# Follow Up on QB Fragility

Submitted by MCalibur on August 6th, 2009 at 11:07 PM

After reading the reactions to the first diary I wrote on the issue of QB injury rates, I thought it’d be worthwhile to go back and address some of the skepticism still lingering.  For the most part, the source of the skepticism was in regards to the number and scope of the data points or in how the QBs were split into their Threat Level categories.

Addressing the first issue was straightforward—get more data. I included the 2004-2007 seasons for a total of 5 seasons including 2008. The scope was all FBS schools and all starting QBs with starters defined as players who averaged more than 17 plays (passes + runs) per game they played in. This level of play was chosen based on the median of the data in that category. One tweak I added was to increase the resolution of the data to include partial games (ie. if a QB gets knocked out in the 3rd quarter and missed the next two games, I counted 2.25 games against him). I also corrected a few errors I found along the way.

Cherry Picking

The issue of the chosen break points was simply a matter of choosing one arbitrary line in the sand over another, I went with the one I thought made the most football sense. I actually considered using the quartiles of the data as break points to ensure that there would be equal numbers of QBs in each category. But, we know that this simply isn’t the case—there are way more Jimmy Clausen/Mark Sanchez/Matt Staffords (Threat Level 0) than there are Pat White/Terrell Pryor/Tim Tebows (Threat Level 3).

The goal of using categories was to group similar players together in an unbiased way. Using whole number denominators (1 run per 2 passes, 3 passes, etc) was a reasonable football-sense based method to do this. Using quartiles would have put the level 3 split data at 2.78 passes per run…huh? Ok so maybe you round up to three then but a) how is that not arbitrary and b) does it make sense to put Steven Threet into the same category as White/Pryor/Tebow? Obviously not…

Regardless of all that, the expanded dataset neutralizes this concern

The Goods

Anyway after all the grunt work was done, not much changed but the extra data does yield more insight. The average injury rate for all QBs was 23.6%. Level 3 running threats came in at a 25.3% rate and level 0 threats were injured at a 23.9 % clip. I’ve left my hypothesis testing skills mercifully nestled amongst the dust and cobwebs of my mind but there’s no statistically significant difference here. The theory that option QBs are more likely to get injured than the average QB is a load of crap.

There does seem to be a significant deviation from the norm in the injury rates of level 1 and level 2 QBs, with 1’s getting injured at a lower than average rate and 2’s getting injured at an above average rate. A possible explanation for this is that Threat level 1 QBs enjoy the safety of the pocket but also have the ability to escape from pressure and not be sitting ducks like threat level 0 QBs. On the other end of the injury spectrum, level 2 QBs leave the pocket but don’t have the speed/agility/field vision/down field linemen necessary to elude defender’s looking to teach them a lesson. This is definitely an eye opener. If you’re going to ask your guy to be a running threat on a regular basis, he had better be a slippery one and you’d better have blockers on the second level. Check and check.

The last interesting note is that 1 out 4 teams can expect to lose a QB for about 3 games; that’s 3 teams each from the B10, SEC, ACC, etc. I would have never guessed that. Another way to look at it is that a recruit can/should be expected to miss 3 games over a 4-year college career.... yikes. Having at least 2 game ready backups is an absolute must, regardless of offensive system.

 Threat Level No. of QBs Injured QBs Lost GMs(% of Total) Avg. Games Lost QB Inj % 2008 3 26 6 7.5% 3.08 23.1% 2 24 10 7.7% 2.33 41.7% 1 46 11 9.2% 4.89 23.9% 0 64 14 4.8% 2.79 21.9% 2007 3 19 7 7.3% 2.5 36.8% 2 26 7 5.5% 2.57 26.9% 1 35 7 6.5% 4.04 20.0% 0 79 19 6.9% 3.61 24.1% 2006 3 14 4 7.4% 3.19 28.6% 2 34 13 10.4% 3.38 38.2% 1 36 4 4.6% 5.31 11.1% 0 60 13 5.1% 2.96 21.7% 2005 3 19 4 5.5% 3.00 21.1% 2 29 8 8.5% 3.47 27.6% 1 36 6 3.0% 2.08 16.7% 0 60 16 6.2% 2.72 26.7% 2004 3 17 3 3.6% 2.33 17.6% 2 28 4 1.6% 1.25 14.3% 1 40 6 3.5% 2.67 15.0% 0 63 16 7.4% 3.39 25.4% All Seasons 3 95 24 6.3% 2.81 25.3% 2 141 42 6.9% 2.81 29.8% 1 193 34 5.6% 3.88 17.6% 0 326 78 6.1% 3.13 23.9% All 755 178 6.1% 3.3 23.6%

Brace for Impact

Given the rhetorical skill of most of the visitors of this site, I’ll take a second to put out a preemptive strike against the “how do we know Tate/Denard/Devon wont be a level 2 Threat?” argument. Simply put, we don’t. But, the categories aren’t an explicit evaluation of the athleticism of the player; they only allow us to infer his physical skills by how he’s deployed by the coaching staff.

Example. Juice Williams ended up as a marginal level 2 QB in 2008 with an R/P ratio of 0.46. The previous 2 seasons he qualified for level 3 status which is not a surprise. His skills didn’t diminish and the offensive scheme of the Illini did not change but for whatever reason, he ran a few time less than before and ended up in the upper end of the same category as Steven Threet. Actually, he ran more often (+2 runs per game ’07 v. ‘08) but passed a lot more (+11 passes per game).

Now on the flipside, does anyone believe that Threet would have had a better chance of staying healthy if we had run him more? The point is that the athleticism of the QB determines his injury risk outside of the pocket, not the number of excursions outside of the pocket he makes.

GET THE TABLES!!!

Seriously, this is excellent work. Thanks for posting it.

Thanks for putting in the hard work. Given that we've all read so much about the death of the spread, it's interesting that the 2008 total of 26 level 3 QBs is so much higher than the 17.25 average from '04-'07.

I understand that using quartiles would group together QB's that weren't really on the same threat level, but did you consider smaller groupings (e.g. 10%, 20%, etc.)? It seems like this might be another interesting way to look at it.

All in all, I really like the analysis.

About the athleticism of the quarterback having more to do with his susceptibility to injury. I'd also add in the measurable intangibles of vision, instinct, and feel for the game. I think a quality QB knows when to dump the ball rather than taking a hit, sliding to avoid tackles, and not "forcing" a play when one is not there.

... and thanks for exposing another myth to the light of facts.

Many fans of the spread have felt this way all along, esepcially your last point; a QB who is out and running should have a better chance of avoiding injury than a stationary one who takes ten or fifteen (or even as many as thirty if he passes a lot and his line sucks) "free shots" a game in the pocket under the current rules.

Great diary. Must have taken a while.

I like this. And it makes sense that it would be safer to either hide in the pocket or be a really, really good runner. Trying to split the difference is just going to get you caught in a situation you're not prepared to handle.

Excellent work.

Good analysis, I think you did about as well as realistically possible. One thing I would think about adding to the analysis is: how did the offense perform after the starting QB went down and how did that compare with the offense's performance prior to the starting QB getting hurt. In other words, are spread option teams more significantly hampered by their starter going down (see Dennis Dixon, Pat White in '07) than an offense with a traditional pocket passer QB going down? Maybe a simple yards/game with the starter compared to yards/game after the starter chart would give a rough approximation.

Your data for this next step could be easily skewed. Pat White and Dixon were both All-American talents, so either getting hurt would be more about losing a star player than just a spread QB. Same thing if Matthew Stafford got hurt last year, the back-up would be a big step down. I guess it would greatly depend on whether it is easier to be a decent spread or pro-style QB, since your depth chart would come into play.

I suppose this begs the question - is it feasible to have a decent back-up QB in a run-based spread option offense or is it just too difficult of an offense to have a back-up QB step into and have the offense not lose much production. In other words, is it much easier for a traditional pocket passer QB to step into the offense than it is for a run-based option QB? Also, given the most prominent examples (Dixon and White), is it more of an issue of these teams not being able to recruit enough highly rated guys to serve as possible back-ups? In other words, does a program like Michigan that can recruit more depth at QB stand a better chance of handling an injury moreso than a program like Oregon or WVU.

Too many variables to control obviously but I'm just throwing out ideas.

I actually thought about this but frankly, I don’t think this warrants a whole lot of further study; we could pretty much guess the answer. This question is not as simple as what type of QB is more replaceable. A QBs success is not just dependent on his physical ability but also on his experience level, reps with the first team, and so on. Also, a hidden assumption is that the backup is equally suited to execute the offense as the starter is.

In Oregon's case, they went from having a Threat Level 2 (Dennis Dixon) to a Threat Level 1 (Jeff Roper, Brady Leaf). I don't show Pat White missing any games due to injury in 2007 but in 2006 and 2008 he missed 1.25 games in both seasons due to concussion. In both instances he was replaced by Jarrett Brown who was also used as a Level 3 Threat but his Yds/Pass and Yds/Run were significantly lower than when Pat White was in the game, no surprise there. WVU won both games he did not play in. Which is, after all, the objective.

So to answer your question properly we need to look at teams with adequate depth with equal talent level and with equal experience level. The starter is the starter for a reason--the coaching staff thinks he's better or more ready to play; probably both. It stands to reason that the offense will be less productive with a back up at the helm especially coming in cold turkey off the bench. In instances when production picks up the argument is that the staff had the wrong guy starting in the first place. Finally the backup will never have as much experience as the starter because...you got it, he's the back up.

But, I will say this. The difference between Oregon/West Virginia and Michigan/Florida is the respective ability to recruit actual depth (similarly talented players) in case of emergency. Schembechler forbid Tate goes down; Denard should be able to keep the offense going, though passing accuracy in the spread option is no trivial matter (see Steven Threet). Going further, Devin Gardner will further stabilize the position next season.

I think its safe to assume that both D. Robinson and Gardner will both be level 3's (1 run for every 2 passes).

But as for Tate...I think with his above average accuracy, that he will not be called upon to run as much and will be kind of forced into Level 2 (similar to what happened to Juice last year, but for different reasons, of course).

I believe that maybe his freshman year, he will be a level 3 because he is still getting to know the passing routes. But by his 3rd year in this offense, he will be passing more than we ever saw from Pat White under RR and Tate will be a level 2, maybe even a level 1 (similar to Colt McCoy, there's that comparison again!).

But I'm not sure that you are measuring the effect that you intend to. (However, I'm not sure how to do what I would propose in a statistical sense.) My complaint centers around the notion that you brought up that Steve Threet is not the same kind of QB that Juice Williams is, but they had similar run/pass ratios last season. I believe that threat level should be determined by the athletic ability of the individual QB, not some statistical measure that could be misleading. I believe that a run/pass ratio for a QB will depend on the system he is running. Rich Rod will never recruit a John Navarre type player at QB, because that skill set is not what RR desires in a QB. Guys like DRob and DGardner, would imo have been recruited at WR under Carr (like Steve Breaston--not that he had other teams chasing him as a QB like these guys did/do). Carr wanted a more traditional (threat level 1 at most) guy at QB in the Peyton Manning/Tom Brady mold.

My hypothesis is that injuries go up when a QB is not suited to the system he is playing in. For example, Threet got injured because he was running around last season doing his best "Pat White imitation". (I know, I couldn't tell either.) I also believe, but have no anecdotal evidence, that a mobile QB stuck in a drop back system will scramble rather than throw the ball away and live to play another down. He'll scramble out of his protection and the defense will get to him more often when he is unprotected leading to more injuries.

Not to beat a dead horse, but this was my complaint with RR and his offensive scheme last season. I understand that he is a "spread guy" and he wanted to run "what we know" for an offense. But he knows a pass-first spread offense like the one he ran at Tulane with Shaun King. Threet is much more like Shaun King than Pat White. I think, especially in the short term, that would have added another win or 4 to the team's record last season. Then with Tate and DRob on campus this season, we have a decision to make at QB, coming off of a stronger season and some real competition. However, it could be argued that with Threet gone, Tate and DRob will get more development in. As a result, these guys will be better prepared this, next, and the following seasons...It might have been worth it.

Good stuff here. I really think that Michigan is not expecting TF to run all that often, but you want the threat of a run. TF really needs to be more like Fran Tarkington (very old reference) in that he almost always was able to scramble to buy a little extra time before he threw the ball. TF needs to make the first rusher miss him, roll out, and occasionally run. These will make the passing system hum like a well oiled machine. A more experienced O-line will also help him greatly.

Mcalibur = hero

They get hurt the least often but when they do they get the worst injuries averaging almost another whole game lost.
I dont' know why but i would guess it's because they only really get hurt when they venture further ouside the pocket than their ability can handle. Their lack of elite athletisicm results in worse injuries. They get hurt less often because they don't leave the pocket as often.

First, lets give this man some points. This is the type of thought-provoking, labor-intensive post that the up arrow was invented for.

Second, I am not totally sold on whether you can draw a lot of conclusion from this data. I would be particularly cautious about drawing conclusions from the "tiers."

For example, the tier 1 quarterbacks could easily be pocket passers that take a lot of sacks (which is usually credited as a run, unless you accounted for that). If you look at something like that and see that they are generally injured for longer, it could easily be because they are getting sacked all the time and not actually trying to run that much more than your average, Tier I pocket passer.

That being said, I was shocked to see how often the quarterback ended up injured. It makes me wonder what the NFL numbers would be for something like this. When you look at a guy like Favre (probably a Tier I guy who did a lot of scrambling to create room to throw downfield) who went so long without missing a game, it makes you wonder what it is that makes some people borderline invincible, willing/able to produce despite an injury, and/or both.

Nice job, MCalibur.
And regarding the +/-6 point deviation from the overall mean in categories 1 and 2, it seems minor enough that I agree with your overall conclusion that the differences aren't statistically significant.

My recollection of past games and watching video makes me think that one of the reasons dual-threat QBs don't get injured more is their willingness and "comfort level" taking a hit --- Of the few times I remember Henne and Mallet (sorry - short term memory) running the football - they either slid to avoid the hit (which puts their heads in the wrong place if a member of the D follows through?) or were otherwise gubbied when they got nailed... all tensed up and etc... I know folks have mentioned this latter part before - but dual-threat quarterbacks - like running backs seem a lot more willing/comfortable in taking a hit (or giving one potentially...) and while I'm sure they tighten up as well, I would wager it isn't quite the same terror factor...

Where did you pull the raw data from?