RB Carousel Analysis
Introduction
During the Indiana game, several people (myself included) were irritated about the constant running back changes. Watching live, it felt like the backs would be pulled right after a solid run, and it seemed like the changes were hindering any back from establishing a rhythm. The RB carousel approach started last year, and while there were mixed feelings about it then, the grumbling has increased this year as the offense has turned into a train wreck. So, I decided to look back and see if there actually is any effect of the constant RB changing on rushing success.
-------------------------
tl;dr
- Michigan changes their running backs a lot (48% of the time)
- Michigan changes their running backs a lot even after successful runs (45% of the time)
- Following both unsuccessful AND successful runs, Michigan gains more yards on average when there IS a running back substitution
- But, Isaac and Higdon do tend to have higher rates of successful runs when they are given more consecutive carries
- Overall, Michigan tends to have a higher frequency of explosive runs on a back's first carry, but potentially at the cost of lower odds of sustained running success
- IMO, there is no evidence from this review to suggest that the RB carousel does more damage than good
-------------------------
Some notes on the methodology
- I looked at running plays with regards to yards gained and play success (at least 40% of first down yards, 70% of yards to gain on second, 100% of the yards on 3rd and 4th down)
- I only looked at running plays, so this does not track substitutions that occurred when going from a running play to a passing one, nor does it account for yardage or success/lack thereof on RB receptions
- I also did not consider a running play from Hammering Panda FB dives or WR runs as an RB change, nor did I include any garbage-time drives [not charted on UFR]
- This only looks at RBs substitutions, yardage, and success. It does not account for play type, blocking, or game context. However, neither do the criticisms of the RB carousel
-------------------------
Summary
Summary |
Rushes |
Changes |
Change Rate |
Overall |
196 |
94 |
47.96% |
Successful Runs |
82 |
37 |
45.12% |
Unsuccessful Runs |
114 |
57 |
50.00% |
Discussion
Michigan substituted their running back on 48% of running plays. I don’t have this number for any other team, but that is definitely a staggeringly high rate. And there certainly is plenty of fodder for potential questioning, as the running back got pulled after a successful run 45% of the time. Given the high substitution rate, the number of times the substitution happened due to fatigue is negligibly few (i.e. the Isaac long run than where he just stepped out of bounds), so this was almost always a strategic decision.
As would be expected, and probably on any team, there is a higher rate of substitutions after an unsuccessful running play than successful one. But I wouldn't be surprised if the difference in rate of substitution is much smaller for Michigan than most other teams.
-------------------------
Substitution Effects
So, we’ve established that Michigan changes RBs a lot (which we already knew), and that they often do it even if the previous run was successful (which confirms half of what we already thought). Now, we need to see whether or not these substitutions actually affect the success of the following play:
Classification |
Rushes |
Next Run Avg. Yard |
Next Run SR% |
Unsuccessful & no change |
57 |
4.47 |
38.60% |
Unsuccessful & RB change |
57 |
5.33 |
31.58% |
Successful & no change |
45 |
4.11 |
48.89% |
Successful & RB change |
37 |
6.86 |
48.65% |
Discussion
This chart categorizes runs as either successful or unsuccessful, and then looks at the next play and whether or not there was a running back substitution, and then finally the average yardage of the next run and the rate the next run was successful.
As one would expect of the coaches, following an unsuccessful run, there is a higher average yardage gained after a substitution compared to keeping in the same back. However, there is a decrease in the success rate of the following play when there is a substitution (more on that to follow).
Now, the key metric for us armchair coaches is in the comparison of the latter two classifications. If our ire is well-placed, we should see that substitutions lead to fewer average yards gained and lower success rates on the following run. However, we instead see a much higher average following a substitution, and there is no difference in the success rate following substitutions compared to keeping in the previously successful back.
-------------------------
Consecutive Carry Breakdown
With that stunning defeat in mind, we must consider that the previous chart is a limited overview. What about the possibility that the constant substitutions hinder Michigan backs from sustaining momentum?
This is a more difficult question to answer, but one possible insight is to compare the average yardages gained and success rates as running backs are given more consecutive carries prior to substitution.
Here is the overall breakdown:
Total |
||
Consecutive Carry (n) |
Avg. Yards |
Avg. Success |
1st (103) |
6.39 |
40.20% |
2nd (58) |
3.88 |
46.55% |
3rd (23) |
4.00 |
34.78% |
4th (9) |
5.22 |
44.44% |
5th (5) |
3.00 |
40.00% |
6th (2) |
5.50 |
50.00% |
7th (1) |
3.00 |
0.00% |
8th (1) |
-1.00 |
0.00% |
Discussion
(2 MGoPoints to the first poster to correctly guess which back was given 8 consecutive carries, and 2 more for the game.)
This shows that explosive runs come most often on running backs’ first carries. However, the success rate does tend to increase as backs are given more consecutive carries. (The highest consecutive carries have the major caveat that they came in end-of-game, obvious running situations.) This suggests that the RB carousel strategy seems to create more explosive runs, but it may be at the cost of decreased success rate.
-------------------------
Breakdown by RB
Finally, I wanted to break this down further by each running back. Since Isaac and Higdon are more traditional, workhouse backs than Evans, who is a more change-of-pace back, it would seem like they would be more susceptible to having their momentum limited by the frequent substitutions. (I took some liberties in deciding how many consecutive carries to graph, based on sample size and whether or not the highest consecutive carries came in end-of-game situations as mentioned above).
Discussion
Not surprisingly, Evans saw a decrease in yardage productivity as his carries increased, and no effect on success rate. This goes along with the idea that he is more of a change-of-pace runner.
Both Isaac and Higdon are very productive on their first carries, but they do tend to produce more successful runs on average when they are given more consecutive carries. While the data on higher consecutive carries is limited by sample size, the trend of increasing success rate on additional consecutive carries does follow with the idea of those two as more workhorse backs.
-------------------------
Summary
Overall, as stated above, this review suggests that the RB carousel may possibly decrease Michigan’s potential success rate on rushing downs, but it is correlated with more explosive runs and no sacrifice in average yardage gained. There is also no evidence that substituting after a successful run makes Michigan any worse off. It will be interesting to see if the frequency of RB substitutions decreases after Higdon produced the first 200-yard game in a decade. But, if not, I will try my best to keep the grumbling down if I see him get pulled after a big run or two.
October 16th, 2017 at 12:04 PM ^
Great analysis. Love seeing this type of OC.
October 16th, 2017 at 12:22 PM ^
And should get bumped to the Diary section so it stays up on the board for a while. Not a short read but a very insightful one.
Thanks OP for taking the time to pull this together..
October 16th, 2017 at 12:26 PM ^
I think there is a perception that people don't read diary posts. I don't know if it's accurate or not, but there definitely seems to be more commenting on board posts than on diaries. Some of that might be people commenting on board posts without reading a lengthy OP such as this though.
October 16th, 2017 at 2:00 PM ^
Something this good (and addressing a question like the OP, many of of us have) should'nt disapear from view when the surge of "Harbaugh is good and here's why I think so" or "O'Korn is not good and here is why I think so" shit-posts appear and push it off the front page.
My guess is many of us are killing time right now at lunch waiting patiently for Brian's game review post to appear and that's why we're here. But this type of post should remain visable for a while IMO and on the board it's likely to go away fairly soon.
EDIT: The very next post on the board proved my point.
October 16th, 2017 at 12:08 PM ^
I'd be interested to see what other teams are doing with their RB rotation. Correct me if I'm wrong, but doesn't Alabama do something similar? Maybe Georgia too?
I'll admit to being annoyed as the constant rotation, but I'm a banker, not a football coach, so I lean towards trusting what our guys are doing.
That said, I really liked the one carry that Walker got against Indiana. I'd like to see him get several more in upcoming games. Maybe he was just lucky. But maybe he's finally developing into the 4-5* back we all thought he was when he signed his LOI.
October 16th, 2017 at 12:14 PM ^
Yes, Walker. I don't know if it was me, or my wishful thinking but he looked impressive in that run. Seemed to have an extra dimension of speed and agility that I've not seen lately. Wish there was more tape to analyze him further b/c I went from 6 to midnight after that run.
October 16th, 2017 at 1:03 PM ^
Definitely. FWIW, Harbs said in the postgame presser that he might get some more work.
October 16th, 2017 at 2:42 PM ^
October 22nd, 2017 at 8:22 AM ^
October 16th, 2017 at 12:26 PM ^
rotates them more by series rather than by play - I could be wrong on this. So a Bama back will come on the field with the offense, and unless he gets tired, he is out there for that position. Next time the offense comes on the field, it might be a different back out there.
October 16th, 2017 at 12:27 PM ^
Yes, but both 'Bama and UGA have stud backs. We don't.
October 16th, 2017 at 12:55 PM ^
Not sure we can say this definitively about Higdon or even Isaac.
October 16th, 2017 at 2:22 PM ^
October 16th, 2017 at 12:43 PM ^
This seems to be the Pro appoach as well. As an example, the Lions yesterday seemed to do situational substituitions at the RB position. Abdullah, Riddock, and Zenner were all in during many drives based on what they were trying to accomplish; short distance, 1st down, passing plays. I don't think this is a bad thing as long as you aren't giving the play away with a pattern. Last year everyone could guess jet sweep when McDoom came in the game, and Wildcat Peppers was a dead giveaway as to what was going to happen. BTW, it still pisses me off that we never used that series to do anything else. We tried to pass out of it one time but it wasn't there and Peppers took a loss.
October 16th, 2017 at 12:12 PM ^
I think the carousel and use of it depends on who our personnel are. If you have a 1a 1b situation, then I can understand putting in some fresh legs after a huge run. If you have a clear 1 and a very distant 2, then I'd anticipate a work horse. I think there are some intangibles that just aren't there in the numbers to analyze.
October 16th, 2017 at 12:12 PM ^
October 16th, 2017 at 1:17 PM ^
It's also gotta help recruiting, esprit de corps, and, for guys who will go on to the pros, it keeps them from getting beaten down. Take for instance the kid from Northwestern (Jackson?). His draft stock will be impacted because he's had a ton of carries in college.
October 16th, 2017 at 7:21 PM ^
Not too sure about all that. Most guys probably would rather be "the" guy instead of sharing carries with two or three others. Also, a guy like Zeke Elliott had almost 300 carries in back to back years (273 and 289) and he went 4th overall. Good workhorse backs are hard to come by but teams fall all over themselves to get them when they can.
Recruits see Gurley, Fournette, Elliott, etc. They want to be that, not Gillislee, Jones, etc.
October 16th, 2017 at 8:20 PM ^
All true. Why Bama backs have not done well in pros. I think they had too many carries. Not sure if Saban has changed this approach or not.
October 16th, 2017 at 12:13 PM ^
I think that Michigan needs to be more strategic in its changes. For example, rotating while searching for an explosive play makes sense in the first quarter, but not so much when Higdon is getting six yards a pop in the 2nd half against State.
October 16th, 2017 at 12:14 PM ^
One of those guys is the QB. RB by committees can work. Before people say "That's Alabama", I believe our defense has been on par or better than theirs the last two seasons.
The problem is not that we have a RB rotation. The problem is our inconsistent O-line and teams have not respected our passing game all season. Also, it seems like our WR blocking is subpar compared to the Carr years. Basically, I think the RB committee is the least of our worries.
October 16th, 2017 at 12:14 PM ^
Should this be a Diary? Great stuff.
October 16th, 2017 at 12:20 PM ^
But I posted it on the board more as a PSA for discussion, because the post about the RB substitutions following the IU game had a lot of hot takes (again, mine included). Here is another appeal to reason for us fans. This coming from someone who hates the "we must blindly trust the coaches" approach, so frankly I was surprised by these results.
October 16th, 2017 at 12:24 PM ^
It is great stuff, and these are the sorts of posts that make this site stand out among its peers in the conference, but I typically will move it to the diary section once it starts making its way down the page lest anyone start with the "Where did my post go?" as this has happened in the past when I moved it right away.
OP, FYI - this will get moved eventually, but the maximum number of people need to see this excellent work first.
October 16th, 2017 at 12:15 PM ^
You know more than our coaches. Keep up the analysis.
So I can disregard.
October 16th, 2017 at 12:52 PM ^
I thought I knew more than the coaches and then I did this analysis and it turns out my hot takes were wrong.
October 16th, 2017 at 1:02 PM ^
Annoying than people (the guy above you) pretending that these coaches are all Ph.D candidates that know so much about the game and we are all little peasants who are clueless.
Football is honestly not that complicated. Jim and staff are not perfect. Something is not working currently, and there's been nothing to indicate that this collection of minds can correct it in time.
Good work on your content.
October 16th, 2017 at 2:24 PM ^
That is also coming from the guy in the harbaugh presser thread that suggests throwing deep late in the 4th qtr to get that last first down to close out the game.
I applaud your work - it is very well done. The topic also crossed my mind Saturday as Walker got his nice run and then exited.
October 16th, 2017 at 12:15 PM ^
Great analysis.
I'd also like to see how Evans has done inside the tackles vs outside the tackles. Feel like he's been asked to grind out a lot of yards this year. Haven't seen anything that, to me, says he's the best equiped to do that.
October 16th, 2017 at 12:17 PM ^
and don't run every time Isaac is in the game.
If the first contact is two yards in the backfield the probability of success is low.
End of analysis.
October 16th, 2017 at 12:43 PM ^
I think Evans would be more successful if they gave him some carries on power plays. He seems to be an unfortunate casualty in the sense that his carries are almost all zones, which we know how that's been going.
He didn't magically get worse from last year, there just aren't many holes when he's in the game.
October 16th, 2017 at 2:11 PM ^
I like Evans but think he needs to get stronger. If he and McDoom broke a few more tackles they would cause defenses all sorts of extra headaches.
October 16th, 2017 at 2:14 PM ^
Not sure I agre with this. I think Evans risks losing his agility by trying to bulk up, and likely won't ever be a great tackle-breaker. He'd be better off trying to upgrade his speed, which as it stands is ok but inferior to other backs with his profile, and notably inferior to Higdon.
October 16th, 2017 at 2:20 PM ^
Evans is just not the same back this year. He was so quick and explosive last year and adding that bulk (which has not translated to breaking more tackles) may have taken that little extra explosion out of his game. I would also rather he get back to what makes him more electric and accept that as his role. He was a great change up last year to De'veon.
October 16th, 2017 at 2:30 PM ^
October 16th, 2017 at 2:51 PM ^
October 16th, 2017 at 12:18 PM ^
October 16th, 2017 at 2:19 PM ^
Evans appears to be your exception then.
October 16th, 2017 at 4:05 PM ^
October 16th, 2017 at 7:31 PM ^
Not just rhythm but also there's something to be said about getting hit on consecutive plays and having adrenaline pumping as opposed to the up and down of having to settle down on the sideline and going back in, etc. I have experienced both and it's not close. Consecutive runs can begin to sort of numb the little pains that individual runs with gaps between do not.
October 16th, 2017 at 12:20 PM ^
8 consecutive carries has to be Isaac in the Air Force game.
This is a really cool analysis, love it. The flaw is of course exactly what you mentioned. There is a sample size of N=0 for a bell cow situation. No back has had more than 8 carries in a row so we don't have any data to compare to other than this momentum trend extrapolation.
October 16th, 2017 at 12:25 PM ^
the ideal way to see if there is any effect would only exist if Michigan nixed the carousel strategy for a couple games against comparabley ranked opponents, and then going back and doing this similar analysis. Again, I wouldn't be surprised if that happened now that Higdon was so productive last game, and didn't have any fumbles.
October 16th, 2017 at 12:28 PM ^
Then I selfishly hope you do a follow up to this in a few weeks. I assume now that you have the framework set up it wouldn't be too difficult.
October 16th, 2017 at 12:38 PM ^
I am trying to learn how to write a script to collect the data from the UFR (similar to how the other UFR analysis website works) but I have a ways to go for that. For this, I had to manually go through the ESPN play-by-plays to see whether or not there was a substitution. I also collected more information than I ended up needing, so yes, it would be much easier to do again.
October 16th, 2017 at 12:24 PM ^
October 16th, 2017 at 12:35 PM ^
October 16th, 2017 at 3:34 PM ^
October 16th, 2017 at 12:31 PM ^
and make throws to the running backs.
If the wide outs don't want the ball bad enough, give the receivers and TEs more opportunities.
Let the WRs block.
October 16th, 2017 at 12:45 PM ^
I would love to see some 'Minor Rage' type shotgun runs out of the two back set. Put Hill in as one of the backs as he can lead block, rumble for a few yards, and also catch out of the backfield.
October 16th, 2017 at 12:35 PM ^
Neg this post.OP spent a lot of time on this.Wish I could give him a +100