilt had an SRS rating of 35.8 and Temple was at 31.7. In week 1, the Owls defeated the Commodores in Vanderbilt by 30 points. If we believed the pre-week 1 ratings, that would mean Temple would have an implied SRS rating of 68.8 points from this game, since it beat a 35.8-point team by 30 on the road. Therefore, to calculate Temple’s post-week 1 rating, we use 85 parts 31.7 and 15 parts 68.8; that yields a new rating of 37.2. That moved Temple up from 71st to 54th in the SRS ratings.

Does that feel like it strikes the right balance to you? Vanderbilt dropped from 56th to 77th. Of course, most teams saw little movement in the rankings. Here were the 16 teams that jumped at least 5 spots after week one:

Pre-Wk 1 RkTeamPost-Wk 1 RkDiffWk 1 OppWeek 1 Score
78Texas-San Antonio6018Houston27-7
66Minnesota5214Eastern Illinois42-20
75Tennessee6114Utah St38-7
89Western Kentucky7514Bowling Green59-31
18Texas A&M711South Carolina52-28
30Mississippi1911Boise St35-13
57Toledo489New Hampshire54-20
94Arkansas St859Montana St37-10
31Brigham Young238Connecticut35-10
80San Diego St737Northern Arizona38-7
48Penn State435Central Florida26-24

Of course, the biggest move there in reality was Texas A&M; Kenny Hill threw for 511 yards and 3 touchdowns (on 60 passes) to help the Aggies pull off the upset blowout over South Carolina. Obviously week one was not a good one for the state of South Carolina. The Gamecocks fell from 6th to 15th in the SRS, while Clemson dropped from 23rd to 33rd after losing by 24 to Georgia. This next table shows the teams that saw the biggest decline in the rankings in week 1.

Pre-Wk 1 RkTeamPost-Wk 1 RkDiffWk 1 OppWeek 1 Score
58Bowling Green79-21Western Kentucky31-59
62Utah St81-19Tennessee7-38
32Houston45-13Texas-San Antonio7-27
6South Carolina15-9Texas A&M28-52
27Boise St36-9Mississippi13-35
70Connecticut78-8Brigham Young10-35
63Iowa St70-7North Dakota St14-34
95Kent St102-7Ohio U.14-17
107Western Michigan113-6Purdue34-43
24Central Florida29-5Penn State24-26
52Washington St57-5Rutgers38-41
77South Florida82-5Western Carolina36-31
98Wake Forest103-5Louisiana-Monroe10-17

Week 2

Nine FBS teams were off in week 2.3 In addition, all 40 teams that had an SRS rating at least 15 points better than its opponent won in week two. Again, none of these games move the SRS needle, whether it’s Georgia Southern (83-9 over Savannah State), Texas A&M (73-3 over Lamar) or Clemson (73-7 over South Carolina State) beating up cupcakes, or UNLV (13-12 over Northern Colorado), Texas Tech (30-26 over UTEP), Iowa (17-13 over Ball State), or Kansas State (32-28 over Iowa State) squeaking by inferior opponents.4 Of those 40 games, 18 came against FBS opponents; as a result, 67 teams will see no ratings change in week two.

For the other 61 teams, we again calculate single-game SRS ratings; then, to calculate post-week 2 ratings, we take 85% of the post-week 1 rating and 15% of the single-game SRS rating from week 2. Let’s use the Virginia Tech-Ohio State game as an example. Prior to week 2, the Buckeyes had an SRS rating of 52.3, and they hosted Virginia Tech (Post-Week 1 SRS rating of 45.6). Since Virginia Tech won by 14 in Columbus, the Hokies get a rating of 69.3 for that game. As a result, Virginia Tech’s post-week 1 rating increased to 49.2. You might think Virginia Tech’s rating would have increased by more than just 3.6 points, but Ohio State was simply not regarded as that good of a team. As for OSU, they receive an SRS rating of 28.6 for the game, dropping them to 48.8 for the year.5

Here’s a list of the teams that moved up the most after week two. The other notable increase was Brigham Young, jumping from 23rd to 11th after humiliating Texas for the second consecutive year. The table below shows all team to jump by 5 spots after week 2:

Post-Wk 1 RkTeamPost-Wk 2 RkDiffWk 2 OppWeek 2 Score
51West Virginia3417Towson54-0
90North Texas7317SMU43-6
101Louisiana Tech8417Louisiana-Lafayette48-20
23Brigham Young1112Texas41-7
32Virginia Tech239Ohio State35-21
22Notre Dame148Michigan31-0
40Utah328Fresno St59-27
91Central Michigan838Purdue38-17
69Nevada645Washington St24-13
86Illinois815Western Kentucky42-34

Okay, one potential flaw in the metric. West Virginia crushed FCS Towson, but still moves up significantly in the ratings. Why? Well, Towson was an excellent team last year: the Tigers were #5 in the FCS regular season ratings, and then made it to the FCS Championship Game. So West Virginia looks like they beat an excellent FCS team by 54 points, which would be worthy of such a jump in the ratings.6 In general, this probably isn’t good, but I’m okay with it because West Virginia received no credit for playing Alabama close in week 1 (that game was discarded because Alabama was the much better team, and won). The Mountaineers then won in week 3 in Maryland, so it does appear as though West Virginia probably is a top-40 team.

The next table shows the teams that dropped the most in week 2. As you probably know, Week 2 was not a good week for the Big Ten. Michigan State fell from 10 to 15 after losing in Oregon; Ohio State dropped from 14 to 24, and Michigan fell from 28 to 39 after being blanked by Notre Dame. Even Purdue managed to drop 10 spots, falling from 10 to 110 after losing by 21 to Central Michigan.

Post-Wk 1 RkTeamPost-Wk 2 RkDiffWk 2 OppWeek 2 Score
71Louisiana-Lafayette91-20Louisiana Tech20-48
88SMU103-15North Texas6-43
30Texas43-13Brigham Young7-41
28Michigan39-11Notre Dame0-31
14Ohio State24-10Virginia Tech21-35
100Purdue110-10Central Michigan17-38
67Fresno St76-9Utah27-59
78Connecticut85-7Stony Brook19-16
10Michigan St15-5Oregon27-46
57Washington St62-5Nevada13-24
85Arkansas St90-5Tennessee19-34

That brings us to week 3. This week, 19 teams were off, with Florida State, Auburn, Clemson, Michigan State, Kansas State, Wisconsin, and Oregon State among the more notable absences. There were 29 games between teams with a gap of greater than 15 SRS points; unlike in prior weeks, the underdogs fared better on Saturday, winning 4 times.

  • FCS Indiana State defeated Ball State, 27-20. Ball State went 10-2 in the regular season last year, and was up 13-3 against Iowa two weeks ago with three minutes left before losing. Now, I think that game is yet another ugly mark against the B1G.
  • East Carolina upset Virginia Tech. Good work by the Solid Verbal, a fun podcast that had been talking up ECU before the season. Oh, and this is yet another bad result for the B1G.
  • Iowa lost 20-17 to Iowa State. At home. In two home games against Iowa State and FCS North Iowa, the Hawkeyes are +5. I think you can figure out what this means for the Big Ten.
  • USC, fresh off a monster win against Stanford, lost after traveling across the country to face Boston College. The crazy stat from this game: the Eagles rushed for 452 yards on 54 carries, while the Trojans rushed for 29 yards (including sacks) on 20 carries.

Three of the 29 mismatches in week 3 came against FCS schools, which means 26 FBS schools were heavy underdogs (and with ECU, Iowa State, and BC pulling off upsets). That means there were 67 teams (19 teams off, 25 heavy favorites won, 23 heavy dogs lost) whose ratings won’t change in week three.

For the other 61 teams, we again calculate single-game SRS ratings; then, to calculate post-week 3 ratings, we take 85% of the post-week 2 rating and 15% of the single-game SRS rating from week 2. Let’s use Missouri as an example. The Tigers had a rating of 53.1 entering week 3, and hosted Central Florida on Saturday. UCF had a rating of 46.2 entering week 3; since MIZZOU won 38-10, that gives the Tigers an implied rating of 71.2. As a result, Missouri’s post-week 3 rating jumps to 55.8.

Post-Wk 2 RkTeamPost-Wk 3 RkDiffWk 3 OppWeek 3 Score
68Arkansas5216Texas Tech49-28
66Syracuse5115Central Michigan40-3
72North Carolina St5715South Florida49-17
97Florida Atlantic8215Tulsa50-21
38Nebraska2612Fresno St55-19
67East Carolina5512Virginia Tech28-21
84Louisiana Tech7311North Texas42-21
86Colorado St779UC-Davis49-21
92Tulane839SE Louisiana St35-20
13Missouri67Central Florida38-10
70Iowa St646Iowa20-17
48Houston435Brigham Young25-33
75Boston College705Southern Cal37-31
95Middle Tennessee St905Western Kentucky50-47
113Western Michigan1085Idaho45-33

After a discarded week 1 game against FCS South Dakota State (which only resulted in a 20-point win), Missouri has now vaulted in the rankings after two dominant wins against Central Florida and Toledo (in retrospect, both of those schools may have been a bit overrated coming in to the year). The Tigers get Indiana at home next week, so we’ll have to wait two weeks — when they travel to South Carolina — to find out if Missouri is for real.

The biggest decliners in week 3?

Post-Wk 2 RkTeamPost-Wk 3 RkDiffWk 3 OppWeek 3 Score
40Texas Tech54-14Arkansas28-49
78Tulsa92-14Florida Atlantic21-50
82South Florida95-13North Carolina St17-49
83Central Michigan96-13Syracuse3-40
73North Texas86-13Louisiana Tech21-42
76Fresno St87-11Nebraska19-55
89Ball St97-8Indiana St20-27
27Iowa35-8Iowa St17-20
29Central Florida37-8Missouri10-38
10Southern Cal18-8Boston College31-37
23Virginia Tech31-8East Carolina21-28
63Marshall69-6Ohio U.44-14
54Indiana60-6Bowling Green42-45
62Washington St67-5Portland St59-21
61Rutgers66-5Penn State10-13

USC falls from 10th to 18th with the loss in Chestnut Hill. Virginia Tech, Iowa, and UCF drop from the 20s to the 30s, while Texas Tech’s bad play finally catches up with the Red Raiders, who fall to 54th.

  1. Using the following formula, based on a regression (R^2 = 0.98): SRS score = 34.2 + 104.2 * FEI rating. For non-FBS teams, I used their ratings at the end of the 2013 season. []
  2. Florida and Idaho had their game suspended after just one play; they joined Cincinnati, South Alabama, Army, and Kansas with de facto week 1 byes. []
  3. The nine: Georgia, Central Florida, TCU, Indiana, Syracuse, Rice, Western Michigan, Texas State, and…. Cincinnati. The Bearcats, due to some scheduling mishaps, took the first two weeks off. []
  4. A couple of notes: First, Georgia Southern is being massively underrated here. The Eagles lost at NC State and at Georgia Tech (in week 3) by a combined 5 points. In the abstract, beating Savannah State by 72 doesn’t mean much, but combined with the close games against ACC schools, and we can safely conclude that GSU is actually a pretty decent team. Texas Tech is not very good. The Red Raiders beat FCS Central Arkansas and UTEP by a combined 11 points; despite those games being unimpressive, they did not move in these ratings because they managed to win. You will not be surprised to learn that Texas Tech then lost by 21 at home to Arkansas in week 3. []
  5. One might quibble with Virginia Tech being only percentage points ahead of OSU. If one wants to quibble over one result in a simple system based on rough approximation, that’s what the comments are for! []
  6. As it turns out, Towson may not be very good this year. They lost to Central Connecticut State in week 1, and Central Connecticut State was a bad FCS team last year. One solution would be for me to downgrade Towson after losing to an FCS team in week 1, but there’s a limit to the amount of time I was willing to spend on this exercise. Yes, I’m pulling the Good Enougher card. []
  • Ty

    I’d be interested in seeing how North Dakota State is. I think they would be a top 50 FBS team.

  • I don’t think it would be too hard to set up an Elo system in Excel with all if the NCAA teams. I finally figured out a simple way to do it for the 32 pro teams, and it takes about 5 minutes to set up and another 5 to maintain each week. You could try to come up with approximate ratings for FCS teams, or you could just have a dummy variable (each FCS is a 1000 where 1500 is average).

    I did one using FO’s DAVE and last year’s DVOA, and my results are really similar to 538’s. I’d be happy to share my personal methodology if you’re interested.