Misc Advanced Stats Among Rivals
tl;dr - no you do have time to read all this and it's not too long
Last year I posted a weekly chart of advanced stats (FEI / S&P+) for UM and conf / non conf opponents + a few other teams nationally to see how we stacked up. I find these data points far more interesting than the NCAA stats which only rank a defense or offense on "yards gained or lost" regardless of conf or strength of schedule.
These advanced stats don't really kick in until week 6-7 so it's still a few weeks away from really being a great tell. Still, with the S&P+ data out there - and with caveats of small sample size - I thought I'd take a look at our schedule in relation to 2 rivals thru the lens of degree of difficult of offenses and defenses all 3 teams have played.
A) How do the 3 teams stack up in S&P+ offense and defense rankings - again, it will be a few more weeks before these fully kick out sample size issues:
| Off S&P+ | Def S&P+ | |
| UM | 48 | 4 |
| OSU | 33 | 6 |
| MSU | 19 | 26 |
So in a vacuum the data says MSU has the best performing offense and worst performing defense. OSU offensive "struggles" (vs expectation at least) are well documented and UM obviously is not a high powered offense right now and basically took 2 halves off (UNLV, BYU) playing keep away. Both UM and OSU look to have top notch defenses while MSU is below their normal top 10 rank.
B) How do the offenses and defenses UM, OSU, and MSU faced thru 4 weeks rank?
| Offenses Faced | |||||||
| Opp S&P+ | Opp S&P+ | Opp S&P+ | |||||
| Utah | 43 | VATech | 39 | W. MI | 69 | ||
| Oregon St | 94 | Hawaii | 118 | Oregon | 21 | ||
| UNLV | 111 | N. Ill | 84 | Air Force | 28 | ||
| BYU | 56 | W. MI | 69 | C. MI | 92 | ||
| 76.0 | 77.5 | 52.5 | |||||
| Defenses Faced | |||||||
| Opp S&P+ | Opp S&P+ | Opp S&P+ | |||||
| Utah | 36 | VATech | 60 | W. MI | 112 | ||
| Oregon St | 50 | Hawaii | 71 | Oregon | 89 | ||
| UNLV | 80 | N. Ill | 30 | Air Force | 73 | ||
| BYU | 39 | W. MI | 112 | C. MI | 77 | ||
| 51.3 | 68.3 | 87.8 |
Current narrative in EL is MSU has played a ton of great offenses so stop hating on our defensive stats! Advanced stats would bear out MSU has faced the better average offense but still a middling 52.5 rank. UM and OSU have faced an even worse array of offenses. Main difference in these ranks aside from Oregon is Air Force offense came in surprisingly top 30.
Meanwhile - not mentioned in EL is MSU has played a set of 4 disasters on defense with an ave rank of 88 (!!). Meanwhile UM has faced the best average defense. So we can grade some of that variance between UM offensive rank and MSU offensive rank on a curve.
C) What would this look like in 2 weeks - when each team gets to play 2 mediocre Big 10 teams (this would assume the S&P+ ranks for those teams don't change - which of course they will)
| Offenses Faced | |||||||
| Opp S&P+ | Opp S&P+ | Opp S&P+ | |||||
| Utah | 43 | VATech | 39 | W. MI | 69 | ||
| Oregon St | 94 | Hawaii | 118 | Oregon | 21 | ||
| UNLV | 111 | N. Ill | 84 | Air Force | 28 | ||
| BYU | 56 | W. MI | 69 | C. MI | 92 | ||
| Maryland | 73 | Ind | 25 | Purdue | 80 | ||
| NW | 105 | Maryland | 73 | rutgers | 86 | ||
| 80.3 | 68.0 | 62.7 | |||||
| Defenses Faced | |||||||
| Opp S&P+ | Opp S&P+ | Opp S&P+ | |||||
| Utah | 36 | VATech | 60 | W. MI | 112 | ||
| Oregon St | 50 | Hawaii | 71 | Oregon | 89 | ||
| UNLV | 80 | N. Ill | 30 | Air Force | 73 | ||
| BYU | 39 | W. MI | 112 | C. MI | 77 | ||
| Maryland | 68 | Ind | 101 | Purdue | 84 | ||
| NW | 10 | Maryland | 68 | rutgers | 105 | ||
| 47.2 | 73.7 | 90.0 |
Somehow MSU's average defensive opponent faced will drop (hah) from 88 to 90! So we can take their offensive stats with a lot of grains of salt. Yes its a solid offense but level of competition thru 6 games will be AWFUL. (they wont face a single top 70 defense all yr until UM week. Meanwhile UM's defense gets the same benefit - facing 2 awful offenses its average offense faced rank will drop an already easy 76 to 80.
Meanwhile UM will continue to face the best average defense as NWs 10 rank will roll in. And OSU will only the second top 60 offense on its schedule thus far in Indiana.
Conclusion? None of this matters on game day when we play MSU and OSU but it is interesting to see the data. In general UM's offense is facing the most difficult defenses (3 in the top 40) while MSU is playing some of the most horrid in the nation. Meanwhile UM's defense gets that same benefit (lots of horrid offenses). OSU likewise is only facing 1 defense in the top 60 thru 6 weeks so UM's D will be a startling difference for them.
September 29th, 2015 at 2:00 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 29th, 2015 at 2:03 PM ^
September 29th, 2015 at 3:53 PM ^
I may be wrong but I think at least one of either S&P and FEI have a component that factors in last year's performance which is slowly diminished and starting in week 7 last year's "effect" is completely gone, meant to somewhat counter the small sample size.
September 29th, 2015 at 2:05 PM ^
outstanding on defense by the end of the year and solid on offense. That said, I think maybe Sparty is in for a 4 loss season. We'll see.
September 29th, 2015 at 2:25 PM ^
That's crazy talk. State always starts out slow and improves throughout the season. They haven't had a bad loss since Iowa in 2012.
September 29th, 2015 at 2:39 PM ^
2013 ND was a pretty bad loss for them. 5 PI penalties cost them the game to an eventually 9-4 ND team.
September 29th, 2015 at 4:08 PM ^
I don't know if I would losing to a 9-4 team a bad loss.
September 30th, 2015 at 3:56 PM ^
To lose essentially by way of Pass Interference calls (like 100 yards of PI penalties) is a bad loss.
September 29th, 2015 at 2:26 PM ^
Well, they're not the #2 team in the country, but where do you get 4 losses from? OSU, UM, and what other two games? Please do tell...
September 29th, 2015 at 2:38 PM ^
Indiana (who can score and follows M) and/or PSU (who plays good D and follows OSU) and/or their bowl game.
They have been extremely lucky with injuries these last few years and they have been both skilled and lucky with finding lower ranked recruits can play.
I suspect they are returning to the mean in terms of the expected performance of the players they've recruited. They beat a good (not great) Oregon team at home in a toss up game and they have been pressed by 3 mediocre opponents.
September 29th, 2015 at 2:43 PM ^
There are only 2 "sure" wins left on the schedule, Maryland and Rutgers. Minnesota, PSU, IU, and Northwestern will be games that we should be favored in but still competitive. Our record in those contests will determine the fate of the season. Go 1-3 and we are headed towards 6-6 which would be a disappointment even in Harbaugh's first season. Go 3-1 and we have are at worst 8-4 which seemed to be the consensus for preseason record predictions.
September 29th, 2015 at 4:20 PM ^
Who are there 4 losses? Lets say we beat them (I think a lot of people are over confident, there DL scares me and I think forces Rudock into some bad turnovers) and OSU wins. That gives them 2 losses. You really believe they will lose 2 of
Purdue
Rutgers
Indiana
Nebraska
Maryland
Penn State.
The only even close games on that schedule would be Nebraska and Penn State and Nebraska hasn't looked very good and Penn State still gave up 8 sacks to Temple. Even if Indiana somehow won (they won't) I still don't see them getting to 4 losses. It just isn't happening to what is still a top 20 team.
September 29th, 2015 at 2:09 PM ^
Why can't the rankings account strength of opponent?
September 29th, 2015 at 2:09 PM ^
September 29th, 2015 at 2:29 PM ^
Held Stanford to 6. They were horrid that day but Stanford put up 40 on Oregon State and USC. Held Duke to 12 (?) and a baby seal to 0 in a 3rd game.
Per S&P+ the Big 10 defenses are stacked right now
- UM 4
- OSU 6
- NW 10
- Minn 11
- PSU 13
- Wisc 14
This is partly due to the level of competition situation and I expect will continue as so many bad QBs and offenses in the Big 10 but it wouldnt be surprising to see 6-7 of the top 25 defenses in the Big 10 when you consider what the average SEC, Pac 12, or Big 12 offense looks like vs the average Big 10 one does. All these B10 defenses get to play rutgers, maryland, minnesota, ill, northwestern, purdue type of offenses so will feast.
Advanced stats will only partially try to compensate for the firepower of the offenses faced.
September 29th, 2015 at 2:13 PM ^
Based on your numbers, I like the fact that MSU won't face an elite defense until we host them. It's always better, in my view, to face some bigger challenges before your rivalry games. It leaves open the possibility of players being a bit shell-shocked.
(I expect that MSU will already be shell-shocked to learn that they actually have to play at Michigan Stadium every few years - thanks again DB!)
Nice work as usual, alum96.
September 29th, 2015 at 2:14 PM ^
in helping tell the story you expected to tell in the first place. We use these because how else can you compare teams until they play each other. I do think that playing one rival half way through the season before facing the other in the last regular season game is a greater advantage in terms of recovery opportunity.
However, if you believe that facing two teams of similar capability two weeks in a row means you will be better battle-tested for both, then getting them at the end might seem like the greater advantage.
Whatever. Michigan still has home-field and that's always worth about 4 points before kickoff.
September 29th, 2015 at 2:30 PM ^
What on Earth does this mean: "Numbers are only relevant in helping tell the story you expected to tell in the first place"?
September 29th, 2015 at 2:43 PM ^
Wordier way of saying "stats say whatever you want them to say"
September 29th, 2015 at 2:47 PM ^
As someone who has spent much of his career doing stats, let me just say... UGGGGGHHHHHH. :-)
September 29th, 2015 at 2:19 PM ^
Your stats are very interesting and helpful. They support two things for me:
- I think, in general, Rudock and the rest of our offense are being a bit tested. This is good. I look forward to seeing how they do against NW. Regardless, I think that Rudock looks a bit worse than he is, because of the level of defense playing against. And that this will only help him improve over the course of the season.
- I am concerned that our defense isn't facing a really stiff smash mouthed offense with a really strong OL that can also pass. I'm not so concerned about MSU. I am concerned that nothing Michigan has faced (or will face in the next 6 games) will prepare them for OSU.
September 29th, 2015 at 2:35 PM ^
There is really nothing anyone can do to prepare for OSU in the Big 10 unless you schedule say TCU or Texas A&M etc out of conference. There is no like offense with high level OL play + actual passing competency.
In fact if you are just looking to play against passing competency (forget a smash mouth line) it is difficult to find in the Big 10 - you basically have MSU, OSU, Indiana and crickets. Maybe Iowa this year?? They put up 62 last week and their Qb is looking quite solid. Air Ferentz?
I agree on your other pts - NW will help the offense prepare for MSU and then MSU, Minn, PSU will help it prepare for OSU.
Lost in the love for OSU O, is the job Chris Ash has done as DC for OSU (well co-DC) - Fickell's D's were a lot like Carr's -loaded with NFL prospects but didnt dominate as they should. Ash's D the past 2 years has been very opportunistic and quite good.
September 29th, 2015 at 2:36 PM ^
To further rain on the parade, I'm really happy with OSU's defense. It's the best since 2009, quite frankly.
I think Harbugh may out-scheme them on a few different plays (for example, that double fake screen TE release pass would have scored from 100 yards away against us) but in general, I don't think Michigan has a consistent shot to move the ball. I think this is where Rudock limits you guys. At least Gardner could run around and avoid sacks. I don't think Rudock has the escapability.
On defense, your corners and safties will need to get off blocks to help corral the edge running game, because your linebackers will be beaten to the edge, and if the secondary starts pressing up, I'm hopeful Cardale is accurate on his deep ball that day.
That being said, speculating about a game in two months now is silly. Someone important's knees could explode tomorrow!
September 29th, 2015 at 2:48 PM ^
Yep with the level of offenses UM plays combined with our D - Iowa 2014 rudock for half the game and then not so great rudock can suffice for 6 conf games. But for MSU and OSU I think Iowa 2014 rudock needs to be here all 60 minutes. I realize we go to conservative offense in the 2nd half but the announcers still pinned rudock down a lot for missing open guys and not reading things well in those 2nd halves.
I thought the BYU first half would give him a lot of confidence so these continued missed opportunities (even within the realm of a boring 2ndd half offense) have me concerned. He cannot check out like that for 20-30 minutes against the high end teams.
And I agree with you on the D as I noted above. Just like MSU's D gets all the pub the real story at MSU has been their stellar OL play the past 2 years. And with OSU the D last year and so far this (adjusting for bad opponents) has been pretty damn good - and overlooked.
Also agree on LBs - LBs had an easy day vs BYU so no test there. Other than Indiana not sure another offense outside MSU/OSU will test them on the edge though in this conference. Again the benefit of playing a lot of run based plodding offenses with mediocre QBs. MSU has built a reputation exploiting that style.
September 29th, 2015 at 2:47 PM ^
Gardner could run around and avoid sacks
Missed that last year, were you watching the same season?
It was usually run around, get sacked, fumble or throw a pick.
Mobility is a non-issue. This is a pro-pocket offense where pressure is relieve by dumping to a TE. You blitz - the TE is open.
I agree with you on the linebacker issue. I do worry about speed on the edge.
September 29th, 2015 at 2:55 PM ^
Most OSU fans prob remember Gardner for the games they saw which were prob OSU 2013, ND 2013 and OSU 2014. (maybe they forgot ND 2014!!)
If that was your sample size he looked like a great scrambler who extended drives and at times was Vince Young 2.0. Those who had to suffer for the other 18-19 games of those 2 years saw all the times he created sacks or did crop circles backwards to get sacked and create 2nd and 19.
Not that he had much help.
Also I'd opine the explosiveness of 2013 in Gardner's legs was just not there in 2014 - if it was injuries, PTSD or whatever. He was beating OLBs to the edge most of the time in 2013. Not in 2014.
All in all he was elusive but his decision making cost as many yards as he helped create with his legs. Esp after game 4-5 of 2013 excluding the OSU and Indiana games where he was Vince Young.
September 29th, 2015 at 2:59 PM ^
True, the casual observer only saw a few night games when he did well.
The rest of the stink was on the BTN.
It is amazing how far the OL has advanced. It's like there was no OL coaching at all last year.
September 29th, 2015 at 3:34 PM ^
What drove me crazy about the Borges offense (and 1 year of Nuss) was the criminal underutilization of the running backs in the passing game. They couldn't pass block (except for Vincent Smith,) yet were expected to stay back and pick up blitzers instead of leaking out for short passes. Harbaugh is using every one of his running backs and his H-backs and his tight ends in the passing game. Rudock has thrown to something like 17 different players already this year. That's amazing.
September 29th, 2015 at 3:57 PM ^
I watched him play us three times.
Also, I'm really not worried about an offense who's main option is tight end. I know Jake Butt is a pretty good tight end, but at the end of the day, he's not Gronk, and scary offenses don't run through tight ends, unless of course you have the single best QB and TE to ever play the sport.
Plus, I don't think OSU will have to blitz that much, their front four does a nice job getting pressure, and they usually only blitz Lee anyway. Although, Raekwon has a 0% chance of covering Butt 1-on-1. Just a 0% chance.
September 29th, 2015 at 4:35 PM ^
Butt has the potential to be Gronk in 5 years, IMO. And scary offenses can run through a lot of TE plays - Stanford - and oh the Patriots with that Michigan guy who is running the same type offense he ran here that JH is trying to re-install and upgrade.
I think The Game comes down to which offense and QB play improves the most in the next two months. I am hoping your coach keeps flip flopping the QBs - that never seems to work out well :)
September 29th, 2015 at 6:32 PM ^
Michigan's offense will struggle unless Rudock starts connecting down the field, but im less concerned about OSU's running game against our defense. If Urban wants to make his bet on being able to run on Michigan all day, I'm happy to take that action.
You can ask Devontae Booker (69 yards, 3.1 YPC) what it was like trying to run to the edge on Michigan:
https://youtu.be/1FmYl7i2xJ0?t=7m
September 29th, 2015 at 9:07 PM ^
Utah's rush offense isn't on the same planet as OSU's. If you don't think your LBs struggle to get wide and make plays in space, I don't know what to tell you. OSU isn't going to square up and try and manball it up the gut, their rush offense stresses defense in all directions. Utah attempts that as well, but not at the level of OSU.
September 29th, 2015 at 9:56 PM ^
OSU has 926 yards rushing over four games to Utah's 816 (including a season low 127 against Michigan) . OSU may still be a bit better, but those certainly feel like they are in the same planetary system to me.
Im not even as exuberant about where we are as most of the fans on this board, but the idea putting up a bunch of points against Michigan on the ground seems... optimistic.
I agree that some of our LBs lack ideal speed, but if you are planning on beating this guy in space all day, feel free to try your luck:
September 29th, 2015 at 11:33 PM ^
I think our defense matches up quite well with OSU. Durkin is basically going with a nickel base with peppers playing a corner linebacker hybrid spot. I know elliot is fast but I think peppers will manage. I worry about runs up the middle and deep passes. On O Drake Johnson was chewing up yards in last years game before he went down.. I imagine with the way our oline is improving we will be able to run a bit on you. Will will still need Rudock to make plays and if I had to guess I just don't see him being able to play at the level we will need. Still happy that we are improving enough that I don't have to assume a loss..I think we have a chance.
September 29th, 2015 at 2:19 PM ^
It makes me curious what their offensive ratings would be if you removed the stats from the game they played against Michigan.
September 29th, 2015 at 2:36 PM ^
That is partially why advanced stats work better deeper in the year. 1 outlier game out of 4 changes the data a lot more than 1 out of 8.
September 29th, 2015 at 2:28 PM ^
I appreciate the leg work. Thanks for that.
HOWEVA, you can't just do this now. There is a reason these statistics are not used until week 7. While you can disclaim you know that, the inaccuracies underly your entire post and make it almost meaningless. To wit:
1. Michigan is going to have faced bad offenses, because those offenses are evaluated based on 1/3 or 1/4 of their games being against UM's defense. This is a huge problem with small sample size. You can't start using these when at least 25% of the info is based on the team you are trying to evaluate. Caveats can't correct that.
2. Similarly, how does UNLV have such a low offensive score? Is it because the stats against non-FBS do not count? So, one of the main reasons they are 111 is because they played Michigan?
3. This is obviously true for the other rival schedules. Has MSU played very good offenses or does their defense suck?
Once again, nice effort, but these are not usable stats at this point. Not even close.
September 29th, 2015 at 2:43 PM ^
Obviously statistical models become more accurate with more data, but S&P+ can start to be used after week 4-5 usually because it's model is on a play-by-play analysis, as opposed to FEI, which uses drive-by-drive analysis. FEI starts week 7 when the sample size gets big enough, but if you use play-by-play you get a larger sample size quicker. Also, S&P+ uses data from the previous season for each team with diminishing amount through the first 4 weeks to help with sample size. I don't know if this is the last week that data is included (albeit very little), or if this was the first week only this season's data was included. For some reason, I think it throws out garbage time stuff, but I don't know if that is absolutely correct or how they define it in their algorithm.
September 29th, 2015 at 2:50 PM ^
S&P+ does throw out the garbage time stats. See lower on page where I cut and paste it.
September 29th, 2015 at 4:03 PM ^
September 30th, 2015 at 4:32 AM ^
1. Caveats aren't supposed to correct it. They're there so we don't take it as seriously as we'd like. It's not a matter of being usable or not. The question is how usable.
2. Why wouldn't they be so low? They did absolutely nothing offensively against UCLA and Michigan. The stats adjust for opponent strength, so even if FCS teams count, blowing them out doesn't count for a whole lot.
September 29th, 2015 at 2:30 PM ^
Correct me if I'm wrong, but aren't S&P+ already adjusted for opponent strength? So looking at, say MSU's offensive S&P+ then looking at their opponents defensive S&P+ and saying MSU's S&P+ can be taking with a grain of salt is not a logical conclusion, no? S&P+ looks at how MSU did against bad defenses and concludes their offense acted like the number 19 offense in the nation should act against those bad defenses, thus caveats about their schedule shouldn't apply here. Sample size does. It would be one thing if MSU had the number 4 offense in the country against terrible things, and S&P found them to be #30 or something.
September 29th, 2015 at 2:41 PM ^
Both FEI and S&P+ try to adjust in diff ways. Key is try. There is no perfect way to adjust everything to even keel but they have different ways to try to adjust.
So for example here are 2 adjustments S&P+ tries to do - are they pefect? no. do they help to a degree try to adjust? yes. Do they make everything exist in a pefect vacuum? no - nothing does. So the short answer is "yes they adjust to a degree but it's certainly not perfect - but a lot better than the unadjusted NCAA stats".
- 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.
September 29th, 2015 at 6:33 PM ^
as the comment you responded to, and it seems like you confirmed the criticism by pointing out that it is opponent adjusted already.
I do think it is very misleading to show the overall opponent adjusted numbers, show the opponents themselves, then discount the overall numbers that have already been adjusted for opponent. I'm not sure what we are taking with a grain of salt here: the overall advanced metrics or the traditional tempo-biased, non-opponent-adjusted stats? Because if it's the latter, I totally agree.
Appreciate the effort put in. Thank you.
September 29th, 2015 at 3:00 PM ^
tl;dr would have been useful
September 29th, 2015 at 3:17 PM ^
September 29th, 2015 at 6:42 PM ^
had already taken a big step back defensively according to the advanced stats: 22nd in S&P+ and 40th in FEI.
September 29th, 2015 at 4:36 PM ^