"I am insane. I am childish. I once drank an entire bottle of wine while watching Michigan and God's favored child Denard Robinson lose in Ohio Stadium to an undefeated Ohio State team BANNED FROM BOWL COMPETITION FYI JUST WANTED TO MENTION THAT FOR NO REASON."
There was once a dream that was called Denard Robinson: Accurate Passer. You could only whisper it; anything more than a whisper and it would vanish…it was so fragile. And I fear that it will not survive the fall.
Last Saturday while watching Andrew Maxwell derf another derpity doo, I half-rhetorically asked the assembled a room full of Spartan fans who's the best passing quarterback in our conference this year. Answers, in order of appearance:
"CHHIIIRP CHIRRRP" –insects with that leg-rubbing noise
"Howl" –wolf in the distance
"How about that little dude on Northwestern?" – a Sparty
"No he graduated." –another Sparty
"Shit. Really?" –first Sparty
After three weeks the stats (min=25 attempts) say it's Denard and ol' Tyranno-arm:
His interceptions are dragging down the passer rating, but half are explained by an accurate throw Vincent Smith deflected, and Roundtree getting shoved into last Tuesday by a Bama cornerback. It's just three games in, and the Big Ten competition this year isn't exactly the NFC South, but raise your hand if four weeks ago you thought there might be even a flimsy statistical case for saying "Denard is the best passer in the Big Ten right now."
We've been over his higher efficiency as a runner from the gun ad nauseum, and charted his regression last year as of December, but is he really a better passer when dropping back? Brian's suggested explanations were Pressure, Situation, or Luck (ie sample size). Let's dig into the UFR database and see if there's an answer.
[ROLL'D OUT AFTER THE JUMP]
First, the passing plays charted where Denard was the quarterback:*
* 2012 is through Air Force. 2011 is missing the bowl game; 2010 is missing Ohio State and the bowl game; 2009 is missing Ohio State.
Right away you can see there's not a lot of under-center throws to go on—only 10 total before last season, and even then less than 1/4 of Denard's passes were not from the shotgun. This year it's 80-20. We're cuttin' out pre-2011 data from here on out. We're also noting marked year-to-year improvement already…
Bad Read (BR)
Bad Read X (BRX)
…that actually started after the Iowa game last year and coincides with the shift back to the shotgun. While we're housekeeping, note that Denard threw only once from a goal line set, scrambled out of the only Fritz passing play, and had just three called passes from the Pro-Set—those are charted as I-form, I-form, and Ace, respectively. DSR=downfield success rating, which is catchable and above divided by attempts, with the "Other" stuff (scrambles, throwaways, batted, pressure) not counted. On to the…
Is there actually a difference chart:
Shotgun (260 attempts)
I-form (45 attempts)
Ace (33 attempts)
TOTAL (338 attempts)
Catchable (CA or CA+)
Inaccurate (IN or INX)
Bad Read (BR or BRX)
Yards per play
Downfield Success (DSR)
I bolded a few outliers in there. Before we get to the big thing, notable for both under-center formations is the scrambling goes way up. I sadly don't have data as to whether this is a quirk of shotgun formations but if you close your eyes and imagine a pass rush on a shotgun versus an I-form you can visualize the veracity of this. I would guess this has more to do with defenses familiar with Michigan's shotgun-spread running game being more responsible against the base formation. I don't have proof so I shrug, but I plan to watch more closely in the future to see if defenses are likelier to open up running lanes when Michigan goes to more Manballian offenses.
The big thing: Holy Ace on a bender Batman! that ridiculous downfield success rating from Ace sets is ridiculous. The inaccurate balls and bad reads are way down, and there were no balls thrown to covered receivers because of pressure. Yards per play were way up (in the I-form they're down from average). This was done in just 33 tries (less than 3 per game), but note the I-form attempts are generally in line with with Denard's shotgun efficiency. From here on out we're not talking about under-center, but specifically Ace sets as the formations that turn Denard Robinson into Drew Brees.
The comfort comes from Robinson's high school career, when he took snaps under center around 60 percent of the time in a Wing-T offense and would often run a waggle play from which he would pass.
Waggles is it? We chart waggles. Waggles:
All Plays (338)
Not waggles, except there's a lot more scrambling (makes sense). It also makes sense that this knocks down his accuracy since the thing all Denard observers can tell you is he's not all that accurate on the run.
Maybe it's the Pressure Getting to Him?
The idea here is that shotgun passes get more pass rush because that's the formation used when the offense must pass and the defense knows this. Protection metrics for each formation on a per-play basis:
It looks like the Ace formation faced more rushers total (equivalent to an extra blitzer every 8 or 9 plays or so); it's just that the offensive line has done a far better job of picking up that rush. That may just be on the type of personnel deployed for each (personnel only shown for pass plays):
So the Ace is where Michigan is deploying two tight ends more often than not, or at least it was last year. It should mean, in general, a pass play from the Ace will have an extra blocker, hence the higher protection metric.
Interestingly, the one thing the Ace doesn't have going for it is the threat of running. While the shotgun makes the defense defend against the bread 'n butter zone reads and Denard runs, and the I-form will faithfully throw Rock 56% of the time, lining up in the Ace formation usually means pass. This is probably why it's getting such a high pass rush.
I've got another theory for why Denard's passing goes up in the Ace, though, and it comes back to what Borges was saying in the presser that Rothstein was quoting: opponents aren't preparing their designer blitzes for Michigan's third-most used formation. The Ace in this offense seems to be performing at an extremely high level because it's a changeup. That goes to situation as well. Here's the formations on distances of 6 or more for called Robinson pass plays:
The heavy lifting is being done by the shotgun, and when Michigan throws from the Ace it's on 1st or 2nd down. It was deployed just twice in 4th quarters—both times early last year. One was a play-action deep post against San Diego State that Denard put just a bit too far ahead of an open Roundtree; the second was the go-ahead throwback screen to Vincent Smith that always works:
So that's my final answer. Drilling on his footwork seems to be helping Denard's accuracy reach Henne-an efficiency overall, but it's not being under center so much as being in the West Coast offense's dedicated pass-protection formation on 1st and 2nd downs. More time in the pocket means better chances to collect yards with scrambles, and more passes that Robinson has time to step into. For about three plays per game, Borges has found Denard that time by trading in the threat of his quarterback's legs for the built-in passer protections that made the Ace Two-TE formations a favorite among WCO aficionados.
It's not often enough to distract defenses from focusing on stopping the core shotgun spread offense, but it's enough to act as an extra weapon deployed where it can do maximum damage. And from the results above, you'd have to admit this particular course of Borges-Robinson fusion cuisine has been exquisite!
I feel like I know this offense a lot better now; but wouldn't the lack of experianced tight end blockers and the fact that non-denard run game isn't happening this year very much limit us to spread offense and 2-rb sets this year dictating that Denard continues to face pressure in the pocket against traditional big ten defenses? I feel as if the Lazer screen is a desperate attempt to keep this pressure off of Denard, if only because Borges inexplicably hates this play.
Anyways I hope the emergance of Lazer, Funchess and other checkdown throws help negate the de-evolution of our pocket protection and deep ball. GO BLUE!
hoke:"[Freshman] A.J. Williams a guy who is mainly on the line of scrimmage, learning how to play college football. He's learning how to block fundamentally and technically. He's got great hands, and he can run." still picking things up by the tone of that qoute. I think there's a reason he's not making a big impact yet. I agree he's an asset but he's still a freshman blocker; I mean Kalis hasn't even seen the feild it's a bit early to expect williams to control the edge rushers on the line.
Borges is calling the formation. Also, just because you have two tight ends in formation, doesn't mean they will block for very long if at all. The oh wide open play to Funchess, he was tight, on the line and the LB just didn't see him immediately run off the line on a pass route. And the tight ends last year weren't the best blockers. So I guess I am wondering why you think Borges isn't calling the play when the stats exist?
He's averaging 117 yds/gm. That would be 1521 yards on the year. I would argue that, given our current coaching staff and their propensity for the WCO, we are lucky that his rushing is even this high so far.
Exactly. I really wish that the Air Force game wasn't so tight so they could have experimented a little more against a decent, but not great opponent. I am not sure what they did in the UMass game because it isn't reflected above and I'm not sure that would have been the best test subject anyway. I would like to see them try this a little more on some weaker opponents down the stretch, if they get up a couple of scores. The problem is Illinois and Minnesota seem to be the only teams that fit the bill the rest of the season.
"And when we play as a team, when the old season is over, you and I know, it's gonna be Michigan again, Michigan."
Looking at the chart, I don't see why Denard is on top. His rating is worse, his completion percentage is worse, and his number of TD's and picks are worse. It should be organized by QB rating; even if some picks weren't his fault, they were still picks.
I sorted by yards per attempt. Rating would have been the other but as I mentioned below rating is thrown off by having four INTs already when he's really responsible for like 1.5. When we were comparing Tate to Threetsheridammit one thing that jumped out to me was how the YPA matched what we saw as better passers much more so than other stats.
Assoc. Editor & Business Manager, MGOBLOG email me for advertising | Alias: @Misopogon
If you're going to dismiss 2 or 2.5 of Denard's interceptions, then I think you need to watch every other interception thrown in the Big Ten and figure out how many were the other quarterbacks' faults, too.
Personally, I don't think those interceptions should be dismissed. The throw to Smith was high, and the Roundtree thing happens. He threw 15 picks last year; it's not like he's been super accurate and now we need to explain away an abberant glitch.
This early in the season we have to be very careful about which stats are being thrown off by the things that "just happen." QB rating strongly values touchdowns and interceptions, and those are good stats after 8 games to judge by, however after three games when half of the stats are known to be just plumb bad luck why bother putting any countenance into them?
There's a lot of other crap in these stats right now that are playing havoc, for example Reilly O'Toole putting up Peyton Manning numbers against Charleston Southern (which accounts for most of his performance so far). At least with YPA there have been hundreds of trials already so that the things that muss it up (example: a screen that goes for 84 yards) are buttressed by 700 other passing plays.
I'm not trying to explain away stats after the fact, but find a plausible answer to "who's the best passer in the conference right now," and sorting by metrics that are not useful until the middle of the season doesn't help with that. If you figure UMass offsets Alabama the YPA of Denard is the most solid number I can pull so far.
Assoc. Editor & Business Manager, MGOBLOG email me for advertising | Alias: @Misopogon
I understand that, but he was a turnover machine as a freshman. He raised his completion percentage as a sophomore in 2010, but a HUGE part of that was the simplicity of Rodriguez's play calls (bubble screen, rollout hitch, play action slant...and that's about it). Then he was a turnover machine in 2011 and had a low completion percentage, followed by being a turnover machine in 2012 with a low completion percentage (so far).
Past performance suggests that his four interceptions and low completion percentage are not aberrations; that's simply who he is as a quarterback, and it's why nobody projects him to play quarterback in the NFL.
So while I know your stats only refer to the current season, I think it's being very...uhh...homer-ish/selective to suggest that he's the best passer in the Big Ten and/or that these mistakes won't continue.
If all we had to go on were the raw statistics I'd agree with you. But then that's why we have UFR -- so we can go beyond the raw and chart each pass by its accuracy and put that in context. The question isn't just about stats but about repeatability, and since the late part of last year the better stats show a marked improvement from Denard. You seem to be hanging your hat on the completion % and interceptions while I'm pointing at this glorious wealth of much more detailed passing metrics that show a clear improvement in accuracy since Denard started getting more comfortable with Borges and vice versa.
I was as resigned as anybody else to the "that's who he is" (I've got plenty of Spartans around me who rarely let a conversation go by without dissing Denard's accuracy).
More data are needed and I'm not ready to bet my NFL franchise* on Denard becoming the next Michael Vick, Donovan McNabb or Steve McNair (all of those guys in college were excellent at throwing on the run, for one, and more importantly they could all float an accurate deep ball which Denard hasn't shown yet he can consistently do) but there is a whisper of hope, from the myriad deadly accurate passes to the remarkably improved field vision that suggests he is good enough to be the best passer in a conference full of what looks to be terrible passers. That's all that thing at the top says (and why I bracketed the chart with a quote from Gladiator and a knock on Big Ten QBs).
*Actually I am but that's only to get rid of the Fords.
Assoc. Editor & Business Manager, MGOBLOG email me for advertising | Alias: @Misopogon
Well, I guess that might be where our ideas diverge. I don't always agree with Brian's UFR and how he doles out positives, negatives, catchability ratings, etc. But I'm also not going to ding a quarterback for straight-up drops like Roy Roundtree had serious issues with toward the end of the 2010 season.
I have stated in the past that I think Denard is "wildly accurate." He often has put throws in places that are catchable but don't allow for receivers to run after the catch. In 2010 he had a good completion percentage (somewhere around 63%, I believe), but so many of his bubble screen throws made guys jump and dive to catch that it really didn't do much good. For example, the Vincent Smith throw/INT vs. Air Force was "makeable" but would have been very difficult. To me that play is 90% on Denard, but others (including you, apparently, and I believe also Brian) think that the 5'6" Smith should be able to reel in high passes over the middle.
I know there are sayings that "If you can get one hand on it, you can get two hands on it" and "If it hits your hands, you should catch it" but those sayings are coachspeak (to which I don't subscribe) and not very realistic.
TDs/INTs I can see, but why is completion percentage more important than YPA? Consider the following hypothetical stat lines:
14-20, 140 yds
10-20, 180 yds
Here, the first QB is completing 70% of his throws to 50% for the second, but the second one is getting nine yards per pass to seven for the first. Do you really think the first is having the better game?
Craig Ross is shaking his fist at you right now. He considers YPA to be the single most important offensive statistic there is, let alone for a QB.
I think people are focusing too much on the ranking of Denard as #1 and not enough on the top 9 put up by Seth and the Sparties. The point is, no B1G QB is covering himself in glory right now, so if you squint at the data hard enough you can sort of make a case for Denard's being the brightest of all the tires on fire.
Yards per attempt is probably the least important QB statistic on that list
Wow, I strongly disagree here. I think YPA is the best measure out there. It answers a very straightforward question: how productive is a QB when he throws a pass? QBs with high YPA are throwing - and completing - challenging throws. Because those throws may be difficult, his completion percentage may not be great, but the payoff when passes are completed is large. One 50-yard completion moves the ball more than six 7-yard completions.
Completion percentage, in contrast, can often be dependent on play calling. If a QB is simply asked to throw dumpoffs and screens all day, he can put up a pretty good completion percentage but may not be doing enough to move the chains. MSU's Andrew Maxwell to date is a prime example of this. In the ND game, he completed over half his attempts, but averaged just four yards per attempt, which is horrible. Time and again he completed passes short of the first-down marker. If it's 3rd and 7 and you complete a four-yarder, that helps your completion percentage but doesn't do much for the team.
Before we get to the big thing, notable for both under-center formations is the scrambling goes way up. I sadly don't have data as to whether this is a quirk of shotgun formations but if you close your eyes and imagine a pass rush on a shotgun versus an I-form you can visualize the veracity of this.
We can also add the factor of time. When in I-form and Ace sets, there is often more protection (as the OP says). When defenders are rushing past 5-apples, they tend to lose their gap discipline (unless they are on Alabama's team). Once gap discipline is lost (one rusher gets loose--Denard sidesteps--or instead a defender get's pushed around by a double-team--now there is a running lane) Denard scrambles and gets loose in the secondary.
Also, blitzes that get picked up (defenses rush more players against Ace) can also lead to running lanes. This could be by alignment--once the blitz gets picked up, there is a hole in the defense which can be exploited--or because of unrealized expectations--defenses know they are blitzing and they expect to get there. That is, they expect to get the QB or force a quick throw. When the blitz doesn't get there, it is more of a broken play from a defensive perspective so more players might ad lib.
Excellent piece of work. My gut feel without looking at the numbers is that Denard is throwing better than he ever has. His mechanics seem better and so are his reads. He is still a work in progress but the trend is definitely looking good. Saturday night will tell us a lot more.
I'm still confused on the usage of yards/attempt as the most important statistic in judging who the best passing QB is and in that, he's only leading by .3
Obviously, the Michigan schedule with the best team in the land and the worst team in the land is enough to throw off any chart.
When looking at passing, I'd look at either completion %, passer rating, or even total yards before I'd look at yards/attempt.
When I ask Sparties who they think is the best passing QB in the big10 (or even the best QB in the big 10), the ones that I know have no hesitation. Their first answer is "NOT DENARD!!" and their 2nd answer is usually Braxton Miller, even though they can come up with no apparent reason for feeling that way.
I don't think the Michigan schedule can possibly work against Denard here. Pretty much every Big Ten team has played at least one terrible opponent. UMass is not that much worse than New Hampshire, CMU, etc. OTOH, Denard has faced by far the toughest opponent of any QB in the conference in Alabama.
How were the projections calculated? Is there a reason his numbers aren't projected linearly? Are we expecting more passes in B10 play compared to the first 3 games? I'm apprehensive that we have seen enough to say that Denard's made marked improvement from last year, but hopefully he can prove it in the rest of the season.
So...you're giving him credit for a touchdown pass that wasn't a touchdown, and you're saying he's the best quarterback in the country despite all his turnovers, two pick-sixes, and poor completion percentage? But hey, he ran all over Air Force and UMass, so that's all that matters...
Sorry, I misspoke, I meant to say Denard is the best player in the country not just the best QB.
Not only that, Denard is the best player I've ever seen to don the maize and blue. Denard does things Desmond and Chuck could only dream about.
Devin Gardner (God bless his maize and blue heart), takes it a step further: Calling Denard: ,"best player I've ever seen", after Denard's improbable right angle, physics defying, incredibly athletic cut against UMass, leading to yet another spectacular football play by Denard. The Stuff Of Legends
I think there are multiple issues involved that have unfortunately been combined.
It looks to me like your measure of success is Denard's accuracy not the play result.
Denard appears to be more accurate in the Ace rather than in I form or in shotgun. While I can see the logic behind splitting under center and shotgun apart to evaluate accuracy; the footwork is different, the sightlines, etc. I do not see the logic behind splitting I form and Ace and judging accuracy. The only difference is the other personel which should not affect Denard's intrinsic accuracy.
If you wish to split formations like that then I think you have use as your measure of success the result of the play because formation involves the other players and play call and not just Denard's mechanics.
It is my belief that if you look at I form plays you will see big plays and TDs go up, which explains Denard's apparent decrease in accuracy. He is going for a bigger chunk on those plays.
Since this blog is so insistent on criticizing I form, I wonder if the data hasn't been subconsciously massaged to reflect that belief.
If, as you say, the I-form was making up for the accuracy inefficiencies with more big plays then that should be reflected in the YPA. It's not. And that's with Michigan running out of that formation 56% of time. I tried them together, and I tried them split up, and when I split them up I immediately saw that the yards per attempt and the accuracy rating for the I-form matched the Shotgun, while the Ace rocketed off.
That was when I realized it's probably not footwork but more time to throw that was probably causing it. Look down at the protection metric and see how clean the O-line kept Denard when we went Ace, despite more pass rushers.
I kind of had an inkling with this since I've done the Shotgun vs. Ace vs. I-form thing from the UFR database three times in the last year and knew the Ace had some weirdly high effectualness (if that's a word).
When I did the "Football hmmm" series over the summer I delved deeply into what I thought were the strengths of an I-form "MANBALL" offense. The great glory of the I-form is that it gets a running back moving forward with a lead blocker and momentum. If you have Aaron Shea lead-blocking for Anthony Thomas while betting that Steve Hutchinson and David Brandt are coming out ahead in their one-on-one matchups, then wheeeeee until somehone hits A-Train and he falls forward (he always fell forward) for an extra two yards to give Michigan its 1 millionth 2nd and 4. If you're trying this with Vincent Smith and Stephen Hopkins...
I don't like that you've accused me of 'subconsciously massaging' the data, for one because I'm not Brian and have my own damn set of biases and zealotry, and two because I was very consciously going in trying to find evidence that MANBALL was actually working, or specifically that the footwork coaches talk so much about--cadenced by drop-back passing--was perhaps a key to Denard accuracy. Instead I got surprised: it's the West Coast stuff that's actually having great success.
I think Borges is with me on this. Think back to that goal line formation we used so much early last year, when Michigan would line up in the I-form and then motion Smith to the flanker, thus turning the formation into an Ace set with Hopkins as the 1-back. What Michigan was doing with that was making the defense go "Big" with its personnel, but then not running Big because we don't have that personnel yet. This has been the crux of most of the formation hand-wringing from both me and Brian, not that the offense itself sucks but that most of Michigan's starting personnel from 2011 through 2013 are guys whose strengths tend toward a spread running game, and whose weaknesses are more apparent in an ISO game. If we had 5 Lewans and De'Veon Bell, or, say, Michigan's 1999 personnel, I'd be singing hymns of rock-pounding. What we've got is Toussaint/Smith, Mealer/Schofield/Barnum/Omameh and, you know, Denard.
And most importantly, please note that the entire crux of this article in the end is that Borges is doing a magnificent job with the fusion cuisine. He has come much further toward the spread than he's asked the offense to come to him. And also note that this particular very effective changeup we're discussing is the exact strategy that he is an expert in: West Coast (not MANBALL). Michigan is 75% the offense that the players were recruited for, 12% the offense that Borges literally wrote the book on, and 13% the offense that Brady Hoke would have us become again someday. If this offense was a pitcher, it's a guy with a wicked unhittable 2-seamer, a changeup with perfect mechanics, and a regular fastball that on its own wouldn't scare anybody but in context works as well as a the 2-seamer. In other words it's Max Scherzer; is there any problem with saying that Scherzer shouldn't try to be Justin Verlander?
Assoc. Editor & Business Manager, MGOBLOG email me for advertising | Alias: @Misopogon
In a normal academic situation I wouldn't have overtly suggested a misreading of the stats or suggested my own hypothesis because that was rude. I might have just asked why you split the I form and Ace and if you had a hypothesis for why that would increase accuracy. I apologize for that. I only did it because it is my experience that you have to put stuff out there to get people to respond. Unfortunately, the internet is a little less civil in that way.
I think that by seperating I form and Ace and using accuracy as a read out you are observing the effect of another factor's effect on accuracy. In other words, I cannot think of a reason why Denard's accuracy would improve with that formation switch unless you were measuring something else--such as pressure (as you suggest) or maybe play action or situation.
Knowing from the original article that Denard has thrown 12 TDs from under center (and only 20 total TDs in 2011), I recall many of those under center TDs coming specifically out of I form. I believe that Borges knows Michigan throws better from under center and runs better from shotgun but also knows he can't let formation tip his hand.
I am surprised that YPA is not higher with I form but what about YPC? Maybe the lower completion rate is masking that effect. I don't know, I'm just kind of waving my hands here.
It's not just tendencies, but also capabilities that matter to a DC. Let's say you're the OC and I'm the DC. If you line up in 2 TE's on 1st and/or 2nd down, my main priority is to stop the run. I'm curious as to how many of those passes from Ace were play action.
In other words:
Run Down + Run Formation + Play Pass = High Success Rate
Yards per attempt might be the best metric, but QB rating might also be, for which Denard is 7th. INTs are INTs, and Denard amasses them in large numbers every year. Not convinced by your argument ... Yet. But I am hoping you're right. And I can see that Denard is a better passer this year as I watch the games.