I did this with the defense previously, and finally got around to the offense. Bear with me, Excel Geniuses, as I struggle to make the vlookup to index(match) transition. A lot of the lessons were discussed so often in the UFRs that it's not worth going over them again. Split Zone was their base running play, and they struggled all season to find appropriate counters for what defenses prepared against it because the natural counter for it—QB keeper on zone read—was not something they could get/were coaching(?) McNamara to do. The officiating got me all house-stompy again. Hassan Haskins was a steely-eyed missile man. Anyway this exercise is about finding things that were new.

Favorite Plays?

They were a Power and West Coast sorta team, which is to say, pretty Harbaugh.

Type of Play	Plays	YPP	EPA/P	EPA(T)
Power Run	255	5.3	+0.2	+55
Pass Short	219	5.9	+0.2	+41
Inside Zone	163	5.1	+0.2	+26
Pass Deep+Sack	114	10.8	+0.5	+59
Screen/RSO	33	4.3	-0.2	-6
Speed to the edge	33	9.8	+0.5	+16
Play-Action	31	7.2	+0.3	+9
Outside Zone	28	7.7	+0.3	+8
QB Run/Keeper	17	7.0	+0.4	+6
Scramble	12	8.8	+0.4	+5
Trick Play	10	20.4	+1.6	+16
Total	915	6.6	+0.3	+236

From the above alone you might say they probably didn't go to their counters enough. Later in the season they started taking more shots downfield at Roman Wilson, who rewarded them for it. Their RB and TE deep routes were also astonishingly effective for how little they were deployed. Even adding all the sacks, dropback passing was one of the most effective tools in their bag. Play-action wasn't used very often—Gattis seemed to prefer working RPOs into the short passing game rather than just fake something.

They were evenly split between power and zone running earlier in the season but leaned on power more as it went on, and mostly ran off their tight ends regardless of what the offensive line was doing. That wasn't a bad plan.

I tried to identify individual heroes but the OL winning their battles varied game by game. Haskins was the constant, rarely getting less than half a point on any carry, and often hurdling fools downfield.

All told the breakdown was 50% called runs (for 5.65 YPC, +0.2 EPA/play), 37% called passes (7.94 YPA, +0.3 EPA/play), and then 6% play-action, 6% run-pass options, and 2% screens.

[After THE JUMP: You won't believe which player was super duper clutch on passing downs, except you will because you read the UFRs.]

---------------------

Favorite Formations?

Michigan liked to use Pistol formations on standard downs to get their RBs going downhill.

To get nittier I need to define some terms:

FORMATIONS

Offset is a running back at pistol depth but not directly behind the QB, ie a cross between Pistol and Shotgun. Ace in this context is under-center with a single back behind the QB, while I-form means under center with a 2nd back at fullback depth. Goal line means they were under center and had three TEs (or 2 TEs and an extra OL) on the field.

DOWN SITUATIONS

I do it a bit differently than Connelly or Fremeau. Standard Downs in general are 1st down and 10-, 2nd and 6- or 3rd/4th and 4-, with the over being Passing Downs. I also classify all plays within 2 minutes of the end of a half as passing downs (my data show they act exactly like it); just assume when I say Passing Downs that 2-minute drill players are included. I also break out Short downs, which are 3rd or 4th down and 1-, or any play inside the opponent's 2 yard line. I didn't remove garbage time because I would stop charting whenever I felt the data were no longer useful.

Here's 2021 Michigan's general formation choices in each situation.

QB Formation	Standard	Pass/2MD	Short
Gun	201 (37%)	196 (55%)	8 (11%)
Pistol	224 (42%)	48 (13%)	19 (25%)
Offset (Gun)	85 (16%)	52 (15%)	26 (35%)
Empty (Gun)	17 (3%)	53 (15%)	(0%)
Goal Line	3	4	16 (21%)
Ace	6	1	3
I-form	1	1	3
Wildcat	1	2	-

We have charted a fair amount of opponents over the years. Even with their numbers normalized by greater trials, Michigan was pretty extreme in being a pistol/offset team. More than any previous Michigan team we've charted, they were a pistol team, though on standard downs that split evenly between the gun and pistol. When they went under center usually there was some kind of razzle dazzle.

They seemed to shift over the course of the season from pure pistol to more offset stuff.

The difference is in the RB's aiming point. From the gun, a handoff to the RB puts him on an aiming point to the edge, whereas the pistol will have the RB step to either side and threaten either A gap quickly. Offset formations are something between that: the RB will definitely be aiming at a specific side of the formation, but still coming downhill and threatening inside gaps.

I'm pasting the next table as a graphic so you can see when they transitioned to more Offset stuff:

Note that I count formations at the snap, so for example if they motioned from Offset or Pistol to Empty that will go under Gun/Empty.

The transition to more Offset stuff clearly occurred in the bye week before Northwestern. It was used less in the next two games, was their primary look in the late season, then less so in the postseason. There aren't enough trials per game to really get an idea what was effective, but Michigan averaged 9.1 yards per play with their Gun formations, 5.7 YPP in Offset, 7.1 from the Pistol, and 2.2 on the rest (including most of the goal line situations). That all seems like sampling error to me since the schedule had some monsters at the back end.

Did it change anything? I looked at yards per play and expected points per play…

…and no it doesn't look like more offset was more effective. If I had to guess what made them change their approach, it was the development of Zinter and Stueber as their best run blockers, whereas inside they had Vastardis (not a mauler) and injuries causing a rotation at left guard.

Personnel?

They used more 3-wide personnel on standard downs after the MSU game for some reason.

Michigan had a second tight end on the field about half the time, including when All or someone lined up as a quasi-fullback. This only varied somewhat per game.

	ON FIELD				WRs on Field
Opponent	WRs	TEs	RBs		Standard	Passing
Western Michigan	2.34	1.58	0.98		2.18	2.87
Washington	2.21	1.63	1.04		2.18	2.64
NIU	2.21	1.73	1.00		2.26	2.38
Rutgers	2.15	1.77	1.00		2.26	2.50
Wisconsin	2.49	1.49	1.01		2.36	2.68
Nebraska	2.53	1.44	1.02		2.40	2.82
Northwestern	2.43	1.54	1.02		2.16	2.83
Michigan State	2.60	1.35	1.00		2.46	2.75
Indiana	2.43	1.46	1.00		2.37	2.87
Penn State	2.19	1.63	1.00		2.16	2.42
Maryland	2.62	1.36	1.00		2.39	2.92
Ohio State	2.44	1.54	1.02		2.40	2.78
Iowa	2.63	1.33	1.00		2.59	2.78
Georgia	2.66	1.24	1.00		2.59	2.79
TOTAL	2.44	1.50	1.01		2.34	2.74

There wasn't much variation though Michigan did start putting more receivers on the field on standard downs after the MSU game, which I presume was a reflection of Erick All's injury (which he exacerbated against PSU) and Andrel Anthony's emergence. The exception to that pattern was the Penn State game, which I think was a matchup choice to take advantage of PSU's smaller/faster LBs.

I can also sort through which players I tagged for each play. We're going to ignore QBs because they often got named as the player of record on sacks and turnovers, but not for any of their good works. It's also best to look at these guys by position, since usage affects their success rates. Total EPA can be read as the number of points they added over the course of the season. EPA counts things like turnovers as well as scores and converted 3rd downs if you're wondering why there's such a discrepancy:

Running Back	Plays	YPP	EPA/P	EPA
Hassan Haskins	294	5.1	0.24	+71
Blake Corum	174	6.5	0.28	+48
Donovan Edwards	58	8.2	0.11	+7
Tavierre Dunlap	2	4.0	0.10	-
Leon Franklin	1	5.0	0.08	-

No surprises here. Haskins and Corum were two of the most effective backs in the country.

Receiver	Plays	YPP	EPA/P	EPA
Cornelius Johnson	78	15.4	0.38	+30
Roman Wilson	46	16.8	0.37	+17
Daylen Baldwin	35	15.6	0.29	+10
Mike Sainristil	35	12.4	0.38	+13
AJ Henning	27	12.7	0.56	+15
Andrel Anthony	24	19.5	0.97	+23
Ronnie Bell	3	54.5	2.04	+6
Cristian Dixon	3	7	-1.27	-4

These numbers include yards gained on penalties, or the times these guys ran the ball. Ronnie Bell's WMU game aside, look what happens when a play goes to Andrel Anthony. Henning had a few huge end-arounds. Cornelius Johnson and Roman Wilson would be an excellent top pairing on any team.

Tight End	Plays	YPP	EPA/P	EPA
Erick All	54	11.5	0.21	+11
Luke Schoonmaker	30	9.7	0.23	+7
Carter Selzer	5	6	-0.1	-1
Joel Honigford	4	10	-0.29	-1

Again, nothing we don't know about these guys. Those numbers for both of the starters are excellent; it's a bit of a surprise that Schoonmaker was so close to All, though my scoring for their blocking numbers creates more separation.

Expected Points

This play was worth +7.05 EPA, meaning Michigan turned a situation when their opponent was more likely to have the next score into LOL.

I put a fuller explanation of Expected Points Added into the defensive breakdown. For every down/distance/field position there's a value based on how many points the offense is expected to score this drive or the opponent's next drive. So like, if it's 1st and goal on the opponent's 1 yard line, you're expected to get 6.1 points. If you kick a field goal right there, you get 3 points, but the EPA for that play is –3.1 points. Get it?

Let's start with how they performed on various down/distance combinations because I found this weird:

Situation	Plays	Avg Yds	Avg EPA	Total EPA
Standard	627	7.5	+0.08	+50
Passing	361	16.1	+0.42	+153
Short	99	5.9	+0.50	+49
Total	1087	9.9	+0.23	+252

That is an extraordinary amount of points being generated on passing downs. I don't know what this means but it probably suggests that Michigan was over-conservative, or at least chose to be conservative, when they on schedule.

	Plays				Avg EPA				Total EPA
Opponent	Standard	Passing	Short		Standard	Passing	Short		Standard	Passing	Short
Western Michigan	37	20	6		+0.61	+0.40	-0.09		+22	+8	-1
Washington	40	29	8		+0.13	+0.58	+0.24		+5	+17	+2
NIU	47	16	7		+0.60	+0.07	+1.46		+28	+1	+10
Rutgers	33	20	10		+0.35	+0.10	-0.12		+12	+2	-1
Wisconsin	39	32	10		-0.23	+0.43	+0.86		-9	+14	+9
Nebraska	62	28	4		+0.17	+0.22	+0.51		+11	+6	+2
Northwestern	62	28	8		-0.01	+0.34	+0.38		-0	+10	+3
Michigan State	52	35	5		-0.26	+0.63	+0.06		-13	+22	+0
Indiana	42	28	7		+0.03	+0.39	+0.60		+1	+11	+4
Penn State	43	31	7		-0.03	+0.26	+0.10		-1	+8	+1
Maryland	47	29	10		-0.04	+0.73	+0.34		-2	+21	+3
Ohio State	42	12	12		+0.29	+0.24	+0.97		+12	+3	+12
Iowa	43	22	4		+0.29	+0.68	+1.02		+13	+15	+4
Georgia	38	31	1		-0.74	+0.48	+1.15		-28	+15	+1
TOTAL	627	361	99		+0.08	+0.42	+0.50		+50	+153	+49

So that's a surprise. I thought I was going to be able to explain this away with Michigan playing overly conservative against outmatched opponents, but they didn't really get to many passing downs against those teams. The tougher defenses, with the exceptions of Ohio State and Iowa, contributed to the pattern of Michigan struggling on their standard downs then making clutch plays in crunch moments.

The MSU and Georgia games were the most extreme examples so I reexamined those games. Michigan only had seven EPA-positive plays against Georgia on standard downs, the best of them an Arc Read on a late 3rd and 3 that McCarthy converted. To the bad was the uncalled PI at the end of the first half (-4.9 EPA), two more turnovers, and a pair of sacks. I also wrote a Neck Sharpies about an average (–0.11 in EPA) standard downs play that I thought demonstrated Kirby Smart knew Michigan's tendencies and wouldn't let them get away with it:

That's the same play that Corum broke for a TD against Iowa. Georgia defended it pretty straight-up, clogging the point of attack instead of asking their safeties to take on Haskins. They also canceled Michigan's flea flicker and Michigan learned they couldn't edge Nakobe Dean. Things that worked against Georgia were bombing it to Edwards (+3.73), bombing it to Anthony (+3.61), and a YOUGOTTABEKIDDINGME throw to Erick All (+3.44).

The MSU game was a similar story. The Andrel Anthony catch and run (+7.15) was a 3rd and 5. Sainristil behind the defense (+4.2) was a 3rd and 8. The sidearm sling to Mikey (+3.2) to make it 30-14 was another 3rd and 8.

The Mystery of Clutch

That got me wondering if a higher success rate on passing downs was a season-long phenomenon. Success Rate, as defined at Football Outsiders, is more janky than EPA as an efficiency metric, but it's effective at capturing the goal-oriented nature of possession football. A "success" means you got 50% of the yards to gain on 1st down, 70% on 2nd down, or 100% on all other downs.

Intuitively, I would expect passing downs to have a much lower success rate: You need more yards to register a success, and the defense is often using specialized personnel and strategy to prevent you from achieving narrowed goals. One of the defining characteristics of 2021 Michigan, however, was their success rate on passing downs was often just as good, if not better, than their standard down success rate.

SUCCESS RATE BY GAME

Opponent	Standard	Passing	Short
Western Michigan	21/37 (57%)	3/17 (18%)	4/5
Washington	18/40 (45%)	6/23 (26%)	4/8
NIU	23/39 (59%)	5/9 (56%)	6/6
Rutgers	13/32 (41%)	2/13 (15%)	4/8
Wisconsin	7/37 (19%)	8/27 (30%)	6/8
Nebraska	29/60 (48%)	7/23 (30%)	2/3
Northwestern	30/59 (51%)	7/26 (27%)	6/7
Michigan State	23/51 (45%)	10/31 (32%)	3/4
Indiana	16/42 (38%)	7/22 (32%)	4/6
Penn State	13/41 (32%)	7/26 (27%)	5/6
Maryland	21/39 (54%)	4/17 (24%)	7/10
Ohio State	25/40 (63%)	2/9 (22%)	11/12
Iowa	24/43 (56%)	6/18 (33%)	3/3
Georgia	8/38 (21%)	12/28 (43%)	1/1

That comes out to a 45% success rate on standard downs I charted, 30% on passing downs, and 76% on short downs. I have (still imperfect) data from the whole season that I've put together for HTTV previewing purposes, and that gets slightly different numbers because my charting has a lot (but not all) of garbage time in it. Taking all that out and comparing with their peers, we can start to see Michigan really did seem to excel at converting passing downs but not the others.

Parentheticals are Michigan's ranking in each:

 

Comparison	Success Rt	Standard	Passing	Short
Michigan	41% (336/826)	43% (235/546)	29% (63/221)	76% (38/59)
B10 Avg	38% (5th/14)	43% (5th)	24% (3rd)	68% (11th)
Pwr5 Avg	40% (28th/65)	45% (40th)	25% (14th)	67% (39th)
NCAA Avg	39% (45th/130)	30% (68th)	24% (22nd)	61% (82nd)

2021 FBS vs FBS, garbage time removed

The data from my charting is a bit rosier, I guess because I went too deep into garbage time. The greater data show they were just an average FBS offense on standard downs, not good at short downs (due to that spate of costly early season whiffs), and pretty good at passing downs.

Who was responsible for Michigan being so clutch on their passing downs? I guess we can go to the UFR…

Chart?

Sure why not.

Passing Downs Chart!

Offensive Line
Player	+	-	T	Notes
Hayes	12.5	25	-12.5	The Penn State game, mostly. Remember these are passing downs.
Keegan	11	20	-9	Elite DTs of UGA and MSU gave him trouble.
Vastardis	16.5	29.5	-13	Rough time vs UGA, IU, MSU, and Rutgers.
Filiaga	5.5	4	+1.5	Got to miss some of the rougher types.
Stueber	11.5	14	-2.5	The Wall.
Zinter	10.5	18	-7.5	Struggles were all early in the year when club-handed.
Barnhart	1	16	-15.0	That explains the Rutgers outlier.
Jones	0	2	-2.0	Ignore.
Atteberry	0	1	-1.0	Ditto.
Crippen	0	0	0.0	Detto.
All	40	5	+35.0	MY DUDE!!!!!
Schoonmaker	14	11	+3.0	And there you see the difference between TEs.
Honigford	1	7	-6.0	Kinda pointless in these situations.
Selzer	0	1	-1.0	Ignore.
Hibner	1	0	+1.0	Ignore.
TOTAL	124.5	153.5	-29.0	Faced some very good DL; ran to a lot of positives.
Backs
Player	+	-	T	Notes
McNamara	115	63	+52	Included the passing in here.
McCarthy	27.5	7	+20.5	Wow.
Villari	0	0	0	DNP
Haskins	23.5	6	+17.5	Pretty good but…
Corum	32.5	8	+24.5	Game-changer.
Edwards	9.5	3	+6.5	Game-breaker.
Dunlap	0	1	-1	A play.
TOTAL	208	88	+120	Clutch City.
Receivers
Player	+	-	T	Notes
Johnson	4	5	-1.0	I wasn't crediting WRs for catches earlier in the year.
Sainristil	11	5	+6.0	But this one stands out.
Henning	1.5	7	-5.5	And this one does too.
Wilson	11	5.5	+5.5	BOMBER.
Baldwin	4	0	+4.0	We'll always have the JJ throw vs WMU.
Anthony	12	2	+10.0	Playmaker.
TOTAL	43.5	24.5	+19.0	Not a fair accounting, but a good one.
Metrics
Player	+	-	T	Notes
Protection	219	323	68%	Passing downs were almost all of the pass pro problems.
RPS	61	62	-1.0	RPS on Successes was +25.

I'm sure most offensive lines have issues on passing downs, as opponents are freed to stunt, bring more interesting blitzes, and play their better pass rushers without consideration to their rushing defense. Stueber's relative success is the reason he's going to be one of the better OTs in the draft this week.

But of course we noticed a few guys who popped in these situations, the first being Erick "Clutch Viking" All.

That second was an incredible quarterback throw, true, but the yards after catch in these situations were a constant theme of throws to Erick All on passing downs. One of Michigan's favorite escapes from their 3rd and long pickles was to have All leak out after the edge defender committed to his pass rush. It got them a couple of huge conversions in the Nebraska game, and was so effective because All was already an effective blocker. Another thing they were doing early in the season was Split Flow Counter, that play that looks like split zone and becomes a pass in the flat to the TE that I hated. All managed to convert a couple of those for first downs by breaking cornerback tackles after taking contact at the line of scrimmage.

Then there was 2nd and 9 versus Penn State.

Passing down.

Both quarterbacks were very good, bordering on excellent, in these situations.

	Good				Neutral			Bad					TOT
QB	DO	CA	SCR		PR	MA		BA	TA	IN	BR		DSR
McNamara	29	63(8)	4		14	8		11	10x	16xxx	9xxxx		68%
McCarthy	7	8(3)	2		1	3		-	4x	4	2x		63%

I didn't even show the +'s, which signify when the QB was throwing under duress. I had McNamara down for 22 DO+ or CA+ events, and McCarthy 5. In other words, find 27 minuses from the pressure metric above and replace them with a quarterback who dealt while avoiding a free rusher.

I think this became an underrated part of McNamara's game over the season because he was so smooth about it. Watch how quickly he processed these reads; you can't beat this.

JJ had his own…uh, method, but as he got better at actually staying in the pocket he could make up for reading things a little slower than McNamara by making the football go faster.

Some of the receivers really popped in these opportunities. Sainristil of all people seemed to have the best hands, and came through on the first 4th down slant against MSU. But there was also a bona fide 3rd down playmaker among the receivers in Andrel Anthony. They went to him ten times last year on passing downs, coming back with three TDs and three more first downs, with another 1st down called back because an offensive lineman was a micrometer too far downfield. One of those plays was the best catch of the year*.

*That counted. Still mad they called back Ronnie Bell's one-hander.

Even when he didn't pick up enough yards for the play to be called a success, Anthony was adding value, such as this end-around that should have ended in the backfield.

He should be on the field every passing down this year.

Nobody exactly struggled on passing downs, but AJ Henning came the closest. I think having Ronnie Bell functionally replace him is going to make a difference, and that this is an offense on the verge of an explosion.

Conclude This!

There's a lot of ground to cover with their standard down success rate, and to make up with the losses of Haskins and Stueber, but the danger of that passing game should allow them to go to the air more often, and increase the frequency that backs are able to pick up intermediate chunks. Last year safeties were charging down early and often, and running plays either got 4 yards (two from Haskins falling through that guy) or 40 (from Corum dodging/Haskins hurdling that fool).