a terrible blight on our fine country
I’ve been wondering lately about Michigan’s modern coaches (i.e., from Bo on), and how their winning percentages stack up when we consider the quality of the opponents they’ve played. Just off the top of my head, it seems to me that:
- Lloyd Carr was a much better coach against ranked opponents than we give him credit for, but lost more games than he should have to unranked teams.
- Brady Hoke thusfar has done well in winning the games he should win (i.e., vs. unranked opponents), or at least has done better than Lloyd Carr did.
- Gary Moeller and Lloyd Carr went through an absolute gauntlet of a schedule in the 1990s.
But I wanted to see if those notions are actually supported by the stats. So I started looking at Michigan’s coaching from 1969 to 2012 against ranked teams. I should start out by explaining that “ranked” here means the AP Top 20, as the AP did not rank teams 21-25 until 1989. I’ve therefore disregarded all 21-25 rankings of opponents.
Here are our heroes’ (and anti-hero's) overall M records:
*All Hoke stats are through 2012 only, as we don't have final AP rankings for 2013 yet.
Here’s how U-M’s coaches have stacked up against teams ranked at game time:
|vs. AP 1-10||vs. AP 11-20||vs. Unranked|
|Bo||16-19-1 (.458)||21-12 (.636)||159-16-4 (.899)|
|Mo||7-5-1 (.577)||5-1-1 (.786)||32-7-1 (.813)|
|Lloyd||19-9 (.679)||19-11 (.633)||84-20 (.808)|
|RR||1-5 (.167)||1-4 (.200)||13-13 (.500)|
|Brady||0-2 (.000)||1-2 (.333)||18-3 (.857)|
Some obvious points jump out:
- RichRod and Hoke, and even Moeller to some extent, have small sample sizes. Keep this in the back of your mind for all that follows.
- Lloyd had a VERY impressive record against the AP Top 10. In fact, he started out on a huge roll. From 1995-2002, Lloyd was 11-1 against the Top 10 at game time.
- Lloyd lost a substantial number of games against unranked teams. Brady’s done pretty well against the unranked. Bo really killed the teams he should have killed.
I know, I know, pre-season bias in rankings, especially early. But let’s not completely discount game-time rankings. Though some of them, particularly early in the season, are just plain wrong, some might actually be good indications of a team’s quality as of the time you played. For example: some of Glen Mason’s Minnesota teams were pretty good as of Michigan week, but then plummeted through the rest of their seasons, perhaps from psychological issues, after losing their red-letter games for the Jug. Were those teams better when Michigan played them in week 5 or 6 than those teams’ final unranked status would indicate? Very probably.
But of course, rankings at game time, particularly early in the season, don’t tell the full story about the quality of the team you’ve played. The season’s final rankings are probably most illustrative (except for Minnesota-like situations as described above):
|vs. AP Final 1-10||vs. AP Final 11-20||vs. Final Unranked|
|Bo||6-25-1 (.203)||16-8 (.667)||172-15-4 (.911)|
|Mo||2-7-1 (.250)||7-2 (.777)||35-4-2 (.878)|
|Lloyd||10-14 (.417)||14-6 (.700)||98-20 (.831)|
|RR||0-8 (.000)||0-3 (.000)||15-11 (.577)|
|Brady||0-3 (.000)||1-1 (.500)||18-3 (.857)|
Final rankings may give you a better overall picture by removing most of the pre-season bias, but with final rankings a big caveat also applies: When you beat a team, its final ranking drops. When you lose to a team, its final ranking rises. Beating a good team makes it look worse to the poll voters. So in general, the coaches should have worse winning percentages against teams in the final rankings than they'd have against teams ranked at game time.
- Bo’s REALLY kicking ass against the unranked dregs (Mo and Hoke too), but not doing so hot against teams having great seasons. It makes sense that a coach’s record against the final Top 10 should not be great, but Bo’s was pretty dismal. All of the bowl losses certainly didn’t help him.
- Lloyd’s 10-14 against the Final Top 10 is still pretty darned impressive, and his combined 24-20 against the Final Top 20 is fairly impressive as well. But again, he’s lagging behind in beating teams having unranked seasons.
- RR of course was absolutely dismal against teams finishing in the Top 20. Wisconsin’s 2008 team was not able to crack the final Top 20 to put a single mark on the board for Coach Rod.
So it looks as if my first two initial thoughts were generally right. Thusfar Lloyd was a pretty good big-game coach, taking his whole career into account (I’ll save comparisons of early career vs. late career for another piece). And Brady’s not losing to a lot of teams that he has no business losing to.
But what about the schedule strength? On average Bo faced fewer ranked teams in the days before 85-scholarship parity (in 1970 he didn’t face a single team ranked at game time or in the final rankings) and before the resumption of the Notre Dame series and the scheduling of Miami and Florida State. Also, even in some of his best seasons the Big Ten didn’t let him go to a bowl game. How many of their games on average did our coaches play against ranked teams?
Based on the game-time rankings:
|% of Games Played vs. AP 1-10||% of Games Played vs. AP 11-20||% of Games Played vs. Unranked|
Based on final rankings:
|% of Games Played vs. AP Final 1-10||% of Games Played vs. AP Final 11-20||% of Games Played vs. Final Unranked|
- It looks like Mo’s schedules were indeed murder, whether you look at the game-time rankings or the final rankings.
- Bo’s and Lloyd’s schedules got significantly easier when you look at the final rankings.
- RichRod’s schedules became brutal when you look at the number of teams he faced that finished in the Top 10. But again, if you lose games, the teams you play look better in the final standings. And of course it’s a small sample size; if RR manages to beat Utah in 2008, Penn State in 2009, Iowa in 2009, or Wisconsin in 2010, those numbers look different.
- Hoke hasn’t played a lot of high quality teams. Thanks, down Big Ten and watered-down non-con scheduling.
Given the disparity in schedule strength, let’s look at the coaches’ winning percentages as if Lloyd’s 13-year schedule is the measuring stick for schedule strength. I think that’s fair, as 13 years is a pretty good sample size, Lloyd had both some really good and some bad seasons, and Lloyd’s tenure was the time in which Michigan’s schedule entered into our current era of weaker non-conference scheduling and greater parity as the effect of the 85-scholarship limit has fully set in.
For example, we’re going to take Bo’s .458 against the Top 10 at game time and assume that he’d played as many games against the Top 10, on a percentage basis, as Lloyd did, and so on. How do our coaches’ career winning percentages stack up then?
Based on the game-time rankings:
|Actual Win %||Win % Adjusted to Lloyd's Strength of Schedule||Change|
Based on the final rankings:
|Actual Win %||Win % Adjusted to Lloyd's Strength of Schedule||Change|
- When adjusting for schedule strength, Lloyd suddenly looks pretty good. He’s only around 20 points lower than Bo’s storied/heralded/legendary career. A swing of just four games in Lloyd’s career would’ve put him above Bo. If Lloyd goes 126-36 instead of 122-40, Lloyd becomes the Michigan coaching king when percentages are adjusted for schedule strength. And that’s not too big a stretch at all. Think about it: if Michigan had gotten few breaks in the 2000 season (i.e., Hayden Epstein doesn’t miss an extra point and a 24-yard field goal against UCLA; Michigan scores more than a field goal in the second half at Purdue; the A-Train holds on to the ball at Northwestern), and if Michigan hadn't given up just one of the several fourth-quarter blown leads in 2005, Lloyd would’ve had a better career winning percentage than Bo, equalized for schedule strength. But more on this later.
- Again, Mo’s schedules were murder. His percentage rises when compared to Lloyd’s schedule strength.
- RichRod’s also up in the final rankings, again because of the many opponents he played that wound up in the final Top 10.
- Hoke doesn’t look so hot. But rebuilding and fusion cuisine and all that.
In the comparison of Lloyd to Bo above, we’ve adjusted Bo’s 1-10 percentage and 11-20 percentage to Lloyd’s frequency of playing in those games. But playing the national No. 1 is usually a much tougher game than playing the national No. 10, or even No. 3. I haven’t taken the time to adjust the winning percentages by frequency of playing every single spot in the rankings, but I have collected some info on how those guys did against the Top 2 and against the Top 5 at game time:
|Record vs. AP 1-2||Record vs. AP 1-5|
|Bo||3-6-1 (.350)||8-12-1 (.405)|
|Mo||0-3 (.000)||2-4-1 (.357)|
|Lloyd||3-3 (.500)||8-3 (.727)|
|Brady||0-1 (.000)||0-2 (.000)|
Those are some great numbers from Lloyd, but all of those games are 2003 or earlier, when he started 11-1 against the Top 10, aside from games against No. 2 Notre Dame and No. 1 Ohio State in 2006. For what it’s worth, U-M hasn’t beaten an AP No. 1 since Miami in 1984.
So how often did the coaches play in really big games (using game-time rankings)?
|% of Games Played vs. AP 1-2||% of Games Played vs. AP 1-5|
So Bo played really big games just slightly more often than Lloyd did, and Mo’s murderous schedules are apparent here as well. I’m not sure this difference between Bo and Lloyd is significant. I’m still comfortable saying that Lloyd is only a touch behind Bo in terms of impressiveness of career. Sacrilege? Perhaps. But in terms of wins, losses, and quality of opponents, I think that’s a fair statement. Now, that's not saying anything about what Lloyd could have achieved, or how maddening it was to watch some of his games. Of course, Bo sustained that level of success eight seasons longer than Lloyd did, and there's something to be said for that.
What do we get from all of this? I think the stats support the wisdom of the ideas set out above: Lloyd Carr was pretty darned good against ranked teams, but not as hot as he should’ve been otherwise; blame the 85-scholarship parity era for that. Brady Hoke has generally won the games he should win. And the 1990s schedules, particularly Gary Moeller’s, were potentially the most brutal stretch Michigan has ever faced.
And don't worry, I'll say it for myself: Cool story, bro.
[Edited to correct Brady Hoke's 2-year record to 19-7 instead of 18-7. All stats accordingly corrected.]
Lloyd Carr officially started his tenure on the NCAA Committee on Infractions today.
If you're reading this site, you're a true Michigan fan. The type that yells "Go Blue" at anyone with two legs and block M on her shirt.
Your commitment goes beyond packing the Big House on Saturdays each fall. You still remember how the sun set over the Rose Bowl in 1998 as Michigan beat Washington State to capture a national title.
Michigan football has rewarded your commitment by winning a lot of football games. The program has the most wins and best winning percentage among all college football programs. This success has been consistent, well, except for a recent 3 year stretch.
The bottom panel shows how a computer algorithm viewed Michigan football each year. The Power Rank algorithm takes a team's margin of victory in each game and adjusts it for their schedule. It makes a difference whether a team plays in the Big Ten or MAC. Last year, Michigan was rated higher (26th) than Northern Illinois (44th) despite having a worse record.
The rating for each team gives an expected margin of victory against an average FBS team. The difference in ratings of two teams gives a predicted margin of victory on a neutral field. For example, Michigan was predicted to beat Michigan State by 5.7 points (including 3 points for a home game at the Big House) last season. Michigan won 12-10. While the visual shows year end ratings, the calculations from before bowl season have predicted 62.8% of bowl game winners over the last 11 years.
These insights into Michigan football jump out from the visual.
23 Years of Sustained Excellence
In 1984, sophomore QB Jim Harbaugh got hurt in the fifth game of the season. Bo didn't have a suitable replacement. Michigan struggled to 6-6 record, finishing 36th in The Power Rank.
For the next 23 years, Michigan football never finished out of the top 25 of the rankings. The teams coached by Bo Schembechler, Gary Moeller and Lloyd Carr showed remarkable consistency.
The Rich Rodriguez Years
This consistency came to an abrupt halt when Carr retired and Rich Rodriguez took over as coach. The dip in rating over these 3 years looks like the Grand Canyon compared with Michigan's results on both sides of this era. Rodriguez's teams won more games as the offense picked up his spread scheme. However, the poor defense kept team rating negative during those 3 years.
Michigan had a terrible time with turnovers under Rodriguez. Turnover margin in football is like flipping a coin. The randomness implies that a team with poor turnover margin should do better the following season. However, regression to the mean does not rescue every team. Rodriguez's teams had a consistently terrible turnover margin, with 10, 12, and 10 more giveaways than takeaways in his 3 years.
Boring wins football games
Lloyd Carr did not play the most exciting brand of football. Run, run, pass on offense. Very predictable and boring.
But Lloyd Carr won 122 football games in his 13 years as head coach. He claimed 5 Big Ten titles and a national championship in 1997.
How did he do it? Craig Ross, author of The Obscene Diaries of a Michigan Fan, told me that Carr attempted to "minimize the vagaries of talent and injuries". He probably had turnovers on his mind as well.
To a mathematician like myself, this quote means he understood randomness and tried to minimize its impact on his team. The calculated ratings from 1995 to 2007 show the consistent results from this philosophy. Similar to the San Antonio Spurs of the NBA, Carr put his team in position to win every year. He broke through with a national championship in 1997. Who cares that the algorithm thinks the Big Ten had a down year?
And for anyone who doubts boring wins football, just remember what happened when the exciting spread offense showed up after Carr retired.
The hidden strength of 2005 team
The remarkable 23 years in the top 25 of the rankings includes 2005. Most fans will not remember the 7-5 season fondly, but Michigan finished 10th in The Power Rank.
How can a team with 5 losses get ranked so highly? The Power Rank considers margin of victory and strength of schedule in ranking teams. A team gets credit for staying close with good teams. In 2005, Michigan lost by 4 points to 3rd ranked Ohio State, 7 points to 6th ranked Notre Dame, and 3 points to 19th ranked Wisconsin.
The 2005 team was much better than their record indicated. The Power Rank rated them two touchdowns better than the average FBS team. The core of Jake Long, Chad Henne and Mario Manningham along with a healthy Mike Hart would lead Michigan to an 11-0 start the following year.
Bo's best team was in 1988
Of the last 7 years of Bo Schembechler's coaching tenure, which team was the beat? The 1985 team that beat Nebraska in the Fiesta Bowl? Or the 1986 team that finished with 11 wins?
Actually, the ranking algorithm gives a slight edge to the 1988 team. Michigan opened the season with a 2 point loss at Notre Dame. The following week, the Wolverines lost an agonizing 1 point game to Miami when the Hurricanes recovered an onside kick to set up a winning field goal. However, The Power Rank considers margin of victory and strength of schedule in rankings teams. Miami and Notre Dame would finish the season 1st and 2nd (Notre Dame won the national title with an undefeated season).
Michigan went on beat USC in the Rose Bowl and finish 4th in the rankings. The algorithm states they were a point and a half better than the 1985 team. However, the algorithm does not make any kind of definitive statement on the best team. To put this in perspective, the 1988 has a 53% chance of beating the 1985 team on a neutral field.
Get a free postcard of the Michigan visual
As Michigan enters the third year of the Brady Hoke era, the program appears to be climbing out of the Grand Canyon of the Rodriguez years. The Power Rank will continue to use analytics and visualization to break down the program in detail. For example, I apply the algorithm to yards per play to account for strength of schedule in ranking offense and defense.
The best way to keep up to date with this analysis is my free email newsletter. If you sign up, I'll send a postcard of the Michigan visual to you and the next biggest Michigan fan you know. To check it out, click here.
Lloyd Carr and Bobby Cremins were appointed to the NCAA Committee on Infractions. Carr could have a pivotal role in the penalties handed down in some of the high profile cases on the docket (Miami, Oregon, UNC, etc.).
Lloyd Carr, Bobby Cremins and former Minnesota AD Joel Maturi appointed to the NCAA Committee on Infractions. Interesting.
Hard to believe, but this team still had players recruited by Carr. Kenny Demens, Brandon Moore, and Elliot Mealer are the final holdovers. Seems like decades ago that Carr was the coach; it's hard to believe he recruited those guys.
Edit: So those 3 were the last that made their verbal commitment under Carr.
- Will Campbell commited under Carr, decommitted, then recommitted under Rodriguez.
- JT Floyd was offered by Carr and was supposedly a silent commit to Carr's Michigan.
- Marvin Robinson was offered by Carr
I'm sure Carr had a hand in at least some part of the recruiting process for a handful of other players on the team, but ultimately the players above are the last ones to commit while expecting to play under Carr.
There is a myth that lives on this board that Denard was a better passer in 2010. This post is not meant to excuse Al Borges' playcalling, or bash Rich Rod, or elevate Lloyd Carr's run-run-run-punt strategy. It's just a look at the falsity that Denard was a better passer in 2010.
The unfortunate, painful truth that this diary reveals is that our passing offense is not much better than it was in 2010, when it wasn't very good at all (when it mattered).
Let's throw out the garbage games and focus on Michigan's games against opponents that had respectable defenses in 2010:
- Ohio (3rd in total yds)
- Iowa (16th in total yds)
- Wisconsin (23rd in total yds)
- Michigan State (32nd in total yds)
You might be wondering, "Where is Notre Dame and Penn State on that list?" Well, I'm glad you asked. They were 46th and 48th...behind powerhouses like San Diego State, Hawaii, and ILLINOIS!!! (the team we scored 67 points against). So they sucked. But we still lost to Penn State. Even though they sucked. Because our defense was, well, worser.
I don't need to lay out the stats from the ohio game. They trounced us, and Denard got pulled in favor of Forcier at the end of the game. We couldn't move the ball at all, and scored only 7 points.
Let's move on to Iowa...
Their defense was ranked 16th in 2010, and yet we were able to score 28 points. This is actually the best comparable for this weekend's Notre Dame game, since ND is ranked 17th in total defense right now. Yes, we lost the game by a score of 28-38, and those four TDs sure do look good...but only because you either don't remember what happened or judge a book (or score) by it's cover (or...score). Here are some relevant stats:
- Denard 13/18, 98 yds, 1 TD, 1 INT
- Forcier 17/26, 239, 1 TD, 2 INT
But here's the most important stat: We only scored 7 points when Denard was on the field. Denard get could get yards (108 on 18 carries) but not points. Iowa was stacking the box, and all the offense could muster was a TD on a drive when Denard threw three passes: one was incomplete, one was for a 6 yd. loss, and the last was a screen to Smith for an 8 yd. TD. Denard got hurt in the 3rd quarter and in came Forcier.
It was Forcier that brought the team back in that game, and Forcier that sealed our fate with his INTs. It's worth noting that completing passes underneath when you're behind by 21 points is MUCH easier. In fact, that leads to lots of confusion about the effectiveness of Denard's passing and the 2010 offense in general: we got loads of "soft" yards because we were hopelessly behind and our opponents played softer coverages and lighter fronts.
Wisconsin has a similar storyline, except that Denard played much more that game. We scored exactly ZERO points in the first half (although we did miss a 30 yd. field goal). With a 24 point lead, Wisconsin converted to prevent defense, and allowed us back in the game. Denard stayed in this time, and racked up a nice, meaningless statline: 16/25 for 239 yds, 2 TDs, and, of course, 1 INT. The important part: Denard was 4/9 for 22 yds passing in the first half. When Wisconsin was playing their base defense, Denard couldn't pass. Only the gooey butter cake version of Wiscy's D allowed DR some meaningless passing yardage. Further proof of this came in the fourth quarter, when we had come back to make it a 21-31 game. Denard couldn't move the ball anymore.
The final example is, perhaps, the most damning. Michigan State had a good-but-not-great defense in 2010. Their success was largely a result of their schedule and some good defensive coaching. They lost badly to Iowa (and 'Bama), snuck by a pretty lousy ND team in overtime, and narrowly edged out a VERY average Penn State team. Their only quality win was against Wisconsin, and that game was played in East Lansing. Despite their easy schedule, the Spartan defense was still only ranked 32nd in total yds. Michigan actually had the lead twice in this game, up 3-0 in the first quarter and 10-7 in the second. Denard was 6/8 for 51 yds in the first quarter, but threw an INT in the endzone. In the second quarter, Denard shined again. He was 4/6 for 81 yds and a TD. At the half, Michigan was down 10-17.
The second half was a very, very different story. Denard was 7/15 for 82yds and 2 INTs. The same guy we saw against ND. Only against a defense that wasn't nearly as good. And we were at home. The 4th quarter TD was only scored after MSU had rung-up a 21 point lead.
So here's the bottom line: Denard has never been a good passer, or even an average passer. And against good defenses, we won't win until he's able to throw the ball somewhat effectively. Maybe that's why Borges keeps making him throw, especially before the B1G season starts.
So what's the difference between now and 2010? The defense. Because our Greg defense is not our GERG defense, we are in every game, and teams don't stop stacking the box against Denard. They don't stop blitzing. They don't play soft coverage. So Denard never gets to ring-up his stats, and looks even worse.
I certainly won't excuse Borges' playcalling on Saturday--it needed to be better. But the fact is that our only quality wins have come when Denard has been able to make plays in the passing game (or Hemingway was able to bail out Denard) and I expect it stay that way. If Denard can't pass, we're screwed, and 4 or 5 losses is our best case scenario.