Writers of Pro Football Prospectus 2008

Most Recent FO Features

JohnsonCal09.jpg

» Scramble Over/Unders: the Norths

The league's northern divisions pose a number of meaty questions, such as: "Is the Bears' offense due for a repeat performance?" "Why do the Lions have such pronounced splits?" and "Has Johnny Manziel made the Cleveland brass even crazier?"

07 Oct 2008

FEI Week 6 Ratings

by Brian Fremeau

The Fremeau Efficiency Index principles and methodology can be found here. Like DVOA, FEI rewards playing well against good teams, win or lose, and punishes losing to poor teams more harshly than it rewards defeating poor teams. Unlike DVOA, it is drive-based, not play-by-play based, and it is specifically engineered to measure the college game.

FEI is the opponent-adjusted value of Game Efficiency, a measurement of the success rate of a team scoring and preventing opponent scoring throughout the non-garbage-time possessions of a game. Like DVOA, it represents a team's efficiency value over average.

Only games between FBS teams are considered. Since limited data is available at the beginning of the season, the ratings to date are a function of both actual games played and projected outcomes based on the 2008 Projected FEI Ratings. The weight given to projected outcomes is reduced each week until mid-October, at which point the projections will be eliminated entirely.

Rank Team Record FEI Last Week vs. Top 10 vs. Top 40
1 Missouri 4-0 0.274 9 0-0 1-0
2 Penn State 5-0 0.265 2 0-0 1-0
3 Oklahoma 4-0 0.258 1 0-0 1-0
4 USC 3-1 0.246 5 0-0 1-0
5 Virginia Tech 4-1 0.220 4 1-0 2-1
6 Alabama 6-0 0.216 7 0-0 2-0
7 Georgia Tech 3-1 0.216 17 0-1 0-1
8 Texas 5-0 0.210 6 0-0 0-0
9 Florida 4-1 0.211 3 0-0 0-1
10 Vanderbilt 5-0 0.210 14 0-0 3-0
11 North Carolina 3-1 0.190 22 0-1 0-1
12 Georgia 3-1 0.189 18 0-1 1-1
Rank Team Record FEI Last Week vs. Top 10 vs. Top 40
13 Ball State 5-0 0.186 32 0-0 1-0
14 Wake Forest 3-1 0.184 8 0-0 2-1
15 Mississippi 2-3 0.176 12 1-1 1-3
16 Texas Tech 3-0 0.174 19 0-0 0-0
17 LSU 3-0 0.164 16 0-0 1-0
18 Oklahoma State 4-0 0.164 20 0-0 0-0
19 Northwestern 4-0 0.154 23 0-0 1-0
20 Tulsa 4-0 0.152 28 0-0 0-0
21 South Carolina 3-2 0.145 46 0-1 1-2
22 Utah 5-0 0.144 15 0-0 0-0
23 Pittsburgh 4-1 0.143 50 0-0 2-0
24 BYU 4-0 0.138 11 0-0 0-0
25 Boise State 3-0 0.138 10 0-0 0-0

The complete Week 6 FEI Ratings for all 120 FBS teams can be found here.

Dear Oklahoma, Texas, Florida, LSU, Missouri, and Oklahoma State: Please make this the signature weekend of the 2008 college football season. As much as we have enjoyed watching six weeks of hyped-up showdowns flop and fizzle into blowout yawners, it's time for two top teams to show up in a game at the same time. Love, football fans everywhere.

Since 2003, teams from the year-end FEI top 10 have squared off 48 times, 37 times in the regular season. Unless you are an ACC homer, you're probably not counting on Virginia Tech and Georgia Tech to both be there at the top come bowl season this year. The September 20 LSU/Auburn game probably won't qualify either. Alabama/Georgia (September 27) and USC/Ohio State (September 13) may all rank in the end-of-year top 10, but those contests weren't exactly the stuff that instant classics are made of. Five Big 12 teams are currently undefeated and ranked in the top 20, and four of them square off this Saturday. Here's hoping they deliver.

Strength of Schedule

I have always had a problem with the conventional ways in which strength of schedule is measured. The NCAA produces a "Toughest Schedule" report, updated weekly, that simply totals the number of wins and losses recorded by each team's opponents and ranks teams according to opponent win percentage. A slightly more sophisticated variation on the same theme was employed by the BCS formula until 2003, combining both opponent win percentage and opponent-opponent win percentage to produce a schedule strength rating. The problem, of course, is that it is obvious to anyone who follows college football that win-loss records are not in and of themselves indicative of a team's strength. Even before last year's Sugar Bowl, did anyone actually believe Hawaii's undefeated regular season in 2007 suggested the Warriors were better than two-loss Georgia? Why, then, would playing Hawaii be deemed "tougher" than playing Georgia? Win-loss records are not devoid of meaning, but strength of schedule measures that rely exclusively on wins and losses as descriptive of a team's power are inherently flawed.

Many computer-based systems report schedule strength as the average rating of each team's set of opponents. This is much more palatable, but still unsatisfactory. First, as has been discussed in a previous column, FCS game data isn't as reliable as FBS data, and substituting a place-setter FCS rating into the equation only muddles the issue. Second, whether counting or discounting FCS games, the total number of games played by teams can be variable over the course of the season, particularly with the addition of conference championships and bowl games at season's end. If a team with an average schedule plays a below-average bowl opponent, its average strength of opponent rating goes down. Does it make sense that its schedule became "easier" by playing an additional game? Isn't playing an extra game against even below-average opponent more difficult (at least to some degree) than having a bye?

Even when the number of games isn't an issue, the question of schedule difficulty is more complicated than might be expected. Take a moment to consider two hypothetical schedules:

  • Schedule A: Six games against top 20 teams and six games against bottom 20 teams.
  • Schedule B: Twelve games against average teams.

Which hypothetical schedule is more difficult? It actually depends on who you ask. An Average team might argue that the schedules on the whole are relatively equal, and would probably expect to post a .500 record against either slate. An Elite team would consider Schedule A to be tougher because it would line up more often against capable competition, whereas an Elite team would likely dominate Schedule B. An Awful team, however, would consider Schedule A to be easier since it would get multiple chances to play teams of its caliber. In other words, even when playing identical schedules, every team might have a different perspective on how difficult that schedule truly is. So who is right?

Last year, I proposed that team schedule strength ought to be compared on an opponent-by-opponent basis from the top down. Games against better competition are given more weight in the FEI formula according to this principle. Does it make sense to take a privileged as opposed to average perspective on schedule strength? I think so. For one thing, most fans talk about difficult schedules almost exclusively in the context of top opponents (i.e. "Our team played two top 10 opponents, and six of the top 25"). Plus, schedule difficulty is intuitively a measure of how many challenges/pitfalls/roadblocks stand in the way of a team and an undefeated season. Since only the best teams ever do run the table, an Elite team perspective on schedule seems most appropriate.

The following Strength of Schedule (SOS) ratings were produced by calculating the Projected Win Expectation (PWE) of a hypothetical Elite team versus each opponent on every team's schedule. The principles for calculating PWE are detailed here. The SOS rating represents the likelihood of an Elite team running the table against the given team's schedule, therefore, a lower SOS rating represents a stronger schedule.


Top 10 Past SOS Top 10 Future SOS Top 10 Total SOS
Rank Team SOS FEI Rank Rank Team SOS FEI Rank Rank Team SOS FEI Rank
1 Oregon State 0.290 47 1 Texas 0.166 8 1 Baylor 0.091 86
2 Mississippi 0.293 15 2 Kansas 0.166 30 2 Arkansas 0.096 117
3 Illinois 0.314 37 3 Oklahoma State 0.169 18 3 Georgia 0.101 12
4 Arkansas 0.334 117 4 Georgia 0.192 12 4 Duke 0.114 48
5 Nebraska 0.355 49 5 Baylor 0.209 86 5 Virginia 0.124 108
6 Rice 0.375 92 6 LSU 0.215 17 6 Illinois 0.126 37
7 Syracuse 0.375 99 7 Virginia 0.217 108 7 Nebraska 0.126 49
8 South Carolina 0.383 21 8 Tennessee 0.240 57 8 Kansas 0.130 30
9 Western Kentucky 0.395 111 9 Duke 0.242 48 9 Mississippi 0.134 15
10 Nevada 0.416 88 10 Florida 0.244 9 10 Tennessee 0.134 57

FEI Top 25 Strength of Schedule Ratings
Rank Team Record FEI SOS to date SOS remain SOS total
1 Missouri 4-0 0.274 96 38 59
2 Penn State 5-0 0.265 93 45 62
3 Oklahoma 4-0 0.258 105 27 43
4 USC 3-1 0.246 84 75 87
5 Virginia Tech 4-1 0.220 12 87 46
6 Alabama 6-0 0.216 40 49 45
7 Georgia Tech 3-1 0.216 38 41 40
8 Texas 5-0 0.210 115 1 14
9 Florida 4-1 0.211 68 10 19
10 Vanderbilt 5-0 0.210 27 17 15
11 North Carolina 3-1 0.190 50 52 56
12 Georgia 3-1 0.189 26 4 3
Rank Team Record FEI SOS to date SOS remain SOS total
13 Ball State 5-0 0.186 98 105 116
14 Wake Forest 3-1 0.184 34 55 50
15 Mississippi 2-3 0.176 2 44 9
16 Texas Tech 3-0 0.174 120 12 38
17 LSU 3-0 0.164 109 6 25
18 Oklahoma State 4-0 0.164 110 3 13
19 Northwestern 4-0 0.154 89 53 71
20 Tulsa 4-0 0.152 119 107 120
21 South Carolina 3-2 0.145 8 39 20
22 Utah 5-0 0.144 97 88 103
23 Pittsburgh 4-1 0.143 62 89 78
24 BYU 4-0 0.138 118 74 111
25 Boise State 3-0 0.138 114 108 119

The complete list of strength of schedule rankings for all 120 FBS teams can be found here.

In summary: If Ball State, Tulsa, Boise State, Utah, and BYU are really good, they should have no trouble at all with the remainder of their opponents (with the exception of the Utah/BYU showdown at the end of the year). The SEC and Big 12 conferences account for 12 of the 15 toughest schedules over the remainder of the season. Georgia's already well-worn road isn't getting any easier, and this doesn't even take into account a potential SEC championship and BCS bowl game at the end of the year if they manage to qualify. I'll update the SOS numbers throughout the season as the landscape changes, and I welcome any feedback on this new methodology.

Posted by: Brian Fremeau on 07 Oct 2008

14 comments, Last at 14 Oct 2008, 12:39pm by emcee fleshy .SD-ATL.

Comments

1
by Becephalus :: Wed, 10/08/2008 - 4:30pm

nice post again, I would be worried if I were a Texas fan...

The Wire should win the Nobel prize for literature.

2
by Will :: Wed, 10/08/2008 - 6:53pm

You can tell it's still early in the season when you see this:

19 Northwestern

Will

5
by BrixtonBear (not verified) :: Thu, 10/09/2008 - 3:48pm

Our team starts 5-0 for the first time since Moses was a sophomore and you take it as an opportunity to piss on our bonfire? You can't even let us enjoy the sure-to-be-fleeting moment?

Tell me something, you get off on kicking over little kids' sand castles too?

7
by Will :: Fri, 10/10/2008 - 2:30am

Not pissing! My alma mater (Cornell) is 3-0 and I'm psyched about it, so I can definitely relate! You know it's early in the season when the Big Red are undefeated!

Will

3
by D :: Wed, 10/08/2008 - 7:50pm

U of I has the 6th hardest schedule in country and the hardest in the Big 10. At least we already played Missouri.

4
by Tom Gower :: Thu, 10/09/2008 - 2:14pm

I've often said that your PWE method of measuring strength of schedule is a better one, and have argued DVOA should be done the same way. Two questions, though:
1. Where did you come up with your measure of a hypothetical "Elite" team? Is this top 5, top 10, top 15, top 20? Any of these is potentially valuable, but they're not the same.
2. One problem with this SOS measure is it doesn't tell you how many games you expect to lose. Playing a top 5 team and the rest patsies (Hawaii this year, hypotehtically) is different from playing a couple top 15 teams and the rest ok. Same SOS potentially, but a greater number of potential losses. So, while you're changing the parameters of the problem you highlighted in your article, you're not eliminating it.

6
by Brian Fremeau :: Thu, 10/09/2008 - 5:21pm

NewsToTom, I should have included that definition of Elite in the column.

1. The Projected Win Expectations are a function of the power advantage of one team over the other, adjusted for home field advantage. The power advantage as described in a previous column is defined by the standard deviation of FEI (historically). The hypothetical Elite team is defined as two standard deviations better than average (0.000), or, an FEI rating of 0.291. This week, that team would be ranked #1. Since 2003, seven teams have finished with a higher FEI rating. Last year, LSU finished #1 with an FEI rating of 0.285, ninth-best all-time.

2. You're right, the SOS measure is a measure of the liklihood of an Elite team going undefeated. I could also run a projection on the expected record of an Elite team against the schedule as well, using PWE. Or are you interested in the expected record of the team itself?

10
by Tom Gower :: Fri, 10/10/2008 - 6:53pm

Brian, thanks for the response.

On #1, I admit I have no statistical basis for doing so, but I think that's too strong a definition of an Elite team. What'd be more useful, I think, is how would an "average top 5 team" fare against that kind of schedule, with this hypothetical "top 5 team" being based on average rating of the FEI top 5 for the past 3 or 5 years.

On #2, the two measures I'm most interested in are (1) the likelihood of an Elite team (however) defined going undefeated, and (2) the expected number of losses for that Elite team.

8
by bradluen :: Fri, 10/10/2008 - 4:11am

So what's the numerical SOS for Boise State (and BYU, and Utah...)? Seems like that would be useful information in determining whether they deserve a BCS bid or not.

9
by Brian Fremeau :: Fri, 10/10/2008 - 8:57am

The SOS for the five non-BCS teams in this week's FEI are all listed in the chart above. There is also a link to the SOS rankings for all teams.

11
by Joe Pisarcik Magnate (not verified) :: Sat, 10/11/2008 - 10:08am

What does (h) mean in the weekly forecasts? I use them in a Pick 'Em League that picks most of the games involving the Top 25. Unfortunately, that seems like a more efficient market than Wall Street these days.

12
by Brian Fremeau :: Sat, 10/11/2008 - 3:01pm

Joe --
I list the home team in all-caps. When the road team's name is an all-caps abbreviation (i.e. UCLA, UAB, SMU, etc ...), I designate the actual home team with (h).

13
by Joe Pisarcik Magnate (not verified) :: Sun, 10/12/2008 - 12:58pm

Thanks, Brian.

14
by emcee fleshy .S... :: Tue, 10/14/2008 - 12:39pm

Such a pleasure to see GaTech, my Alma Mater, severely over-rated for once. Beating Gardner-Webb by three should take care of that, right?

---
ATL's FOMBC Wasn't My Fault!