...says Denzel Valentine of Big Ten Tourney favorite MSU, which is 5-7 in its last 12 games. Cumong, man.
Even with all the turnovers, Michigan is ranked #25 in scoring offense and #38 in scoring defense. These rankings give U-M better than an 80% probability of finishing the year 9-3 or better. If Michigan can fix the turnover problems, this year should be exceptional (10-2 or better). If not, this year could be a repeat of last year (8-4) ugh. [Teams with a – 0.5 TOM/Game or worse have just an 11% probability of finishing the year with and 8-4 record or better.]
We Wants It . . . We Needs It . . . . Must Have The Precious!: Each football game in a season is precious. Almost every year, the outcome of a single game determines which teams play for conference championships and which teams play for the MNC. The difference between success and failure is razor thin and even one performance based turnover in a season can be what controls your fate. Every coach knows this, every player knows this, and every fan knows this. That is why turnover analysis is so important.
Discussions about whether turnovers are random miss the point entirely. IMHO no one actually believes that ALL turnovers are random and no one actually believes that ALL turnovers are controllable. In fact, everyone agrees that turnovers are a combination of luck (i.e. random and not controllable) and performance (i.e. not random and controllable). But, if even half of all turnovers are controllable (and far more than half are), their impact is often the difference between an excellent season (M would have been 10-2 without the TOs in the ND and ohio games last year) and a mediocre one (M was 8-4 last year). Simply shrugging your shoulders and saying, "Oh well, I know the reason we lost the ND and ohio games last year was because of turnovers but we were just unlucky. We should have better luck this year" seems quite insane to me.
A Brief History: I started this diary during the end of the 2009 season and just looked at total turnovers per year. However, it was obvious that any rational turnover analysis must be based on a game by game examination rather than just total TOs per year. For 2010, the diary analyzed TOs for each game but used an average value of 4 points per turnover to determine impacts. This did not recognize that the situation that exists before each turnover is critical to the importance of and the value determined for the turnover. Therefore, in 2011, the analysis evolved to use Expected Points (EP) to determine impacts. Using EP, each turnover may have a value from 0.0 to 11.6 "points" depending on the situational analysis.
A Decade of Michigan Quarterbacks: The decade started with 5 straight years of positive turnover margins and national rankings of #47 or better. This corresponded with quarterbacks throwing few interceptions except in 2007 when injured Henne and Mallet threw a bunch but were bailed out by takeaways that were outstanding (ranked #21). Then, the era of "he who shall not be mentioned" brought 3 straight years of just horrible giveaways (both interceptions and fumbles), takeaways, and TOM. Since Brady Hoke took over, quarterback interceptions have not improved. The only reason 2011 was not a negative TOM year was because the defense was terrific and ranked #27 for takeaways. Quarterback interceptions are the key to controlling turnovers.
New Turnover Data: In 2012, the folks at Football Outsiders – FEI did a weekly "Revisionist Box Scores" that striped out TOs, Special Teams, and Field Position. FEI calculates the value generated by each drive and then lost on the drive up until the turnover, as if the drive had concluded at that spot on the field. They looked at all 696 FBS games. The FEI results corroborated the results that I had calculated in these diaries (yea!). A more detailed review unleashed some rather unexpected results. As posted last year, TOs are the determining factor in the outcome of just 16% of games. But, for a 12 game regular season, this results in the average team having 2 games determined by TOs. In 2012, there were 107 FBS teams that had at least one game determined by TOs – that is 86% of the 124 FBS teams. While 27% of all FBS teams had just a single game decided by TOs, 59% of all FBS teams had 2 or more games decided by TOs, and one team (South Florida) had an unbelievable total of 7 games decided by TOs!
Obviously, if a team has more than one game decided by TOs, they will have a net of multiple wins, multiple losses, or a combination of wins and losses. South Florida fumbled their total of 7 games decided by TOs into a net of 3 losses. The biggest loser was Marshall – managing to turn all 4 games decided by TOs into losses. The other teams who lost 3 games due to TOs are Southern Mississippi, South Alabama, and Maryland. The biggest winner was Ohio University turning all 4 games decided by TOs into wins. Louisville, Northwestern, and Middle Tennessee were the three teams that won 3 games due to turnovers. Without turnovers, Louisville would have had a record of 7-5 instead of 10-2 and Northwestern would have ended up 6-6 instead of 9-3.
As you can see from this chart, teams that end the year with a negative TOM are unlikely to have a winning season. Teams with a + 0.5 TOM/Game or better have a 70% chance of having a WLM of + 4 or better (WLM = Win/Loss Margin = Wins – Losses). For a 12 game season, 0.5 TOM/Game is equal to a total TOM of 6. In 2012, Michigan defied the odds by posting an 8-4 record (+4 WLM) with a – 0.7 TOM/game.
Last year I concluded that Michigan had a total of 3 games determined by TOs although my actual EP calculations only resulted in 2 games determined by TOs. The exception was the MSU game. The calculations resulted in a value of just 1.5 EP for the Kovacs interception with 2:03 left in the third quarter because it was third down and MSU would have punted to Michigan on the next play anyway. Since M won by 2 points, the exact numbers do not indicate that TOs were the determining factor. This is where EP (as well as any purely mathematical analysis) falls short. Being at the game, there is no doubt in my mind that the TO was a major factor in the win.
For the folks at FEI, Michigan lost a net of 2 games due to TOs. For me, we lost a net of just 1 game due to TOs. Using the FEI data, M was one of just 14 teams that lost a net of 2 or more games due to TOs last year.
Final Thoughts: Please decide which of the following four quotes from this week are true and which are made up.
A) Brady Hoke: "Turnovers? I am surprised that anyone would even ask such a question. Everyone knows that turnovers are random events and no amount of coaching or player preparation can have any affect on the number of giveaways or takeaways."
B) Brady Hoke: "Obviously, you can't give the ball away. Right now we've got a major league problem where we've got to fix it. Because that's not going to win any championships."
C) MGoBrian: "I am surprised that so many people are concerned about Devin Gardner's turnovers. Yes, his 6 turnovers in four games last year and his 10 turnovers in four games this year are the worst in the FBS, but they are random events that neither he nor anyone else has any control over. Probability tells us it will simply fix itself."
D) MGoBrian: "Gardner hardly ever puts the ball away. That's a QB sneak on which Gardner has one hand on the ball, and the fumble Gods strike." and "Gardner didn't set his feet because he was getting some pressure up the middle, and while that is suboptimal he had some room to work with that he did not use."
CALM BEFORE THE CONFERENCE SCHEDULE
NOTE: The format will alter slightly next week – conference averages will become the focus of the remaining entries in this series with cumulative averages and the OOC / Conference breakdown probably being the format of the entry which comes immediately after the regular season ends in December.
This past weekend was a relatively light one for the Big Ten, with only six teams active, four of them playing against a fellow conference opponent (i.e., two conference games). Altogether, it makes the record of the conference 1-1 in the non-conference games, with Iowa and Ohio State being the victors in the two in-conference games.
SCORING OFFENSE AND DEFENSE:
There hasn’t been much of a shift on, well, most metrics, but we’ll go over these briefly all the same. Ohio State still tends to be the most prolific scoring team in the Big Ten, with Indiana and the schedule of teams that play even less defense than even they do trailing by only a few points. Once again, the two stingiest teams are Michigan State and Penn State, with Purdue being the stingiest with itself, as you can see on the differential graph.
TOTAL OFFENSE AND DEFENSE:
Once again, not much movement here due to the light week – nine teams still average over 400 yards per game on offense, with Michigan being the ninth of those nine. The more notable division here is in yards allowed, where Michigan sits in the top six with some fairly respectable numbers then there is Minnesota followed by air in a lot of cases. On the yards per play differentials, you can see that it is basically still Wisconsin, some guys and then Purdue.
RUSHING OFFENSE AND DEFENSE:
There are not really any shocking twists in the Big Ten rushing game right now – Wisconsin, Ohio State and Nebraska run the ball quite a bit normally and do it very well, so the top quartile right now is expected. Michigan is ninth, but with a number that actually isn’t terrible if it is a seasonal average. When it comes to stopping the run, Purdue is bad at it, but Indiana is worse and the Boilermakers must at least enjoy that.
PASSING OFFENSE AND DEFENSE:
The passing averages are another area where it is one team – Indiana in this case – and then a large chunk of the conference within earshot of one another, all followed by an underperformer – Minnesota here. As for defending the pass, Nebraska and Northwestern are the worst at it so far, and as you might guess, Michigan State is the most successful at it to date.
When we’re talking about getting third downs, there are four teams in the Big Ten that are converting half of theirs or better right now – Illinois, Iowa, Northwestern and Indiana, with Michigan at a reasonable 48.1%. On the flip side of the ball, only two teams are allow more than 40% of third downs to be converted, and those would be Northwestern and Purdue. On the differential graph, you can see the ongoing progress of Boilerquest as well.
SPECIAL TEAMS ITEMS:
with this post, THE KNOWLEDGE officially announces the 2013 edition of THE KNOWLEDGE CHALLENGE
THE KNOWLEDGE specifically delayed the introduction of THE CHALLENGE for this year to give the participants a better idea of the capabilities of various teams before entering the competition
as long term followers of THE KNOWLEDGE of these very pages know, there are several important awards associated with THE CHALLENGE
each week, THE CHALLENGE will be to correctly predict the score of the Michigan game (THE KNOWLEDGE will provide helpful pointers) and become the Protege of the Week (POTW)
at the end of the year, the most consistent performer will be awarded the top honor of this blog - TOP FRIEND OF THE KNOWLEDGE
the second most prestigious award is the CO-SOARER OF THE KNOWLEDGE - awarded to the person who correctly answers the question below:
which two teams will play for the National Championship and which team will win it?
to help seekers of the CO-SOARER title, let it be noted that the winner will be from among the following teams: Oregon, Georgia, Alabama, Stanford, Oklahoma, Fresno State
now, THE KNOWLEDGE shall reminisce a bit about the season review THE KNOWLEDGE provided before the season
THE KNOWLEDGE revealed that Michigan will end the season 13-1 and will go 4-0 in the OOC games
however, when Akron came close to winning in the final minute, many doubted THE KNOWLEDGE
when U Conn took a 21-7 lead, some continued to doubt THE KNOWLEDGE
these people have been left in the dust
as THE KNOWLEDGE has soared and basked in glory
correctly predict the score of the Minnesote game to be the POTW
pointers to the game:
- Michigan will win
- Minnesota will lose
- Gardner will play better than the last 2 games
- Minnesota will surprisingly play well
I have what I think is a super-awesome-ncaa-fixing idea which I have mentioned to a few friends in conversation and most seem to think it is a pretty good one. I'd be curious if y'all have any thoughts on it:
The obvious problem is that many NCAA athletes contribute way more to a school than they are compensated by way of an athletic scholarship. Of course, any NCAA commercial will tell you that the value of the student athlete's education is beyond measure. Meanwhile, for every Denard Robinson that seems to squeeze every ounce of value out of his college experience, there are 12 [insert one and done from Kentucky here]'s who have no interest in what a college education has to offer.
A degree from USC, or even a year of free education from USC, had no value at all to OJ Mayo, I'm sure. At Michigan, I remember knowing of several classes which were specifically known to be 'football classes,' (at risk of pissing someone off from the Ojibwe department, I won't mention that Ojibwa was definitely one such class). So here's the thing: let's not force athletes to rack up 120 credits in Ojibwe to complete their degree. It's a joke, and it does no one any good. Another issue with the degree is some kids come from such worthless high school backgrounds, that they are completely unprepared for college level courses, so they don't get as much out of the free education as they should.
Let's make the Michigan football experience what it actually is: a serious education in multi-million dollar industry which has just as many career opportunities as linguistics, history, medicine, or engineering (ok, maybe not engineering/medicine). Make athletics a major. Film study and off-season workouts? Make them classes. Are they not learning how to be players or coaches or fitness experts or nutritionists? Is there any less opportunity in these fields than there are in traditional college majors? Also, it's a cool way for coaches to enforce attendance rules on what used to be 'optional' workouts. If the kid doesn't do summer workouts, they fail the class, and then their grades are not good enough to participate in the sport.
All sorts of majors have to satisfy basic requirements, so I am not suggesting they take no English or history or math. I am only asking that they be given course credit for the 40 hours a week they put into mastering their craft, just like a music performance major might. And for those kids I mentioned previously who come to college unprepared, let's allow the 'school of football/baseball/whatever' to offer some *truly* remedial courses. I am *not* suggesting watering down the degree. I am suggesting that we make players receive fantastic, personalized education that meets their needs. Some athletes are crazy smart and have a strong high school background. I am not suggesting that they have to get 120 credits of remedial reading, I can think of all sorts of cool/advanced courses. How sweet would it be to get to teach a game theory course, coaching 423-Expected Value and the Punt?
Also, since this 'school of athletics' (or whatever better name someone comes up with), is a bit of a special case, I would say that athletes should be allowed to dual enroll in another school if they choose. So, speaking of the crazy smart athletes above, (like Jordan Morgan and Devin Gardner) let's still let them enroll in social work, or engineering if they choose. Honestly, Jordan Morgan has been working his ass off for 4.5 years at basketball and school, he totally deserves to have 2 degrees. Or a volleyball player or a swimmer might wisely choose to dual enroll in athletics and education, for example, since she knows her field has a few less opportunities than football or basketball. But still, she is learning a lot of the same fitness/nutrition/competition/management skills the football players are, and she should receive a degree that reflects that.
I think it would be really cool if a few schools pioneered an idea like this. "Come to Michigan, the first University to ever have a school of football. Lloyd Carr teaches handling the media, and Mike Barwis teaches how to get paralyzed people to walk again."
Obviously, this does not address every issue with the amount of money that there is in NCAA sports. But at the same time, I feel most pay-for-play options being considered have a lot more drawbacks than my idea. Instead of rehash them all, I will simply say that to me the most compelling anti pay for play argument is in a quick comparison of attendance at college football games vs. attendance at arena football league games (or whatever your minor league system of choice is). People love cheering for these kids that lived in the same dorms, went to the same classes, dealt with the same ridiculous weather and long walks, etc. I love Michigan football because I had the best time of my life there. As soon as athletes are legit superstar millionaires walking around campus, those kids have *nothing* in common with me, and my love/association with Michigan football will definitely be diminished. There is a reason college sports are the only ones who approach the professional leagues in terms of popularity, let's try not to mess with that.
Sorry, this got really long, but I would love some feedback on why this idea won't work, as I feel it's pretty unique and the best way to deal with the problem that I have heard.
After the past two games, much discussion has centered around a rapid transition from guarded optimism to total panic in regards to our offense. The relative merits of our interior lineman, in particular, have been debated widely in platitudes as well as UFR minutiae. While Miller is facing a bit of a talent and size deficiency, we all return to the inexperience of these (and other young) Wolverines as a large factor in our offensive struggles. Though not speaking exclusively on the OL, ST3 hammers the matter of "youth" home in his most recent Inside the Box Score.
It’s widely accepted that an experienced line correlates with a successful offense. I didn’t expect to have to dig deep into an MGoSearch to find some statistical evidence accompanied by glorious charts, but the hunt turned up empty other than a 2009 Unverified Voracity linking to a WSJ article confirming the strong correlation. This particular evaluation used combined OL starts as a metric, determining that “offensive-line experience is one of the telltale predictors of success in college football.” I sought out to see how this correlation might look for Michigan and its immediate cohorts: the Big Ten member teams, Notre Dame, and next year’s new kids Rutgers/Maryland. I’ve dubbed this the B1G+.
So how would a lurking, stat-friendly but non-mathletic blog poster make some evaluations? Without data on career starts, I used eligibility year (per rivals depth charts as of 9/26/13) as a metric for experience of an offensive line. True freshman are a 1, redshirt seniors a 5. Herein lies an obvious limitation: "age” and “experience” can be quite different in matters of football.
Given that I’m interested in the effect of a young OL, my metric for success was an offense's yards per play; see Ron Utah's recent diary for another breakdown of how our offense stacks up based on yards per play. The WSJ study used AP poll result to measure success; see LSAClassOf2000 question the legitimacy of this measure. I included data on team RPI to give some sense of overall team strength.
Scientists: I got a B.A. in Psych from LSA and something called “arts and ideas” from the Residential College, so forgive me for my sins. If I understand your process, I'm testing the hypothesis that offensive lines with a greater average age will produce more yards per play. If I understand your caveats, it’s unlikely that my data set is a large enough sample to draw significant conclusions. But I've got a nifty heat map:
(Green = 1+ standard deviation above average, Orange = within 1 standard deviation either way, Red = 1+ standard deviation below average)
The hypothesis would suggest we see a lot more green on the top half of the map (other than SOS, which is mostly for reference). Of teams with older than average B1G+ OLs, Ohio State fits the hypothesis best with Wisconsin a close second. To be fair, 3.8 years into eligibility per OL in Madison is probably closer to 4+ anywhere else. What do they put in the cheese up there?
MSU and Purdue are extreme outliers against expectation. Michigan St. may be explained by the effect of the "age does not equal experience" limitation. If I recall correctly, they have shuffled guys to the line from other positions out of necessity. Purdue... I don't know anything about the makeup of that line, but to be fair their SOS is tops in the B1G+. Note that Michigan is the epitome of an average team across the entire row, including SOS.
On the bottom half of the map, there are several overperforming young offensive lines. Maryland is cranking out more yards per play than anyone but Wisconsin despite having the youngest OL in the sample. Indiana is having no problem moving the ball against a schedule more difficult than MIchigan. Same for Illinois, though the rush numbers are right on the fringe of going "red," leaving them an average overall offense. Notre Dame's rushing attack is a minor anomaly. How about a scatter plot?
At a glance, the hypothesis is bogus through four weeks of B1G+ action. That's clearly a negative correlation, both across and within quadrants. On the other hand, the trend line looks about right if you throw out Purdue, MSU and Maryland. Michigan is to the B1G+ as David is to man, but Minnesota will be out to prove they are the more perfectly mediocre offense from the most perfectly mediocre conference next weekend.
The tone of the blog after UConn has shifted towards acceptance of our averageness rather than extreme panic or outdated optimism. If nothing else, these cute visuals may lend credence to that MGoStageOfGrieving. Sure, we're not that "young," but we're not that bad either, independently or relative to age/competition.
"I'm unhappy because we sucked." - Al Borges did not say this, but was thinking it.
As we continue our transition to "MANBALL," I was curious to see, statistically, how that transition is going. The questions I'm trying to answer are: "What is this team good at? What are they bad at? What is the logic behind the play-calling? Are we ready to be a MANBALL team?"
What follows is a chart (based on Brian's UFR) of all the formations used against UConn, the type of plays that were run, and the averages. It's a big chart. It's also copied from my post in the UFR thread, as are most of my comments below it. A few notes:
- Plays that had a pre-snap penalty or penalty other than pass interference are not counted.
- Pass interference is counted, since it is assumed the play was successful enough to draw a penalty
- Sacks are rightfully categorized as passing yards
- Yes, I'm aware that this analysis has limited variables and misses important data points. If you want to add something, please do.
|Ace H twins||1||0||0.0||1||0||0.0|
|Ace twin TE||2||17||8.5||1||0||0.0||3||17||5.7|
|Ace twins stack||1||0||0.0||1||0||0.0|
|Ace twins twin TE||2||16||8.0||1||-16||-16.0||3||0||0.0|
|I-Form twins stack||1||2||2.0||1||2||2.0|
|Pistol FB twins||1||-1||-1.0||1||-1||-1.0|
|Shotgun 2TE twins||1||9||9.0||1||9||9.0|
|Shotgun 3-wide jet||2||14||7.0||2||14||7.0|
|Shotgun 4-wide tight||2||14||7.0||2||14||7.0|
|Shotgun double stacks||2||20||10.0||2||20||10.0|
|Shotgun empty TE||1||6||6.0||1||6||6.0|
|Shotgun trips TE||5||27||5.4||5||27||5.4|
|Shotgun twin TE||1||0||0.0||1||0||0.0|
While it doesn't take into account some easy missed plays and some heroic efforts to make something out of nothing, the chart does show that we seem to be much more successful when we're not under center. We ran 35 of our 68 plays from the pistol or shotgun, and the shotgun was our best bet.
I agree with Brian's conclusions that this team benefits greatly from being in the gun. I'd love to see more MANBALL out of the Pistol, but the under center stuff didn't work for most of the game.
That said, the Ace formation gave us critical rushing yards during our comeback. I believe it was effective because UConn feared we might actually pass when we were behind in the 4th quarter. When they know we're going to run, the under center stuff just doesn't work.
For those of you calling for more simplicity--you have a point. We used 26 different formations for 68 plays.
Some interesting data points:
- We are really efficient in the goal line set. That's because DG is running, and he's good at it.
- The Ace set worked fine for running (mostly late), but the passing ruined it. Some of that is on DG, so this set might improve.
- The I-Form was generally bad, and the Big set was terrible. A big play on a PA pass was missed by DG though, so it's not quite as bad as it looks.
- Shotgun was our most common set with 31 plays.
- Not much Pistol at all, and from the plays we did run, it doesn't look like we're practicing this much.