I thought that myself when I read that article that talked about a Data Scientist(tm)
spread is dead
Note: Most of of my long "here's a handy guide for Borges complaining" post got deleted. Here's a few bits of it.
So I did the thing where I update (finally) the UFR database, then self-UFR the last game because this damn column comes before Brian does the UFR for the latest one, and then I say things about all the things. Let's just cut to the things:
Yards per play (# attempts in parentheses) of normal downs by formation.
|Opponent||Shotgun||Ace||I-Form||Pistol||Tackle Over||Goal line||Total YPP|
|Central Michigan||6.7 (17)||7.7 (21)||7 (19)||3 (2)||-||1.5 (2)||6.8|
|Notre Dame||7.4 (14)||10.6 (22)||3.2 (13)||4.4 (14)||-||-||7.0|
|Akron||21.8 (11)||5.6 (19)||4.5 (18)||-1.8 (5)||-||-||7.6|
|Connecticut||5 (25)||2.8 (19)||1.8 (12)||3.5 (4)||-||4 (1)||3.6|
|Minnesota||6.1 (9)||4.9 (7)||1.7 (7)||-||5.8 (20)||-||5.0|
|Penn State||3.6 (16)||7.7 (22)||3.8 (9)||6.4 (7)||0.8 (11)||-||4.8|
|Indiana||7.2 (34)||14.1 (14)||8.8 (16)||4 (1)||11 (8)||-||9.2|
|Michigan State||5.3 (23)||7.4 (7)||1.8 (4)||0.1 (12)||-||-||4.0|
|Nebraska||4 (23)||0.3 (7)||1.8 (12)||6.3 (6)||-||0 (1)||3.1|
|Northwestern||5.4 (16)||4.7 (21)||5 (24)||11 (3)||-||-1 (2)||5.1|
|Iowa||2.7 (23)||1.2 (9)||5 (13)||4 (2)||-||-||3.1|
|Total||6.1 (33%)||6.7 (27%)||4.6 (23%)||3.6 (9%)||5.5 (6%)||0.8 (1%)||5.6|
What's left out are 2-minute drills, 3rd/4th down and longer than 6, and any short situations.
And yards per play by how spread, i.e. the # of receivers in formation, they got each game, with % of plays they lined up that way in parentheses:
|Central Michigan||2.33||1.5 (3%)||2 (3%)||8.1 (56%)||7 (33%)||-1.3 (5%)||-|
|Notre Dame||2.29||-||-0.8 (6%)||8.6 (59%)||5.5 (35%)||-||-|
|Akron||2.23||-||9 (4%)||5.5 (70%)||13.5 (26%)||-||-|
|Connecticut||2.33||4 (2%)||2.5 (7%)||2.9 (49%)||4.5 (43%)||-||-|
|Minnesota||1.70||2 (5%)||4 (35%)||5.4 (47%)||7.7 (14%)||-||-|
|Penn State||2.14||-||5.7 (14%)||5.2 (58%)||3.7 (28%)||-||-|
|Indiana||2.66||-0.5 (3%)||16.2 (7%)||9.7 (40%)||10.7 (23%)||6.5 (27%)||-|
|Michigan State||3.00||-||-8 (2%)||4.1 (26%)||4.8 (41%)||3.6 (30%)||-|
|Nebraska||2.47||0 (2%)||3.7 (6%)||0.5 (43%)||6.3 (41%)||1.3 (8%)||-|
|Northwestern||2.15||-1 (3%)||2.2 (8%)||4.9 (64%)||5.7 (23%)||18.0 (3%)||-|
|Iowa||2.91||-||-1 (2%)||3.6 (38%)||2.1 (30%)||3.8 (26%)||3.5 (4%)|
|All games||2.38||0.8 (2%)||4.6 (8%)||5.7 (51%)||6.2 (30%)||4.8 (9%)||3.5 (<1%)|
My base assumption was that when Michigan goes more spread they're putting the B+ receiving threat of Dileo, or the C+ receiving of Chesson on the field, and usually moving the C- blocking/D- receiving of A.J. Williams or Joe Kerridge off of it.
The Fetal Position Theory of Offense
However going to wider looks didn't seem to do much good against Iowa or Michigan State (that one at least M was behind for a good portion). That's because of a lot of things, one of those being that despite spreading it out, Michigan's been leaving those guys anyway. Both players are probably the best blockers of their position groups, but that's not saying much. Meanwhile they give up pretty much any threat of doing something other than blocking, and opponents have used that opportunity to tee off. Since neither is good enough to pick up a majority of those blitzes, there've been a lot of messes in the backfield as a result.
Iowa blew up Michigan's penultimate drive with back-to-back A-gap blitzes. On the first Kerridge was in to pass block and got lit up by the blitzer, who was immediately into Gardner. On the second they had Green in there and had him run a pattern that the defense ignored. With immediate pressure Gardner ignored Green and chucked a pass into an unready Funchess's back. That is progress, but the lesson is just doing the thing you ought to be good at doesn't fix the problem; you have to practice doing it as well.
But even when they do spread like a boss, there's a lot of things going wrong. Look that this play, the penultimate (so rare you get to use that word twice in a day) offensive one of the game for Michigan:
Starts at 0:53:38 if browser player isn't working.
There's so many ways to win here, but nothing comes of them.
1. There's bubble action. Though of course they don't throw it despite it being open because this isn't a check (Michigan's checks are only to ISO or the pistol speed option). And Funchess, not Gallon or—infinite ARRRGHHH—Norfleet, is still the designated bubble screen guy. Anyway with the safety deep and bailing, the bubble is 7-12 free yards if Michigan can recognize it and throw it, but that has to be built into the offense. The way Borges has been using the bubble screen is on called plays. It was cool that he threatened it out of a more open look—previously it seemed his capitulation to this one play was predicated solely on its usefulness for running from heavy sets. The way Rodriguez used—and the way Urban will deploy it against us on Saturday—is it was as an instant check to things opponents did to hamper his zone read game.
|Fuller captured one of the ultra-rare instances of a shotgun-Kerridge play that wasn't him blocking.|
2. Tipped zone blitz. That end spread out and the obviously blitzing linebackers suggests there's going to be a zone blitz pre-snap but there's just 6 seconds on the playclock at that moment so they don't really get to adjust to that. Still, this is a win for an outside run, since that DE is going to drop back and stand where a great block would normally deposit him.
3. Center and guard versus linebackers. This ought to be a win. Those LBs stunt their blitz a bit so that the first gets a 2-for-1 and the second can slice in free. By coming up pre-snap they made it harder for Glasgow to get off the combo and pick up his guy, but he just has to come off the double and take a half-step sideways to block that gap. Here's where agility in a spread center helps you, and where the lack of it hurts Glasgow, even when he knows what he's supposed to do.
4. Magnuson starts the play far to the playside of that tackle. That's a big advantage for the offense. The DT indeed slants into Mags, then chucks him and gets to the outside, totally blowing that advantage.
5. It's Devin Gardner and Fitz Toussaint in the backfield, so chances of a missed tackle are pretty good. However Gardner is at about 45% right now because he's been beat up so badly in the last few weeks, and Toussaint's pass blocking problems are part of what's inviting these interior blitzes, since the downside for the defense (having that guy violently cut to the ground while a receiver slants into the unoccupied territory) is unlikely from Michigan's offense.
Also Jackson's is hesitant with his block, and is set up to spring Funchess to the outside, so that nickel guy is going to be free to tackle after just a few.
So going to a spread isn't going to fix everything. The formation did give them more room and opened up the bubble, but Michigan can't access those yards because they come to the line too late to see anything in the defense and adjust to it, and hasn't practiced doing that. The defensive playcall made Glasgow's athleticism the key to the key block, but that's not Glasgow's strength. And crappy blocking elsewhere meant this play was still dead in four ways. Such is Michigan's offense. They're not all good at any one thing, and they don't do the things that some of them are good at, and the end result is a lot of plays where guys are forced to execute the things they're bad at.
The Little Bubble Package: Dead?
Here's how the Bubble-or-Run package has fared:
Iowa was crashing the backside SAM and had their safety ready to pounce on the bubble (even if Funchess didn't drop one). I think there's something you can do about that (dare I say rollout?). Running it against Northwestern was cool, but I would have expected Borges to know by now that's it's scouted and lead off with that third counter. Or add it to the scrap heap with the rest of the fancy things he's tried. I'm sure a picture pages is coming so I won't get further into it.
Background image by mgouser hillhaus
A thing I noticed this offseason while going over the depth and usage of various Michigan defenders is that Mattison used a lot more nickel than we gave him credit for. One thing Ace noted was that we're (finally) recruiting more cornerbacks. We shrugged a bit while losing two more CBs to playing time transferitis this fall, but I don't think we should be shrugging so much.
A little background (skip this if you already know personnel terminology and usage): Defensive coaches tend to match their personnel to the types of players on the field for the offense, NOT the formation. In general the number of backs and tight ends will be matched by linebackers, and the more that come out for receivers the more DBs the defense will send out. Three wide receivers generally means five defensive backs (i.e. nickel), two wide receivers equals four DBs (e.g. 4-3 or 3-4), etc.
The classic personnel shift is on 3rd and long, when the steady rock-pounders make way for the seven-yards-or-bust fellas. But it happens so often despite the situation that it's more accurate to see the game of matching personnel as another strategic aspect of the master's football game.
The offensive personnel is usually expressed in three digits meaning # of RBs, # of tight ends, and # of receivers, respectively. So 113 means 1 RB, 1 TE, and three WRs. Sometimes they'll call that same "eleven" personnel, referring to the first two digits. Examples below; click embiggerates.
How the matching up occurs is up to the coach. You could, for example, play a run-first OLB whenever a fullback is in, and sub him for a more rangy linebacker when the the fullback runs off the field for a tight end who's a known receiving threat. This happens all the time, but it's hard to track the defenses' reactions since we can't tell one linebacker in a formation from another in UFR. We do have data from which we can determine how many receivers were out there at any given time, and it's clear from these data that the more receivers the more defensive backs.
|WRs in Game||DL||LBs||DBs|
The last row is important because it shows Michigan left its base 4-3 Under set for an extra defensive back far more often than otherwise, usually at the expense of a linebacker. We didn't go to a nickel every time three receivers stepped on the field, in fact there were 22 plays charted where Mattison put his 4-3 personnel against four-wide (mostly against Northwestern and Purdue). But the charts not only say that Michigan was forced out of its base 4-3 set often; it says we played more Nickel downs than 4-3.
|Receivers in Formation|
If I remove 4th quarters and all plays that occurred when Michigan was up by more than one score, the 4-3 just barely edges the Nickel, 147 to 140. This isn't opponents trying to play catch-up. It's two things: the personnel that Mattison inherited, and the spread offense forcing Michigan to adapt to it.
Why all the nickel and diming? The first part is a story about outside linebacker. Early in the 2011 season Michigan played Brandon Herron and Brandin Hawthorne at WILL, while at SAM we lost Cam Gordon to injury and his backup was a redshirt freshman. That freshman, Jake Ryan, was earning his way toward more playing time, but in the meantime we still had Carvin Johnson taking snaps at free safety while Thomas Gordon was in at the nickel role. Watch what happened at about mid-season:
That is Gordon moving to free safety and splitting time with Woolfolk, while the freshmen linebackers had their usages increase. Greater faith in Jake and Des explains some of the variance, however the real story is matching personnel:
|San Diego State||2.51||4.38||1.88||43.21%||44.44%||6.17%||6.17%|
I pointed out the two extremes on the schedule with boldation: Northwestern used about twice as many receivers in their formations as Iowa did, but there was a limit to how many defensive backs Michigan would counter with. The nickel served as well for 4 WR as for 3, yet accounted for 4 in 5 plays. However when the opposition went to 2 WR (Iowa), Mattison could spend a majority of the game in the 4-3.
When Michigan's on offense. Nothing is out of the ordinary yet, but when we turn the tables and show how defenses have reacted to Michigan's personnel it gets interesting:
|Season||Avg. Receivers in Formation||Avg. DBs in Formation||Difference|
This is not including anything when Michigan was more than a score down, but the season averages counting everything say about the same thing. I went through the plays and even a few youtubes and yes, in 2010 they played one-high against us despite spreading the field to pass as much as Purdue. Michigan went bigger in 2011, and got more defensive backs, which is counterintuitive except for one factor: opponents in 2010 really really really feared the running game, and tempted Michigan to pass.
Okie dokie. | Greg Shamus via ESPN
One more table to break this down by Michigan's opponents last year, 4th quarters and two-plus-score leads excised:
|Opponent||WRs in formation||DBs in formation||Difference|
|San Diego State||2.44||4.89||2.4|
Nothing really jumps out except perhaps more spread in close games, and SD State's apparent paucity of linebackers (weird—didn't they just have that guy who recruits lots of linebackers there?) Actually that's Charlie Strong's 3-3-5, and the GERG numbers from 2010 are similar due to the same effect.
What it means for this year. Alabama and Air Force aren't going to be spread it out—their challenges are elsewhere. However the Big Ten schedule is spread-heavy, with Ohio State joining the ranks of the many-receivered. Due to recent attrition, Michigan goes into 2012 with just six scholarship cornerbacks for three positions that will be filled half the time. It's a good thing the coaching staff has four guys coming in at corner to replace the one expected departure. These days, in order to keep up with the Joneses, that nickelback position has to be considered as much of a starter as, well, a third receiver.
I feel happy!
Every offseason there is someone (often named Gary Danielson) who goes on record proclaiming the doom of the spread offense and a return to the paleolithic days when quarterbacks were pale and made of granite. The best and dumbest remains this from the Cleveland Plain Dealer:
This may sound strange when coach Mike Leach's version of the spread has Texas Tech near a national title game, but Michigan's struggles this season while Rodriguez has implemented his system into college football's winningest program might be a sign: The spread, in fact, is dead.
The scheme was designed to give underdogs some hope, when a team could open up the field by recruiting a smaller quarterback with a sharp mind and a quick release, and a handful of speedy receivers. But the offense intended to confound the big boys has now been adopted by the big boys, and that may have started its demise.
But that was two years ago.
This year's evidence centered heavily on…
Texas abandoning the vestigal Vince Young-y bits from its offense after the graduation of Colt McCoy and ascension of monolithic Garrett Gilbert to the helm:
With the exit of Colt McCoy, so goes the shotgun spread for the Texas Longhorns. For the 2010 season, Mack Brown and offensive coordinator Greg Davis have decided to go under center with starting quarterback Garrett Gilbert.
Going under center could mean the beginning of the end for the spread, a style that was made popular by powerhouse SEC programs and then picked up by other conferences.
Florida abandoning the Tebow offense in favor of a conventional pocket passer:
Meyer and offensive coordinator Steve Addazio tweaked the spread offense to tailor Brantley’s strengths, putting him under center more and eliminating many designed quarterback runs.
The effectiveness of Alabama's traditional battering ram of an offense featuring returning Heisman winner Mark Ingram:
When Alabama prevailed last season, it was with gnarly defense and a vanilla offensive scheme — albeit led by Heisman Trophy-winning back Mark Ingram.
That profile in turn had ripples for Texas, a 37-21 loser to the Crimson Tide in the title game, that perhaps suggest a shift in the broader landscape.
and spread 'n' shred HQ Michigan sucking:
How are these memes working out so far?
Texas fans are livid that Mack Brown's handpicked talent couldn't manage a meaningful touchdown against UCLA:
What is the Texas offensive scheme? My answer- We have a spread that we pass out of 80% of the time, and an under-center formation we run out of 80% of the time. We use the spread 70 – 80% of the time against quality opposition. We call very few running plays for the QB- just a couple of called QB draws per game. We don’t run zone read or lead option, which were core plays for us the last several years. Our offense has an H-back that can block on running plays or be a receiving option on pass plays.
The proposed short term solution is to utilize "more zone reads and option runs" and use whichever quarterback has the best combination of running and throwing ability.
Florida fans were clawing their eyes out after managing just over 200 yards of total offense against Miami (Not That Miami) and just over 300 against Tennessee (Also Pretty Much Not That Tennessee) but found joy in the redzone in the form of one Trey Burton:
The freshman scored six touchdowns in Florida's 48-14 victory over Kentucky, including five rushing as a quarterback in the Wildcat formation. The feat broke Tebow's old record of five touchdowns against South Carolina in 2007. … On Wednesday, UF offensive coordinator Steve Addazio said Burton's role as a quarterback in the Wildcat package likely will expand as the season progresses. Burton's role might be similar to the role Tebow played as a freshman, when he was a changeup to starter Chris Leak, who led the Gators to the BCS national title in 2006.
Alabama's grinding non-spread attack is sixth in total offense and just took out their most difficult competition to date by doing this with Mark Ingram and Trent Richardson:
Ingram took eight handoffs out of the wildcat, nine from the pistol, three from shotgun and four when the quarterback was under center. Richardson only took eight handoffs, with his two biggest gains, 53 and 10, out of shotgun.
For those counting, Mark Ingram took four of 24 snaps from a conventional I-form against a top ten foe on the road.
Finally, no one's laughing at half of Michigan's team now:
Also there is Cam Newton, though Auburn highlight technology has a decidedly Soviet feel to it. FWIW four weeks into the season (almost nothing), three of the top four offenses in the country are dyed-in-the-wool spreads that feature a ton of quarterback runs: Michigan, Oregon, and Nevada.
We now return you to your regular programming, and Gary Danielson to the alternate universe he spends six days a week in.
Beat that dead horse beat it beat it beat that dead horse yeah
Gary Danielson keeps banging the anti-spread drum, although that may be because he's the only man in America you can call for a quote about how the spread is dumb. Some guy in West Virginia did—complete with Rodriguez slam, natch—and got a litany of quotes to the same effect.
I only bring it up because this seems like the exact worst argument you could ever make about anything:
Danielson said the spread's weakness was displayed late in the Illinois-Missouri season opener when Mizzou needed one more first down to seal the win, "and on third-and-3 they had nobody in the backfield to run the ball except (Heisman Trophy-candidate QB) Chase Daniel.
These are the ways in which this argument is the worst argument ever:
- This event never happened. The only Mizzou third and three in the fourth quarter came with just under 13 minutes left on the clock. (Daniel threw incomplete.)
- At no point was Illinois within a score of Missouri, so "sealing the win" isn't exactly of paramount importance.
- This game between two spread teams (with garbage defenses, sure) featured 94 points and over 1000 total yards.
Oh, wait, this might be worse:
"I don't mean we're going back to grind-it-out football. I think every team will have to have their four-receiver sets, but I think in the future coaches are going to realize they have to be able to hand the ball to the tailback, too."
West Virginia ran 76% of the time last year, Northwestern, etc etc etc.
A few days ago when I pointed out that nine of the top ten offenses in the country were "spread" offenses some commenters protested that any grouping of offenses that included Illinois and Texas Tech was too broad to be meaningful. I agree with that. HOWEVA, Danielson groups Missouri and West Virginia and Michigan all together; anything in a shotgun with more than two wide receivers is the "spread." This makes his argument the "spread" is on the way out obviously untrue.
If Danielson was specifically addressing the Rodriguez-WVU spread there might be a case to make, but he'd have to make it in a significantly less dumb fashion. A fashion like this:
When Rodriguez got to Tulane with Tommy Bowden they threw the ball all over the place, but (a) it was in Conference USA, (b) they were excellent at the 3-step passing game, but defenses are better at defending against those passes now than they were a decade ago, and (c) his downfield passing game left something to be desired. And in the years since, it's not that Rodriguez is at heart a running guy, it's just that was what worked and it masked some of the passing game deficiencies. When I study the route combinations, they do not appear to be designed conceptually, and instead are a kind of grab-bag of a few routes here or there. You don't see his schemes organized of horizontal, vertical, and triangle stretches.
That's Smart Football, and that's something to be legitimately concerned about. I'm not sure if we'll get a read on whether or not he's progressing in this area with these quarterbacks and this offensive line, but I plan on pinging Smart Football's proprietor Chris after the year to find out if he's detected any adaptations.
The always-incorrect Gary Danielson:
"I said it before the season -- and I was out there by myself -- I think we've seen the spread has peaked, like the wishbone did in the mid-70s," Danielson said Thursday.
The top ten teams in total offense so far, spread teams in bold:
- Texas Tech
- Louisiana Lafayette
- Nevada (pistol variety)
- Oklahoma State
- Penn State
Might want to cancel the funeral.