The Evolution of Jim Harbaugh’s Running Game by Adjusted Line Yards

January 3rd, 2011. The Orange Bowl: Miami, Florida; #4 Stanford vs. #13 Virginia Tech. If you didn’t know what Jim Harbaugh had built in Palo Alto, you were about to find out.

The 2010 Cardinal had a litany of future NFL stars like Andrew Luck and Richard Sherman, but to me, what stands out watching the game today is the offensive line play. Anchored by two future NFL starters in Jonathan Martin and David DeCastro, Stanford mauled the vaunted Virginia Tech defense up front to the tune of 247 rushing yards at eight yards per carry. The game was an absolute masterclass in power football; a throwback to a different time, before the era of the dual-threat quarterback.

When the game zigged, Harbaugh zagged; doubling down on size, power, and physicality in a college football world increasingly moving towards the up-tempo spread. Perhaps it’s in his DNA, having been coached by Bo Schembechler at Michigan and Mike Ditka in Chicago, or perhaps he saw something the masses had missed.

Regardless of where it came from, Harbaugh established himself as one of the elites that night. He would go on to carry the downtrodden San Francisco 49ers to three consecutive NFC title game appearances and to the precipice of a Super Bowl win.

Things looked very different that evening in Harbaugh’s hometown of Ann Arbor. The morning after the Cardinal trounced Virginia Tech, Rich Rodriguez was dismissed as the head coach of the University of Michigan. A tenure that began with a ham-handed coaching search, endured a nonsensical witch hunt by the Detroit Free Press, and ultimately saw its defensive ineptitude outweigh the euphoric highs of Denard Robinson’s playmaking was finally over.

As far as the Michigan brass were concerned, the spread experiment had run its course. It was time to go back to the old way: pro-style, Schembechlerian football.

Despite glaring deficiencies in his resume, the athletic department quickly announced the hiring of Brady Hoke. Hoke was 47-50 as a head coach at that point and had never been a coordinator at any level. I’ll save Michigan fans the experience of reliving that era and fast forward to December 2014.

The return of Jim Harbaugh to Michigan

By now, most of us know the story of how Jim Harbaugh came back to Michigan.

Against all odds and the mockery of national pundits who were dead certain that Harbaugh would never – under any circumstances – return to the college ranks, Harbaugh turned down contracts from pro franchises that would have made him the richest man in football to restore his alma mater to greatness.

The program that Harbaugh led to victories over Ohio State as a young quarterback was now his. And after two seasons, the restoration is ahead of schedule.

In 2015, Harbaugh elevated a roster that had gone 5-7 the season prior to 10-3, including a 41-7 throttling of Florida in the Citrus Bowl. In 2016, with the addition of Don Brown, the excellent defensive coordinator who had built elite units at Boston College with primarily two and three star recruits, the team went 10-3 again, though this time around Michigan never slipped out of the top-ten in human polls and never out of the top-five in most computer rankings. Their three losses were by a combined two regulation points (including a double-overtime loss at #2 Ohio State that saw the Buckeyes benefit from dubious officiating).

A stolen game aside, Michigan played at an extremely high level in 2016 and will only continue to get better as the roster becomes further stocked with players hand selected by Harbaugh. The one missing piece from last season, however, has proven to be the one simultaneously most desired and most elusive to the program: a devastating run game.

Supporters of Michigan watched Harbaugh’s Stanford teams impose their will against opposing front sevens and couldn’t help but fantasize about the Maize and Blue once again doing the same. By the end of his time in Palo Alto, Harbaugh’s Cardinal offenses would first confuse helpless defenses with a dizzying array of pre-snap motions and personnel packages, then mashed them physically, wearing down exhausted linemen and linebackers with a merciless rushing attack. “Win With Character, Win With Cruelty,” signs inside of the Stanford Football facilities read.

Michigan isn’t nearly to the level of offensive proficiency that Stanford had achieved during the Harbaugh & Andrew Luck era, but it takes time. How much time is the question.

Is Michigan Stanford yet?

Harbaugh is second to none in the arena of quarterback development. Look at his accomplishments:

  • Built a future number one draft pick in Luck at Stanford
  • Rescued Alex Smith’s career from the scrapheap in San Francisco
  • Saw what nobody else did in Colin Kaepernick en route to a Super Bowl
  • Turned Iowa castaway Jake Rudock into a second-day draft pick

We can sleep soundly about the long-term health of the Wolverines’ passing attack.

But correcting the running game is a more difficult endeavor, as it isn’t centered around one man’s physical ability and decision making like the passing game, but requires five very large gentlemen and one smaller (and much faster) one working together in perfect harmony.

If one lineman misses an assignment and blocks the wrong man, or worse, blocks no one at all, it doesn’t matter if the other four did their jobs flawlessly – the play fails. And even if all five linemen are perfect on an individual play, the running back has to play his part as well – that means hitting the right opening created by the line at the right time.

So it makes sense that Michigan’s current rushing attack is lagging behind where Stanford was at by 2010, but by how much? And how do you separate the performance of the offensive line from the running back’s? If a back takes a handoff from his own ten-yard line and dances through the defense all the way to the goal line, the offensive line assuredly did its job. But (to the chagrin of the big men up front) they should not get credit for what’s happening after the back is propelled past the linebackers and into the secondary, where all that separates him from paydirt is similarly-sized defensive backs and lots of green grass.

One way to do study the impact of an offensive line is by examining a stat created by Football Outsiders called Line Yardage. A partial description of the statistic is as follows:

Based on regression analysis, the Line Yardage formula takes all running back carries and assigns responsibility to the offensive line based on the following percentages:

  • Losses: 120% value
  • 0-4 Yards: 100% value
  • 5-10 Yards: 50% value
  • 11+ yards: 0% value

For example, a loss of two yards equates to (-2) * (1.2) = -2.4 line yards. A gain of six yards nets the offensive line 4 yards for the first 4 yards, then half of the next two yards, which totals 5 offensive line yards.

I have charted the yardage of every run play of the Harbaugh era at Stanford as well as his first two seasons at Michigan and fit this to Football Outsiders’ Line Yardage formula. The following chart shows each of his six offensive lines at Stanford and Michigan.

(It should be noted that the adjustment Football Outsiders does with these numbers differs from my own, as we will only be comparing one program’s trajectory to a historical precedent instead of trying to rank every offensive line in the country based on their run-blocking ability. The numbers I am presenting are unweighted, as there were no substantial differences in the strength of schedule between any of the six years examined.)

A few caveats here:

  • Sacks were not considered, as a play that results in a sack was not a designed run play.
  • Run plays that were not the direct result of conventional offensive line blocking, such as jet sweeps or quarterback scrambles, are included, as I had to look through the play by play to chart this which don’t go into enough detail to parse these types of plays out.
  • Perhaps the most important caveat: the actual line yardage number for each season is meaningless outside of comparing it with other seasons. An individual line yardage number itself doesn’t directly translate to what you see happening on the field, so it is not advisable to make any literal interpretations from them.

As we can see, Stanford’s offensive lines (at least in terms of run blocking) followed a smooth, upward trajectory from mediocre to elite while Michigan’s first Harbaugh-led offensive line was above national average yet stayed at about the same level between years one and two.

Based on the Football Outsiders’ adjusted line yardage for 2014 (the last year of the Hoke era) vs. 2015 and 2016 (and these numbers are adjusted by F.O. based on game situation and strength of opponent), the Michigan offensive line performed at right around the same level, at least in terms of run blocking, in each of the three seasons. While a bit disappointing, this lends a lot of credence to the reasonably-common theory passed around Michigan circles that the development of ’16 seniors Kyle Kalis, Erik Magnusson, and Ben Braden was stunted by the previous regime. Michigan had a program-record eleven players selected in the 2017 NFL Draft, yet none of the draft-eligible trio of seniors were taken by professional franchises.

Though Michigan’s offensive line will lack experience this season, it is still anchored by all-conference player Mason Cole. He’ll be accompanied by younger linemen (unlike in 2015 and 2016), but likely ones with greater athletic ability and, perhaps more importantly, ones that are a clean slate for Harbaugh and his staff to mold into the type of bulldozers he had at Stanford.

Next, let’s look at another couple charts; both of which are promising for Michigan supporters. First, the adjusted offensive rushing efficiencies for the four Harbaugh years at Stanford and his first two at Michigan, both by The Power Rank’s yards per carry adjusted for strength of schedule (TPR Rushing) and Bill Connelly’s S&P. The last column shows total offense by S&P.

team The Power Rank S&P Rushing S&P Offense
2007 Stanford 48th 37th 83rd
2008 Stanford 4th 6th 31st
2009 Stanford 10th 12th 6th
2010 Stanford 29th 30th 3rd
2015 Michigan 79th 61st 38th
2016 Michigan 65th 49th 40th

You’ll notice that even as the Stanford offensive line was still a work in progress, the run game took a massive step forward in 2008. This goes to show the impact that a back like Toby Gerhart (2009 Heisman finalist and first-team All American) can have on an offense. Remember that line yardage doesn’t give the offensive line total credit for big, explosive plays like the long runs Gerhart ripped off with regularity at Stanford during the 2008 and 2009 seasons.

Fortunately, Michigan has a young back in Chris Evans that, while being much different than Gerhart stylistically, has the potential to be similarly explosive. As a freshman, the Indianapolis product ran for 7.0 yards per carry on 88 touches; he’ll have even more of an impact as a sophomore taking the lion’s share of the carries.

You’ll also notice that the Stanford run game dropped off from excellent to merely good between 2009 and 2010, and yet, the offense as a whole continued its meteoric rise. Of course, Stanford had a two-time first-team All American and future NFL Pro Bowler at quarterback in Andrew Luck during Harbaugh’s final season in Palo Alto.

Given Harbaugh’s track record, I wouldn’t bet against Harbaugh having a quarterback playing at a similar level in Ann Arbor before too long. And once that young man is accompanied by an offensive line like Stanford’s in 2010, the sky is the limit for Michigan.

You can follow Tony Kaminski on Twitter and read his previous work at Big House Analytics.

202 words that perfectly summarize ESPN in 2017

Right around the time the ink dried on a $15.2 billion deal to broadcast the NFL, subscribers began fleeing cable television in droves — not because of anything the Worldwide Leader did wrong, but because of secular changes in the way broadcast and video works. Phones, Twitter, and YouTube began instantaneously delivering highlights and entire games to fans, obviating the need for anyone to watch SportsCenter, or any other news shows, to catch up on what happened in sports, or even, in some cases, to watch live games. Terrestrial ad revenue never migrated online, and the revenue to be found there was largely eaten up by Facebook and Google, leaving little to pay those new reporters.

ESPN is still wildly profitable — the operating income of Disney’s media networks (of which ESPN plays the largest role) was $1.36 billion in the 2016 fourth quarter — but it’s less profitable than it used to be, and projects to be far less so in the future. With its latest cuts, ESPN isn’t just trying to stanch the bleeding and/or to be seen by investors as attempting to do so: They’re also laying out what the network will look like over the next five years and beyond.

Deadspin’s Kevin Draper wrote these words after ESPN laid off 100 employees, and they capture the essence of the problems at the Worldwide Leader in Sports.

While I’ve been following the ESPN story since the layoffs, I got re-interested again when I saw Jason Whitlock’s editorial in the Wall Street Journal. He suggested these layoffs represent a victory of Deadspin over ESPN, part of the “politics hurt ESPN” narrative.

As much as I like Deadspin, both as a reader and contributor, the editorial is awful. There’s no data to back up that ESPN’s left leaning politics has hurt the network.

The truth of ESPN is in those 202 words of Draper.

Podcast: Super Bowl Preview of New England vs Atlanta

On this week’s show, I break down the Super Bowl match up between New England and Atlanta. Topics discussed include:

  • How the Super Bowl in 1991 between the Giants and Bills might forecast Bill Belichick’s strategy for this game
  • The success rate of this Atlanta Falcons offense, and how it stacks up against the best NFL offenses of the past 16 years
  • The one statistic, adjusted for strength of schedule, in which neither team excels
  • The relative importance of passing versus rushing for NFL playoff and Super Bowl teams

To listen on iTunes, click here.

To listen here, click on the play button.

2016 MLB ensemble win totals

These ensemble win totals combine the predictions of 5 different quants and the markets (Bookmaker on April 2nd, 2016). The idea is that merging predictions tends to cancel out the errors made by each, leading to more accurate predictions.

1. Chicago Cubs, 95.3
2. Los Angeles Dodgers, 91.1
3. New York Mets, 88.7
4. Washington, 87.9
5. San Francisco, 87.7
6. Houston, 87.3
7. Boston, 87.0
8. Toronto, 86.6
9. Cleveland, 86.2
10. St. Louis, 84.7
11. Pittsburgh, 83.9
12. Seattle, 83.9
13. New York Yankees, 83.8
14. Tampa Bay, 83.2
15. Chicago White Sox, 81.8
16. Texas, 81.2
17. Detroit, 81.1
18. Arizona, 81.0
19. Los Angeles Angels, 80.3
20. Kansas City, 79.9
21. Miami, 79.8
22. Oakland, 78.5
23. Minnesota, 77.9
24. Baltimore, 77.8
25. San Diego, 72.7
26. Colorado, 72.1
27. Milwaukee, 71.3
28. Cincinnati, 70.3
29. Atlanta, 67.6
30. Philadelphia, 65.3

A few thoughts.

Kansas City

What to make of the defending World Series champions?

The markets predict 84 wins, but the Royals do poorly in the computers (70 wins in Davenport, 75 wins for Baseball Prospectus).

The Royals had the best cluster luck in 2015 (+72 runs during the regular season), and it remains to be seen whether their high contact, strong bullpen approach can continue to defy the numbers.


Every predictor had Boston with 86, 87 or 88 wins. Guess it’s certain that this team will bounce back from a disappointing 78 win campaign from last year.

List of 5 analytics predictions for win totals

These sources were consulted for win total predictions.

The top 5 ways numbers go against Carolina, a 2016 Super Bowl preview

I wrote this preview for members of The Power Rank four days before the game. It went public the day after the game.

Predicting Super Bowl 50 is the ultimate battle between eyes and numbers.

By the eye test, Carolina looks like the clear favorite over Denver. The Panthers have a stellar 17-1 record, and they destroyed two of the NFL’s best teams, Seattle and Arizona, to make the Super Bowl.

The eye test also favors Carolina at the quarterback position. Cam Newton has had an Most Valuable Player caliber season, a touchdown machine at the pinnacle of his game.

In contrast, Denver QB Peyton Manning looks old. He missed a multitude of throws against New England in the AFC Championship game. His new label “game manager” shows how fast one of the game’s greats has succumbed to age.

However, a multitude of numbers go against Carolina, a 5.5 point favorite in the markets, in the Super Bowl. Let’s break down the top 5 numbers.

1. Carolina’s Strength of Schedule

The Panthers have a remarkable 17-1 record. However, it’s less impressive once you look at their competition.

They play in a weak NFC South division, which accounts for 6 games. Out of division, they drew the NFC East and the AFC South, two divisions in which the winner had 9 wins. That’s another 8 games.

Below, I’ve listed my NFL team rankings for members which combine a number of metrics based on points and yards per play. The teams in italics played the Panthers.

1. Seattle, 8.77.
2. Arizona, 7.24.
3. Cincinnati, 7.18.
4. Pittsburgh, 7.12.
5. Carolina, 6.90.
6. Denver, 5.84.
7. New England, 5.70.
8. Kansas City, 5.68.
9. Green Bay, 2.37.
10. Minnesota, 1.19.
11. New York Jets, 0.94.
12. Buffalo, -0.00.
13. Baltimore, -0.01.
14. Oakland, -0.22.
15. St. Louis, -0.71.
16. Houston, -0.85.
17. Chicago, -0.96.
18. Detroit, -1.00.
19. Philadelphia, -1.66.
20. Tampa Bay, -2.07.
21. Atlanta, -2.27.
22. San Diego, -2.27.
23. Washington, -2.44.
24. New York Giants, -3.22.
25. Dallas, -3.60.
26. Indianapolis, -3.76.
27. New Orleans, -4.21.
28. Jacksonville, -4.77.
29. Miami, -5.19.
30. San Francisco, -5.68.
31. Cleveland, -6.52.
32. Tennessee, -7.53.

Carolina played 3 above average teams this year. Yes, they just beat the top two teams in these rankings, but they went +8 in turnover margin in those games.

Speaking of which…

2. Turnover margin

Carolina is +20 in turnover margin for the season. As I mentioned in the previous section, +8 of this margin came in the last two games against their strongest competition.

Randomness plays a huge role in turnovers. In my research, I’ve found no variables that correlate with fumble rates in college football.

For NFL quarterbacks, completion percentage seems to predict interception rate. However, this shouldn’t help Carolina, as Cam Newton has a career 59.5% completion percentage, about the NFL average.

The randomness of turnovers implies that the Panthers’ +20 turnover margin this season has little to no ability to predict Carolina’s turnover margin in the Super Bowl against Denver.

3. Carolina’s Strength of Schedule, Part II

In case the previous section didn’t make my point about Carolina’s putrid strength of schedule, let’s look at the pass defenses Cam Newton faced this year. These rankings are based on my yards per pass attempt adjusted for strength of schedule.

1. Denver, 5.13.
2. Carolina, 5.49.
3. Seattle, 5.51.
4. Kansas City, 5.59.
5. Cincinnati, 5.60.
6. Houston, 5.78.
7. Green Bay, 5.89.
8. New England, 5.96.
9. Oakland, 6.01.
10. St. Louis, 6.06.
11. New York Jets, 6.06.
12. Minnesota, 6.10.
13. Pittsburgh, 6.15.
14. Philadelphia, 6.23.
15. Arizona, 6.26.
16. Baltimore, 6.28.
17. Buffalo, 6.56.
18. Chicago, 6.58.
19. Tampa Bay, 6.60.
20. Indianapolis, 6.68.
21. Detroit, 6.69.
22. Dallas, 6.72.
23. Washington, 6.84.
24. Atlanta, 6.85.
25. Tennessee, 6.85.
26. Jacksonville, 6.91.
27. San Francisco, 6.97.
28. San Diego, 7.02.
29. Miami, 7.15.
30. New York Giants, 7.24.
31. Cleveland, 7.26.
32. New Orleans, 7.88.

Carolina has faced 3 good pass defenses this year. Three!

I should note that Arizona’s pass defense would have made a fourth good pass defense before Carolina racked up 11.2 yards per attempt against them in the NFC championship game.

Carolina has thrown for almost 7 yards per attempt, 5th best in the NFL this season. However, strength of schedule adjustments drop Carolina to 11th in my pass offense rankings.

Carolina’s pass offense seems potent partially because of highlights. I’ve seen a multitude of picture perfect Cam Newton throws for touchdowns this year on ESPN.

These highlights represent a small sample size of throws. In the bigger picture, Newton’s numbers don’t look so good by a powerful efficiency metric: yards per pass attempt.

4. Match ups in passing

The pass defense rankings also suggest Carolina’s pass offense faces an uphill battle in the Super Bowl. Denver has the top ranked pass defense by my numbers, as corners Chris Harris Jr. and Aqib Talib lead an elite secondary.

The below visual shows the match ups for the game by yards per pass attempt adjusted for schedule. (You can access the interactive version of these visuals by clicking here.) Better pass defenses appear further to the right, which facilitates comparisons. The offense or defense further to the right is predicted to have the advantage.

Screen Shot 2016-02-04 at 5.00.11 PM

The visual shows that both defenses have a large advantage. Carolina also has a good secondary, as Josh Norman has blossomed into a shut down corner this season.

Because of these numbers (the rushing numbers also favor the defenses), the market total of 44 makes no sense to me. The NFL has averaged 45.5 points per game this year. With two tremendous defenses, you expect a lower total.

Part of this total comes from the eye test: Carolina has looked dominant the past two games against very good defenses. Still, the season long numbers suggest a low scoring game.

5. Cam Newton’s completion percentage

Cam Newton isn’t an accurate passer by NFL standards. He has completed 59.7% of his passes this season, right near his career average.

Even in his best game of the year against Arizona, Newton missed his fair share of throws. For example, he threw behind Corey Brown in the 2nd quarter. However, his receiver made a great adjustment, then beat the Arizona safety for a touchdown.

Ok, fine, I realize that Newton threw that pass a long way, and he should get a little slack for that. Still, he’s no where near Peyton Manning pre-2015 in accuracy.

Newton does present a significant threat with his legs. Carolina runs the zone read, which makes the defense defend against both the QB and RB in the running game. Denver must defend these plays well.


Eyes versus numbers.

I’ll admit it, the eye test makes me a bit queasy about taking Denver +5.5 over Carolina. My numbers predict a 1 point win for Carolina because of the weakness in the Denver offense.

And Carolina has looked damn good the last two weeks.

But numbers reveal the true picture of a team. Carolina has a stellar defense leading an average passer at quarterback. The offense should regress from their performance last week against Arizona, just like Iowa and Michigan State learned hard lessons in their bowl games this year.

It doesn’t always work to trust the numbers. There’s at least a 40% chance Carolina covers the 5.5 points. Numbers kept saying Kansas City would lose their next series in the MLB playoffs this fall. It never happened.

Nevertheless, numbers make a convincing argument against Carolina. The under looks particularly juicy.