Advanced Stats for Opponents & Others
Football Outsiders has 2 advanced stats (S&P+, FEI) metrics that become more useful as the year goes by and more data comes in. S&P is play based, FEI is drive based. 6 weeks in S&P generally is pretty useful; FEI won't be released until next week. I post weekly data on our opponents, other Big 10 teams, and other misc teams of note.
Biggest movements in general this week:
- Maryland improved on both offense and defense quite a bit despite the lopsided loss to OSU. They had a low bar to improve from, esp on offense.
- Oregon State's data got demolished due to the disembowling by the flying rich rods.
Note - MSU is not that diff from the mike riley machine on S&P+ stats, yet the 2 teams have completely diff records. Then again if Oregon, Purdue, and rutgers could make 1-2 more plays in the last 2 minutes of a game MSU would be at .500.
S&P+ likes BYU more than MSU on both sides of the ball. Interesting.
Note - Indiana is viewed better on both offense and defense than a "program defining" team like Oregon.
dFEI | dS&P+ | oFEI | oS&P+ | ||
UM | 1 | 46 | |||
OSU | 26 | 42 | |||
MSU | 47 | 37 | |||
NWestern | 4 | 93 | |||
PSU | 17 | 61 | |||
Minnesota | 11 | 85 | |||
Maryland | 43 | 74 | |||
Indiana | 83 | 15 | |||
Rutgers | 109 | 83 | |||
Nebraska | 53 | 39 | |||
Wisconsin | 16 | 71 | |||
Iowa | 21 | 44 | |||
Purdue | 75 | 82 | |||
Illinois | 10 | 79 | |||
Utah | 24 | 23 | |||
BYU | 33 | 36 | |||
Oregon St | 73 | 87 | |||
UNLV | 59 | 101 | |||
ND | 32 | 6 | |||
Arizona | 110 | 18 | |||
Bama | 2 | 17 | |||
LSU | 40 | 12 | |||
TCU | 64 | 5 | |||
Baylor | 100 | 2 | |||
Oregon | 97 | 27 |
October 11th, 2015 at 8:51 PM ^
The Fremeau Efficiency Index (FEI) considers each of the nearly 20,000 possessions every season in major college football. All drives are filtered to eliminate first-half clock-kills and end-of-game garbage drives and scores. A scoring rate analysis of the remaining possessions then determines the baseline possession efficiency expectations against which each team is measured. A team is rewarded for playing well against good teams, win or lose, and is punished more severely for playing poorly against bad teams than it is rewarded for playing well against bad teams.
The S&P+ Ratings are a college football ratings system derived from both play-by-play and drive data from all 800+ of a season's FBS college football games (and 140,000+ plays). The components for S&P+ reflect the components of four of what Bill Connelly has deemed the Five Factors of college football: efficiency), explosiveness, field position, and finishing drives. (A fifth factor, turnovers, is informed marginally by sack rates, the only quality-based statistic that has a consistent relationship with turnover margins.)
October 11th, 2015 at 9:38 PM ^
It's the opposite situation of the Rich Rod years. Rich Rod's offensive would light the world on fire only in the second half after we were down a few touchdowns. I think because the opposing defenses would basically concede medium yardage plays. I wrote a big diary on this in 2010 that pissed everyone off.
October 11th, 2015 at 9:40 PM ^
Alum 96
thank you for this explanation- i was pretty perplexed. I am assuming it is a ranking index of all the CFB teams so each team's number is its relative rank compared to others
October 11th, 2015 at 9:55 PM ^
Yes, 127 FBS teams this year (down from 128 last year) so the lower # the better. 60 of those teams are P5.
Someone posted a "combined" S&P+ rating last week which has UM #3 but I cant make head or tails how it works.
I prefer these measures that show how a side of the ball ranks vs other teams side of the ball.
Generally I'd say any rank top 20 is pretty damn good to great, 21-40 is solid, 41-60 average to mediocre (esp if you are in P5) and then it gets bad from there, esp if you are a P5 team... i.e. a lot of Big 10 offenses are just bad. Wisconsin dropped off a cliff, they were one of only 3 Top 50 offenses last year in the conf (as Indiana died once Sudfeld was out for year).
While technically "61st" is middle of the pack (source: math) you see a lot of bad units on bad teams in the 70s, 80s, 90s so a well oiled P5 team should not be having a unit that low considering you are competing with the UNLVs, Charlottes, Idahos, Arkansas States of the world. So you see a lot of fanbases beat chests and say "74 ain't that bad, middle of FBS!!!" when indeed its well into mediocrity.
October 11th, 2015 at 8:59 PM ^
Another important thing to remember is that it is still early enough for teams to be affected by playing michigan. In other words, if we agree that Michigan's D is very elite then every offense we have played will appear statistically worse because they had to play us for one of their 5 or 6 games thus far. Despite that, BYU is right around MSU level. That being said, I think that this is underrating MSU a reasonable amount. Nevertheless, Its nice to be on the right of the numbers.
October 11th, 2015 at 9:09 PM ^
Of course, 16.7%% of your score will be affected by any 1 opponent this week. Less so go forward each week. 83.3% of data by non UM opponents is a pretty good tell for opposing offenses at this point.
NW's S&P+ offense actually improved 7 slots (was 100 last week) despite the shutout. Because NWs other 5 opponents continue to play and their data changes as well - i.e. Duke's defense is looking pretty damn good too hence making NWs struggles on offense more explainable - they'd played 2 top 10 defenses in 6 weeks. And Minn's D is supposedly "good" too.
By the way Duke beat Army 44-3 this week. The same Army that gave PSU all it could handle.
October 11th, 2015 at 9:01 PM ^
October 11th, 2015 at 9:31 PM ^
But Harbaugh was never much on sharing.
October 11th, 2015 at 9:02 PM ^
+1 Informative
October 11th, 2015 at 9:07 PM ^
I've been looking at these for a few years, and I like FEI more at a fundamental level.
I do think that drives (from the POV that, outside of a few end-of-half circumstances that can easily be controlled for, teams try to maximize both points & yards every time they have the ball) are the best level to analyze football at, as opposed to plays.
FWIW, I am in the data analytics space for work and on the side, I've been playing around this year with my own drive-based model. Not claiming it's any good, and I am interested to see how it compares vs. FEI when we see the FEI data next week. But FWIW what I have is below (rank among all FBS teams). These numbers include data all through Week 6:
(drive-by-drive data is actually not that hard to obtain, it's only about 2 hours of box score searching onlie on Sunday AMs)
Michigan: Offense 13 , Defense 2 (Clemson is #1 Defense).
Ohio State: Offense 58 (!!!), Defense 35
Michigan State: Offense 32, Defense 48
Penn State: Offense 104, Defense 8
Northwestern: Offense 78, Defense 16
BYU: Offense 28, Defense 43
Utah: Offense 15, Defense 12
October 11th, 2015 at 9:11 PM ^
I like FEI more too but later in the season (week 9ish). But we don't have it yet so S&P+ it is. I think its important to show both together because sometimes there is a 10-15 slot difference so I like to go with an average - one can raise fuss with both of course in any week for any 1 team. I am deathly curious if UM can be #1 in FEI D too. (last year UM's ranks on D in both were 35-40 by the way, which I felt was more real than the NCAA stat of #7).
October 11th, 2015 at 9:40 PM ^
As you said, best to average several of them together to get a better flavor of things. No model can be perfect.
I think --- think (hope) --- that FEI becomes fully available after Week 7. I'll be hitting refresh on Football Outsiders regularly. :-)
October 11th, 2015 at 9:13 PM ^
October 11th, 2015 at 9:19 PM ^
By 16.66666% each. :D
October 11th, 2015 at 9:45 PM ^
(e.g., pretended it didn't exist)
Without it: Michigan has the 29th ranked offense, 3rd ranked defense. Northwestern is 74 and 2.
While the U-M defense was excellent yesterday, I do think the MORE impressive thing was Michigan offense's early efficiency. 2 long touchdown drives against a very good defense on their first 2 touches.
21-0 with 48:00 left on the clock but that was really the ball game.
October 11th, 2015 at 9:59 PM ^
Yep the offense was more surprising then the defense yest. Main challenge for D was stop Jackson and they stopped him cold. NW had a top 25 ranked rush offense - once that was gone they were neutered. Their pass O is putrid - NW QB is a youngin and he doesnt win games - just tries to not lose them.
The O scoring 24 on NW was pretty damn good, esp considering UM basically packed it in, in the 3rd quarter (yet again).
October 12th, 2015 at 9:50 AM ^
Harbuagh's game planning is absolutely amazing. This is an awesome victory this week when some teams might be caught looking ahead to next week.
October 11th, 2015 at 9:51 PM ^
That's very interesting. Out of curiosity, where do you get your data to create the model? And what software platform do you use?
October 11th, 2015 at 10:13 PM ^
Play-by-play data from the 2005-2013 era is available for free on the web. The 2014 data is not available for free, but I've been able to re-create it.
http://datahub.io/dataset/college-football-statistics-2005-2013
For 2015 data, I just go to the various team websites every Sunday AM, I have scripts that can take a boxscore (the boxscores are almost universally in 1 of 3 different formats) and turn it into drive-by-drive data in Excel. It only takes me ~2 hours to get the ~60-ish games played every week that involve an FBS team.
The historical data are good for comparing the "actual result" versus the "expected results", given the time and place from which a team starts a drive. For instance, Michigan started their 1st offensive drive yesterday from their own 41. 2347 times in the 2005-2014 era has a team started a drive from their own 41. 27.1% of the time does a team score 7 (or more, e.g., 8) points on a drive when starting from their own 41.
FWIW, ONE time did a team start at their own 41 yard line and manage to score -2 points (a safety). UCF early in the 4th quarter of their 2008 20-12 loss to Miami FLA. Kind of fun to sometimes mine the data for historically weird results such as that.
October 11th, 2015 at 9:08 PM ^
OSU's offensive ranking will undoubtedly trend up, but they really don't seem as dynamic as they were in years past. UNLV looked decent against SJSU yesterday; that is looking more and more like a quality-ish win.
October 11th, 2015 at 9:14 PM ^
OSU was #1 in S&P+ offense and #7 in FEI last year. And that was with breaking in a new QB, new OL, Barrett struggling a few games. They are certainly nowhere near last year's levels at this point.
By the way if anyone had said before the year UM's advanced offenses stats would be within 15 ... forget 5 points of OSU you'd think we'd be putting up 45 a game!
October 11th, 2015 at 9:09 PM ^
October 11th, 2015 at 9:18 PM ^
Yep, and it concurs with eye test. They've still put up 34+ in 5 of the 6 games but they'd also played some pretty bad defenses to do so.
October 11th, 2015 at 9:22 PM ^
Sent from MGoBlog HD for iPhone & iPad
October 11th, 2015 at 9:44 PM ^
October 11th, 2015 at 9:34 PM ^
October 11th, 2015 at 9:18 PM ^
October 11th, 2015 at 9:22 PM ^
Sent from MGoBlog HD for iPhone & iPad
October 11th, 2015 at 9:48 PM ^
October 11th, 2015 at 10:05 PM ^
No they don't.
Here are the S&P+ data points broken down more directly.
- Success Rate: A common Football Outsiders tool used to measure efficiency by determining whether every play of a given game was successful or not. The terms of success in college football: 50 percent of necessary yardage on first down, 70 percent on second down, and 100 percent on third and fourth down.
- IsoPPP: An explosiveness measure derived from determining the equivalent point value of every yard line (based on the expected number of points an offense could expect to score from that yard line) and, therefore, every play of a given game. IsoPPP looks at only the per-play value of a team's successful plays (as defined by the Success Rate definition above); its goal is to separate the explosiveness component from the efficiency component altogether. For more information about IsoPPP, click here.
- Redzone S&P+: This measures drive-finishing ability by looking at the success rate and IsoPPP measures for only plays that come after a first down inside the opponent's 40-yard line. Coaches start adjusting their play-calling for a shrinking field closer to the 40 than the 20, and there is more separation between good and bad offenses if you look at plays in this range instead of plays inside the 20-yard line (as the redzone is commonly defined).
- FP+: This is an opponent-adjusted measure of your ability to create field position advantages. This is based on drive data instead of per-play data. For an offense, it looks at field position you create for your defense (with help from special teams, which is not yet stripped out of these numbers); for a defense, it looks at the opposite.
- Opponent adjustments: Each team's output for a given category (Success Rate, IsoPPP, and split stats like rushing, passing, redzone, standard downs, passing downs, etc.) is compared to the expected output based upon their opponents. This is a schedule-based adjustment designed to reward tougher schedules and punish weaker ones. In the tables below, the "+" designation is for measures that are adjusted for opponent.
- Garbage time adjustments: The S&P+ figures used in the tables below only look at the plays that took place while a game was deemed competitive. Garbage-time plays and possessions have been filtered out of the calculations. The criteria for "garbage time" are as follows: a game is not within 28 points in the first quarter, 24 points in the second quarter, 21 points in the third quarter, or 16 points in the fourth quarter.
October 11th, 2015 at 9:53 PM ^
Add Clemson to your "teams of note." They're legit.
October 11th, 2015 at 10:03 PM ^
Man it's hard ever to take Clemson seriously when they are so famous for derping an easy game. They do have a friggin easy schedule the rest of the way if they can get past FSU!
Advanced stats love Clemson - #10 offense, #6 defense. Will add 'em next week but then we can expect them to lose to NC State or something!
October 11th, 2015 at 10:34 PM ^
October 11th, 2015 at 11:05 PM ^
It just hasn't been updated for week 6 and will still include preseason data through week 7.
October 12th, 2015 at 8:41 AM ^
Not the unit FEI (off and def)
October 12th, 2015 at 9:46 AM ^
It is really helpful to have all of this information put into one place.
October 12th, 2015 at 9:46 AM ^
It is really helpful to have all of this information put into one place.
October 12th, 2015 at 9:46 AM ^
It is really helpful to have all of this information put into one place.
October 12th, 2015 at 9:46 AM ^
It is really helpful to have all of this information put into one place.
October 12th, 2015 at 9:46 AM ^
It is really helpful to have all of this information put into one place.
October 12th, 2015 at 9:46 AM ^
It is really helpful to have all of this information put into one place.