talk to caris yo
People are worried about the defense, and with good reason. The worrying bit isn't so much the best quarterback in the state averaging 5.9 YPA and being forced into two turnovers by getting clobbered, but rather Western Michigan running for almost 5 YPC with guards they picked up at a yard sale in Jackson.
I have good news and bad news about this. The good news: a major reason for these issues was a true freshman in his first game who made obvious errors. He fixed some of those errors. The bad news: he fixed those errors so hard he made the opposite error. More bad news: he wasn't the only culprit.
We're looking at two successful first-half counters run by the Broncos. Here's the first. It's second and two on the Michigan 47 on Western's second drive of the day. Western's all like "look, ma, I'm the 2010 Michigan offense" and Michigan brings out its aggressive one-high press man for the first time:
You see the 3-4 front with three tight corners. Kovacs is out of the picture deep. The slot "corner" is Thomas Gordon. The LBs from top to bottom are
Herron Jones, Johnson, Demens, and Beyer, with Roh/Martin/Van Bergen the DL. Your key players are the bottom three guys in the front seven: Beyer, RVB, and Demens.
A moment after the snap:
The tackle blocks down on RVB, leaving Beyer free to fly into the backfield. This is an Admiral Ackbar situation that Beyer is too pumped up on adrenaline and youthful stupidity to recognize. He's all like "gonna get me some QB."
Meanwhile, the RB is moving right, but check out that OL directly in front of the QB: he's pulling left. This is a counter.
A moment later Beyer is recognizing his DERP far too late. He's already three yards into the backfield and his momentum is stopped as he tries to change direction now that the QB doesn't have the ball. the pulling G is going to hammer him.
Not all is lost, though: Demens has read it and is moving into the hole. And you see a lot more of Van Bergen's jersey, don't you?
RVB has given about a yard but now has his helmet across his blocker. Beyer defeats the OG's block and would have a shot at a tackle if he hadn't flown upfield so fast. There's that lead blocker and a lot of room for Demens to close down but he could…
…just about turn it back inside to RVB, who has now totally defeated his block, or he could…
…turn into Jonas Mouton and lose leverage.
That's 25 yards before Kovacs can come up and save the bacon.
Video, with annotation!
I learned this from Spielman. There are two main ways to defend the power play: "squeeze" and "spill." Squeezing is getting into the guard upfield a bit so that the RB has to take it inside into a more restricted hole. Beyer would have to be a yard or two closer to the LOS and to the inside to be squeezing. From that spot he can make a play, or at least make it harder to burst outside that LB.
Spilling is kind of a scrape exchange type deal where the playside DE roars down the line at the pulling G and cuts his ass to the ground. This is intended to create a pile that takes out the other lead blocker and forces the running back to bounce outside, where a linebacker scraping over the top should clean stuff up. Beyer would have had to shoot directly at the G as soon as he reads the pull.
Obviously, he does neither and gets kicked out of a very large hole. If he's in the right position he's dealt with the block well enough to make a tackle. He's not.
Demens did Mouton it. He's got a tough job here with the fullback and a big hole, but letting the guy outside of you is a cardinal sin—unfortunately, one we're all too familiar with. If Demens gets outside that fullback WMU might get a big run anyway but "losing leverage" (the jargon) guarantees it.
Another quiet Van Bergen plus. This is the kind of thing I am talking about when I say RVB is good but the things he does often go for naught. Here he beats a downblock, which is tough, to show up in the hole and potentially rescue Johnson, who you may note ran ridiculously playside and ends up farther away from the play than double-teamed NT Martin. Demens loses the plot and Van Bergen's reward is just a UFR plus and a chase downfield.
Ugh Johnson. To reiterate: the guard directly in front of Johnson's face pulls and he ends up yards away from relevance.
Kovacs. He tackles. He does not not tackle. Here he sort of misses, but this was very rare. This may not hold up against Big Ten teams but there were plenty of opportunities for the Broncos to pick up a touchdown that they could not because Kovacs tackled them.
Yesterday I put up an analysis of a simple iso that cut back behind Mike Martin and picked up ten yards. In the comments Magnus mentioned he thought this must be a mistake on someone's part because when you have the DT and MLB both heading to the playside A gap your defense is no longer "gap sound"—ie does not have one guy in every place a tailback can go—and things like ten yard iso plays result.
This resulted in some discussion about how the MLB's job in the 3-3-5 is to "make the nose right", IE fill the other A gap depending on what the nose does. This is a phrase unleashed to the world by Jeff Casteel's 3-3-5 DVD, and I think it's what West Virginia does with its middle linebacker. It's evidently not what Michigan is doing with Demens when he's aligned in what I've come to think of as the Gergbacker position. Demens doesn't have time to make anyone right because he's too close to the play; he picks a side of the line and goes into the guard.
Another commenter complained that I shouldn't criticize Van Bergen for getting locked out upfield on this particular play because I can't be sure what his assignment is. That's true and a frustration I often have but amongst Wisconsin's brain-melting array of second half runs there is a serendipitous iso that Michigan stops that demonstrates the trends from yesterday's post and suggests that the key guy on a cutback is indeed the backside DE.
It's first and ten on the Michigan 41 in the midst of Wisconsin's first soul-crushing ground based touchdown drive of the second half. Wisconsin comes out in the same I-form they showed on the play featured yesterday. Michigan goes with basically the same stack look as well, though they've flipped Kovacs and Avery. The backup DL (Banks, Patterson, Black) is in:
A moment after the snap we see a difference: the backside tackle is releasing downfield instead of blocking Banks out of the play. That's left to the TE. He gets slanted under:
A moment later we see that Patterson is getting playside of the center… and Demens is shooting into the same gap to take on whoever shows up. Banks is sliding down the line behind them; also note that Jibreel Black has beaten the block of the RT and is coming upfield.
At the handoff point Patterson is beating his guy and Demens is about to slam into a guard at the LOS. In doing so he halts all progress from both the G bubbled over him and the FB. Massive cutback lane would result, except Banks is right on the center's hip. Black is now through the tackle totally and converging; tailback has nowhere to go:
Wad of bodies…
…and two yards.
So. To continue the Week of Defensiveness, usually these plays are picked because they illustrate a larger trend—Kenny Demens runs at the playside guard all day and eats facemask, and I'm pretty sure the design of this defense has a backside DE assigned to an A-gap. My choice here was between criticizing Van Bergen for getting locked out so easily or Greg Robinson for putting him in a tough position. The right answer is some of both, probably.
Object lesson type objects:
- This is a slight variation on the play yesterday. Yesterday Wisconsin kept the backside T in to block Van Bergen and ended up blocking Mouton with a guard. Here the guard attempts to slide over on Patterson and the T is assigned Mouton. These seem like subtly different playcalls with the first designed to cut back and the second to go straight upfield.
- Kenny Demens really does just run to the playside A gap all game, where he enjoys a scone with the DT. Here it works, though the next play is a 12 yard Down G run, the play after a four yard power play, and the play after that a 23-yard Down G touchdown.
- So that means your options on the cutback are backside DE or no one. Here Banks gets a relatively easy task since the guy lined up over him heads downfield and he can just slide along the line; Van Bergen had that guy blocking him. Still, the results were not so good and were repeated on a number of other runs.
- I'm pretty sure this is a bad idea. And not just on general principles! Having the backside DE clean up behind the NT seems like a thing that would work in the 4-3 where the backside DE is actually a DT inside of the tackle. In this scheme he releases downfield or he's got what seems like easy work to seal out a guy who's supposed to be an A-gap player.
- How about Jibreel Black beating a block and being useful on a run play? Woo progress!
So… Greg Robinson doesn't know how to run a 3-3-5. For whatever reason, Michigan is running a 3-3-5. This results in situations like this first quarter run for the Badgers that echoes several themes from the This Is Not A Stack post: by lining up his MLB just behind his nose tackle he dooms that guy to take one step to the playside, whereupon he is eaten by a guard who has no one lined up over him. Even if Michigan successfully plugs that hole they are crazy vulnerable to cutbacks and counters.
It's second and three on Wisconsin's second drive of the day. they come out in an I-form with twins to the field side. Michigan aligns in its stack formation with Jordan Kovacs—supposedly the bandit—aligned to the strong side of the formation, with Cam Gordon hovering over the wide receiver stack to the outside:
Wisconsin's going to run a simple iso play that's designed to go off the right side of the line. You can see in shot two Demens's alignment just three yards off the LOS:
Martin's getting momentarily doubled in the shot below but the G peels off quickly to block the rolled-up Demens. On the backside Ryan Van Bergen has gotten kicked out and Mouton is staring down a free release from the backside G:
Demens is swallowed. I think the idea here is to force Wisconsin to come off their double of Martin quickly, allowing him to run free and eat people, but don't quote me on that. Ezeh is attacking the FB, and there's nothing on the frontside:
Because Martin has slanted past the center and Van Bergen has gotten kicked out there's a big cutback lane. Mouton is in a bad situation, essentially standing still as a guard comes out on him. Demens is getting blocked; the OG has his arm around his back. This never gets called holding but he's being held:
Mouton gets blasted three yards downfield and gives up the inside. Courtney Avery was filling from his overhang spot and could have maybe held this under three yards but once Mouton gets hit in that position the RB cut past him and it's up to Demens and/or Martin to spin of blocks and close it down.
They can't. Vinopal is forced to tackle ten yards downfield.
Object lesson type objects:
- Theory as to the deployment of Kenny Demens two feet from his NT. Demens threatens to shoot into the backfield immediately on plays like this, which seriously reduces the time Wisconsin guards have to double Mike Martin. This allows Martin to use his quickness to slant under the center, get to the playside, and close off holes.
- Problem with the deployment of Kenny Demens two feet from his NT. Once you're engaged with an OL he is going to grab you and delay you and let go before he gets a flag, so you can attempt to get off him and close down the massive cutback hole you've opened up by shooting both your MLB and NT to the playside but you're probably not going to make it.
- Second problem with the deployment. I imagine it has something to do with opponents' consistent ability to hit balls over the linebackers and in front of the safeties; having your linebackers five yards off the LOS gives them more time to reduce throwing windows. I've charted basically all of Tolzein's throws and three or four could have been shut down if the linebackers had been a yard or three deeper.
- WTF Van Bergen? When Michigan is aligned in this fashion the overhang guy—in this case Courtney Avery—is in great position to shut down anything that bounces all the way outside to the short side. With both Demens and Martin headed playside RVB should be flowing down the line, relying on Kovacs to bounce anything that gets behind him and Avery to clean up. Instead he gets upfield and seals himself, basically, leaving Mouton in acres of space with a guard bubbled over him.
- Mouton could do better here, too. It's never good when you're taking on an OL three yards downfield and that OL is moving while you're not.
Ironically, I think this MLB deployment would have worked out okay for Ezeh, who's a big guy with some pop but terribly indecisive. Here there's not much of a decision. Line goes one way, you pound the playside guard ASAP. It seems like a waste for Demens, who has displayed good read and react skills in his brief career as a starter.
I was looking for an adjustment here where Michigan would defense something like this by not having Demens right at the LOS but haven't found it yet. I've seen a lot of small guys getting battered and crappy pursuit angles. I'm not sure if my haziness on what the appropriate play is is my fault or the defense's fault; it seems like Michigan players are making basic errors but it could be a shifting scheme in which a guy like JB Fitzgerald's attempt to defeat a downblock sees him go from the LOS at the snap to six yards off the LOS outside the hashmarks.
More than anything this seems like another example where the scheme is either incoherent or the players don't know what they're doing. Van Bergen getting upfield is the killer here and makes no sense given the alignment of the D.
[Sorry about the delay. Internet issues this morning.]
To continue a theme on Michigan's suddenly mediocre run game: dude, it was all Purdue's cover zero approach. Here's a play that's beautifully blocked all around and still ends up with meh results. It's Michigan's final drive of the first half; the Wolverines have the ball first and ten on their own 43.
Standard three wide from Michigan; Purdue moves their nickelback over the slot receiver and has their safeties in that no-mans-land between deep and shallow:
By the snap they're still in that shallow range; the deeper guy is only seven yards off the LOS—that's where WVU's stack keeps its middle linebacker. Michigan's going to run a stretch:
As usual the key block is the one in the center of the field, where Schilling and Molk are working to seal the playside DT. you can see above that the playside DE is not slanting inside this time and will get kicked out. Michigan is doing something weird, though: they are blocking the backside end. They've done this a lot this year but in almost all cases they've done it on inside zone plays where a cutback is one of the main ways to gain yards. On a stretch this guy is usually set free to fruitlessly chase.
At the mesh point Robinson gets contain from the safety no jk lol linebacker and hands off. Molk has already gotten across the playside DT, who is dead:
That is a dead DT. With Lewan getting a kickout Smith will have a hole to cut into. Schilling releases downfield and gets a good block on the playside LB; Koger heads outside and pulls the other safety no jk lol lb with him. Hurray yards?
Yes, but issue in the middle of the field wearing #45:
This linebacker is the guy Huyge would have released into if he wasn't blocking Kerrigan. He's totally unblocked and can run down the line as Smith zooms past dead DT and Omameh drives the backside guy yards downfield; Schilling and Koger have picked up eliminating blocks on the other LBs.
Through the hole, Smith is one on one with the LB…
The end result is a play where all six Michigan offensive linemen get blocks between good and great… and Michigan gets six yards.
Object lesson type objects:
- Generic stretch stuff. This is another example of how the stretch usually works when it works: Molk seals that DT, Lewan kicks the DE, Schilling releases into a linebacker, a lead blocker takes another one outside, and running back hits a big gap.
- Cover zero problem reinforcement. In this particular instance blocking that backside end is a bad idea, but with eight guys in the box Michigan knows they'll leave contain to a safety and send Kerrigan tearing down the line. If Smith has to cut behind Molk he will be swallowed at the LOS. There's a nonzero chance Kerrigan manages to grab Smith from behind even if the blocking plays out like this. Michigan has a choice between leaving the MLB free or leaving a guy they know will crash down the LOS free. In this particular instance getting a block on the MLB, even a crappy-push-by-the-RB-forcing-a-cutback block, is a touchdown unless Smith gets run down from behind.
The fundamental math is still the same: there is going to be an unblocked guy in the box and Purdue can slant and shift its defense in an effort to get the ballcarrier to an unblocked guy. On the previous example Purdue did this effectively; here they hang on by the skin of their teeth but do get it done.
- Weather allows cover zero. BWS did a post on Michigan's only successful gotcha play, the fourth quarter completion to Kevin Koger that jump-started Michigan's final touchdown drive and was so wide open Denard could throw a wobbly underthrown duck and still have it easily complete. The pass on that play and about a dozen other hilariously misthrown balls by all four quarterbacks goes a long way towards justifying Michigan's resignation to pounding its head into the wall repeatedly. I'm sure Michigan won't do this again—not that they'll have an opportunity against Wisconsin and Ohio State, two teams that won't run cover zero much if at all.
UFRing the Purdue game was a blast from the past, and in a frustrating way. After the first drive Purdue spent much of the day with eight guys in the box, and Michigan ran at them anyway. Since Michigan's always based out of a three-wide formation this is the equivalent of having nine guys against a tradition I-form; Purdue spent the day showing really soft man zero that may have morphed into cover three after the snap but still should have provided Michigan ample opportunity to exploit the big chunks of space Purdue was leaving open.
Look at Roundtree here:
That was his whole day: sitting by himself and never getting a bubble. There were a lot of reasons for this—primarily the weather and Michigan's reluctance to do anything risky against a Purdue offense that's much worse than Michigan's defense—that were in retrospect correct. At the time it was very frustrating.
Anyway, this Picture Pages is about what this eighth guy in the box allowed Purdue to do. Going into the UFR I was hoping I'd see something that would explain why the offensive line seemed to get whipped so badly, and I think this is it. So that's the setup above. It's Michigan's third drive of the second half. They start on the twenty, it's first and ten, and Purdue has eight in the box. Michigan runs a basic stretch at them.
A moment after the snap you can see that the Purdue linemen are slanting away from the stretch instead of flowing with it. This is not something you see often:
At the mesh point Robinson sees the slot guy containing and Kerrigan moving upfield past Huyge, who's releasing, so he hands off.
A moment after the handoff we see that Molk has completely sealed the playside DT. Normally on a stretch play this means the opponent is dead meat. That's because the playside DE has to maintain contain and sets up outside the OT, which means running a good distance outside, which means there's a huge lane for the tailback and whichever guard is playside gets a free release at one linebacker in a lot of space.
Here the DE has not maintained contain. He's slanted inside Lewan and threatens to get upfield for a TFL. Also the MLB is driving hard to the outside. Schilling either aborts his release to rub the DE or just gets caught up on the Lewan block in an effort to get out on the charging LB:
A moment later we see that Molk has erased both DTs with the seal but the playside DE is sitting in that hole. Lewan knows he's lost the battle and starts shoving him past the tailback. Smith has to go outside, where he's got a lead blocker in Hopkins against two Purdue linebackers. Schilling has no chance on the MLB since he shot for this exact hole at the snap:
A moment later Hopkins kicks one guy, Lewan shoves the DE, and Schilling is following the other linebacker into the hole. Smith's cutting up because he doesn't have much of a choice.
Linebacker is now the blur between Schilling and Smith. Schilling's managed to get him to run past the play a bit and he's got to make a diving ankle tackle…
…but he does:
Michigan receives zero yards.
Object lesson type objects:
- Purdue can only do this because they have eight guys in the box against six blockers. One goes with Denard, so that's seven on six. Normally you see playside DEs set up outside on the stretch because if they don't that lead blocker to the outside threatens to pound a single linebacker and send the tailback into the secondary. Here Purdue outnumbers Michigan, which allows them to slant that DE inside and still get two guys on the perimeter when Lewan pushes the DE past the back. Purdue consistently answered the "one safety or zero" question with zero, and these were the results. Here they get the play to go exactly where they want to and kill it.
- I don't think anyone blocking did anything wrong. The only block in question is the one on the playside DE where the guy gets under Lewan because he's slanting inside. If that DE gets past Lewan into the backfield that's a major issue but a main principle of zone blocking is you take the guy where he wants to go faster than he wants to go. Guy wanted to go inside, Lewan shoves him inside and opens up a crease at the LOS. Extra linebacker makes the play. The only thing I think Michigan could have done here is a weird anti-scoop where Schilling shoves the DE outside of Lewan instead of shoving the DT inside of Molk. I don't know if anyone's ever tried that so it's hard to blame the players.
- This is actually close to breaking for some yardage maybe? Despite all this Schilling's good-faith effort to do something with that filling linebacker and Lewan's ability to create a decent hole sees Smith almost cut past the charging LB, whereupon he'd get somewhere between five and many yards. He can't.
That's not a serious knock on Smith on his most effective day as a Wolverine. A back with the ability to make the cut he does on this sloppy field and the speed/power to run through that ankle tackle attempt would be a special guy indeed; hopefully Demetrius Hart can be that guy.
So, yes, I think I did find an explanation in what Purdue was doing that partially exonerated the offensive line. Michigan saw this front all day and kept running into it, which resulted in a crappy day on the ground. In this case the "crappy day on the ground" is 4.3 YPC excluding sacks, which is basically what good DeBord teams averaged on the ground. Since DeBord absolutely loved to run "away" from the extra guy and out-execute in the face of herculean odds, this makes sense.
Michigan's offense is going to net a huge RPS minus in UFR because of this rock, rock, rock playcalling, but don't take that too seriously. I get why they did it when Denard threw two horrible interceptions and a lot of his simple hitch routes to the sidelines were fluttering ducks. The conditions affected his throwing significantly, and allowed Purdue to spend 80% of the game running cover zero. When they stopped this on Michigan's late third-quarter drive from their own four, Michigan went right down the field until a clipping penalty on Lewan put them in second and forever.
More tomorrow when the offense UFR drops.
This pair of plays is striking only in comparison to the defense. They're both inside zone runs Michigan gets about a yard on, but one's a third and goal from the one so that's all you can get from there. The first is in the second overtime; Michigan has just executed a throwback wheel of its own for a first down at the eleven. They come out in a three-wide.
Illinois does their usual bit with a linebacker over a slot receiver, but this time he's going to walk down and blitz:
Michigan's running an inside zone. This can go anywhere but on this play the hole opens up behind everything as Illinois is slanting away from their blitz. By the mesh point something odd is happening:
The backside defensive end is headed directly upfield. Most of the time the DE will either sit and contain, shuffle down the LOS, or roar down it. Here he's getting way upfield. Two reasons for this, I think: 1) he really really has to contain because of the corner blitz and must not let the QB outside of him, and 2) by doing this he guarantees a handoff.
Meanwhile, Webb's pulling to the backside of the play to get a block on whoever the cutback guy is for Illinois. Because of the switch this is not the DE but the slot LB. Webb does not realize this:
[Update: commenters point out this is actually Koger, which it is. Apologies to Webb.]
Oops. With Webb blocking the guy who the zone read is supposed to option off Michigan's left a free hitter on the backside. He tracks…
Michigan gets two yards and eventually has to resort to a Houdini escape when Terry Hawthorne jumps a third and eight slant. The deflection miraculously bounces to Hemingway.
The second play is Michigan's second to last play of the game. It's third and goal from the one. On first down an inside zone run was stuffed when Schilling got slanted under. On second down a QB lead draw was stuffed when Forcier went airborne unnecessarily. On third down Michigan comes out in a three-wide package. From the one.
Before the snap the corner walks down again:
Illinois runs the same curve, shooting the DE directly upfield after several plays where he was tearing downhill at the RB; he made the tackle on first and goal. Webb, however, is taking a much different angle:
Huyge and Omameh are blocking downfield but not that well. Spence chucks Omameh to the ground and shows up in the hole; Webb is still headed for the slot LB:
Shaw hits it directly upfield in this useless screenshot and does get the ball over the plane, but here's Tate celebrating:
So… that's an adjustment to an Illinois adjustment. After getting fooled by this the one(!) time it came out earlier, Michigan goes to the sideline for mere moments—the offense has to come right back out. In these moments someone grabs Webb and says "if the DE goes straight upfield your assignment is the linebacker crashing in from the slot." Then Michigan seems to invite that very play by coming out in a three-wide formation on third and goal from the one. They want that DE to erase himself and for a LB to get singled up on the 260 pound Webb. His block provides the extra momentum that barely gets Shaw over the plane.
The reason I'm so down on the defense and high on the offense is that these things seem to happen on one side of the ball but not another. It should be clear that no one's on the keeper when Scheelhaase takes it for big yardage, but then he does it again on the exact same play. It should be clear that something's not right in the way you're defending the option but nothing really changes. Etc. Michigan's offense adapts in ways the defense doesn't, and I don't think it's youth when the players who can't contain a keeper are Roh and Mouton.
Chitownblue asked about an RPS plus on the final play of the game, but that's an example of what I'm talking about. Illinois runs the same play they got their earlier conversion on and Michigan runs the same defense. You've got a rub route against man coverage and a seven-man protection you're blitzing into. If Illinois had gotten to pick Michigan's defense on that final play, they would have picked a man-zero all out blitz. They got it, but Jonas Mouton saved the day by making a great play. If he gets cut or just blocked Scheelhaase rolls away from Roh and has a receiver wide open for a score.
It is in these ways that Michigan's defense is different from its offense.
Bonus: you know that corner who jumped the slant in the second OT and should have had Michigan in fourth and long but for some Notre Dame-level BS? Michigan ran a circle route (a fake slant to an out) on the two point conversion and got Hemingway wide open.