FEI: Guaranteed Points
by Brian Fremeau
The South Carolina Gamecocks took down No. 1 Alabama on Saturday in impressive fashion, an outcome that may have a dramatic effect on the rest of the college football season. As Bill Connelly and Rob Weintraub have pointed out, the Crimson Tide were beaten physically right out of the gate and South Carolina's Stephen Garcia quarterbacked a near-perfect game. Alabama's prolific offense turned one-dimensional and sputtered down the stretch. They weren't destroyed or run out of the building, but Alabama was stripped of its invincibility armor and its iron-clad grip on a BCS Championship Game berth. At least a dozen other programs have improved chances for a national championship run.
The Oregon Ducks take over the top spot in this week's FEI ratings, but the Boise State Broncos are right on their heels. This is the final week in which preseason data is included in the ratings (roughly accounting for 10 percent of each team's current FEI rating), and it's hard to argue against the Broncos' resume to date. The obstacles Boise State faces in terms of earning national respect due to its strength of schedule have been well documented in this column and elsewhere. What the Broncos will need to do to maintain their FEI rating is dominate the remainder of their schedule.
I ran a set of projections based on the current FEI ratings for all teams, and Boise State will likely have to improve in the second half of the season in order to stay ahead of most of the power conference teams. The SEC, Big 12, Pac-10, and Big Ten contenders will all receive schedule boosts that the Broncos can't match. That said, those power conference teams aren't nearly as likely to run the table as Boise. Based on current FEI ratings and projected win expectations, there is now a 66 percent chance the Broncos will finish 12-0. That dwarfs the undefeated likelihoods for Oregon (22 percent), Auburn (14 pct.), LSU (12 pct.), TCU (57 pct.), Nebraska (29 pct.), Oklahoma (22 pct.), and Ohio State (29 pct.) -- not to mention any of the other undefeated teams not ranked in the FEI Top 10.
South Carolina ranks right behind Boise State this week and has a 31 percent chance of running the table the rest of the way. If they can complete that run and beat the SEC West champ in the conference title game, they'll be a very appealing candidate to surpass Boise State in the BCS standings. But none of the undefeated likelihoods mentioned here include potential conference championship game opponents. Whoever emerges from the SEC West contenders -- Auburn, LSU, and Alabama -- would slash South Carolina's chances at a 12-1 record in half. Ditto for Oklahoma and Nebraska, the likely participants in the Big 12 championship.
The highest ranked two-loss team (that is, two losses to FBS opponents, of course) is Oregon State. The Beavers happen to have the nation's No. 1 strength of schedule, not only for the entire season but also to date. Oregon State was competitive in losses to Boise State and TCU and knocked off No. 16 Arizona on Saturday -- all of which were road games. They have an opportunity to put together a run in the next month or so, and under Mike Riley, they've had a history of shrugging off slow starts. But they'll wrap up the year at No. 4 Stanford and at home (finally) against No. 1 Oregon. An elite team would have only a 47 percent chance of navigating that schedule with two or fewer losses. If the FEI ratings hold, Oregon State's schedule may rate as the toughest I've ever measured.
As Oregon State and many others experienced over the weekend, field goal kicking is a critical component of any team's blueprint for victory. Oregon State beat Arizona by two points, 29-27. Each team attempted a field goal in the game. The Beavers made theirs, and the Wildcats missed. There were plenty of other moments that could have changed the outcome -- turnovers, missed extra points, etc -- but there's nothing else quite like an "automatic" three points to make all the difference in the world.
Next week, I'll debut the first offensive and defensive efficiency ratings for the season based on drive data collected to date. One of the key components of those ratings is the way I assign a value to the success of an offense in a series versus the value that should be attributed elsewhere. An offense that reaches the end zone has "earned" 6.96 points on the drive, equal to the national average value of all touchdown-scoring possessions. The extra point earned by the kicking game is worth .04 points on the scoreboard. A missed extra point costs a team .96 points earned by the offense.
For drives that do not reach the end zone but conclude within field goal range, I calculate the national average success rate for converting those opportunities into field goals. This is not strictly determined by a field goal percentage regression, however. We must also account for coaching decisions to punt or go for it in field goal range. The national drive ending value of the opponent's 35-yard-line, for instance, is equal to the number of successful 52-yard field goals kicked in college football, divided by the sum of the field goal attempts, punts and fourth-down attempts from that yard line.
Let's take a closer look at the two field goals attempted in the Oregon State vs. Arizona game. With the game knotted at seven points apiece in the first quarter, the Beavers drove from their own 24-yard line to the Wildcats' 6-yard line. The offense earned a drive-ending value of 2.64 points. The Oregon State field goal unit and Beavers kicker Justin Kahut completed the three-point play by earning 0.36 points with a successful kick. At the end of the first half, Arizona drove from its own 2-yard line to the Oregon State 20-yard line, earning a drive-ending value of 2.18 points. The failed kick by Arizona's field goal unit and kicker Alex Zendejas cost the Wildcats 2.18 points.
(Note that the drive-ending value earned by the offense is only part of the calculation of a team's offensive efficiency. Drive-ending value earned over starting field position value is the formula for offensive efficiency).
We can thusly calculate the value that every field goal kicker/unit contributes to his team per attempt. Though the kicker himself is most accountable for the success or failure of a kick, other members of the unit can have an impact on success rate through poor snaps, holds, or blocking. The following Field Goal Value (FGV) ratings are most accurately assigned to the team and not the individual kicker(s) involved.
|Field Goal Value (FGV) Per Attempt|
|2||Middle Tennessee||4||1.00||1.16||42||East Carolina||6||.83||.25||82||Utah State||9||.56||-.27|
|3||Southern Mississippi||12||1.00||1.08||43||Tulane||5||.80||.24||83||Kent State||4||.50||-.29|
|5||Kansas State||6||1.00||1.02||45||Ohio State||13||.77||.19||85||North Carolina||4||.75||-.31|
|7||Georgia Tech||9||1.00||.95||47||Northern Illinois||10||.80||.19||87||Tulsa||13||.62||-.34|
|10||Kansas||1||1.00||.82||50||Penn State||12||.83||.19||90||Arizona State||10||.60||-.39|
|14||Wyoming||2||1.00||.67||54||Iowa State||7||.71||.14||94||Ball State||4||.50||-.52|
|16||Oregon||4||1.00||.65||56||Florida International||4||.75||.13||96||Texas Tech||7||.43||-.58|
|19||Baylor||10||.90||.63||59||Troy||13||.77||.09||99||San Diego State||6||.50||-.70|
|20||Eastern Michigan||3||1.00||.62||60||Washington State||4||.50||.08||100||Buffalo||6||.50||-.71|
|21||Nebraska||2||1.00||.62||61||North Carolina State||9||.78||.08||101||New Mexico||2||.50||-.73|
|23||Washington||8||.88||.59||63||New Mexico State||5||.80||.04||103||Mississippi State||5||.40||-.81|
|25||Georgia||12||.92||.56||65||Western Michigan||5||.80||.02||105||Bowling Green||8||.50||-.86|
|32||Boise State||10||.90||.43||72||Vanderbilt||4||.75||-.08||112||Western Kentucky||5||.20||-1.12|
|38||Virginia Tech||7||.86||.28||78||San Jose State||6||.67||-.23||118||Louisiana Lafayette||6||.17||-1.52|
|40||North Texas||7||.86||.25||80||Arkansas State||9||.67||-.26||120||Virginia||3||.00||-1.70|
Two things not taken into consideration here are weather and wind conditions. Nevertheless, I believe this is the best way to evaluate and compare field goal success rates across college football -- certainly better than looking at field goal percentage alone. I'll update the FGV data and include it in the weekly team tables going forward this year.
Three and Out
This section of the weekly FEI column features a set of three offensive and defensive Top 10 data tables sliced from the raw possession efficiency data I collect each week. None of these splits are explicit factors used in FEI, but they may provide a unique perspective on the drive success rates in college football.
Week 3: Three-and-outs, Available Yards, and Explosive Drives
Week 4: Reaching the Red Zone, Methodical Drives, and Late and Close Efficiency
Week 5: Converting 10+ Yard Drives Into Scores, Points Per Possession, and Scoring After Three-and-outs
The following tables include only non-garbage drives from FBS games.
|Yards Per Touchdown Drive|
|Offensive Leaders||Defensive Leaders|
|Louisiana Monroe||6||427||71.2||Fresno State||16||1125||70.3|
I've typically been providing the Top 10 in each category in Three and Out, but I thought this particular metric was interesting on both ends. The average touchdown drive is 62.6 yards. Clearly, teams like Central Michigan and Missouri have had to drive far farther on their touchdowns than Army, Memphis, or Rutgers. On the defensive end, the differences are even more stark. Iowa's defense has surrendered only two non-garbage touchdowns this year, both to Arizona, on drives of 72 and 8 yards.
|Playing With A Two-Score lead||Playing With a Three-Score Lead|
|Boise State||56||93||.602||Ohio State||49||124||.395|
|North Carolina State||68||123||.553||Utah||32||97||.330|
|Florida State||50||95||.526||Florida State||29||95||.305|
Instead of breaking down offenses and defenses separately, I thought it would be interesting to see which teams are playing with a comfortable lead offensively or defensively. A two-score lead is any lead of at least nine points. A three-score lead is any lead of at least 17 points.
|Third Downs Per First Down Series|
|Offensive Leaders||Defensive Leaders|
Per 3rd Downs
Per 3rd Downs
|Oklahoma State||188||65||.346||Texas A&M||147||84||.571|
This table is slightly outside the scope of the typical Three and Out material because it includes data not exclusively produced from drive data. I asked Bill Connelly to pull this breakdown from his play-by-play data and he gratefully obliged. This table includes all game data, including FBS vs. FCS games and garbage possessions. Which might explain why Syracuse looks so good defensively, having played two FCS opponents to date. Nevertheless, its a good glimpse of which teams are working for first downs harder than others, and which ones are forcing their opponents to do the same.
As always, if you have a suggestion for a future Three and Out featured table, please add a comment, or drop me a note on Twitter or via e-mail. The most popular tables will be updated and republished in future weeks.
FEI Week 6 Top 25
The principles of the Fremeau Efficiency Index (FEI) can be found here. FEI rewards playing well against good teams, win or lose, and punishes losing to poor teams more harshly than it rewards defeating poor teams. FEI 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 (GE), a measurement of the success rate of a team scoring and preventing opponent scoring throughout the non-garbage-time possessions of a game. FEI represents a team's efficiency value over average. Strength of Schedule (SOS) is calculated as the likelihood that an "elite team" (two standard deviations above average) would win every game on the given team's schedule to date. SOS listed here includes future games scheduled.
Mean Wins (FBS MW) represent the average total games a team with the given FEI rating should expect to win against its complete schedule of FBS opponents. Remaining Mean Wins (FBS RMW) represent the average expected team wins for games scheduled but not yet played.
Only games between FBS teams are considered in the FEI calculations. Since limited data is available in the early part of the season, preseason projections are factored into the current ratings. The weight given to projected data will be reduced each week until Week 7, when it will be eliminated entirely. Offensive and defensive FEI ratings will also debut in Week 7. The FEI ratings published here are a function of the results of games played through October 9.
FEI ratings for all 120 FBS teams are listed in the stats page section of FootballOutsiders.com. Click here for current ratings; the pull-down menu in the stats section directs you to 2007 through 2009 ratings.
|15||North Carolina State||4-1||.180||22||.130||25||.233||43||8.1||4.4|
16 comments, Last at 16 Oct 2010, 8:31pm
#1 by Will Allen // Oct 13, 2010 - 1:13pm
FEI sux! What does New Mexico State have to do, to get rated higher than New Mexico!!?? Do you even watch the games, Fremeau? Sheesh!
#2 by DJ Any Reason (not verified) // Oct 13, 2010 - 1:58pm
New Mexico State is clearly ranked too low because Fremeau doesn't even watch the games! The AP Coaches Poll is way better than this. Sheesh!
#3 by cfn_ms // Oct 13, 2010 - 2:04pm
does not create any kind of meaningful separation. The rest of the resume must be outweighing that one result. I'm not sure I agree with the result, but it doesn't strike me as crazy either.
#6 by Will Allen // Oct 13, 2010 - 2:16pm
New Mexico State had the swagger!! You and Fremeau obviously have never played the game!! Swagger rulz!!
#10 by CuseFanInSoCal // Oct 13, 2010 - 9:02pm
And South Florida is ranked higher than Syracuse, despite a head to head loss in a game 'Cuse dominated (13-9 final score notwithstanding; we failed on a 4th and 1 on the goal line and their touchdown was from a kick return) and the Orange having fewer losses. What's up with that?
obDisclaimer: Yes, I'm kidding.
#4 by cfn_ms // Oct 13, 2010 - 2:05pm
is #1 here, #30 on S&P. Either of the authors care to comment on why there's such a huge gap for that team, as well as any of the other model differences?
#5 by Brian Fremeau // Oct 13, 2010 - 2:13pm
I discussed this phenomenon in a piece earlier in the year about the differences in our two systems:
There are three contributing factors in the differences between our systems:
1. Drive data and Play-by-Play data variances. Can a team be successful at one and relatively unsuccessful at the other?
2. Which data "counts"? The FBS vs. FCS question comes into play here, as does the separate ways we calculate and treat garbage time plays and possessions.
3. Opponent adjustments. Our respective formulas are very different, even if based on similar principles.
(A fourth factor at this time of year is the way we handle preseason projected data as well.)
Personally, I think factor No. 3 is probably the most significant one.
#8 by cfn_ms // Oct 13, 2010 - 2:25pm
Yeah, I do remember that being talked about as the general differences between the models. I was curious what was driving such a monumental difference in this case. I'd agree with you that opponent adjustments are the likely cause, but that's just a guess.
I still think that having Oregon #30 is pretty much insane, and at least at first glance strongly suggests a material flaw with the S&P model.
#7 by bingo spiele (not verified) // Oct 13, 2010 - 2:19pm
YOur love for game is fantastic and all of this require pasion.
Alabama has got many chances but they are just could not convert.
#9 by Kevin from Philly // Oct 13, 2010 - 4:11pm
rj? Is that you?
#11 by TV_Pete (not verified) // Oct 14, 2010 - 11:25am
First, let me state that the Sagarin ranking/predictor and Football Outsider's FEI ranking are my favorites. I tend to prefer these to the polls, although the polls have their place in the BCS determination of the "best two teams for the BCS Championship". Of course, if I had my way the BCS formula would be used on January 2nd or January 3rd, after all of the "classic" Bowl Games to then determine the final pairing. It would only be 1 extra game for 2 teams and this game could be played about the same time and location as the current BCS championship. Ticket sales and travel might be the only major issue.
I remember a study a number of years ago that examined the effect of night games in the NFL. It was determined that West Coast teams had a significant advantage playing night games against East Coast teams. I believe it roughly doubled or negated the home field advantage. Why is there such an advantage? People normally function better, mentally and physically, earlier in the day. For West Coast teams they are playing until 7:00 or 8:00 PM, while East Coast teams may be playing after they are normally asleep.
With Alabama's opponents' bye week schedule and the Florida-Georgia history (I believe that Georgia has not beaten Florida in 20 years except after they had a bye week and Florida did not) I wonder if opponent bye week scheduling could be considered for part of a predictor. I suspect it is a minimal advantage (say 1 point), although it can benefit more (say 3+ points) to a team facing an extremely unusual offense. This would help explain why Georgia Tech's offense seems to do poorly in bowl games: they get a 1-point advantage for extra preparation before bowl game, but their opponent gets 3+ points. I don't know that the determination of "extremely unusual" is something that could be easily determined by a model or algorithm.
#12 by Will // Oct 15, 2010 - 11:51am
#13 by FO Fan (not verified) // Oct 15, 2010 - 3:03pm
I'm not sure if this post is more about Nevada being underrated or TCU being overrated ... but how is it that Nevada has essentially identical efficiency stats to TCU, with similar schedules, and yet suffers a .1 deficiency in FEI on the whole? That seems like an enormous discrepancy to be accounted for by just 10% of a preseason ranking, so is there something else at work here? I'm in Reno personally and don't really think 28 is too unrealistic a ranking for UNR, just curious about the methodology that makes a similar team like TCU rank so much higher.
#14 by Brian Fremeau // Oct 15, 2010 - 3:28pm
Two things. The SOS ratings listed here are season-long SOS. To date, TCU's schedule rates as somewhat tougher than Nevada's. Second, roughly 10 percent of the current ratings are impacted by preseason projected data and there was a huge gap between the two teams to start the year: No. 23 TCU, No. 80 Nevada.
#15 by Iron (not verified) // Oct 16, 2010 - 2:22pm
First off, I like the basis of the model. It seems more capable of capturing the true talent and quality of a team than other models as well as eliminating the "legacy of the team/conference" biases present in traditional rankings. Nice work.
Two questions: One about the model and the other about the outcome.
1. It would appear the model overweights winning (or projected wins) which is contrary to the stated purpose of the overall rank being about "playing well." SOS, Mean Wins and Remaining Mean Wins all win-based measures. How does the model ensure the "playing well" intent holds sway over the "win-based" elements?
2. Oregon and Boise State both ended the week with a .264 FEI. Why not a tie for first instead of ranking one above the other? What's the tie-breaker criterion.
#16 by Brian Fremeau // Oct 16, 2010 - 8:31pm
Thanks for the feedback.
1. Not sure what you mean here exactly. The only factor in the FEI ratings are Game Efficiency adjusted for opposition faced. The "win-based" metrics listed here are observations from the FEI ratings, not inputs.
2. Oregon and Boise State rounded to .264, but weren't precisely equal. I do this for other teams further down the ratings as well.