How FEI and S+P see the game: Oregon at Stanford 2013

FEI is the Fremeau Efficiency Index, created by Brian Fremeau. Brian Fremeau is an author at Football Outsiders, ESPN and BCFToys. FEI is an advanced statistical measure for college football that tracks drive efficiency instead of per-play success.

S+P is created by Bill Connelly. Bill Connelly is an author at SBNation, RockMNation, Football Study Hall and Football Outsiders. S+P is an advanced statistical measure which combines success rate, explosiveness per play and opponent adjustments.

A foreword

I'd really like to thank the members of Tomahawk Nation and the fans of Florida State for coming by last week and having an absolutely stellar discussion about the merits of FSU and Oregon, statistical systems and the horribleness of the BCS. That was a lot of fun and really informative and useful. Thanks so much for contributing!

I also want to link to last year's analysis of the Stanford game, as this was a game where advanced stats barely saw anything like this coming save in a couple of important parts. There are some reasons to discount the advanced stats this time as well - and I'll get into those - but it's also important to understand where advanced stats would have warned us. In particular, FEI did a very good job of indicating that Oregon would struggle on offense against a very strong Stanford D. But we'll get into that in a bit.

How S+P sees the game:

There are some new stats from Bill Connelly this year.

Play Efficiency: the success per play based on the down and distance of the play.

Drive Efficiency: the success of scoring based on the field position created.

Difference in Net Points (DNP): the average of the points an offense scores on a given drive compared to the points it would be expected to score based on starting field position.

And the old ones:

Passing downs: second down and 8 or more, or 3rd/4th down and 5 or more.

	OVERALL		When Oregon Has the Ball ...		When Stanford has the ball...
Category	Oregon	STAN	ORE Off	STAN Def	ORE Def	STAN Off
F/+ Rk	4	3
S&P+	5(267.4)	9(260.4)	6(137.0)	6(148.2)	15 (127.7)	31 (111.1)
Play Efficiency			3 (146.4)	6(133.9)	33(112.0)	27(113.7)
Rushing S&P+			2 (151.6)	22(120.5)	38(110.3)	53(107.6)
Passing S&P+			7(142.6)	4(151.0)	44(108.3)	20(127.8)
Std. Downs S&P+			1 (148.1)	7(133.7)	47(105.0)	37 (110.9)
Pass. Downs S&P+			9 (142.3)	4(149.6)	18(123.7)	18(129.9)
Drive Efficiency			10 (127.7)	5(162.5)	12(143.4)	35(108.5)
Difference in Net Points			4(1.83)	47(-0.74)	3(-2.59)	43(0.19)

As I said in the FSU preview, Oregon's numbers went up slightly after beating UCLA but the rankings went down. Looking into it, it's because the passing ratings are going entirely insane. Oregon got more efficient on passing downs and went down 5 spots because the top number is an absurd 222.6 points.

I'm still surprised given Mark Helfrich and a change in our run/pass ratios (we were doing a 65/35 split; now we're closer to 60/40) how similar we are to last year. Our run game is still dominant. Our pass game is still excellent. Our overall score is remarkably similar, as is our style. This is not the case with Alabama or Stanford. The consistent play from year to year Oregon has had is one of the most remarkable things in college football; consistency is the landmark most schools aim for, and even the best programs tend to have off years. Oregon has had 5 years where they were no worse than the top 10 in the nation. Even Alabama can't say that.

Similarity scores for Stanford

Washington (27th) is very close to Stanford in ranking and in behavior. Both Stanford and Washington are only okay at standard downs and good on passing downs. Stanford is much better passing the ball and not as good running the ball as Washington. Both are very similar on drive efficiency, play efficiency and net points. And really...that's it. The next closest team Oregon's played is UCLA (38th); Stanford is significantly better on standard downs and worse at passing, just barely.

Just like in the FSU preview we see that Oregon has faced nothing as good as Stanford on D. (and in fact, from a statistical perspective Stanford and FSU are neck and neck). The closest is still UCLA (23rd) and Washington (17th), but both are a full 20 points below Stanford. Washington and Stanford are otherwise fairly similar in tendencies and strengths; both are stellar against the pass, great on passing downs, slightly less great on standard downs and only good against the run. In fact, Washington is slightly better against the run than Stanford. The main advantage Stanford has is on drive efficiency; Stanford simply does not, for whatever reason, allow drives to complete. This should sound familiar to you, as it is what Oregon has done all year as well.

Similarity scores for Oregon

For the first time all year, I can not say that Stanford has not faced anything like Oregon. Arizona State is very close to Oregon in S+P (7th). Oregon is significantly better at running the ball and in standard downs than ASU. ASU is slightly better passing but much, much better on passing downs - an absurd 208.7. I honestly think this is a bug in S+P right now.

On defense Oregon is just slightly better than UCLA (23rd) and Washington (17th). Oregon is nothing like them on a per play basis, being significantly worse in all categories. However, there are two places Oregon is much better - at drive efficiency and overwhelmingly in difference in net points. The DNP advantage, I believe, comes from Oregon's offense as much as the defense; I'll let Mark Seymour from Baylor's Our Daily Bears take it away:

I have a theory related to that DNP rating that might seem paradoxical at first glance, so stick with me: it is a function of our offense as much as the defense. What I mean by that is our offense, as good as it is, puts so much pressure on opposing teams to score touchdowns, not field goals, that they often go for it on fourth downs in our end of the field at the expense of just getting points, giving our defense the opportunity in an advantageous situation to stop them and suffer no nominal damage.

This should sound pretty familiar to Oregon fans. Think about Arizona last year, where 8 trips inside the Oregon 40 yielded zero points. Think about throwing deep into coverage on 3rd and 20, or arm puntapalooza.

Oregon's offense vs. Stanford's defense

Stanford and FSU are so similar here I could cut and paste what I wrote there. Oregon has a remarkable advantage when running the ball - over 30 points. If you're thinking that this is what I said last year, that's sort of true - but only sort of. The advantage last year was only 13 points. Oregon has been better running the ball this year compared to last, at least so far. More importantly, Stanford has been decidedly worse at defending against the run. Both Washington and UCLA are supposedly better against the run than Stanford is (though not by much). In addition Oregon has a small advantage on standard downs (14 points). On passing downs and against the pass it is a dead heat, which usually favors the defense. But on standard downs and running the ball? Oregon has a lot of room to run, so to speak.

Now, the two teams with great defenses but only so-so run defenses Oregon has faced tried two different approaches, and neither ended up well for them. Washington loaded up against the box and attempted to first play man, then zone against Oregon. This didn't work out all that well, as Marcus Mariota first simply ran the ball and then picked apart the zone. UCLA, as I predicted, played a lot more man and had more safety help, so Oregon just steamrolled over them on the ground and made Byron Marshall the next great thing. The important factor here is that neither approach worked that well. Selling out to stop the run opens up the pass - and unlike prior years, Oregon is lethal passing. Staying to stop the pass and hopefully let your line deal with the run doesn't work either. Contrast that with last year's Stanford, which was worse against the pass but better against the run. I went back and looked; in general, teams that are good defenders against the run give Oregon significantly harder times than teams that are good against the pass.

The numbers don't think Stanford will do a good job stopping Oregon from getting yards, but they do think that Stanford will do a good job of stopping Oregon from getting points. There's a 30 point differential in Stanford's favor on drive efficiency - meaning that Oregon is likely to end drives badly against Stanford. We'll see that again in the FEI section. Ultimately this means that Stanford will try for the big hit, the big fumble, the big sack, and attempt to end drives or let the other team get a first down. What's odd is that Stanford isn't that great at causing turnovers in general, so it's more likely Stanford will get stops in other ways. That being said, Oregon has had some trouble with that this year; discounting the California game (which we'll see is a very reasonable thing to do) Oregon is still fumbling about twice a game, on average. And as we'll see, putting Oregon into a situation where they need to kick field goals to get points? That's essentially as good as a stop.

Keep in mind that while this is what the numbers say, there are some issues that counteract those numbers - namely injury. Losing your star DE and captain of the defense to a season-ending injury the game before is about as big a blow as Stanford could have had. If Oregon is able to run all over Stanford - moreso than against Washington or UCLA - this is almost certainly a big reason why.

Oregon's Defense vs Stanford's offense

Last year Stanford's offense, statistically, looked dreadful before playing Oregon. It of course wasn't; Stanford threw for a good amount of yards, ran well, and did fine against a fairly strong defense. But the S+P numbers wouldn't have thought this for an instant. Why did S+P get it so wrong? Well, Stanford switched QBs midseason from Josh Nunes to Kevin Hogan. Kevin Hogan was remarkable in his starts. At the time I though Hogan would not be able to deal with the pressure of Autzen; I was dead wrong. What I should have done is looked at the more recent stats, extrapolated and tried to fit.

Fortunately I don't have to do that this game, as Hogan has been the starter all season - and Stanford's offense is just okay. Well, no - they're good. But they're not great, much less orbital laser beam wonderful like Baylor or FSU. In particular they are average at running the ball despite a dedication to run Tyler Gaffney whenever humanly possible. They are good on passing downs and passing the ball, but their tendencies are to run the ball a lot more than pass (a 60-40 split in favor of running) and use play action and the threat of the run to open up receiver threats. The largest advantage Stanford has is a 20 point advantage on passing. That tends to indicate that Stanford will get a fair amount of yards passing the ball. As jtlight is fond of saying, however, that doesn't mean that our pass defense stinks. It means that we're willing to trade yards for favorable stopping chances and turnover opportunities. Much like Stanford's D vs. Oregon's O, Oregon has a 30 point differential in drive stopping. Turnovers, penalties, and sacks are the order of the day here. Stanford is dead even on turnover margin, though they don't tend to turn it over or cause many turnovers (only 10 total the entire season). By comparison, Oregon is +13 - and most of that is because of interceptions, not fumbles. If Oregon doesn't get the ball via turnover it may mean a stop and a field goal try. Unfortunately, Stanford has a competent kicker - meaning that long drives may end up in some points. If Oregon and Stanford trade blows like that I don't see it going well for Oregon. I also don't see it going that way in this game.

How FEI sees the game:

Some definitions from the FEI site. For offense, these are the drives that the offense does. For defense, these are the drives that the defense has allowed. Also note that these are not weighted by defensive strength or anything like that, so they correspond best to raw overall numbers.

First down rate: the % of drives that result in at least one first down.

Available Yards: the ratio of yards gained by total yards to go

Explosive drives: the % of drives that average at least 10 yards per play.

Methodical drives: the % of drives that take 10 or more plays

Value Drives: the % of drives that start on their side of the field and make it to the opposing 30 yard line or better.

Field Position Advantage (FPA): the share of the value of total starting field position earned by each team against their opponents.

	OVERALL		When Oregon Has the Ball ...		When Stanford has the ball...
Category	Oregon	STAN	ORE Off	STAN Def	ORE Def	STAN Off
F/+ Rk	4	3
FEI Rk	4(.279)	3 (.284)	10(.572)	2(-.799)	9(-.575)	35(.249)
Field Position	9(.553)	2(.591)
Raw OE/DE			5 (.810)	19(-.424)	2(-.716)	59(.091)
First Down rate			5 (.827)	13(.565)	16(.578)	36(.723)
Available Yards rate			4(.683)	21 (.368)	11(.345)	32(.523)
Explosive Drives			10(.247)	9(.059)	15(.072)	17(.205)
Methodical Drives			116 (.074)	121(.235)	1 (.048)	79(.133)
Value Drives			1(.662)	20(.299)	3 (.224)	51(.420)
Special Team rank	42(1.008)	2(5.046)
Field Goal efficiency	104(-.458)	64(.132)
Punt Return efficiency	1(.487)	20(.101)
Kickoff return efficiency	32(-.075)	1(.529)
punt efficiency	114(.155)	53(-.074)
kickoff efficiency	82 (-.112)	15(-.305)

Oregon's special teams got even worse after UCLA; not surprising. Other than that, Oregon went up significantly. FEI loves it when good teams play each other, and Oregon crushed UCLA handily. Oregon's offense went down slightly, but the defense went back to being a top 10 program thanks to holding UCLA to 60 yards passing and no scores in the second half. Across the board FEI loves Oregon save in special teams, where apparently there is something called 'field goal kicking' that Oregon should look into. It's all the rage with the kids.

Stanford is interesting. FEI loves Stanford in the same way that FEI loved Kansas State last year. Kansas State was efficient on their drives but more importantly got a ton of extra yards from field position and special teams play. This naturally was why Oregon took the opening kick for a TD. Nope, sorry, that's not quite right. There's some cause to think that FEI overrates Stanford as a team - but don't mistake that for overrating Stanford's defense. Just be aware that as far as FEI is concerned, Stanford has gained a lot of their success on field position and special teams play, and were it not for those would have at least one additional loss (specifically, the Washington game).

New! The best and worst of the teams

One of the frustrating things looking back last year was seeing games act as huge one-off values that oddly affected a team's play. While FEI does weigh each game differently depending on how relevant it is, it's also important to understand the really large highs and the really deep lows. A good example of this is Oregon's crushing of Arizona last year. Oregon scored 25.0 points from their defense that game by FEI standards - but if you look at the rest of the season, their defense only scored a total of 40.1 points for the rest of the season and averaged less than 4 points a game coming from their defense. That Arizona game is a clear outlier, and one that implied Oregon's defense was much better than it was. While I'm not going to recalculate all the games for FEI based on these highs and lows, thanks to having FEI's Game Factors we can see some of them and make educated guesses as to how good a team can play - and how good a team is likely to play. In this piece, when I say something like 'above 100' keep in mind that the way FEI works is that it ranks every single score from every single team. As an example - Oregon's offensive performance against Washington was the 9th best performance by any offense recorded so far this season. Oregon's offensive performance against Cal was the 662nd best performance by any offense recorded so far this season. These numbers are weighted by opponent as well. There is more explanation and examples over at Football Outsiders.

Oregon's highs and lows

high on offense: Washington (8th overall, 2.183)

low on offense: California (656th, -.082)

Standard deviation (throwing out the Cal game): 108

high on defense: California (64th overall, -.702)

low on defense: Colorado (329th overall, .031)

Standard deviation (throwing out the Cal game): 112

Keep in mind that FEI does factor some of this down - and the California game is the second least weighted game of Oregon's season. Even so, California for a variety of reasons is an extreme sample that should likely be removed; it is unlikely Oregon will cause 8 fumbles in any game not played in a monsoon. Other than that, Oregon has been remarkably consistent on offense - 4 games that were ranked 69th or better. So what were the bad games? well, throwing out Cal we have Colorado (148th) and Virginia (293rd). Neither were horrible, but neither were outstanding. There's not much else to see there; neither team was particularly good on defense. One thing to note - both were on the road. The Cal game being this big odd game here is also a sign that even though FEI does not weigh the Cal game heavily FEI probably underestimates Oregon's offense.

On defense we have slightly more variability. We have the Cal game as the high, but we should throw that out for the same reason we throw it out on offense as the low; it's not likely we are going to be playing in a monsoon. The next highest is the Washington game. We range from 93rd to 329th (against Colorado). That's not a ton of variability. There are 3 games between 64 and 108, 1 200, and then 3 300s. The worst game was on the road - but the next two worst games were at home, and the best game was also on the road. The three worst games also happen to be against the worst opponents - Colorado, Tennessee and Washington State. The best were against Washington and UCLA. What this tells me is that Oregon's defense is a bit inconsistent but is not being propped up by a one off game. And we tend to play up to good competition. That's good! It means that the numbers for Oregon are probably close to what we'll get, and if you look at the median scores on defense it means we'll probably have something like how we did against Virginia or UCLA as the baseline. That being said - we could have a big game, or we could have a big letdown. Hard to say. It is unlikely that we will play as poorly as we did against Colorado, but it's possible. It also means that FEI likely overestimates Oregon's defense due to that Cal game.

It's shocking how consistent Oregon is from week to week - you'll see an inconsistent team here in a second.

Stanford's highs and lows

high on offense: Arizona State(60th overall, 1.536)

low on offense: Army (751st, -.331)

standard deviation: 194

high on defense: Arizona State(2nd overall,-1.737)

low on defense: San Jose State (169th overall, -.375)

standard deviation: 49

That Arizona State game sure looks like a big one-off, doesn't it? And it is...sorta. It is the only game that Stanford was in the top 150 on offense, and is the best game on defense by a huge margin. However - on defense, Stanford has had another top 31 game (Oregon State) and 4 others in the top 165. They have no games over 200. While this means that they aren't likely to play at the same level as they did against ASU, it doesn't mean they're a bad defense - just potentially a somewhat overrated but consistently good one, and the standard deviation here indicates that nicely. Median score wise, their defense is between Army (101) and Utah (135). On offense, however, they're highly variable and often bad; they had all two games below 200 (ASU and UCLA) and the rest were significantly worse, including the 751st worst game against Army. Their median score would be between Utah (327th) and Washington State (377th). Do not expect big fireworks from their offense, ever. The Arizona State game implies that Stanford's defense and offense both may be overrated in the same way Oregon's was last year.

(For you FSU fans reading along, FSU is really variable here. They have 4 offensive games in the top 40 and then a 208, a 233 and a 436. On D, they have a 28, 46,, a 95, and then one that goes all the way to 427. It's clear that Florida State can play a game that is incredible - but they aren't hugely consistent.)

Similarity scores for Stanford

FEI does not see Stanford as a great offensive power. Yeah, I know, you're shocked. Washington (11th) is significantly better - both from a raw offensive power and from facing a harder schedule. UCLA (44th) is the next closest the other way, but that's a pretty big difference - though UCLA has faced even harder opponents.

Defense is where Stanford shines. As does, oddly, most of the Pac-12. Stanford is ranked #2, USC is #3, Oregon #9, Arizona State #12, UCLA #23. Huh. In any case, as you might expect there's nothing close that Oregon has faced. UCLA is the closest but the gulf is almost 400 points - which is the difference between UCLA and Tennessee (61st). This is very similar to what it was like last year as well. Last year we dismissed this as being FEI going all crazy. This year we'll pay attention to it more. Stanford has also faced the 5th hardest schedule on defense - and that's before facing us.

Similarity scores for Oregon

Oregon is closest to Arizona State (3rd) and Washington (11th) on offense. Arizona State has faced a massively hard schedule on offense (6th overall) compared to Oregon's 90th; this is why their rating is so high. Washington is similar, facing the 10th hardest overall. This is a good test to see whether or not rating teams that are blowing out competition left and right are better or worse (and by how much) compared to teams that are facing adversity.

Oregon is closest to ASU on defense (12th). The primary difference between the two is that ASU has faced significantly harder teams than Oregon has. Oregon is the second-most efficient team in the nation on defense without opponent adjustments; the first is Michigan State. This again goes to jtlight's oft-stated point: it is very difficult to gauge how good we are given our plans against fairly bleh opponents. Stanford, mind you, won't help that all that much.

Oregon's offense vs Stanford's defense

Stanford is scary efficient everywhere on defense save one place - almost a quarter of all the drives against Stanford have been 10 or more plays. Remember how we were talking about drive efficiency, and how Stanford would try for the big hits? That aligns here; if Stanford doesn't get that big hit, drives just keep going and going. Oregon doesn't tend to do that kind of drive; the last two weeks Oregon has not had a single drive take longer than 9 plays. But they might have to here. This would go well with Oregon's strength being running the ball against an only good Stanford running defense. Stanford is much better than anyone Oregon's faced at stopping explosive plays, so don't expect a whole lot of highlight tape this time around. Stanford's advantage in overall FEI is essentially the same as it was last year. From a numerical standpoint I don't expect significant changes from last year's results, at least on offense.

Oregon's defense vs. Stanford's offense

Oregon should absolutely crush the hell out of Stanford here. While Stanford has a 100 point differential in their favor on defense, Oregon has a 260 point differential in their favor when Oregon's on D. This is close to what Stanford had over UCLA. Or what Oregon had against UCLA, for that matter. Unless Oregon makes significant mistakes I do not see Stanford scoring all that much. Much like UCLA, I would expect Stanford to struggle with longer fields and only get good scoring opportunities on short fields.

Special Teams

Stanford is the second best team in the nation in special teams play. In particular, they are the best at kick returns by a large margin. They are also excellent at kick coverage. Oregon...is not. Oregon is the best at punt returns but is otherwise at best mediocre everywhere else. While Stanford is just mediocre at kicking field goals, Oregon is 104th in the nation.

Here's a good time to talk about some of the other work Brian Fremeau does on BCF toys. One of the things he publishes is the values based on alternative measures. What if you didn't take into account things like turnovers? Or special teams play? He recalculates FEI and rankings based on that. If you're curious, Stanford's FEI drops from .291 to 261 if you exclude special teams play - this is one of the biggest drops out there. If you ignore field position Stanford goes even further, dropping from 291 to 242. (Oregon, meanwhile, becomes the best team in the nation). And this is an open question: is special teams predictive? Is field position predictive?

One data point suggests a big fat 'no' to that - the Fiesta Bowl. Kansas State had a great .279 FEI at the end of the season - but if you take their special teams out of it, they drop to .215. We know the result of that game, so we can see that Kansas State was not at the time close to Oregon - but according to FEI they were. This is a long way of saying that I think special teams and field position are a bit too weighed in final scores for FEI and is less predictive due to it being, well, fairly stochastic, one time events. Stanford has returned 3 kicks for TDs this season, which is amazing. It does imply that Stanford will get great yards on their kick returns; it doesn't imply that Oregon will give up one to the house. Field position is similar; field position is hugely descriptive as to why a team is winning, but due to all sorts of things may not be particularly predictive. A 20 yard bouncing ball can result in a huge difference in field position. So can some odd penalty at the wrong time. If Stanford goes crazy like they did against Washington, Oregon will have a significantly harder time of it. This seems like an obvious statement, but it's no less true.

So what does this all mean?

Last year I was concerned that Hogan being a different and far more competent QB meant Stanford's offensive numbers were not indicative of current performance. That turned out to be a fairly good concern, as Hogan did enough against a very good Oregon defense to keep Stanford in the game. Counting numbers - even ones that are opponent-adjusted and take into consideration the results and context of the play - do not do a stellar job of seeing how a team does with injuries, changes in game plan or QB and the like. So I downplayed how good Stanford looked on defense, overplayed Oregon's success on defense and thought it'd be a game close to Arizona and Arizona State.

I was wrong.

It's hard to say that I'm wrong. But it's important to say, and it's even more important to look into why. I didn't listen to what FEI was stating - that Stanford is for whatever reason a stellar defense and will get stops. I didn't look past the S+P numbers to see trending or to see how Hogan had done. Every game isn't defined by the numbers before it, but there were signs that that loss was coming - signs I didn't pay enough attention to. This year I vowed that I'd do more to understand those things and not let my personal biases interpret the facts. The FSU-Oregon analysis was the first step; this is the next.

The numbers for this year currently favor Stanford, slightly. Their defense is strong and will get stops. Oregon is not good at capitalizing on drives that don't end in the end zone. Stanford gets better field position and gets more return yards. Oregon has a very good advantage in running the ball, but Stanford has an advantage in passing. Stanford should get the edge when looking at just the raw numbers.

But.

Stanford has significant injuries: Ben Gardner is the biggest - though they get Henry Anderson this week. Devon Cajuste is another blow that has made their offense trend down in recent weeks, but they'lre getting him back. Their statistical dominance relies heavily on their dominating win against Arizona State in both FEI and S+P, and they have not replicated that level of excellence in any of their games or really come close to it. Oregon, meanwhile, only looks better when removing their outliers and is getting players back (notably De'Anthony Thomas).

I do think it'll be a close game - closer than many Duck fans would like it to be. I think that one of the biggest keys will be how Oregon's defense holds up against Tyler Gaffney and how many turnovers are caused. But I also don't see Stanford holding Oregon's run game in check like they did last year; a 30 point differential + a key injury means that Oregon has a very large advantage doing the thing that they do best and most often.

The current line is Oregon -10 and may move even higher.I don't see that being something Oregon will cover. Oregon could blow out Stanford with the sort of warp speed scoring that Oregon does - but I believe that there will be more counterpunching from Stanford, and Oregon's sloppy play will cost them in similar ways to the Stanford/Washington game. I think Oregon wins - but does not cover.