Total yardage and yards per play were almost identical. Michigan had a significant edge in efficiency. So how did Notre Dame control the game and come away with the win last weekend over Jim Harbaugh and company?
Confused? Check out this handy-dandy advanced stats glossary.
The Basics
As you can see, the Irish and Michigan were nearly dead even on all of the high-level categories here. These numbers here exclude only end of half / game kneel downs, as there was no garbage time in this one. The points per drive stand out – in part because the Michigan kickoff return touchdown isn’t captured here (it seems weird to capture that as a “possession” or “drive”, so it will later be included as part of evaluating field position).
Efficiency
The Wolverines were significantly more efficient that the Irish, driven by a solid run success rate and much better passing efficiency than Notre Dame. Both offenses struggled mightily on passing downs – not a surprise given the strength of the pass defenses on the field. These success rates in some ways mask the struggles of the Michigan offense – the Wolverines moved the chains decently, but were unable to sustain drives or pick up chunks of yardage at a time. They were limited to a death by a thousand cuts offense, and found out that the Irish defense was essentially rhinoceros skin as the field shrank.
Seven of Michigan’s ten possessions crossed mid-field, and yet the Wolverine offense could only muster up 10 points. Clark Lea’s defensive scheme definitely leaned toward “bend but don’t break” – relying on Notre Dame’s ability to get pressure on Shea Patterson mostly rushing four and giving up few long plays. “Clutch” is an adjective more frequently ascribed to offenses than defenses, but the ND defense rose to the occasion throughout the opener:
-
- Michigan’s first drive started at the 35 after a kickoff out of bounds, quickly picked up a first down, and then entered Irish territory with an 8-yard completion on 1st down. Then the Irish stuffed Karan Higdon for a 1-yard loss on second down, and limited him to a 1-yard gain on 3rd and 3. Harbaugh, coaching like a nervous bird, punted on 4th and 2 from the Irish 43.
- The Wolverines picked up three first downs on their second drive, moving to the Irish 29. After two unsuccessful plays on 1st and 2nd down,a Khalid Kareem sack for a 16-yard loss moved Michigan out of field goal range.
- Again, Michigan’s third drive moved all the way to the Irish 5. On 2nd and goal from the 2, the Irish sniffed out a play-action pass (Drue Tranquill picking up the fullback slipping out) and Jerry Tillery picked up a sack, followed by a clutch Alohi Gillman pass breakup on 3rd down. A field goal was a huge loss for UM.
- The fourth real Michigan offensive drive, an eternity later due to the return TD and a last end of half possession, was jump started by a blown coverage and 52 yard completion. After arriving at the Irish 16, Patterson went incomplete, incomplete, incomplete, and the field goal attempt was botched.
- These are all still in a row! The fifth UM possession started at midfield after Wimbush’s interception. Asmar Bilal came up with a four yard TFL on 1st down, and the defense held on 4th and 4 from the ND 45.
The Irish offensive efficiency numbers are non-glamorous, in part because of a fairly conservative approach with a lead against a very good defense. Don Brown’s defense, especially in the second half, basically dared the Irish to throw more into one-on-one matchups. And while Chip Long obliged a few times to mix things up, he mostly said no thanks, including closing the game by running on seven of the last nine passing downs.
It’d be easy to nitpick the run and pass success rates, but again, it’s important to remember that this Michigan defense is probably very good. The Wolverines led the nation in opponent success rate last year, so the Irish numbers in this game were actually better across the board compared to Michigan’s 2017 stats. Brown’s defense allowed just a 27.7% passing success rate last year, and 33.3% on the ground. The Irish disadvantage in leverage rate and average 3rd down distance- facing a ton of passing downs. But that was overcome by strong 3rd down performances, highlighted by multiple Wimbush scrambles and some nice throws in those scenarios.
Explosiveness
This is how you make up for a disadvantage in efficiency – when successful, make it count. And while the yards per play battle was close to even, Notre Dame had dramatic splits from the early and late drives. Through the first four Irish drives, which ended with Jafar Armstrong’s second touchdown, the Irish averaged 6.41 yards per play. After that point, with a double-digit lead for most of the game? Just 2.89 yards per play.Some of that was offensive conservatism, but some was also the Michigan defense playing better and the Irish offense executing worse. Brian Kelly pointed out as much in his Tuesday press conference, stating that there was a lesson to be learned in putting a team away.
Overall, I think it’s tough to come away with much besides cautious optimism for the Notre Dame offense. The line struggled a bit against an elite DL, but wasn’t dominated and flashed on several occasions in the run game. Chase Claypool, Miles Boykin, and Alize Mack each had big contested catches. The stable of running backs showed promise but room for improvement.
Brandon Wimbush was a steady leader. It wasn’t a perfect performance, but his excellent start to the game throwing the ball likely made early progress on the ground possible. He hit several touch throws and slants under pressure that were troublesome last year, and constantly made Michigan pay with his legs even when they saw it coming. Like the offense as a whole, there were missed opportunities, but generally the decisions were good and mistakes didn’t snowball. The spectacular plays were enough to win, and silencing the Ian Book discussion for a while is probably a good thing for this team’s ceiling.
Every single defense from here on out should be easier, and Wimbush should get plenty of live work in the passing game between now and Stanford. It’s easy to forget that Miles Boykin only played around 100 snaps last year, and that there’s still rapport-building to establish between Wimbush and some of these receivers.
Winning the explosiveness battle was just as much about the defense, though. Michigan’s long run on the night was just ten yards. Patterson had to work very hard to hit tight ends and receivers on crossing routes while he was on the run himself, and the first tackle was made virtually every time. Every time it seemed like the Wolverine offense was finally going to break through, the defense generated another critical negative play.
After leading the nation in long pass plays, allowing zero of 50+ yards, all last season Nico Collins did snag a 52-yarder on a blown coverage (miscommunication between Julian Love and Jalen Elliott). Fire Lea!
Finishing Drives, Field Position, & Turnovers
As mentioned earlier, many hard-fought Michigan yards ended up squandered. Three Michigan drives of 47+ yards resulted in zero points, while Notre Dame scored on each drive that ventured inside the Wolverine 40. The Irish were almost perfect, as a penalty negated a beautiful Boykin touchdown reception on the drive that later ended with an almost-equally beautiful Justin Yoon kick right down the middle.
Field position (and inter-related special teams) could have been extremely costly. The average starting field position differential ending up only being about a yard in favor of Michigan, thanks to the game-ending fumble recovery, but could have been far worse. The lowlights included the kick return touchdown and another kickoff out of bounds. Tyler Newsome had some excellent punts but also a 25-yarder that gave the Wolverines the ball on the ND 41. It didn’t cost Notre Dame, but this continues to feel like a dumb category for a team this talented to struggle in.
From an advanced stats perspective the Irish were a little lucky with turnovers since they recovered the only fumble of the game (100% recovery rate!). Watching the game, however, Wimbush put the ball in less danger than Patterson. There were a number of pass deflections and poor passes Patterson threw that he was lucky not to finish with more than one pick. His pocket presence was exceptional considering that Michigan defensive ends were quickly coming around the edge on most drop-backs.
Final Thoughts
On paper I predicted Michigan would be the toughest opponent the Irish would face all year. And they still could be! But the potential things advanced stats either dinged the Wolverines for or didn’t know were on full display in the loss. The offensive line looked just as bad as 2017, the offensive scheme and weird play-calling arrangement seemed poor, and the defense was very good but not good enough.
The performance by the Irish defense was exceptional, with great signs in Clark Lea’s first game as DC. It will be fascinating to see how the ND defense adjusts to different matchups, including better offensive lines. Will Lea bring more pressure against other teams? Or is the “bend but don’t break” that worked against the Wolverines what Lea really wants to do? How dominant can this unit look against Ball State, Vanderbilt, and Wake Forest?
The Demon Deacons in particular shouldn’t be overlooked, but there’s the potential for more spotlights wins (or crippling losses! Miami wasn’t that long ago) less than a month away. Stanford has to get through USC and Oregon before coming to South Bend, but even splitting those games they will be a top-20 opponent. Virginia Tech can’t lose a trap game against Duke, but other than that has cupcakes prior to a game in Blacksburg that could be a top-10 matchup.
All I know before I get too far ahead of myself is that this was a win to enjoy, and I refuse to let “eh Michigan isn’t good under Harbaugh anyway” or “Wimbush is still bad, we’ll lose later” speculation dampen that. A win over a team with that much NFL talent, that needed and wanted a win so very badly, is enormous. It’s still early, but the growing track record of the “new” Kelly-Long-Lea(Elko)-Balis era is something worth getting excited about, and this adds another top-10 quality performance to their resume.
I absolutely love the explosiveness graphic you chose. Nice touch!
I realized during the game that Michigan didn’t have any answers on offense when I watched them try to run option and get stuffed. I really hope this game showed off how talented our defense is more than how inept Michigan’s offense is.
A few things…
i forgot what leverage rate is…and it’s not in Connelly’s glossary
i was shocked to see michigan’s avg gain on first down was “only” 5.5…i was complaining all game that we lost nearly every first down.
Considering how conservative the offense got with a lead can the “not putting a team away” really be on the players and not the coaches? Sure there’s some responsibility in both columns but the offense could have been set up to succeed by actually running the offense.
i think Finke deserves to be in that short list of receivers who had big contested catches.
As for the defensive strategy going forward, i hope the aggressiveness of the Citrus bowl is the norm and BBDB is used less often. It seems that Pride stuck to his man pretty well, but maybe his guy wasn’t targeted often? No PBUs for either Pride or Love 🙁 Love seemed a little out of place playing so far off his man. His strength seems to be sticking close and playing the ball. He doesn’t seem to have been used that way against michigan.
Good call on Finke – he’s my new twitter avatar but left him out here somehow. What a play.
Leverage rate is your percentage of standard downs vs passing downs when on offense. Will have to check but believe this would have been the lowest leverage rate in any game in 2017
Thanks!
“All I know before I get too far ahead of myself is that this was a win to enjoy, and I refuse to let “eh Michigan isn’t good under Harbaugh anyway” or “Wimbush is still bad, we’ll lose later” speculation dampen that.”
Don’t dampen your enthusiasm, but those are still things that we need to address. Overreaction in either direction is silly–the fact that ESPN has us 4th (!!) in their power rankings and something like 4 of their 5 playoff projections have us in the playoff seems more ludicrous to me at this point as saying we’ll lose later because Wimbush is still bad.
What these stats tell me is that we found a way to beat one of the 4 biggest hurdles on our schedule (UM, Stanford, VT, and USC, dropping FSU after this weekend because WOOO BOY THAT O-LINE SUCKS WORSE THAN MICHIGAN’S), so our floor raises and it appears more likely we’ll have a solid season with a good bowl game; we can be reasonably confident we’ll beat the teams we should beat, barring unforeseen injuries, etc.
BUT, we didn’t show any particular dominance that suggests a definite advantage over the remaining 3 toughest opponents, or that we could hang with an Alabama or a Clemson. That’s not to say we have disadvantages or I think we’ll lose–the way we won doesn’t necessarily engender confidence that we’ll definitely win. We’re a solid team, probably top 10, and a 1-2 split with those remaining tough opponents (giving us a 2-2 split overall) puts us at 10-2 which most people suggested in the poll. I picked 9-3, anticipating a loss to Michigan, so I guess I lean towards 10-2 now as well.
(Also, the only opponent that dominated what was thought to be a good team was VT, and I think we may find out that FSU is not a particularly good team at all. Stanford and USC were fine but big wins against outclassed week 1 opponents should be what they are doing, unlike last year /glares at Stanford).
Any week 1 pronouncements seem inconclusive to me, other than hey that Tua kid is pretty good, isn’t he? Blow out Ball State, blow out Vandy, and beat Wake decisively, and it will be easy to attribute the way we won over Michigan to “first game, just needed to do enough to win, who cares.” Look shaky against any or all of them, and it’ll be awfully hard to think we’re a playoff contender.
TL;DR: MOAR DATA
Also, I’m so torn: On the one hand, I want to believe we that beat a good team in a tough game, and that this “IS HARBAUGH ON THE HOT SEAT?!” talk is truly the week 1 hot taeks they appear to be. On the other hand, it’s Michigan, and I hope they never get it together and lose every game by 40.
Hope to do more than beat Wake decisively. They beat Tulane by 6 points and had a number of turnovers. They are definitely in a rebuilding year. I have them on the same level as Vanderbilt this year. Vanderbilt has Nevada before us and Wake Forest has BC. Should learn a little about each team from those opponents. However, as long as ND stays focused, both teams should be out of contention in the first half.
The three game stretch of Stanford, VT, Pitt scares me. Loose focus in prep for anyone of those and the game is going to be closer than it should. VT looked like beasts against FSU.
Wake lost their starting left tackle for the season against Tulane, and if there’s a position I wouldn’t want to go into a game against ND at it’s gotta be offensive line.
Not to project too far ahead (because I certainly believe Stanford and VT are games the Irish can lose), but if ND were undefeated going into the Pitt game, what are the odds they’d be ranked #2? I don’t want to play Pitt if we’re ranked #2.
Let’s hope Pitt gets their yearly scalp this week against Penn State.
I hear you. I think more than anything my intent there was more “let’s live in this moment a while” versus swinging the pendulum too far in any one direction. More data will be good, and I think the USC-Stanford game in particular will be a big indicator on their relative strengths.
VT will be interesting, because FSU may just be a mess early but they also have a very cushy schedule here. The only other currently ranked team on their schedule is Miami, who also may not be that good. I think they’re getting a little too much love right now – they didn’t move the ball all that well on FSU and won by 21 with a 5-0 turnover margin. I still think they may be top-25-35 quality versus better, but they could pull the classic Northwestern where with the schedule and a team of that quality they still hit double-digit wins.
Agreed, VT may be get a nice inflation that makes it more impressive if we beat them. A win against a 10-2 ACC Coastal champ and runner up to Clemson for the conference title would be a nice feather. Miami certainly looks beatable, and BC may be their next toughest competition for the division. And they don’t really have anyone else from the Atlantic that’s scary.
That said they’ll be UNC’s one conference win, or something crazy.
“Don’t dampen your enthusiasm, but those are still things that we need to address. Overreaction in either direction is silly–the fact that ESPN has us 4th (!!) in their power rankings and something like 4 of their 5 playoff projections have us in the playoff seems more ludicrous to me at this point as saying we’ll lose later because Wimbush is still bad. ”
I wouldn’t say it’s an overreaction from ESPN’s perspective, as they had ND at 4th highest odds to make the playoffs before the season started – clearly those odds would have only increased after beating the only team favored to beat ND all year. It’s less Chris Bridges to have them in the Playoffs now than it was 6 days ago.
I was flipping through their college football live show and they had a big graphic showing a 56% chance Notre Dame makes the playoff. I feel like that’s really high, but to the point Michael’s touched on in the article ND could/should be favored in every game here on out if they keep winning. Still seems more like a 10-2 team than a 12-0 one to me, but at this point we should be looking at 9-10 wins on the season, when last week 7-9 was much more in play if a loss to Michigan in Week 1.
So I don’t think there’s any need to over-react, but at the same time it’s still a really, really good, possible program-affirming win to put in your pocket and move forward with.
That is absurdly high. We should only be like a 56% chance of beating Stanford or USC. That doesn’t even take into account no one knows if an 11-1 ND will even make the playoff.
ESPN “Statistics”
@Don’t you dare conflate Bill Connelly with ESPN!@
But seriously, Bill Connelly is awesome. And while the ESPN FPI is not as respected as S&P, it’s still a model that generally provides reasonable predictions, though of course the Vegas lines incorporate important qualitative data that the model doesn’t.
Yeah, Connelly is awesome. Hence my swipe at ESPN’s FPI BS. 56% chance of making the playoff? That’s absurd. That’s Alabama level.
This is as dumb as the whole in-game “projected to win” “metric” which as often as not has a team with a 96% win likelihood with a 7 point lead in the early 3rd quarter. Or ESPN’s BS “QBR” numbers that make no sense. It’s just dumb.
You call it dumb, but I love ESPN’s win probability %. Then again, I love how wildly it swings during college football and basketball games.
Tonight: College kids make mistakes. More to come at 11.
Oh, sure, you’re loving it for the exact reason I’m hating on it. By all means, embrace the crazy! lol.
In his preseason preview, Bill Connelly had ND favored in every game with win probabilities of: Michigan 62%, VT 66% and USC 61%. Next lowest were FSU, Stamford, NW and WF (!!!) at 75-76%
https://www.sbnation.com/college-football/2018/5/8/17316948/notre-dame-football-2018-preview-schedule-roster
I don’t know what goes into his win probability %, but most assuredly, there’s some #maths in there somewhere.
All the win probabilities are basically using their measures (S&P+, FEI, ESPN’s FPI, whatever) that say X team is # of points better than Y, then plugging that in to account for homefield and seeing (with kind of a normal distribution) about what % of the time that should get to a win of any measure. You can get more fancy than that, but it mostly boils down to something like that most of the time.
FEI also week to week projects the score and % chance of a blowout win, close win, close loss, etc., which is pretty interesting. It really reflects what the prediction game is about – there’s a range of outcomes that all can happen if you play these games 100 times, but which of those is most likely? We, as biased humans, are often really bad estimating the probabilities of future events – we are way too swayed by our eyes and experiences. Some of that is relevant data, but most of it is availability and confirmation bias.
Anything less than 99.9% against Stamford would mean we’re terrible! i mean, they’re just a city in Connecticut, not a DI college football program 😛
So in that case it’s not an overreaction to week 1, it’s simply ESPN’s hand-wavey “statistical analysis”
I’m not trying to die on an ‘ESPN has good analysis’ mole hill, but almost all statistical analysis is “hand-wavey.”
And how is your analysis that ND shouldn’t be in the Top 4 quantified? Also somewhat “hand-wavey?”
For one, I’m not the one trying to pass off my assessment as “statistics.” But hey, whatever, we beat Michigan, that means we’ll be in the playoff, let’s inflate that expectation so that if we don’t make it we have something to artificially be angry about. That sounds great.
What is wrong with saying, “We have a model – based on historical data – and this is what the model says?”
No one is arguing that said model the truth. And definitely no one is arguing that a statistical model trained on past data is the truth – those of us in the statistics field know better! But I think the counterargument of, “How is the KG model better? Can that be quantified and tested on historical data?” is a fair response.
This is a sore point for me because at work I deal with this discussion constantly. I have a model with known holes – like how the S&P doesn’t deal with coaching change and doesn’t treat a player lost to the NFL any differently than a player lost to graduation. But it is a logically consistent model that can be tested and refined. Criticizing such a model without offering an alternative system that can be tested is unfair and arbitrary.
Orlok just shutting it down.
{insert: grizzlyadamsnodding.gif}
There’s nothing wrong with saying there is a model, and this is what the model says. What’s wrong with saying I, given the results of one game, don’t yet see what the model is projecting?
There’s two issues here:
One, that I disagree with the conclusions being drawn on the models. I think the responses between myself, Michael, and Brendan have been pretty reasonable on that. It’s not even that I’ve said “OH THEY ARE WRONG” just that we don’t have enough this season yet to really judge, so I’m taking a wait and see approach while others seem to be more gung ho. Cool.
The second issue is that I really dislike ESPN’s models in particular. I haven’t really challenged Bill Connelly’s because I like Bill, and I like S&P+. He does a good job of explaining rationale behind what he does. So that’s not nearly as “hand-wavey,” though I’ll grant the point that all statistical analysis only puts out what you put into it. ESPN’s stuff, however, is much less open, and much more prone to coming out with clearly weird results. Take their QBR ratings, which at one point had a 2010 game by Charlie Batch (12/17, 186 yards, 3 TDs, 2 interceptions) as the highest ever rated with a score of 99.9. They’ve since updated it (it no longer appears in the top 100 results, I just checked) and finally put an explanation in 2017, but it’s still very subjective. Their in-game “predicted chance of win” stats are also problematic. And I’ve long heard people who love S&P+ correspondingly complain about ESPN’s FPI. Hence my comments directly on not trusting SPECIFICALLY ESPN’s model.
I don’t have a “model”–which is obvious, and I’ve never claimed one. I’m not even saying my “gut” is right. All I’ve said is ‘hmm….we don’t know yet if Michigan is good, so I think ESPN personalities rushing to put ND in the playoff is a bit premature.” You yourself said “Completely agree KG” 5 hours ago. Hell, I wouldn’t have reacted to Drick if he didn’t feel the need to make disparaging comments in a different response to me, one that I agreed with. He’s not trying to die on the mole hill, yet here we are.
Fair enough my man – I also suspect these models are overly rosy about ND and your points about ESPN vs. other models are well taken. I wouldn’t call that hand-wavey, but I think I locked in on that set of words which was missing the (random) forest for the (decision) trees.
The whole situation is something of a trigger for me, because leadership will often trash an entire model due to one bad result. Your Charlie Batch story is hilarious, without a doubt. But if this were happening in some divisions I’ve worked in, we would then just have the VP fill out a list of what they thought the QBRs should be and we’d use that instead of any sort of model (and our VP would have played football 30 years ago and not know how modern offenses are supposed to work). It is hilarious but also infuriating to me that I got my PhD and get paid tons of money for a VP to look at my results and say: “You know that automated process we told you to make? We don’t like one number so just do everything by hand.” And this also happens to most if not all of my friends.
So yeah – I agree the models are wrong in this particular case. But I would look at these models very seriously and adjust from them as a baseline. Here, we agree that the revision of playoff probability should be downward.
Oh, I hear you. I was an intelligence officer and policy analyst for 17 years. And yet still had commanders who would be like “okay, fine, that’s what the enemy would really do, but we don’t want them to do that in this training simulation, so do it the way we want it.” Sigh.
This makes me think of the Blue Force in the Millennium Challenge 2002 (mentioned in “blink” by Malcolm Gladwell which I definitely had to flip through to remember the story).
That’s all well and good but SIR I will NOT let you sully the good gd name of Chaz Batch!
He is certainly one of the best average, middling, forgettable QBs of the NFL circa 2010ish.
“All I’ve said is ‘hmm….we don’t know yet if Michigan is good, so I think ESPN personalities rushing to put ND in the playoff is a bit premature.”
I think this is my poorly-worded nit that I’m picking. ESPN didn’t necessarily “rush” to the ND to playoffs conclusion. They were high on ND BEFORE the Michigan game.
Apologies for the pestering, I just find your steamrolling of others a bit abrasive sometimes.
Okay.
Completely agree KG; it’s too soon to say much. But since we’re posting on a football website – let’s speculate!
As sweet as it would be to watch Michigan never win again, I think they’re a better team on offense and defense than they were last year. I have them losing to wisconsin and OSU but beating MSU and PSU. The rest of their games they should win handily – though Iowa is a dragon slayer.
Based on my speculation about Michigan – and also based on his decision making and minimal passes to nowhere – I judge Wimbush to be much improved and our receivers to be no worse (and possibly better) than last year). So the dropoff is going to be in the O-line and the running game, but we get Dexter back the second half of the season.
Wake, Northwestern, and Pitt are all capable teams. The hardest teams remaining are Stanford, USC, and VT (probably in that order). For what it’s worth, FSU’s skill players are dangerous and Syracuse looks improved.
Man, all of these teams are individually very beatable, but have you looked at how much we’re traveling the back half of the season? That is truly terrible and I’m understanding Kelly’s complaints about Jack’s scheduling.
Yeah, the travel schedule is murder. And no one in the national media or on the committee will care or give us credit for it.
Nor should they give any sort of extra credit, especially for this program that isn’t in a conference and has 100% autonomy on who/where they want to play. Notre Dame has no one to blame but themselves for their schedule and traveling more isn’t to make things tougher it’s mainly for the brand.
I know. I e-mailed the university reps about this, and they basically said, “Jack Swarbrick has heard Brian Kelly and doesn’t care.” Actually, they said, “Jack wants us to play premiere games.” But they literally acknowledged (a) the problem and (b) that Jack doesn’t seem to be doing anything about it.
Should it excuse a loss? No. But it should absolutely be recognized if we win those games. They gave Stanford the benefit of the doubt for their body-clock BS a few years ago. Why shouldn’t the difficulty of travel be recognized accordingly if all other things are equal between ND and another team?
Because ND’s travel is because they think playing Navy in San Diego or Syracuse in Yankee Stadium adds luster to those games. No one forced them to do that except themselves. Thus if that trips them up, the athletic department deserves nothing but scorn from the football program and world.
I doubt anyone held a gun to David Shaw’s head and forced him to play Virginia.
Yeah, when was the last time Alabama had “brand” scheduling like this?
@At least we’ll look good in pinstripes…@
All the advanced stats loved our defense and kinda liked our offense (both account for returning production) in their preseason rankings. We were #9 in preseason S&P+, #6 in preseason FEI, #6 in preseason FPI. Our high rankings in each now – #8 in S&P+, #6 in FEI, and #5 in FPI – are merely an incremental progression based on beating a (probably) quality opponent in week 1.
If you want to have some more fun, track this link as we move along:
This guy reverse engineers Connelly’s S&P+ formula to create updated week-by-week projected win probabilities. It’s pretty freaking cool. It has us with a 44% chance of being 11-1 or better (add the last two boxes in the bottom right to get this), up from a 21% chance last week. It also shows a 75% chance of being 10-2 or better, up from 50% last week. The Michigan win was massive for our long-term prognosis in advanced stats, because all the advanced stats viewed it as a matchup of two pretty good teams.
Of course things can change a lot from week to week; for example, our win probability over Florida State jumped from 72% to 87% after our respective games. They played really, really badly against VT, which has an outsized effect on the numbers at the moment. I’m not sure about FEI and FPI, but Connelly says preseason S&P+ relies 85% on projections, and that decreases as the season goes on. So as we get more relevant 2018 data it should hone in on “real” quality.
Anyway, there’s a substantial probability that the reason the advanced stats like us is that we’re actually good. I agree that it will out, of course. And football games are played by human beings, who are notoriously flaky. Nonetheless… Just saying.
Seems a bit “hand-wavey” because it doesn’t align with my gut…
You’re right, how dare I suggest ND might not go to the playoff. ESPN says it after week 1, so it must be so.
Geez man, I haven’t seen anyone on here suggesting we all book tickets for the Cotton Bowl. We’re talking about PREDICTIONS. Those predictions don’t jive with what your instinct says, so clearly, we’re all wrong for having read and discussed them. Apologies. I’ll let ESPN, SBNation, this site and countless others know that they should stop reporting and predicting about college football until the season is complete.
Geez man, that doesn’t seem necessary, it’s much more productive to make fun of me and discourage anyone who might disagree with ESPN’s statistical projections from voicing a contrary opinion.
You said:
This seems like an overreaction because I don’t think ND is that good.
Other posters:
Ok, sure, but here’s where that “overreaction” is coming from.
You:
Nah.
You: Let me respond to another poster with a left-handed personal swipe at KG, because that’s completely helpful and will get my point across.
I am left handed, so, joke’s on you.
Cool
I work in forecasting for a fortune 500 company, and that pretty much summarizes *every* response my team gets from upper management.
Which is fine–I understand the idea that you beat a good (?) opponent, the statistical numbers will move in a positive direction. I may be in the humanities, but I can get that much.
I like Connelly a lot, and listing to PAPN religiously. But I think in our specific case, the heavy early reliance on last year’s numbers to project this year is missing things that I think (won’t know until we play more games) are hurting us, namely the left side of the line being gone. Other things it accounts for (Brandon’s limitations, lack of RB productivity returning) but I’m not convinced yet (again, moar data necessary) that we’re going to be fine.
I guess what it boils down to for me is that Michigan looked bad enough on offense that I’m not convinced they are a good team. They might be, but they could also lose 5 games. Our schedule, while tough, doesn’t have anyone with the possible exception of USC (if their new QB gels, which looks possible) and Stanford (but their defense needs work) that is a test on the level of playoff contenders. Yes, it’s great that we’re favored in all our games and the statistical models project 11-1. That’s better than projecting out to 8-4. My fear is that our 11-1 will be Michigan State-ian and we’d be blown out in the first round of the playoffs. The first half felt like a beatdown of Michigan, but the statistics tell a very different story–we won, but we didn’t dominate them. 12-0 or 11-1 DESERVES a spot in the playoffs, and I’d gladly take it no matter how we get there. But I don’t think we have enough information–either statistically or “eye test”–to suggest we’ll OMG definitely win 11 games and be in the top 4.
I think that’s realistic….this is a team that if it performs well is borderline playoff quality. With bounces the right way I think a one loss or better season is in play if that’s the case. Does that mean that the team will be the best one in the playoff? No way. But you make the playoffs, show up, and see what happens. This team is certainly at a talent disadvantage against the Bama, OSU, Clemson, and UGA division, but the track record in games against that quality of team semi-recently is not like they’ve been slaughtered (UGA last year, OSU in the Fiesta Bowl, Clemson 15).
So sign me up for a Michigan States-ian playoff berth if we can get it, and hope we have a better result is I guess what I’m saying.
Absolutely–I’d rather make the playoff and lose to the eventual champ, even ugly, because then you can say HEY, WE MADE THE PLAYOFF. And the recruiting pitch becomes “we’re one of the few teams that can do this, but we need YOU, 5-star, to help us win it.”
Yeah, I definitely agree with that – calling playoff teams after week 1 is pretty ridiculous. But ESPN has to fill its content needs somehow.
Massey-Peabody, also fairly well respected in their projections, had the following playoff odds:
1. Bama (62%)
2. Clemson (51%)
3. Oklahoma (51%)
4. Ohio State (49%)
5. UGA (42%)
weirdly large gap….
6. ND (18%)
7. Auburn (17%)
8. Washington (13%)
9. Virginia Tech (13%)
10. Michigan (9%)
I think I could be on board with this–it’s reflective of our schedule that we’re the sixth-most likely team to make the playoff, but the % isn’t an huge endorsement of us either. Which makes sense. Those listed below us have to make it through someone listed above us, either in-division or in a conference championship game.
I acknowledge (as a former biostatistician) that all of these projections are hand-wavey…
But the Massey-Peabody projections are a better reflection of how I “feel” the actual odds of ND making the CFP are at this point in time.
But we can always wait for more data.
I actually feel like it’s more than 18% at this point, looking logically. Auburn beating Washington was huge, if ND takes down USC+Stanford, no PAC team has gone undefeated in conference play since 2010. They’ll cannibalize each other a bit and boom bam Notre Dame can be the defacto PAC rep if they get to 11-1. A lot can and will change, it’s only 1 week but so far so good.
Granted, the long odds coming in getting to 11-1 and perhaps that won’t be attained.
M-P has the probability of the Pac-12 as a whole getting a bid down to 29%. Eeesh. I think the next lowest conference is the Big 12 at like 60%.
Michigan at 9%?? Before or after losing to us? Possibly understandable before, silly if afterward
Well, that one matches FPI’s conference win likelihood for michigan 8.6%….although their win out % is only 1.1% So yeah, 9% seems silly since simply winning the Integer won’t be enough if they end up with 2 or more overall losses.
I know Navy just got pantsed by Hawaii, of all places, but something scares me to see that we have a near 99% chance to win that game. This is where “eye test” and past experience doesn’t allow me to agree with that. Navy flat out will come to play and their whole season revolves around ND and Army.
I think it just means that HAWAII is a dominant team that will clearly make a run for the G5 NY6 bowl, and possibly dethrone UCF as National** Champions***
I agree that one stands out. Ball State is more than twice as likely to beat us as Navy? Syracuse 4 times as likely?
On paper Notre Dame should smash Navy if it were a normal game. Because on paper it doesn’t know Navy’s going to try to give ND 6-7 possessions if possible and kill as much of the clock that they can and keep it close and have it “Get late early” as Yogi Berra might say.
A lot easier to hypothetically get a few stops and get up 21-0, and never look back and win going away as they have to change up strategy than it is in reality given the uniqueness of their offense, willingness to go for it on 4th downs and #grit. Can’t measure that on paper yet.
Cool Table, and the floor for ND according to above table is 6 wins….. woo hoo we are going bowl-ing!
The biggest reason ESPN FPI has us so high is simply because we’re fortunate enough to not have to play any of those truly elite teams this year. Stanford is our toughest remaining opponent, and they looked bad against SDSU for more than a half. I also agree with you that 56% chance of making the playoff is high, but I also think our schedule is full of good, not great teams. I don’t see an opponent that will end the season in the top 10 on the schedule; that’s hugely beneficial in the current playoff format.
I wonder with Stanford, they had Love shut down in their first game by SDSU, so started passing on them. I think we have a better defensive backfield than SDSU. This could be the game plan to beat them?
I think that should always be the plan against Stanford, but man they’ve been solid at beating us downfield with jumpballs to big pass catchers. I still think I’d start the game by stacking the box against them, but I’d be ready to shift that strategy pretty quickly if necessary.
One thing i think has been underrated/not talked about enough is how ball hungry our receivers/TEs looked. The first pass there was a 50-50 ball that last year might have been picked off but our guy went and made a play on it. Same with Finke’s TD. and the TD pass that got called back too, i think. Really great to see the combo of improved QB play and improved WR play at the same time. Not to mention anything else. More punch anyone?
I think the most interesting thing to watch the next few weeks will be where the CBs play. As mentioned, they gave an absurd amount of cushion on 1st down (which I wasn’t opposed to against a crappy offense like Michigan’s that would need big plays to beat us). How is Lea going to line things up in preparation for Stanford? Their massive pass-catchers make for a very tough matchup, and I want to know how he’s going to try to slow them down. Will he try to jam them at the line (to throw off route timing), or will he leave plenty of space (knowing that if they catch screens in space they probably aren’t going to beat our tacklers 1-on-1 anyway)? And will he use the next two games as a tune-up, or will he try to disguise his strategy for Stanford? Either way, I think it’s the most interesting storyline for the defense heading toward that potential Top 10 matchup.
I’m very interested to see how Stanford looks against SC this weekend. They looked like trash on both sides of the ball for the first 28 minutes against San Diego State, and if not for a fumble on an interception return they might’ve gone into halftime down 14-2 instead of up 9-7.
They had 50 yards rushing on 28 carries, including 29 on 18 carries by Love. They had 382 yards of total offense and Arcega-Whiteside had 226 of that. He’s good, but if he’s the only reliable weapon they have I think we’ll do just fine.
So I’ll be watching their offense very closely against SC to see if that was an aberration.
I’d like to believe that, but I think 2015 Devin Cajuste has scarred me for life.
Fair enough. Keep in mind though that without the DC-who-shall-not-be-named, Devin Cajuste would be about as memorable as Tavita Pritchard.
I will acknowledge, however, that as Friend of 18S Jamie Uyeyama said recently, it’s reasonable to think that Stanford will find a way to win against us until they don’t. We’ve been the better team on paper each of the last three years and lost all three games.
Everyone is talking about statistical analysis and I’m sitting here thinking, the end zone stripes are going the wrong way.
I figured you were obsessing over the shoes.
That’s me every day.
Your thoughts, sir?
I already posted elsewhere that I kinda liked them.
I like them. Not a big fan of the heavily gold cleats, noticed guys are practicing in them more often this week.
My kids were like “ARE NOTRE DAME’S SHOES MADE OF GOLD?!?!” when they saw the bottoms.
the Murtaugh brand is strong as ever. time to break out the ole MS Paint editor….
You know it!
I’m just hoping “hand wavey” becomes part of the vernacular around here.
Brendan, Orlok, and Bandit conducting “statistical models”:
Downright amazing reference. That was good episode.
Great write up!
I saw elsewhere ND had 0.4 second order wins according to S&P+ (so advanced stats say Michigan should have won?). Michael, would you know what’s driving that?
Yeah, i saw that on Football Outsiders…left me scratching my head. Had it been .5 i wouldn’t have given it a thought other than F Michigan (we’re still not swearing in comments,right?)
From what I can tell, the only time swearing in comments is accepted is if you can claim ignorance of the rule or it’s impossible to get your damn point across without doing it.
Here’s my educated guess – second order wins are going to use the five factors, trying to isolate and strip out in part the “luck” components that can swing a game but are unlikely to replicate themselves. So when calculating this S&P+ is seeing:
-Yardage = dead-even
-Efficiency – significant edge to UM (+10% in success rate)
-Scoring opportunities about even
-ND with big advantage in pts per scoring opportunity (important, but probably weighted less / seen as more random)
-Michigan with decent field position edge (Connelly’s five factors box score shows it at 6 yards of average starting field position, I’m guessing based on some garbage time possessions not included that I wouldn’t really consider garbage time in the 4th)
-ND with the turnover edge BUT considered a little lucky because pass deflections – interceptions were equal, and the one fumble was recovered by us (above the .5 “expected” to recover on 1 fumble. it happens.)
So basically even but probably between the efficiency, field position, and turnover “luck” slight edge to UM. What the stats are blind to, of course, is the sequencing of how the game played out and impacted those numbers. Similarly VT only received 0.7 second order wins over FSU.
Makes sense, thanks for the thoughts!
Late to the party, but my 2cents are with KG:
1. We beat Michigan, supposedly a good team but to the naked eye (which I trust WAY more than any version of stats) they look one dimensional, strong but not strong enough on defense, mediocre on offense.
2. Any stats/projections based on last year’s performance are garbage in/garbage out. Different years, different teams. Not unlike stock market warnings “past performance is not a predictor of future results.”
3. I’m not a great believer in stats being a yardstick for prediction any better than our eyeball/sports instincts are. We all “know” ND will beat Ball State, and Bama will beat Sanford, yet things happen on the field that often defy our gut or the stats. If anything, probably stats are better suited to decoding why a team won or lost, although often it’s just “ turnovers” etc.
Just as a philosophical point, IMO Rockne, Leahy, Bryant, Ara, Lou knew how to win without advanced stats. No sure when or if Saban, Meyer, Swinney started using them, but I’d bet they go more by their deep knowledge of mechanics and what it takes to win than anything else.
None of this is meant to disrespect statistician or forecasting professionals— whole industries rely on them.
Are you serious? This is basically every old school, pre Moneyball argument rehashed. And its all awful logic.
Also I’ll point out it’s not really what I’m saying. I really like advanced stats, as long as it’s not ESPN 🙂 We’re just at a stage in the season (week 1) where we don’t have enough data, either statistical or “eye-ball,” to really know much yet. IF it is the case that Michigan really isn’t that good, then in a few weeks both our eyes and the stats will agree on that. Basically, Bill Connelly himself says it takes a few weeks for S&P+ to be worth much, and I’m agreeing.
I strongly approve of your lack of respect for ESPN. They are such garbage.
The one piece that’s intriguing to me is that their system (FPI) is using expected points added (EPA), which basically looks at given your performance on a down and distance who is mostly likely to score next (and how many points). It uses a lot of historical data, and like all CFB data has to be opponent adjusted to be optimized, but I like that it provides a different perspective than FEI or S&P+.
That said, the way that ESPN doesn’t let anyone behind the curtain on it – at all – and sort of “hand waves” away criticism, to steal from this comments section, is beyond annoying, and why I’m cool with anyone who just doesn’t want to deal with them. It’s like an entrepreneur trying to sell you on some “proprietary” technology that refuses to show you the patent or any detail behind it but wants you to invest.
Rockne would’ve loathed advanced stats but thought it a good idea to skip a game as head coach to go recruiting.
At least he didn’t coach a team across the country via telegram while being the head coach at another program like Pop Warner.
He did that?
1926 Carnegie Tech. Only loss on the season.
Warner accepted the job at Iowa State in 1895, but had other job applications out and also got hired at Georgia. Iowa State started their season before UGA, so he coached Iowa State until it was time for him to start UGA’s season. And, well, it’s easier to just copy Wikipedia:
“Ultimately, not only did Warner end up coaching at Iowa State before his time at Georgia; but while in Athens, he also received weekly updates from Iowa and sent back telegraphs with detailed advice for the following week. One story recounts that in the middle of September (just before leaving for Georgia), Warner took his team north west for a previously agreed-upon game against the Butte Athletic Club of Butte, Montana. Apparently overconfident, Warner bet the entire sum of his Iowa State wages — $150 (Approximately $4,000 in 2016), on his team’s victory. At halftime, his team trailed 10–2 against a clearly superior team. Warner decided to enter the game, filling in at the guard position. Though this had a positive impact, it was not sufficient as his team still lost 12–10. In order to try and make up for losing all of his Iowa State wages, Warner worked out a deal where, for $30 (approximately $800 in 2016), he would stay in contact with Iowa State while at Georgia.
Soon after Warner left for Georgia, Iowa State had its first official college game of the season. In Evanston, just north of Chicago, underdog Iowa State defeated Northwestern 36–0. A Chicago Tribune headline read, “Struck by a Cyclone”. Since then, Iowa State teams have been known as the Cyclones. The team finished with three wins and three losses and, like Georgia, retained Warner for the following season. In 1896, Iowa State had eight wins and two losses. Despite leaving Georgia for Cornell in 1897, Warner remained head coach at Iowa State for another three years, posting winning records.”
So yes–for 2 seasons, he was coach in absentia at Iowa State, while also coaching Georgia. Then did it again while he was at Cornell.
Which part is awful logic? That we don’t know whether UM is really a good team after one game? Or that basing game expectations and projecting finishes on stats from last years teams, which no longer exist as such in the current year is garbage in/garbage out? Or that the stats are better at dissecting why/how rather than predicting outcomes? I doubt that I’ve covered “every old school argument”. Haven’t taken the time to get past the 3 you see.
I know you love the stats, but a good dose of common sense and football knowledge does fine, IMO.
And yes, I’m serious.
This is “Back in my day, we didn’t wear seatbelts or car seats and everyone turned out just fine” territory.
because you’re serious….
1) we don’t know if anyone is really good after one game. it’s too small a sample. anyone who looked good could just be playing an opponent worse than we think. anyone who looked bad could be playing someone better or just have had their worst performance. That’s why looking at the past – last year’s performance (and even further back), recruiting, etc. paint a more accurate picture. To the “naked eye” a few years ago it looked like Texas was back after they beat us, a top-10 team at the time.
2) Again, this is where saying last year’s stats are “garbage in / garbage out” is a gross overstatement. Of course those teams don’t exist in their present state! But if you’re tracking what is predictive of year to year performance, the recent history of the program is as good of a barometer as you have. And the advanced stats do the same things I’m sure your idea of common sense and football knowledge does – factor in returning production (i.e. who has left those teams), talent on the roster, and the track record of the program. UCF was really good last year, but we don’t expect them to replicate it like say, Oklahoma. Why? The longer track record. It’s accounted for.
3) Advanced stats can certainly both dissect why/how as well as predict outcomes. S&P+ for example, consistently hits about 70% of its picks, and between 50-54% vs the spread. This isn’t just making slam dunk top-25 picks, it’s also every single FBS game week to week. Again, there’s lots of data here on how well these projection systems have performed if you have an open mind about them being swayed (it doesn’t seem like you’ve done a lot of research on this, to be honest).
4) The whole point of advanced stats is that paired with what we see, they can explain more than either alone. Advanced stats aren’t intended to ever be the end all be all, just the start of the conversation. Our “eye test” and “common sense” are extremely biased and often wrong. For example, conventional wisdom is that you need to run effectively to set up play action passes. Well, if you look at what actually happens on the field, instead of relying on confirmation bias, analytics has shown that the success of play-action passing actually has zero relation to how often or how well you’ve run the ball in that game.
This isn’t unique to college football. Again, on the whole idea that experts and common sense and the eye test can tell you all you need, I’d encourage you to check this out. It’s fascinating that across fields, especially ones with high subjective interpretation and uncertainty (hello sports!), how often experts are wrong: https://fs.blog/2017/03/algorithms-complex-decision-making/
I bet Rockne, et al, based early season game plans on how the team performed the prior year.
Saban has like 10 full time analysts.
Yeah, I think Rockne’s budget was only like $30m, not Saban’s $50m. Spilt that $20m 10 ways and there are your analysts.
$30m in 1931 is the equivalent of $460 million today. Clearly the administration USED to prioritize football, but now just prioritizes video boards and plastic grass. Harrrrrrumph.
My job involves using advanced stats to estimate the likelihood that an applicant will end up being good on the job. Really like how you explained the value of statistics.
Interesting, mihalko.
I was the top headhunter in the world in my field, paid multiple 7 figures annually, retired 2 years ago (long after seatbelts were invented, Drick). I know the “advanced stats” you’re referring to if you mean you do this at the highest executive levels. I never used them, mostly because they weren’t well established in science, and I didn’t need them to do my job. I also noticed that my colleagues who chose to use them didn’t achieve what I did, and by a wide margin. CEO’s were the voters, with their trust and money.
I’m not disrespecting you the person, just the tools. It’s also possible that you work at a lower level and have tools that are useful. Actually resumes, records of accomplishments, references (old timey tools) and being able to sit across a table from someone and be able to suss out if they are the real deal and a good fit culturally with the client works amazingly well even in this day and age.
Regarding ND-Michigan game, my eyes told me our Dline dominated Michigan’s Oline the whole game. Thus Michigan’s offense scored little. Our Oline did just enough in the first half and we had several very important and lucky catches, especially Finke taking the ball away from the substitute UM safety who replaced the star safety ejected for targeting just before that. Great play by Finke on an underthrown ball.
We had one terrific running play, the first TD run. Otherwise the run game was mediocre.
Overall I think we have what might be a strong D. I say “might” because I think Michigan’s offense really sucks. I definitely hope it’s a 2012 level D, because I think our offense will need lots of help, caveat being that maybe the Michigan defense IS really good. Certainly that defensive end living in our backfield is outstanding, unless our tackles are weaker than we hope. Hopefully we have smooth sailing the next few games to get the offense revved up.
Thanks for engaging, Michael.
Regarding:
1. How many times do teams get highly ranked preseason based on prior year, perceived recruiting success, and returners, yet seriously underperform the projections? Lots, I think we can agree, with the exception of the super-elites, which mainly consists of Bama, lately Clemson, Ohio State. They have such an edge, like ND, FSU, USC have had at times, that they’re pretty easy to predict.
2. Something being the best we have doesn’t necessarily make it all that good. Pick a random group of 10 people and the smartest one in the room isn’t necessarily going to be all that bright, especially with median IQ’s where they are these days. Said another way, the worlds tallest midget isn’t able to play center for the Lakers.
3. I’d guess most of us who follow football closely could pick close to 70% of game winners in the teams we really follow, and don’t care enough about the rest to pay attention. Getting 50-54% of the spreads right says you could toss a coin and come out as well. And you are correct, I haven’t done the level of research you’ve done on advanced stats. They kinda bore me, I’m obliged to say.
4. I think we can agree on most of what you say here, until you get to the “conventional wisdom” part. That’s a pretty nebulous term, one person’s conventional wisdom is another person’s BS, in my experience, and the conventional wisdom followers are usually the sheep that are too lazy to think and analyze for themselves.
Look, I get that you really value these stats, I respect you, enjoy your articles, and appreciate your intellect and sincerity. I am in the habit of thinking critically about things and forming my own judgements and opinions. That’s served me extremely well in life. So we will have to agree to disagree on some things. I’m ok with that and I hope you are too. Peace😉
This is as fired up as I’ve ever seen Mike (he’s an honorary Canadian).
almost as fired up as when someone wouldn’t stop arguing that I should have included Dak Prescott on my Heisman predictor list (a flawed list that started with the assumption no non-QB’s would win, when of course one did that year)