Wow, that is quite an impressive list of quarterbacks. Hopefully Gardner's success continues!
is there such a thing as an etsy genuis? if so, this is it.
Since 2003 only two quarterbacks have been worth two full touchdowns, 14 points, above average* over the course of the entire season. Colt Brennan’s 2006 season at Hawaii was the first and Tim Tebow became the first major conference player to do it in his Heisman Trophy 2007 season at Florida (Michigan did their part to stop him, holding him to a season low +5.9 in Lloyd Carr’s final game).
*adjusted for strength of opponent’s throughout the entire study
Last year in five starts Devin Gardner’s PAN was, you guessed it, over 14 points. Gardner started out on fire, averaging +17.2 in his first three starts at quarterback against Minnesota, Northwestern and Iowa. Against Ohio State and South Carolina, he was still in double digits, but his final average ended at +14.7 for the season. Could this be the prelude to a world class 2013 season for Gardner or just a hot hand off of the bench?
To look into this possibility I looked for every quarterback since 2003 who has produced a five game streak (1AA games excluded) of at least +14 like Devin Gardner did last year. I wanted to understand how common a five game run at this level was and if there was any parallel between a great five game stretch and overall great quarterbacking.
In addition to Gardner, 28 other quarterbacks have accomplished the feat. The others on the list are a virtual Who’s Who of the last decade of college quarterbacks. Three players did it who are still in the NCAA. Johhny Manziel, Marcus Mariota and Tajh Boyd managed the task and are mainstays on preseason All-American lists. Of the 25 other players who have done it and have moved on with their careers, 13 have started an NFL game and 10 are projected starters for the 2013 season. I can’t think of any other college stat that could predict NFL starter status at an over 50% rate. The group of players who have done what Devin Gardner did last season includes the following NFL starters:
Aaron Rodgers, Green Bay (Cal, 2003)
Alex Smith, Kansas City (Utah, 2004)
Ben Roethlisberger, Pittsburgh (Miami (NTM), 2003)
Cam Newton, Carolina (Auburn, 2010)
Colin Kaepernick, San Francisco (Nevada, 2010)
Matthew Stafford, Detroit (Georgia, 2008)
Philip Rivers, San Diego (NC St, 2003)
Robert Griffin, Washington (Baylor, 2011)
Russell Wilson, Seattle (Wisconsin, 2011)
Sam Bradford, St Louis (Oklahoma, 2008)
Denard Robinson, Andrew Luck, Christian Ponder and Mark Sanchez all just missed the cutoff with 5 game runs averaging +13.
Add to them first round selections and former starters Matt Leinart, Tim Tebow and Brady Quinn and you have an over 50% likelihood of becoming an NFL starter based purely on your best five game stretch.
Of the remaining twelve cases, only four came from a Big 5 conference school. Graham Harrell did it in 2008 for Mike Leach at Texas Tech, Nick Florence did it last year for Baylor, Zac Robinson did it in 2007 at Oklahoma State and former Michigan quarterback Ryan Mallett did it in 2009 at Arkansas.
Every player is going to have ups and downs but based on where the other players have gone that have accomplished 5 game streaks on par with Devin Gardner’s five game run as a starter, this is not a fluky performance accomplished by many. Those that have done it at major programs have a high likelihood of an NFL future.
I honestly have no clue. I do feel confident that he should be pretty darn good. Quarterbacks who have had a streak of +14 have maintained very high performance even after such a streak so regression to the mean is still a highly positive outcome in most of these situations.
The hardest thing to get a grasp on how Devin Gardner projects is that his situation is so unique. On the one hand he was a five star quarterback coming out of high school stuck on the bench behind a Michigan icon. On the other hand he never looked great at quarterback until the five game stretch. He is more suited to what Al Borges is looking to do than Denard but is still more dual threat than Borges’ preferred traditional drop back style.
Adding to the uniqueness of Gardner’s situation is the fact that no other player made the list in his first five starts. There are hardly any that had a five game +14 streak in their first year as a starter let alone in their first five starts. Every one of the players on list did in the context of a full season as starter and of course none of them spent the prior eight games at wide receiver.
I do think Gardner should be one of the top couple quarterbacks in the Big Ten next season, at the very least. The big question that this raised is how high is his ceiling. After looking at the company and the challenges that came into it for Gardner I think it’s safe to say the ceiling is officially gone. I have looked at college football statistics every which way for years and no look has stood out to me as much as this one in its ability to translate to future success. Hopefully that bodes well for Michigan’s seasonand Gardner’s pro potential.
Wow, that is quite an impressive list of quarterbacks. Hopefully Gardner's success continues!
Pull up to the scene with my ceiling missing.
After hearing all of the praise Gardner is getting in the offseason, I've been trying to temper my expectations before they go through the roof. Then here you come with this and I'm back to square one.
I am trying to stay grounded, because stuff like this makes my expectations get out of whack.
I'll preface this by saying I'm asking mostly as an attempt to keep my expectations from going full-on tween-girl-at-Bieber-concert.
Does anything in your data suggest that Gardner's performance may have been artificially inflated by a "surprise factor"? The Gardner offense was a significant departure from the Denard offense, and Denard's status was unclear throughout Gardner's run. Even going into the South Carolina game, I think the Gamecocks had to operate under the assumption that a return of Denard was at least plausible.
This would at least in theory make preparing for Gardner more difficult than it would normally be for a QB of comparable skill.
On the flipside of that argument, maybe Gardner's inexperience makes that a wash.
How many (if any) of the other QBs had their high PAN streaks at the beginning of their careers as starters?
He did state that none of the other QBs had the high PAN streaks in their first five games, and very few in their first season as starters.
I guess I missed that on the first read-through. I wonder who the "hardly any" players who had it in their first year are?
Rodgers and Newton did after starting at Jucos. Mallet did it his first season at Arkansas. Leinart did it as USC and Tebow did his first year as a starter but saw lots of time the prior year. I think that's it for the major conference guys.
Manziel and Mariota would also qualify. I don't know if that was assumed, but it does bring the total up to 8 (the two mentioned in the previous sentence, two in the previous post, and Gardner) of the 29 accomplishing the feat in their first year. Out of curiosity, how many players on your list managed to accomplish the feat in multiple seasons or independent 5-game runs?
Teams may have needed to spend time preparing for Robinson even when Gardner was starting, but I'm not sure there would be much of a surprise. Our offense with Gardner under center was a fairly "normal" college offense, so it was probably similar to the offenses our opponents had played against for much of the season. If anything, the Borges/Denard offense was the "surprise" one, since there aren't many teams that feature a pro/spread hybrid with the QB as the primary rushing threat.
Like I said, I'm intentionally reaching a little. But what I'm getting at is that some things you'd do schematically to stop Denard (e.g. bring a safety up) are exactly the things that Gardner could uniquely (for Michigan) exploit (e.g. F IT I'M GOIN' DEEP + Gallon cloaking device).
Sure, the Gardner offense was more "pro style", but who and what you'd key on varied greatly between the two QBs. I feel like that would make it hard to switch back and forth for a defense - it's tough to get out of your gameplan. With Denard, you had to cheat the secondary up to keep him from hitting a home run. The go to receiver was Roundtree, then maybe Gardner and Funchess. With Gardner, the long ball was much more dangerous and you had to focus on Gallon. Denard was coming at you from a tailback alignment more often.
I'd say Alabama the last couple seasons has demonstrated the value of extra prep time - Saban seems to be a master of sculpting his already very good defense to uniquely destroy what you're good at (but it's tougher to do that in the regular season).
So I'm reaching, but it's not like surprise is completely implausible...
Norfleet can be our denard this year...or justice.
I don't think it's as much a "surprise" factor as a simple amount of preparation factor. In college, you only have so much time to prepare a defense for an opponent. When that opponent has to spend a good amount of time preparing for an offense that ultimately wouldn't be used at all, that's a big factor. And remember, before each and every game, Denard's availability was completely under wraps. Nobody knew if he would play, if he could throw, anything. So while Michigan's coaches knew there was no way Denard was playing as a true QB and could practice entirely in packages more suited to Borges' style, opposing teams had to prepare for both. This was a huge advantage of our offense, one that I think has gotten a little lost in the glitter of Gardner's stats over 5 games.
As an aside, a second fact that I think tends to get lost when staring at Gardner's stats, especialy in advanced stats, is that while efficiency stats said Minnesota and either NW or Iowa (I can't remember which) were excellent pass defenses, anyone with eyes could see that they were in fact awful. Big Ten passers were atrocious last year, they made big ten pass defenses seem better than they were, statistically. An average QB would look fabulous when compared to his in-conference brethren. I mean, Matt McGloin tore shit up. Matt McGloin!
I think Gardner has the potential to be really great. But if one wants room for skepticism in his numbers playing 5 games last year, you can find it.
But, I think what counters and trumps the surprise factor is the fact that up until about 11pm the previous Saturday, Gardner was a WR. I know he played the position in high school, but for over 6 months, he was studying and practicing as a WR.
I rewatched the Minnesota game this past weekend. Gardner looked pretty lost in the beginning of that game. That is absolutely not a knock on the guy, because most athletes could not have done what he did. You could actually see his handle on the position return during the course of that game. So, in his first 5 games after playing at WR (and at a high level too), Gardner had this streak.
I think of how sometimes guys really struggle with transitions to just a different spot on the same offensive/defensive line. But, since little nuances specific to each position on the lines exist, that makes sense. What Gardner did does not make sense. That kind of productivity without the focused practice and studying for the QB position is absolutely incredible.
I'm not going to hold him to an unreasonable standard. But, after what he was able to accomplish with so little preparation last season, I feel their is literally no cap on what this guy can do.
this gave me chills... thanks!
HEISMAN BABEH! ...right?
Those are the odds I'm seeing online. I've been to a few books so far that don't take Heisman bets though (guy behind the counter claimed it could be rigged since it is a vote, as if point-shaving or match-fixing are totally impossible in a game). Not sure where you can place an actual bet but Gardner is to me the best option on the board with that kind of big payout.
The NFL sees a guy go off for five games, and remembers, so they'll get more opportunities to start than someone who doesn't go off like that.
That said, these guys have had a lot of success at the highest level, and sustained success in college outside of their streaks, so this is probably not a big issue. An elite QB will make a Rose Bowl berth more likely ...
The problem with using NFL starts... is that the NFL realizes when a guy does well? What?
And that wasn't even the mathlete's main point. His main point was that the QBs who have managed this feat (of which there are very few), they were all extremely successful, some in the NFL but certainly all in college. That was more a "fun fact" than the core of the argument.
and something that has always jumped out at me is that in his final h.s. campaign, his team began the season with three losses and had to win out just to qualify for the playoffs. They did so under his leadership, including busing down to Ohio and beating a team that hadn't lost a home game in something like 65 games. Dividing that number by two, you get roughly a stretch of 12 to 13 seasons in which they enjoyed that home winning streak. That, to me, shows tremendous leadership of the type that Leach, Franklin and Harbaugh also possessed. No we can't guarantee our predictions but this kid is a talent. Even though it was pointed out that he qualifies as a "dual threat," he would have w/o doubt been recruited by Lloyd as well as RR.
Is PAN adjusted for strength of opponent? Because, you know, Iowa.
But still, OSU and S. Carolina.
this Gardner fellow is pretty good?
On one hand, I think those PAN numbers for Gardner are inflated by some lucky plays (eg, the Roundtree catch that had no business being a catch at the end of the NW game).
On the other hand... guys, we are totally going to wreck house this year.
The assumption is this type of thing happened for the others as well. Think Manziel fumble-to-myself-holyshithesopen TD against Alabama
That assumption would make more sense to me if it weren't attached to a 5 game sample. I bet the PAN numbers are really sensitive to a big plays like that. Projecting this level of success just makes me suspicious.
If I wasn't so beaten down by a long day at work, I'd combine UFRs with Mathlete PAN charts and get to the bottom of this.
Well, since PAN is a measure of deviation from the mean, large game values for PAN will almost by definition have to contain some unusual plays. A streak of good luck is almost a necessity.
That said, for "lucky plays" to push you into really high PAN values, you have to be starting from somewhere pretty good to begin with. Which is my guess for why this stat is catching mostly elite QBs - getting a +14 PAN probably requires a combination of consistent very good play combined with a few good breaks.
@Mathlete: might be cool to see a table of all 28 QBs with their best 5 game PAN streak along with their career average PAN. I'd be curious who's closest to the +14 line on average, and for whom their +14 streak was most out of character.
What is this stat?
copied and pasted from another mathlete article:
"Points Above Normal. PAN is essentially how many points above an average FBS team was a team/unit/player worth."
On the other hand he never looked great at quarterback until the five game stretch.
He never had any playing time before the stretch, good or bad. Spring games - playing second string snaps to Denard - should be doused with so much salt as to make them worthless. Beyond that we hardly saw him. Please explain why someone of your understanding of sample sizes continues this meme.
"Spring games - playing second string snaps to Denard - should be doused with so much salt as to make them worthless."
You're right. They should be taken lightly and they probably are (by Mathlete); but that's all we had to go by. It's a fact to hardly argue: he looked poor. It's also a fact in that it was minimal snaps and not fair to use as an end-all-be-all foreshadowing (your point). But to say it's a tiresome meme isn't really fair. I don't think it was always a point to put Gardner down neccessarily but it was a point, nonetheless, as far as trying in any way to figure out "what we had" in our next QB (especially when Denard went down), no matter how little salt should've been sprinkled upon it.
Backing up Denard may have been little more than ancedotal evidence, but it's evidence all the same. And up until the back half of last season, it was the only evidence we had.
The fact of evidence being extant doesn't required it be used. This was very poor evidence, so much so that it should not be used to form a conclusion.
What's more is that it's not even that accurate, beyond sample size limitations. In his first and only game action in 2010, Gardner was 7-10 for 85 yards and a TD, and another 25 yards rushing and a TD. His passer rating was 174.4.
In 2011 he was put in multiple games, but never exceeded 7 pass attempts. This included MSU, where he heaved a pass after having passed the LOS. I think this play caused a lot of angst, but I saw Kirk Cousins make the same mistake in his sophmore or junior year. Also included in that 2011 sample is a critical TD toss against Illinois - a great play under tremendous pressure.
All of that adds up to some good, some bad, not enough to go on.
Even more unique to Devins situation is that he hadn't even been practicing at QB by the 8th game of the season, to the point that he wasn't used at Nebraska when Denard was hurt and Bellomy desperately needed a reprieve, then he promptly goes on a 5 game +14 streak. Game 8? not even considered as a qb option with the B1G championship on the line, game 9? in the running as a top qb nationally.
And add to that he was fully capable of playing qb against Nebraska... In Hoke's time here, that is the one and only thing I question.
"He is more suited to what Al Borges is looking to do than Denard but is still more dual threat than Borges’ preferred traditional drop back style."
I think - perhaps because historically the QBs that can run have been used as run first guys - a lot of fans see rushing ability/athleticism as something that invariably must subtract from passing ability somehow. This idea that Gardner is "more dual threat" than is ideal for the system.... why? Does his scrambling ability take away from his arm strength? Does it make him less accurate? Does it make him shorter? More inconsistent? Less of a team leader? If you ask me, Gardner's athletic ability does one thing: It buys him more time to find open guys. It is only a liability if he tries to run when he shouldn't.
You don't have to be Peyton Manning or Tom Brady to be a good pro-style QB. Aaron Rodgers can move. Andrew Luck can move. Steve Young and John Elway were all kinds of mobile. I'm not saying Gardner is or will be any of those guys. But his legs aren't going to stand in his way if that's where his arm and his head are able to take him.
Now if Borges shrugs his shoulders and says, "Well Devin can run, so I guess we'll just call a bunch of zone reads for another year and not call west coast/pro-style stuff" then that's another story. But is there anyone here that believes this is what we can expect to see in the fall? I don't think so either.
It's a sports trope, but a pretty accurate one. Run-first QBs usually aren't asked to throw much in practice, game situations, or off season workouts as their offense is typically built around the option or even zone read. You're right that "run first" doesn't always mean "can't pass," but if you look back at all of the mobile QBs in college over the last few decades, how many of them flamed out in the NFL or weren't even drafted? How many of them had RGIII's mid-deep ball accuracy or pocket awareness?
Gardner, though, should turn out to be another recent exception to the "rule." He definitely looks comfortable in the pocket and has the athleticism, mobility, and technique to throw on the run. Bottom line, I'm not worried.
I don't think of Gardner as "run first." I think of him as a pro-style QB that happens to be mobile. I guess the short version of my previous post is that a lot of us seem to think running ability and "run first" are the same thing. That's when legs become a perceived liability. But I don't think that's Gardner. I don't think Borges sees Gardner that way. I don't think Gardner sees himself that way.
And yeah, historically athletic QB's have suffered as passers because they play in offenses that are built around their legs. This is less and less the case in the modern game, and won't be the case for the Michigan offense any longer. Coaches are starting to realize that being fast doesn't automatically mean you can't throw and/or don't have the brains to read defenses and find open receivers.
I like your work, Mathlete, but I wish (and it seems from several of these comments that I'm not alone) you would re-explain PAN in the first paragraph of each of these posts or at least link to a previous explanation of the methodology. It's hard for me to critically engage the topic if I'm fumbling halfway in the dark.
I understand this quote superficially ("Points Above Normal. PAN is essentially how many points above an average FBS team was a team/unit/player worth") but the algebraic formula is what I'm wanting.
Maybe it's just because of the positive outlook on our QB situation, but this is one of your best columns, Mathlete. I'm a big fan of your number crunching in general, but this one did a great job of connecting tangible results to a number based probablity of success.
Is there a PAN for dummies? How do you isolate one player in a team sport and figure out how many points he is worth above normal? Not being sarcastic...just slow.
Low quantity of math but high quality of information as always from the Mathlete!
prediction: undefeated and national champs this year.
Don't believe it?
JUST DO THE MATH, IT WORKS OUT
I am also bullish on Gardner this season. I am also a big fan of your body of work, Mathlete. However, I do have some questions about the selection of data used for this analysis.
It's obvious why you chose a 5-game streak as the cutoff point for deciding which quarterbacks to compare Gardner with -- that's the length of Gardner's streak. However, you are lumping the good (those with 5 game streaks like Gardner) with the great (those with presumably longer streaks). The same is true for the +14 PAN cutoff. Once again, Gardner barely makes the cut, but he is compared to those who presumably have much higher PANs over their 5-game streaks.
Perhaps it would be more instructive to look at those quarterbacks who have had streaks of 3 to 7 games, with PANs of +10 to +18. That way, you could look at who makes the cut and then say "we can expect Gardner to be the average of these people."
Maybe I'm making this too complicated, but you're a complicated guy, Mathlete, with a nose for precision. I think you'd be up to the task . . . .
reading that took me from 80s neck roll to full pylon. Season cannot get here soon enough.
The pessimist in me begs the question: Is there any possible way we are over valuing the defensive prowess of our final opponents last season?
I would love to think DG is as valuable as Tebow was in his Heisman season, but man... that seems like a reach.
*Not trying to be a dick to Mathlete because his stuff is sound and great.
Run for yardage and Run to Pass. Devin is a passer first and formost. Think of the play last year where he scrambles around for enough time that Drew Dileo is standing by himself in the endzone 45 yards away. Devin did many things in those 5 games that our offense has been missing. How about drawing the defense offsides with a nice cadence and taking the free shot for a bomb downfield? At least once this season that will be 6 points for free. I am also really excited about the development of DG. This pro style offense will also not allow opposing defenses to put 8-9 guys in the box and will help open up our running game as well. Having a running attack that actually scares people will open up the play action, and that is my favorite part of Borges' offenses. His playaction is deadly, son. Another thing not addressed in the comments above is that all of these games were B1G opponents or better. You have to assume that outside of Notre Dame, DG will tear it up in the non-conference games this year. His first 5 games stats will exceed his last. I expect some crazy fireworks in the opener, and we won't be throwing 6 INT's the following week either. I still can't believe we were even in the ND game last year after turning the ball over that much. (how the hell do you make paragraphs now, btw? Hard return and tab do not work and if I make nice paragraphs in the plain text editor it switches back in rich text...)
One thing I do know is that Gardner but up good numbers without much of a running game. Once we finally establish a running game I think he is going to be great with the play action pass.
It's obvious that most posters here earned a degree from an institution that prizes critical thinking (and that's a good thing), because so many people are trying to blow holes in the data.
NEVER forget, that data is just data, and data itself is NEVER causal, though it may uncover correlaries and causal possibilities.
But look at our schedule: does anyone think that our AVERAGE opponent is going to be worse than the opponents we faced in that five game stretch? No way. Iowa is pretty bad, Minny was decent, NW was good (not great), and Ohio and South Carolina were top tier teams. Those five games were not disproportionately easy. And, based on the figures given by the Mathlete, DG averaged a PAN of approximately 10.95 vs. Ohio and S. Carolina. That's pretty damn good.
Will Devin Gardner be an NFL starter in two years? I don't know, and neither does Mathlete. The point is that the data is extremely compelling, and that it indicates that DG's success was NOT a fluke, but is likely to be sustained, plus-or-minus some deviation. There is NO WAY to organize the data that will give you a definite or guaranteed prediction. SO JUST ENJOY THE DAMN DATA, OR DO YOUR OWN STUDY.
Mathlete, this is an excellent analysis and brilliantly written.
DG makes a run at the Heisman this year, I think he is speacial player and from what I have been reading has worked really hard on his weakness this off season.
Michigan has not has a QB this talented and with a good head on his shoulders since Henne.
I think we all will regret him not starting Hokes first year.
Kinda OT but just saw this on Grantland - mostly fluff but still enjoyable read. I can't post new threads so somebody take it and run with it.