On Thursday, Dan Mullen spoke about quarterback accuracy during his Thursday press conference. Some readers took that as an opportunity to point out to me that we disagree on this topic.
@WillMilesSEC mullen doesnt agree with your accuracy stats on qbs😂 https://t.co/hVuEr0fYK7
— 🐊🐊🐊🐊mullen era🐊🐊🐊🐊 (@joshgatorgann) April 4, 2019
— Unique Productions (@Unique863) April 5, 2019
First off, it’s okay to disagree. I’m not always going to be right. And while he’s forgotten more about football than I’ll probably ever know, Mullen isn’t always going to be right either.
But I don’t really think we disagree here.
Mullen’s point during the press conference was that improvement in accuracy can’t just be measured with completion percentage. Instead, he focuses on ball placement on each throw and how to improve that.
I agree with that assessment, but maybe not for the same exact reasons.
Is completion percentage an important statistic?
I don’t think there’s much debate here.
Completion percentage is a really important indicator of QB success. We can see this clearly if we plot completion percentage versus QB rating from 2005-2018 for QBs with more than 14 attempts per game.
Completion percentage versus QB rating of QBs from 2005-2018. (Will Miles/Read and Reaction)
There’s a pretty clear correlation between the two statistics. Better QB play does track with completion percentage.
Indeed, if we look at the 252 QBs who had QB ratings higher than 150 – which represents very good, but not elite, QB play – only 5 of those QBs (2.0%) had completion percentages below 59.2. Only 12 had completion percentages below 60.
The two red data points are for Feleipe Franks in 2017 (triangle) and 2018 (square). What this indicates is that Franks’ completion percentage in 2017 was completely in-line with his QB rating while he outperformed his completion percentage in 2018. There are QBs who have had lower completion percentages and better QB ratings, but not very many.
The importance of completion percentage also is observed for winners of the Heisman Trophy. Nine of the last ten winners have had a completion percentage above 66.1.
The one player who did not is Lamar Jackson in 2016, when he completed 56.2 percent of his passes. Of course, Jackson also ran for 1571 yards on 260 carries, which is a better season running the ball than Ciatrick Fason had for Florida in 2004 (222 carries, 1267 yards).
The average completion percentage for those past ten Heisman Trophy winners is 67.3 percent. It seems like that indicates that it’s a pretty good stat to keep track of.
Accuracy vs. Decision-Making
Completion percentage is a measure of how often a QB completes a pass, which must mean that it measures accuracy, right?
Well, it depends on how we define accuracy.
If we define accuracy as the ability to give the receiver the ability to run after the catch, completion percentage doesn’t really measure that.
An example of this is shown on this play, where Franks double-clutches and then delivers the ball behind Kadarius Toney. Had the ball hit Toney in stride, this is an easy first down but still is likely only advanced another 3-4 yards down the field. That’s a significant thing as far as the game is concerned, but is not reflected in completion percentage.
Another example is shown here. Franks correctly diagnoses the play and finds a wide open C’yontai Lewis streaking down the sidelines. The ball is thrown on a rope and just slightly too far and Lewis makes a fantastic diving catch. This is a huge play, but had the ball been thrown just slightly better, Lewis might have had an opportunity to cut back inside and score.
In both of these cases, we’re defining accuracy as the QB putting the ball exactly where he wants it so that his receiver is put in an advantageous position. Both of these balls were completed, but they could have been completed in a more accurate fashion.
So that’s one definition of accuracy. But what if we define accuracy as the ability to make completions easy – where ball placement isn’t absolutely critical – instead of having the fit the ball in perfectly.
Well, I do believe that this is highly correlated to completion percentage. And it has a lot more to do with decision-making and reading a defense than it does the ability to throw the ball to an exact spot on a frozen rope.
An example of what I’m talking about can be seen in the above play. Franks is in trouble, throws the ball off-balance, and throws a little bit of a duck. But Josh Hammond is so wide open that even though the ball hangs, it is still a big completion.
Now, had Franks hit Hammond in stride, this may have been a touchdown. But because he was throwing to the correct receiver who would be open against this coverage, he had a much larger margin for error. It’s an interception if it’s not thrown to the correct receiver instead of a completion.
This play is just before the diving catch by Lewis. It’s second-and-6, so there is value to gaining just a few yards. Franks is able to avoid a sack by chucking the ball into the sidelines. But he misses a completely wide-open Lamical Perine as a safety valve. It’s an easy throw, and likely an easy first down.
We can see the same thing against Florida State. On this play, Franks fakes the QB run (which had been effective just the play before) to move the coverage to the left. He then comes back to the right against the zone and hits a wide open Trevon Grimes. Again, there was a large margin for error on this throw.
It is absolutely true that a QB is going to have to hit difficult throws that he must fit into tight spaces and place perfectly. But it is also true that there will be plenty of opportunities where the QB can defeat the defense with his mind – by diagnosing the correct place to go with the ball – instead of with his arm.
Completion percentage has a difficult time measuring the first. I think it does an excellent job of measuring the latter.
Can accuracy and decision-making improve?
I have no doubt that Mullen’s assertion that the way to improve accuracy is to improve ball placement is true. I also suspect that it’s improving a player at the margins because the decision-making aspect of accuracy is an inherent trait rather than a taught one.
Percentage of QB’s with a certain change in completion percentage from high school to college. (Will Miles/Read and Reaction)
The above chart is an analysis of 50 QBs whose high school stats I’ve been able to find. It compares their completion percentage in their senior year of high school versus their college completion percentage. The average high school completion percentage of those 50 QBs was 62.0. The average in college? 62.1.
We do have to recognize that there are outliers. But only 10 percent of those QBs have seen an increase by more than six percent. Based on this sample – admittedly limited – over 60 percent of QBs will end up between 3 points less and 6 points more than their high school completion percentage.
This has been true for Feleipe Franks in his first two seasons. Franks had a high school completion percentage as a senior of 58.9 percent. He has completed 56.8 percent of his passes at Florida, and completed 58.4 percent under Dan Mullen last season, exactly where you would expect him to be.
The argument that he will show improvement in this area like Drew Lock or Matthew Stafford is convincing on its face. Lock’s completion percentage in college improved from 49.0 to 54.6 to 57.8 to 62.9. Stafford’s improved from 52.7 to 55.7 to 61.4. However, Lock had a completion percentage his senior year of high school of 66.0 with Stafford at 64.9.
Basically, those players were finally able to reach their potential while it appears Franks may have already reached his. Of course, many people also point to the dearth of coaching that he received in high school and then until Mullen came into the fold at Florida. Perhaps that means he’ll be an outlier.
After all, one of those outliers was Nick Fitzgerald, who saw an increase in completion percentage of over 10 points.
Takeaway
This may seem like a bunch of semantics, but it’s actually pretty important. That’s especially true when you are trying to identify whether a high school QB that your team is recruiting is going to be a game changer for your program.
I was able to find high school completion percentages for eight of the past ten QBs to win the Heisman Trophy (couldn’t find for Robert Griffin III or Sam Bradford). The relationship between college and high school completion percentages is striking.
Accuracy of eight of the last ten Heisman Trophy winners in high school and college. (Will Miles/Read and Reaction)
I can understand why you maybe didn’t see Baker Mayfield being as good as he turned out to be at Oklahoma. But we should have known that he would be at least a good QB with the ability to put up a QB rating of over 150.
In fact, I got this wrong last year thinking that Oklahoma would take a step back under Kyler Murray. My rationale was that Baker Mayfield was so good that even if Murray was great, he couldn’t be that good. It turns out he was, especially when you factor in his running ability. Had I looked at his high school stats, that would have told me something different.
None of this means that you can’t have a very good QB who doesn’t have an elite completion percentage. But it sure makes things a lot harder.
Mullen is going to make Franks better by working on ball placement and timing. He’s going to make the offense better by using the threat of a running QB to open up other aspects of the offense and make reads easier. All of those will help Florida move the ball and helped the Gators move the ball in 2018.
But regardless of how you want to define accuracy, there’s no doubt that completion percentage as a statistic is clearly related to QB effectiveness.
CGator
Will, I agree with you mostly here. I believe accuracy has a lot to do with body makeup. Everyone is different, and some people have the natural mechanics that lead to accuracy, and some people have to work on it and never have the natural accuracy. You can improve it on the margins, as your stats indicate. That said, I have to ask exactly what Mullen meant. If ball placement means getting the ball to the open guy, yes that can improve completion percentages, but as you show, you can do that with some poorly thrown balls. For instance, teaching a QB to make back shoulder throws can improve completion percentage, but that’s adding a throw to the arsenal to make a more effective passing attack; you still have QBs who make that throw very accurate, and others who are less accurate. But if by ball placement he means the QB consistently putting the ball in a better place to facilitate a completion, and YAC, then isn’t he back to talking about mechanics? QBs with naturally gifted mechanics are better able to put the ball in the right location; teaching those without it To improve seems, generally, to bring improvement on the margins. As your stats show, QBs seldom make big improvements in completion percentage over what they did in high school, suggesting a natural limit, or at least restriction, on their ability to be accurate.
Will Miles
Mullen was talking specifically about ball placement, not the receiver to which the ball is delivered. I do think there is something to what he said about not letting one bad throw turn into three. And while Heisman winners typically have high completion percentages, there are plenty of players at below 60% with QB ratings over 140 who can help win games.
Thomas Moss
What stuck out to me is better ball placement as Mullen describe talking about the mechanics when the QB has time to plant and throw can be improved. Taking that logic. Yards per completion should increase in my opinion. Not sure if that is measuarable. Mullen also stated throwing a ball off balance or while running is more about natural talent.
I had a knowledge HS coach switch my swing in baseball. I could argue it made me worse as my muscle memory since little league was changed. So I agree that improvement after high school is not a likely outcome. Look at Tebow. Maybe his slow release may have had a better outcome in NFL then his updated quick release .
If you have a QB with the talent to play even if he is partying ever evening. I think that QB chance to improve through knowing the offense is more likely. So it may be possible Frank may improve, but I doubt it because we do not have a proven OL. He could be a lot better and I think but his number may go down.
Last year I was more oppomistic than most, but I think I’m more pesstimisic because we may have a kid just out of High School going against an NFL player next year.
The key like you said is definition and objective.
Stebow
It seems derivative to say that completion percentage is an important indicator of QB success based off its correlation with quarterback rating, when completion percentage is the most heavily weighted component of QB rating. I realize the difficulty in finding a metric to compare to completion percentage to determine its contribution to success at the position, but I do not think comparing QB rating to its main components is very meaningful. I wish I could end this comment with a suggestion to improve the analysis, but I can’t think of any readily available data that would be appropriate. This difficulty alone leads me to believe there is a lot of merit in Mullen’s viewpoint that accuracy is a very difficult attribute to quantify.
P.S. I love the high school to college completion percentage comparison. I never would have guessed how little variance is present, especially to the upside.
Will Miles
They are definitely related, but the question I’m asking isn’t really whether they are related, but whether the error bars are big enough that you can be an “elite” QB without the decision-making skill that related to completion percentage. If there were a way to decouple, that would be great. But that’s part of why I included the Heisman-winning players because there is one who won without completing a high percentage and he won because of his running ability. It’s not a perfect analysis. Best I can do with the time I have =).
Tom
Will, I love your article and the stats you put together in your opinion pieces, but I really don’t feel you have a grasp of what developing a QB means with regards to bringing all the other 10 pieces together with the QB so that in the end you have an offense that’s both efficient at moving the ball and finally putting points on the scoreboard. Franks is a flawed QB that if not for the physical measurable’s he possessed coming out of HS…..had no business even sniffing a 4* rating.
Its Franks head that worries more than anything, from a intellectual and psychological standpoint going into next season. …….physically Franks is extremely gifted! There will always be a fine line between the serviceable Franks and the Franks that was potentially going to be benched, if not for the injury to Trask last season. This is where stats for a player like Franks are nothing more than a means to put a QB’s play into perspective. I believe Mullen understand this, and is just another example of his genius in as far as being able to maximize a given QB’s makeup….”the good, the bad, and the ugly” The bad and the ugly will always be just on the cusp of rearing its head with Franks, the lure of his physical tools however is worth the effort to help Franks over come those negative attributes…….and I believe the results prove that out to be true, at least last season.
Ball placement has everything to do with increasing the potential of positive outcomes. I say outcomes because it can mean any number of results, but not necessarily always all 3 every pass play. The 3 outcomes with ball placement break down as follows: 1.) better results with lowering INT potential 2.) greater probability of a receiver simply catching the ball 3.) putting the receiver in better position for adding yards after the catch. I think its important to pint out that 2 and 3 don’t always go hand in hand depending on the specific play. Sometimes based on the coverage all your looking for is the completion, and then other times the completion plus yards after the catch matter……Understanding what matters for a specific situation is about teaching through repetition in as far as ball placement is concerned. This is what I think what Mullen is trying to drum into Franks head, because it doesn’t come natural for Franks
Chris Varney
Nice breakdown Will! Work em silly!
Marshall Printy
1) it would be better if the 50 QBs chosen were somehow representative of starting college QBs as a whole (i.e. a random sampling).
2) The question is whether there is a relationship between HS completion and college completion. In other words what is the r^2 of the data, or what % of the variance is attributable to this relationship. Can you circulate what that r-squared is and also share the scatterplot (vs. just bar graph that you have)
3) Lastly, I think a big factor in completion % is also the scheme an offense runs and what % of passes are shorter passes vs longer passes. Presumably the QBs with higher completion % and lower QBRs are throwing more short-distance passes. More than anything, I want Franks to improve his YPA as I think that’s a more important metric than completion % per se.