An interesting finding regarding FCS QBs

[InfusedStormlight]

Wall of text incoming.... If you only care about the numbers I have tables below, they are fairly self-evident.

I was playing around with some stats and found something interesting regarding my team's QB (Elon), Towson's QB Tom Flacco, and JMU's QB Ben Dinucci, among others. I was thinking about how hard we and other teams like Albany have had a hard time with sacks this year (Elon and Albany both have 34 sacks, although Albany did play three more games). IWU had 40 in their 12 games. But I was curious if we could normalize the number of sacks by number of pass attempts. This makes sense: if you pass more, you can expect to be sacked more, on average. So I divided sacks into pass attempts and came up with the following results for the top 20 QBs by passing attempts, plus Tom Flacco and Davis Cheek.










Name
Attempts
Sacks
Attempts/Sack


Davis Cheek
309
34
9.1


Tom Flacco
356
35
10.2


Jon Copeland
507
40
12.7


Jeff Undercuffler
464
34
13.6


Ben Dinucci
345
24
14.4


Breylin Smith
490
34
14.4


Reece Udinski
479
32
15.0


Dalton Sneed
438
29
15.1


Eric Barriere
438
27
16.2


Zerrick Cooper
439
27
16.3


Kenji Bahar
448
26
17.2


Aqeel Glass
445
24
18.5


Vito Priore
455
24
19.0


Bailey Zappe
560
25
22.4


Shelton Eppler
545
24
22.7


Trey Lance
277
11
25.2


Case Cookus
481
17
28.3


Reid Sinnett
375
13
28.8


Javaughn Craig
406
14
29.0


Chason Virgil
475
16
29.7


Daniel Smith
396
12
33.0


Kevin Thomson
450
13
34.6


Jake Maier
492
13
37.8



Attempts/Sack looks very similar to Comp. %, too:
31298

These results mean that Davis Cheek (Elon), Tom Flacco (Towson), and Jon Copeland (IWU) had the hardest time being sacked, with Cheek only having ~9 passing attempts per sack.

But I wanted to take this further. As you can see in the graph above, I figured that something that's tied to number of sacks is completion percentage, because the more pressure you have the more likely you are to either scramble or throw it away. Now, sacks don't automatically mean you throw the ball away more, you could be taking the sack, which I'm sure does happen sometimes, so this isn't perfect. But I think it does give a sense of how well a QB did at completing passes given the amount of pressure they have, and can give more context to completion percentage. Here's the results:

















Name
Comp%/(Att/S)


Davis Cheek
6.37


Tom Flacco
5.94


Ben Dinucci
4.96


Jon Copeland
4.64


Breylin Smith
4.48


Dalton Sneed
4.34


Reece Udinski
4.27


Jeff Undercuffler
4.14


Kenji Bahar
3.74


Zerrick Cooper
3.64


Eric Barriere
3.63


Aqeel Glass
3.31


Vito Priore
3.14


Bailey Zappe
2.85


Shelton Eppler
2.76


Trey Lance
2.67


Reid Sinnett
2.32


Chason Virgil
2.21


Case Cookus
2.13


Javaughn Craig
1.99


Daniel Smith
1.81


Jake Maier
1.73


Kevin Thomson
1.70



31299

Now, perhaps a better way of looking at these stats is normalized by the field's max and min. What this essentially does is mean that, e.g. Ben Dinucci's 71.3% Comp. % is better than everyone, so he's 1.00, and Jeff Undercuffler's 56.5% is 0.00. Here's the results if you normalize the data, then add Norm. Comp. % and (1 - Norm. Att/S). I subtracted Norm. Att/S from 1 because a lower Norm. Att/S means harder and higher means easier, so it makes sense to invert the results so that the person with the hardest time is at 100 percentile and 1.00. Adding them gives you sort of a composite score for how







Name
(Norm. Comp. %) + (1 - Norm. Att/S)


Ben Dinucci
1.82


Dalton Sneed
1.40


Breylin Smith
1.36


Reece Udinski
1.30


Kenji Bahar
1.26


Tom Flacco
1.23


Trey Lance
1.16


Davis Cheek
1.09


Jon Copeland
1.03


Bailey Zappe
1.03


Reid Sinnett
1.02


Aqeel Glass
1.00


Shelton Eppler
0.94


Zerrick Cooper
0.93


Eric Barriere
0.91


Chason Virgil
0.89


Vito Priore
0.87


Jeff Undercuffler
0.84


Jake Maier
0.60


Case Cookus
0.59


Javaughn Craig
0.38


Daniel Smith
0.38


Kevin Thomson
0.27












Finally, here's the list sorted by highest to lowest for (Norm. Comp. % + (1 - Norm. INTs)) / (Norm. Att/S). This composite score is essentially just a view of how





Name
(Norm. Comp. % + (1 - Norm. INTs)) * (1 - Norm. Att/S)


Ben Dinucci
1.38


Reece Udinski
0.94


Tom Flacco
0.86


Trey Lance
0.76


Davis Cheek
0.72


Kenji Bahar
0.70


Eric Barriere
0.69


Dalton Sneed
0.58


Jon Copeland
0.52


Breylin Smith
0.44


Aqeel Glass
0.43


Reid Sinnett
0.34


Zerrick Cooper
0.32


Jeff Undercuffler
0.32


Vito Priore
0.30


Bailey Zappe
0.30


Shelton Eppler
0.28


Case Cookus
0.27


Chason Virgil
0.24


Daniel Smith
0.10


Javaughn Craig
0.08


Kevin Thomson
0.07


Jake Maier
0.00











This, to me, means that Ben Dinucci is the best QB in the nation at performing well despite his offensive line. Comp%/(Att/S) is a good metric by itself to determine how good a QB is completing passes even though they are pressured a lot, but I think including INTs also helps, since being pressured also tends to make QBs throw more INTs.

I can't think of any other metrics that are directly affected by how often you get sacked, like Comp. % and INTs are. If there were, we could improve this score by adding their normalized data to the last data point above within the parentheses that are multiplied by (1 - Norm. Att/S).

[FormerPokeCenter]

I think you might consider how efficiently QBs spread the ball around. If there are very few people touching the ball, then the QB's sack numbers would be particularly impactful...but if there are lots of people touching the ball, I.E. running backs and recievers, then the impact of the sacks is lessened.

Also, are you adjusting for team that pass a lot, or that run a lot? If primarily run oriented teams have a high sack ratio, I'd offer the opinion that it's because their offensive line might be built for running. In other words, the O line might be heavy handed and prone to firing out, while optimal pass blocking would involve linemen who aren't so heavy handed and who might be leaning back.

If you're going to quantify stuff like this, other data, like average time taken for a play that results in a completion. Are the sacks coverage sacks, i.e., sacks that take a longer time to devlope, or are they qucik strike sacks made because the line isn't blocking.

I'm thinking this sort of analysis is a lot more nuanced than what you're currently tracking....

[InfusedStormlight]

I think you might consider how efficiently QBs spread the ball around. If there are very few people touching the ball, then the QB's sack numbers would be particularly impactful...but if there are lots of people touching the ball, I.E. running backs and recievers, then the impact of the sacks is lessened.

Also, are you adjusting for team that pass a lot, or that run a lot? If primarily run oriented teams have a high sack ratio, I'd offer the opinion that it's because their offensive line might be built for running. In other words, the O line might be heavy handed and prone to firing out, while optimal pass blocking would involve linemen who aren't so heavy handed and who might be leaning back.

If you're going to quantify stuff like this, other data, like average time taken for a play that results in a completion. Are the sacks coverage sacks, i.e., sacks that take a longer time to devlope, or are they qucik strike sacks made because the line isn't blocking.

I'm thinking this sort of analysis is a lot more nuanced than what you're currently tracking....

The truth resists simplicity, for sure. Reality is obviously more nuanced, but without either analyzing each play one-by-one for every QB or having some seriously detailed stats like what you're describing (FCS doesn't have stats like avg time a play with a completion takes) there's no way to get an accurate picture of reality. My analysis at least gives a sense of the general trends with certain QBs, and gives context to stats like Comp. % and INTs.

[FormerPokeCenter]

LOL...there are lies, damned lies....and statistics! ;)

[Noryan34]

Why did you normalize one set of data as the favorable result the 1.00 and the other set the unfavorable result as 1.00. A lower att per sack is worse or unfavorable meaning it would be easier to sack that person.

[BEAR]

Breylin Smith isn’t doing bad for only playing in 4 games his freshman year and getting no protection this year in his sophomore year. I can’t imagine how accurate he would be with some protection and if his receivers would run the right routes! xlolx

[InfusedStormlight]

Why did you normalize one set of data as the favorable result the 1.00 and the other set the unfavorable result as 1.00. A lower att per sack is worse or unfavorable meaning it would be easier to sack that person.

I didn't interpret a lower Att/S as "it's easier to sack them so they must not be as good", I interpreted it as "their offensive line isn't as good which means they are pressured more often". A lower Att/S means the other stats should look more impressive, because they are accomplishing them with more pressure than their peers.

[djohnk]

I'm curious ... what is Trey Lances "Norm. INTs"? With math if you have a zero in the equation it can often mess with results unpredictably.

[Catamount87]

What about tossing in rushing statistics into the mix? With the RPO being so prolific as well as using QBs that are essentially RBs this muddies the waters even further.

[Gangtackle11]

Thanks for the analysis. I’m very happy Dan Smith (47 TDs total) no matter the findings. Kid is a baller. xpeacex

[Herdistheword]

Assuming that a lower attempts per sack means that the line is bad isn’t really correct by itself. The QB could be more responsible for those sacks than the line...Is the QB holding the ball too long by not making quick enough reads, wrong reads, etc?

Sorry, but these metrics mean absolutely nothing by themselves. You are interpreting them with a pretty severe bias as it stands right now.

There is no doubt to anyone who watches NDSU that Trey Lance has a fantastic oline that gives him a ton of time. However, it would be stupid to dock Trey for that when Trey still demonstrates the ability to go through his reads and make correct decisions.

[Lorne_Malvo]

This thread is proof that statistics can be meaningless.

[InfusedStormlight]

I'm curious ... what is Trey Lances "Norm. INTs"? With math if you have a zero in the equation it can often mess with results unpredictably.

Since I'm not using just Norm. INTs, but (1 - Norm. INTs), his (1 - Norm. INTs) is 1.00. Breylin Smith's is 0.00. But also, since I'm just adding this value to Norm. Comp. % instead of dividing by it it doesn't cause any problems.

[InfusedStormlight]

Assuming that a lower attempts per sack means that the line is bad isn’t really correct by itself. The QB could be more responsible for those sacks than the line...Is the QB holding the ball too long by not making quick enough reads, wrong reads, etc?

Sorry, but these metrics mean absolutely nothing by themselves. You are interpreting them with a pretty severe bias as it stands right now.

There is no doubt to anyone who watches NDSU that Trey Lance has a fantastic oline that gives him a ton of time. However, it would be stupid to dock Trey for that when Trey still demonstrates the ability to go through his reads and make correct decisions.

These metrics don't hurt Trey Lance, they simply help some of his competitors, which isn't exactly the same thing. It's not like I would interpret his lower values as bad for him. IMO these metrics are more useful towards the low end of Passing Attempts/Sack, rather than the high end. The innovation here is mostly just using Passing Attempts to normalize the number of sacks someone has, using that as a measure of general pressure, and contextualizing other numbers with it.

[djohnk]

This thread is proof that statistics can be meaningless.
Agreed ... this isn't even really stats, just a creative algorithm with what i suspect has some confirmation bias built in.

[InfusedStormlight]

Now that I think about it, what if we incorporated Yds/Attempt into this? I think of this because if someone is getting pressured more, they're more likely to make shorter throws. But if they are already making shorter and quicker throws like quick outs and fewer long throws, and are STILL being sacked a lot, that says more about the OLine to me. If they are usually making longer throws, that means that the OLine in general can hold up better and the OC trusts them to hold for longer. What do y'all think? Could Yds/Attempt be useful in here?

[FormerPokeCenter]

The University of Louisiana at Lafayette (USL to us old timers) used to tout a formula that used five criteria to evaluate their fan's experience. They called it the Penta-something or other. I always called it the Pentaphlegm...

Basically, they were trying desperately to show that they were on an even footing with the major colleges so somebody devised a system whereby if you stood in a specific place, cocked your head just right and stuck out your tongue, you could potentially view USL as a major college, based on fan experience.

The end result of their math was to rank USL higher than LSU fan experience. To do that, they had to use all of the school's athletic wins, including women's softball, divided by the number of games and number of fans and some other creative math to arrive at a number that somehow compensated for 100,000 people watching LSU dominate in Tiger Stadium.

I'll give them this, it was creative. I mean, I'm sure there's some algorithmic magic that can make those 30 people at a softball game intrinsically equal to 100,000 people at a football game.

I suppose the same sort of creativity can be invoked to arrive at a way to make a QB's sack numbers look like a positive. If this catches on, it won't be long before QB's are intentionally taking sacks to make their other numbers look better ;)

[InfusedStormlight]

Agreed ... this isn't even really stats, just a creative algorithm with what i suspect has some confirmation bias built in.

There is always truth behind statistics, since they are a source of truth. There's a reason they are used. Yes, sometimes they can be useless, but usually I think that's because they have no context and give an inaccurate view of the truth, and common sense can see it. In this scenario, common sense informs the statistics and they are the thing that provides context, so I think they are useful. I'm never going to claim that 4th down conversion % has much to do with Time of Possession, because there's no logical basis for that being a context-driven source of truth about what happened. But Passing Attempts/Sack? Of course that is an indication of pressure. And pressure certainly has an effect on completion % and number of interceptions.

[Gil Dobie]

I'm not seeing opponents defensive rating and strength of schedule factored in. Did I miss something? Lance played against 5 top 25 defenses, DiNucci 2, St Francis and UNI. Not a really great way to us stats to compare QB's, without adding who the opponents are.

Against a common opponent, UNI
DiNucci 19/28, 157 yds, 5.6 avg, 1 TD, 1 Int, 1 sack
Lance 10/18, 145 yds, 8.1 avg, 3 TD, 0 Int, 0 sacks

[InfusedStormlight]

The University of Louisiana at Lafayette (USL to us old timers) used to tout a formula that used five criteria to evaluate their fan's experience. They called it the Penta-something or other. I always called it the Pentaphlegm...

Basically, they were trying desperately to show that they were on an even footing with the major colleges so somebody devised a system whereby if you stood in a specific place, cocked your head just right and stuck out your tongue, you could potentially view USL as a major college, based on fan experience.

The end result of their math was to rank USL higher than LSU fan experience. To do that, they had to use all of the school's athletic wins, including women's softball, divided by the number of games and number of fans and some other creative math to arrive at a number that somehow compensated for 100,000 people watching LSU dominate in Tiger Stadium.

I'll give them this, it was creative. I mean, I'm sure there's some algorithmic magic that can make those 30 people at a softball game intrinsically equal to 100,000 people at a football game.

I suppose the same sort of creativity can be invoked to arrive at a way to make a QB's sack numbers look like a positive. If this catches on, it won't be long before QB's are intentionally taking sacks to make their other numbers look better ;)

Now this is a more reasonable criticism. But again, these numbers (Pass Attempts/Sack) should not be seen as a positive stat for QBs, that's not what I'm arguing. I'm arguing that it gives context to other QB numbers, and the fact that we don't use it to inform our opinions makes some of these other stats disingenuous.

If anyone has any more criticisms of the methodology, I'm all ears. But just saying "stats are meaningless" is ignoring why people say stats are meaningless. They say that because they are lacking context. If these stats provide more context, I see that as a good thing.

[InfusedStormlight]

I'm not seeing opponents defensive rating and strength of schedule factored in. Did I miss something? Lance played against 5 top 25 defenses, DiNucci 2, St Francis and UNI. Not a really great way to us stats to compare QB's, without adding who the opponents are.

Against a common opponent, UNI
DiNucci 19/28, 157 yds, 5.6 avg, 1 TD, 1 Int, 1 sack
Lance 10/18, 145 yds, 8.1 avg, 3 TD, 0 Int, 0 sacks

This is fair, for sure. I'll do some thinking as to how to properly incorporate strength of defense. Thanks :)

[IBleedYellow]

This is literally you picking stats to try and come to the conclusion that you want.

Imma just say this: yikes.

[Noryan34]

I didn't interpret a lower Att/S as "it's easier to sack them so they must not be as good", I interpreted it as "their offensive line isn't as good which means they are pressured more often". A lower Att/S means the other stats should look more impressive, because they are accomplishing them with more pressure than their peers.

Correlation without causation.

Do you have the statistics to prove that sacks are correlated to the quality of the offensive line? if so can you post. if not you then cannot make this assumption

[InfusedStormlight]

Correlation without causation.

Do you have the statistics to prove that sacks are correlated to the quality of the offensive line? if so can you post. if not you then cannot make this assumption

Obviously there's no objective statistic that captures "quality of offensive line". But Att/Sack and TFLs are the best stats we have to directly tie to an offensive line, and sacks are the best stat we have to determine how much pressure is being applied to the QB. Saying "stats are useless" or buck-passing and criticizing without offering an alternative is lazy.

Of course there is some causation between number of sacks and the pressure the QB feels. Do you dispute that?

[Noryan34]

Obviously there's no objective statistic that captures "quality of offensive line". But Att/Sack and TFLs are the best stats we have to directly tie to an offensive line, and sacks are the best stat we have to determine how much pressure is being applied to the QB. Saying "stats are useless" or buck-passing and criticizing without offering an alternative is lazy.

Of course there is some causation between number of sacks and the pressure the QB feels. Do you dispute that?

Without watching tape and grading each play like PFF. It is impossible to determine which sacks are attributable to the offensive line and which are the QB fault.

Your analysis implies that sacks are a offensive line only stat. While thst can be a valid analysis it must be mentioned when talking about your results. Because if you swapped sacks and made them a QB stat your results would most likely be vastly different

[Noryan34]

Another thing simple analysis like this don’t account for is the amount of completions under pressure. Which is impossible to do without watching tale
This is part of the reason I think that PFF has become popular as they actually look at each play

secondly. This analysis punishes a QB for throwing the ball away when under pressure. As it would decreases his completion % but raise his att/sacks

what is the actual goal of your analysis? Determine how easy or hard it is to sack a certain qb? Cuz I think your first analysis is fine att/sack does that

Or is to determine the best an while under pressure?

[PAllen]

Or the QB just doesn't get rid of the ball quick enough.

[caribbeanhen]

college boys gonna college..... I'll just watch games thank you!

[Gangtackle11]

college boys gonna college..... I'll just watch games thank you!

Im with you Carrib.

“Stats are for losers. Final score for winners.”

https://youtu.be/XueXrg6Yj1o

A line that stands the test of time.

[caribbeanhen]

Im with you Carrib.

“Stats are for losers. Final score for winners.”

https://youtu.be/XueXrg6Yj1o

A line that stands the test of time.

I had no idea he said that GT, but in the end you can't really argue with him

[Redbird 4th & short]

Now that I think about it, what if we incorporated Yds/Attempt into this? I think of this because if someone is getting pressured more, they're more likely to make shorter throws. But if they are already making shorter and quicker throws like quick outs and fewer long throws, and are STILL being sacked a lot, that says more about the OLine to me. If they are usually making longer throws, that means that the OLine in general can hold up better and the OC trusts them to hold for longer. What do y'all think? Could Yds/Attempt be useful in here?
yards per attempt are always useful ... but they certainly don't tell the full story. Theyre way better than yards per catch because they better reflect your passing completion % along with yards gained per play (not per catch) .. . and ultimately, are more determinative of the ability to move the chains. What good is going 11 of 20 for 150 yards (impressive 15 yards per catch, 7.5 yards per attempt), especially if that includes 1 catch for 80 yards, meaning your other 19 passes only got you a paltry 70 yards (just 3.5 yards per attempt) with completion % of just 50%. That won't move chains unless you have a great run game.

I think the data is interesting but could use some refinement. But once you collect the data and formulate the metrics ... from there, it always comes down to subjectively assessing how good a QB is in the pocket (feeling the pass rush, amd kaing good throws), how good the OL is at pass blocking, how well you are able keep defenses honest in both the run and pass game, and ultimately whether your OC is play calling to optimize those strengths and weakness.

Too many variables (objective and subjective) and therefore impossible to measure with some subjective analysis to "interpret the results". Identical results across different QBs could be explained 3 of 4 different ways ... our QB is bad wiith pocket pressure but throws fine, our run game sucks, our OC sucks at play calling, etc.

That doesn't mean you shouldn't have metrics like this, it's always good to quantify something .. then you just have to analyze the results.

[veinup]

this data further proves that dalton sneed is an absolute legend, if i’m reading it correctly

[BisonFan02]

Trey Lance is responsible for 41 TDs....28 passing and 13 rushing....0 INTs....****ing ZERO. That is pretty interesting. /thread

[thebootfitter]

Trey Lance is responsible for 41 TDs....28 passing and 13 rushing....0 INTs....****ing ZERO. That is pretty interesting. /thread
Yeah, right now, I don't think there's any FCS QB I'd rather have than Trey Lance. Regardless of what the stats suggest. Of course, I'm a bit biased as a Bison fan, but even the stats point to him being near the top of the heap. And the stats you just posted are the exclamation mark.

[clenz]

yards per attempt are always useful ... but they certainly don't tell the full story. Theyre way better than yards per catch because they better reflect your passing completion % along with yards gained per play (not per catch) .. . and ultimately, are more determinative of the ability to move the chains. What good is going 11 of 20 for 150 yards (impressive 15 yards per catch, 7.5 yards per attempt), especially if that includes 1 catch for 80 yards, meaning your other 19 passes only got you a paltry 70 yards (just 3.5 yards per attempt) with completion % of just 50%. That won't move chains unless you have a great run game.

I think the data is interesting but could use some refinement. But once you collect the data and formulate the metrics ... from there, it always comes down to subjectively assessing how good a QB is in the pocket (feeling the pass rush, amd kaing good throws), how good the OL is at pass blocking, how well you are able keep defenses honest in both the run and pass game, and ultimately whether your OC is play calling to optimize those strengths and weakness.

Too many variables (objective and subjective) and therefore impossible to measure with some subjective analysis to "interpret the results". Identical results across different QBs could be explained 3 of 4 different ways ... our QB is bad wiith pocket pressure but throws fine, our run game sucks, our OC sucks at play calling, etc.

That doesn't mean you shouldn't have metrics like this, it's always good to quantify something .. then you just have to analyze the results.
Largely this.

Will McElvain at UNI had a poor completion percentage this year - though he currently holds the MVFC record for passing yards in a season for a freshman - at 54%. If you don't watch UNI you'd think he is a pretty bad passer. He went 14-27 for 185 per game this year

What you wouldn't know is defenses were dropping 8 into coverage about 85% of the time because UNI couldn't run the ball.
You wouldn't know that of those 13 incompletions UNI averaged 5-6 drops per game. Of his 8 INT 5 of them hit the hands of the WR.
All of a sudden he is up to 70% and assuming no other passes are thrown that weren't already (not logical) he then goes to 19/27 for 225 per game and this changes everything....all with Will doing literally nothing different himself.

His yards per attempt was something like 6.8 this year. If those drops become catches all of a sudden his YPA probably jumps to 10 or so based on the fact that 75% of our passes traveled 15+ yards down field in air.



As someone that has played with stats for the forum for years IBY is right.

This is someone looking to prove a point and playing with a spreadsheet until his point is proven.

[BisonFan02]

Yeah, right now, I don't think there's any FCS QB I'd rather have than Trey Lance. Regardless of what the stats suggest. Of course, I'm a bit biased as a Bison fan, but even the stats point to him being near the top of the heap. And the stats you just posted are the exclamation mark.

Protect the football. NDSU largely (to an extent anyway....been spoiled lately) runs the ball and needs a "game manager" style QB that doesnt turn the ball over. Dont **** it up....run the ball....control TOP and win.

[POD Knows]

Trey Lance is responsible for 41 TDs....28 passing and 13 rushing....0 INTs....****ing ZERO. That is pretty interesting. /threadDamn it, now you jinxed him, never talk about the streak, if he throws a couple pics today I am writing you out of the will. xlolx

[X-Factor]

Damn it, now you jinxed him, never talk about the streak, if he throws a couple pics today I am writing you out of the will. xlolx

But he didn’t. [emoji6]
Maybe the silly superstitions can now be put to bed. They have zero impact on real life

[BisonFan02]

Damn it, now you jinxed him, never talk about the streak, if he throws a couple pics today I am writing you out of the will. xlolx

Dodged a bullet there. :D

[POD Knows]

Dodged a bullet there. :DIt was freaking close dude, I was adding extra space to the casket to store the booty until the last 7 seconds of the game

[IBleedYellow]

So uhh.

How'd all that math work out for ya?