Competent Coaching and RPS Trends
Mods, I intended for this to be a series of Diaries but I don't know how to create one. Could someone move it and educate me please?
During the course of the last season I started wondering if there was an objective way to measure the influence of good coaching beyond the all-encompassing “win/loss” record. Wins and losses are all that matter, but there is so much that goes into that. We’ve seen officiating, weather, a fluke play or a key injury cost us a game. Should that count against the coach? Perhaps, but I wanted something more concrete. This blog does an excellent job of analyzing Xs and Os and grading players, but I wanted a means to grade the coaches. Something to quantify if Coach X is really "better" than Coach Y.
So, is Jim Harbaugh really worth those millions of dollars? What does Don Brown give us that DJ Durkin didn’t? Do certain teams scheme against us better than others? These are questions I had during the 2016 season. That’s when I realized we have the RPS metric from the UFRs. This metric of course is subjective. It’s Brian’s take on when he feels one side is a schematic advantage over the other. It’s not perfect, but I wanted to explore it.
I went back and read every UFR since the 2010 season. I compiled the RPS numbers into a data set and looked at it a few ways. My hypotheses were:
1. There should be a strong positive correlation between RPS and win/loss record.
2. Harbaugh and staff should have a much higher RPS number than their predecessors, proving they are better schematic coaches.
Before I get to the results, one last caveat to the data set:
Several UFRs could not be located. Whether they were never published (ie Ohio State, Bowl games, etc), or simply lost on the interwebs, I have no clue. I wrote the mgoblog staff with a list of missing UFRs but so far no answer if they were able to locate any that I missed. For missing UFRs the RPS is left blank (which is different than receiving a 0 RPS score). A score of zero will factor into the average, a blank score does not. If anyone can find the missing UFRs, I will update the results.
Results
Hypothesis 1: Strong correlation between RPS and Win/Loss
For instances where I am missing either an offensive or defensive UFR, the point is not plotted.
Here are the results of Offensive RPS (x axis) and Defensive RPS (y axis) plotted. Wins and losses are color coded. Results are about what you would expect.
Only 3 times did we win the RPS battle in both offense and defense and still find a way to lose the game. Those three games:
-2013 Iowa. Michigan, down 3 was near field goal range with 2 minutes left when Gardner scrambled left and inexplicably fumbled the ball before going out of bounds), icing the Iowa win.
-2015 Utah. New Coach, new QB against a tough road opponent. 3 INTs, one for a TD doomed this.
-2015 MSU. Michigan did everything they needed to do in order to win that game, except prevent the flukiest of fluke plays.
There were two losses in which the offense graded slightly positive, but the defense negative which resulted in losses:
-2014 Rutgers
-2014 Notre Dame
The rest of the results are as you would expect. Eight of our losses during this time period had both an offensive and defensive RPS value of zero or less. Clearly when it rains, it pours.
Summary:
When we perform a post mortem after a loss it almost certainly comes down to RPS failures, and usually on both sides of the ball. Teams found something to exploit us, and we failed to adjust (MSU double A gap blitzes). Or, we failed to successfully exploit weaknesses against the opponent (Borges screen passes).
On the flip side our wins are all across the board. Our seven largest negative RPS scores on defense were actually wins. We even beat Penn State in 2014 despite an offensive RPS score of -11. We can see in the data that we are able to gain wins without the need for RPS. We can sometimes “out talent” the opponent without needing to get fancy. In fact, I did see this as a comment by Brian in several instances saying something to the effect “RPS was not needed against this opponent”. That particular day we were content just lining up and running vanilla offense/defense without showing our playbook.
To borrow from Brian, “Cackle with glee when”:
-Offensive RPS is +3 or better
-Defensive RPS is +3 or better
-We have simply overmatched our opponent and toss RPS out the window
In the next installment I will show RPS by coaching staff and who earned their $$ and who may be overhyped.
In the final installment I will show RPS trends against opponents. Which teams have our number, and whose milkshakes do we drink regularly?
I love them, personally.
As a soccer and hockey player (more fluid sports) it's fascinating to analyze the matchups/grades while watching the corresponding plays in order to gain a better understanding of what constitutes good football.
I almost can't watch the games with some of my friends anymore because of it.
excited for future installments; thanks for compiling these data!
If Michigan could run the ball, it would've closed out 2015 Michigan State, 2016 Iowa, and 2016 Ohio State in victory formation. Those 3 game account for half of Harbaugh's 6 losses in 2 seasons.
An improved rushing attack led by a dominant OL is what I hope to see the most under Harbaugh's leadership. Truly elite teams can run the ball on opponents even when the opponent knows what's coming. It's one of the hallmarks of a dominant football team.
but I like your analysis.
Thanks for your work. This is a really interesting way to answer the questions you set out to answer!
You can look at the RPS data over time at this link: https://mikegros.shinyapps.io/ufr_analysis/
by going to Single Player -> Team -> RPS for either the offense or defense
The offensive OPS has definitely taken a step forward under Harbaugh, while the D mainly improved after Rich Rod's departure, as you'd expect. The lack of D RPS this last year is likely because the players got most of the on-field credit. Informative plots below:
First with the Offense:
Then with the D:
They look better if you open in a new tab or even better, look at them at the link provided
I'm looking at the website with a computer, not a smartphone app.
The menu in the upper left, which has my username in its top line because I'm logged in, contains a "Create content >" entry. When I mouse over the ">", a submenu with "Diary" or "Forum topic" comes up. Clicking on "Diary" brings up a page with directions and a text entry box. I don't have any actual content, so I didn't go any further.
Thanks for the RPS scatter plot.