Writers of Pro Football Prospectus 2008

Most Recent FO Features

JohnsonKer18.jpg

» Seventh Day Adventure: Week 13

The biggest game this week is the Iron Bowl, where the playoff hopes of Alabama, Auburn, and Georgia hang in the balance.

09 Dec 2015

FEI Week 14: Sparty Yes

by Brian Fremeau

Though I stopped short of suggesting that Michigan State had eliminated itself from playoff contention after losing to Nebraska last month, I did devote the FEI analysis column four weeks ago to the significance of the fourth-quarter comeback the Spartans allowed that night. More than any team this season, MSU has been involved in high-stakes drama with major playoff implications. Aside from that stumble in Lincoln, the Spartans made the plays necessary to finish on the winning end of all of those dramatic situations.

The Big Ten Championship game last Saturday night set the stage for their latest high-wire act. Instead of a special teams score as time expired (as MSU had in wins over both Michigan and Ohio State), the Spartans leaned on their offense to drive for the winning points against the Iowa Hawkeyes. With 9:31 left to play in the game and trailing Iowa 13-9, Michigan State started a drive on its own 18-yard line. The longest drive in terms of yards earned by the Spartans at this point in the game traveled only 47 yards against the stout Iowa defense. On the season, Iowa's defense had faced 45 opponent drives begun at or inside the opponent's own 20-yard line and had given up only one touchdown in those situations all year.

The drive began, and it just kept moving. A 13-yard pass completion on third-and-4 from Michigan State's own 24-yard line produced the first first down. A 4-yard run on third-and-3 pushed the Spartans near midfield. A 16-yard completion on third-and-8 moved the ball into Iowa territory while the clock ticked under five minutes to play. Five plays later, a 2-yard run on third-and-1 gave the Spartans a first down in the red zone. Four plays after that, Michigan State ran for a first down on fourth-and-2 at the Iowa 5-yard line. And three plays later, Spartans running back L.J. Scott powered through tacklers and extended the ball across the goal line for the go-ahead score. The drive went 22 plays and 82 yards, lasted over nine minutes, and earned seven points for the win.

My numbers don't care about the length in terms of plays or game clock of drives -- though there is an indirect reward for ball-control drives that limit the total number of possessions in a game and therefore improve overall game efficiency. But I get pretty geeked out as a football fan regarding the possession details for drives that distinguish themselves in exceptional ways. Michigan State's 22-play drive against Iowa was the single longest drive in an FBS game this season (out of nearly 20,000 recorded to date), and it is tied for the third longest in college football of the last five seasons. And of any of the drives that were similarly long over the last five years, none had a conference championship and postseason playoff bid on the line.

There have been a total of 23 drives in FBS games since the start of the 2011 season that lasted at least 20 plays. Only eight of those drives reached the end zone -- that in and of itself might be kind of surprising, but drives of at least 20 plays are pretty good indicators of an offense that is struggling to get any big chunk plays against a defense, and sputtering in the red zone after that kind of methodical play is more likely than not. Including the Spartans, only ten of the teams that have had a drive last at least 20 plays went on to win the game over the last five years. And of that group, only Michigan State had a 20-plus-play touchdown drive in a win over a defense ranked in the top-30 of my opponent-adjusted DFEI ratings.

In other words, Michigan State's touchdown drive could be considered exceptional in a vacuum -- even without a Big Ten Championship and College Football Playoff berth on the line, the drive was worthy of our attention. Coupled with those stakes and those rewards, it was extraordinary.

FEI Degree of Difficulty through Week 14

FEI Degree of Difficulty (DOD) ratings are based on FEI ratings, but instead of measuring efficiency against schedule, DOD measures record against schedule. How difficult would it be for an elite team (two standard deviations better than average) to play a given team's schedule to date and achieve that team's record? This question is closely aligned with the post-season selection priorities for the College Football Playoff committee.

The table below includes ratings for each team's schedule strength to date (PSOS), the average number of wins an elite team would have against the given team's schedule to date (EMW), and the given team's record against FEI top 15 (v15) and FEI top 30 (v30) opponents. Overall schedule strength (SOS) is also provided. DOD ratings for all teams can be found here.

FEI Degree of Difficulty through Week 14
Rk Team W-L DOD PSOS Rk EMW v15 v30 SOS Rk CFP
1 Clemson 12-0 .309 .309 70 10.9 2-0 3-0 .309 70 1
2 Alabama 11-1 .538 .183 27 10.4 2-1 5-1 .183 27 2
3 Michigan State 12-1 .577 .191 29 11.6 1-0 4-0 .191 29 3
4 Oklahoma 11-1 .703 .270 58 10.8 3-0 4-0 .270 58 4
5 Iowa 11-1 .749 .310 71 11.0 0-1 0-1 .310 71 5
6 Notre Dame 10-2 .759 .091 3 10.2 1-2 2-2 .091 3 8
7 Ohio State 11-1 .775 .377 82 11.1 0-1 1-1 .377 82 7
8 Stanford 11-2 .873 .205 36 11.6 1-0 4-1 .205 36 6
9 Houston 11-1 .904 .541 106 11.4 1-0 2-0 .541 106 18
10 Northwestern 9-2 .912 .286 63 9.9 1-0 1-2 .286 63 13

For the second straight season, my hypothesis that FEI degree of difficulty closely matches the priorities of the College Football Playoff selection committee has proved to be true. The top four teams in the DOD ratings matched the top four teams according to the committee. This year's final rankings weren't very surprising, controversial, or even debatable, so I'm not going to take a victory lap on this one. But I do think this is validation of the DOD approach in terms of striking a balance between the questions of "best" and "most deserving" teams. That priority might not be the best way to pick playoff teams, but it is probably the most palatable for college football fans.

FEI Week 14 Ratings

The Fremeau Efficiency Index (FEI) is a college football rating system based on opponent-adjusted drive efficiency. Approximately 20,000 possessions are contested annually in FBS vs. FBS games. First-half clock-kills and end-of-game garbage drives and scores are filtered out. Unadjusted game efficiency (GE) is a measure of net success on non-garbage possessions, and opponent adjustments are calculated with special emphasis placed on quality performances against good teams, win or lose. Other definitions:

  • SOS: Strength of Schedule, measured as the likelihood of an elite team going undefeated against the given team's full schedule, including conference championship games. Schedule strength data based on FEI ratings and calculated across other dimensions can be explored in this interactive visualization.
  • FBS MW: Mean Wins, the average number of games a team with the given FEI rating would be expected to win against its schedule of FBS opponents.
  • FBS RMW: Remaining Mean Wins, the average number of games a team with the given FEI rating would be expected to win against the remainder of its schedule of FBS opponents.
  • OFEI: Opponent-adjusted Offensive Efficiency value generated per possession.
  • DFEI: Opponent-adjusted Defensive Efficiency value generated per opponent possession.
  • STE: Special Teams Efficiency value generated per game possession.
  • FVE: Field Value Efficiency value generated per game possession.

Preseason projection data receives no weight in this week's ratings. Ratings for all FBS teams can be found here.

Rk Team FBS
Rec
FEI LW GE Rk SOS Rk FBS
MW
FBS
RMW
OFEI Rk DFEI Rk STE Rk FVE Rk
1 Alabama 11-1 .285 1 .207 6 .183 27 10.7 .0 .49 29 1.07 2 -.02 83 .12 16
2 Clemson 12-0 .260 3 .205 7 .309 70 10.9 .0 .78 12 .85 7 -.03 89 -.12 111
3 Oklahoma 11-1 .249 2 .230 3 .270 58 10.6 .0 .74 15 .95 3 -.01 75 .08 27
4 Stanford 11-2 .247 4 .216 5 .205 36 11.3 .0 1.33 4 .08 60 .12 5 .17 7
5 Michigan State 12-1 .234 6 .139 21 .191 29 11.1 .0 .61 23 .66 15 -.06 108 .09 21
6 Notre Dame 10-2 .229 7 .166 15 .091 3 9.6 .0 1.07 6 .19 52 .10 9 .13 11
7 Ohio State 11-1 .227 5 .244 2 .377 82 10.7 .0 .54 26 .70 10 .04 39 .21 2
8 Mississippi 8-3 .188 8 .131 24 .091 4 7.9 .0 .81 10 .40 37 -.04 101 .00 69
9 North Carolina 9-2 .179 10 .195 8 .330 76 9.1 .0 .82 9 .12 56 .10 10 .09 20
10 TCU 9-2 .175 11 .110 28 .227 42 8.6 .0 .31 38 .53 22 .06 24 .07 35
11 Houston 11-1 .173 14 .219 4 .541 106 10.5 .0 .64 20 .49 27 .05 29 .18 6
12 Tennessee 7-4 .164 12 .129 25 .104 7 7.7 .0 .62 22 .21 50 .15 2 .19 5
13 Navy 8-2 .161 15 .187 9 .221 40 8.4 1.0 1.34 3 -.03 69 .00 69 -.03 77
14 Arkansas 6-5 .160 16 .060 46 .070 1 6.9 .0 1.28 5 -.43 101 -.05 105 .06 41
15 Baylor 8-3 .159 9 .186 10 .286 62 8.4 .0 .79 11 .49 28 -.04 96 -.04 82
Rk Team FBS
Rec
FEI LW GE Rk SOS Rk FBS
MW
FBS
RMW
OFEI Rk DFEI Rk STE Rk FVE Rk
16 Florida State 9-2 .158 17 .174 13 .243 47 8.6 .0 .53 27 .50 26 .04 36 .02 61
17 USC 8-5 .154 13 .114 27 .081 2 8.5 .0 .74 14 .47 30 .01 65 .05 47
18 LSU 8-3 .147 19 .082 38 .096 6 7.1 .0 .66 17 .36 39 -.03 88 -.02 76
19 Iowa 11-1 .147 18 .141 20 .310 71 9.2 .0 .39 31 .51 24 .04 37 .06 36
20 Memphis 8-3 .144 20 .129 26 .237 43 7.6 .0 .75 13 .18 53 .11 7 .08 24
21 Toledo 9-2 .144 21 .155 17 .454 97 8.9 .0 .25 41 .67 12 -.02 82 .04 51
22 Western Kentucky 11-2 .140 27 .258 1 .577 110 11.3 .0 1.46 1 .03 62 .03 42 .06 37
23 Michigan 9-3 .138 22 .172 14 .264 54 8.6 .0 .36 33 .58 20 .07 16 .06 38
24 Oregon 8-3 .135 25 .050 49 .106 9 6.9 .0 .88 8 -.23 87 .01 66 .05 46
25 Utah 9-3 .133 24 .097 33 .300 66 8.6 .0 -.17 82 .82 9 .10 11 .10 17
26 Mississippi State 7-4 .128 28 .056 48 .165 21 6.8 .0 .62 21 .20 51 .02 56 -.01 70
27 Washington 5-6 .128 26 .081 40 .178 26 7.1 .0 -.46 103 1.14 1 .07 15 .08 26
28 Bowling Green 10-3 .127 29 .178 11 .447 95 10.3 .0 1.41 2 -.11 79 -.09 119 -.04 78
29 Florida 10-3 .127 23 .081 41 .144 16 8.7 .0 -.03 60 .85 6 -.02 87 .20 4
30 Oklahoma State 9-2 .112 30 .092 34 .293 64 7.2 .0 .58 24 .22 49 .05 30 .07 33

Posted by: Brian Fremeau on 09 Dec 2015

9 comments, Last at 11 Dec 2015, 2:56am by Eleutheria

Comments

1
by techvet :: Wed, 12/09/2015 - 11:01pm

So with regards to Houston, why does FEI DOD vary so much with how the committee looked at the team? Also, since ND is ranked higher in FEI DOD than Stanford, does FEI DOD have any ability to accommodate head-to-head outcomes, or does it throw them aside? Obviously the committee considered head-to-head competition in ranking Stanford higher.

2
by Brian Fremeau :: Thu, 12/10/2015 - 9:44am

Good questions on FEI DOD. Through two seasons, I think FEI DOD is most closely aligned with how the committee treats the very top of their rankings, but doesn't necessarily align with how the rest of their rankings shake out. That's understandable. The committee may literally be applying different criteria when they discuss teams in the 7-15 or 16-25 range since due to their process: http://www.collegefootballplayoff.com/selection-committee-protocol

I also didn't design FEI DOD as a committee prediction tool, but rather wanted to produce a "record vs schedule" rating and then hypothesized that at the top, this would mirror the committee's priorities. If I were to adjust FEI DOD as a committee prediction tool, I probably would need to add in a factor for head-to-head results and conference championships. I'd also probably add in a factor to change the scale used to evaluate non-Power 5 teams since those teams have consistently been ranked by the committee somewhat lower than FEI DOD suggests they should be.

3
by RoninX :: Thu, 12/10/2015 - 12:21pm

Thanks for taking time to put that drive in context. As you say, add in the circumstance and how Iowa's defense had been playing to that point (and Cook's apparent shoulder injury reaggravation) and it becomes truly mind blowing.

4
by Eleutheria :: Thu, 12/10/2015 - 3:44pm

For those who are interested:

The R^2 of the top 25 DOD vs CFP (unranked teams are assumed to be ranked 28th so that they could be included in the data):
2014: .568
2015: .548
both: .546

A correlation of 0.74 is pretty strong and might be a nice starting point for a model on to predict CFP rankings. Other potential variables:
Weighted FEI (as I think CFP cares less about week 1 losses then week 10 losses)
Conference Championships
record vs top 25 DOD teams
anything else?

5
by skunkfunk :: Thu, 12/10/2015 - 4:08pm

Head to head results, perhaps a value for conference prestige or some such.

6
by Eleutheria :: Thu, 12/10/2015 - 7:20pm

Record against top 25 teams is head to head results.

7
by Brian Fremeau :: Thu, 12/10/2015 - 10:25pm

The suggestion might be to give extra weight to a head-to-head result for teams that otherwise have a similar profile (i.e. Stanford and Notre Dame have similar records and FEI DOD ratings, bump Stanford over ND due to head-to-head result).

9
by Eleutheria :: Fri, 12/11/2015 - 2:56am

but I assume that's a little too subjective for a statistical model.

How close in DOD does two teams have to be for head-to-head to take effect? And should head-to-head outside of those being included or just weighted less?

I think it would probably be much simpler, and not that less predictive, to just do total record between top teams, rather then looking exclusively at teams that are closely ranked.

I could be mistaken though.

8
by Brian Fremeau :: Thu, 12/10/2015 - 10:26pm

Interesting stuff, and I have been meaning to play around with Weighted FEI this off-season.