For what it's worth, Josh Allen stole the show at the Senior Bowl

[ToddPhin]

To answer a couple of your questions, the cutoff is 97 because of data availability, I think.

As to how it works, the exact details are not public, but this is what it says in the article:



What you mention about character is important, the model tries to incorporate it by including expected draft position (major red flags will drive a prospect's stock down). But if you do know something almost no one or no one else does, you might actually beat the model -as hard as it seems, since it's accuracy seems to be higher than 90% at the lower and higher end.

Still, the model has great limitations. It is very powerful only with the prospects it really likes or the ones it really hates. Maybe one day AI will do a better job than humans at scouting, but that day is still far off.

Even if there were no cutoff b/c of data availability, a cutoff would still have been needed anyways b/c of the introduction of the Air Raid and Spread in the late 90's that altered the landscape of the college passing game. It changed things from apples to oranges, which means it would be pure folly to continue comparing these new oranges to those old apples.

If, however, the data cutoff had fallen well before 1997, then the "1 out of 31" stat that WildBill posted would look entirely different and would certainly not be so incriminating against Josh Allen, seeing as many pro QBs did in fact complete less than 60% of their passes in college prior to the late 90's. If there were any pre-1997 data available, THAT'S what you'd be comparing Allen too...... and there'd be nothing about Allen's completion percentage compared to that data that would suggest he'd fail in the NFL due to "accuracy issues".

 

[where's th'fish]

Well, completion percentage has gone up by a huge amount in the NFL and I imagine it has, too, in college. But it's very likely they adjust their data by era, like they do for their other stats, like DVOA. Not to mention completion% is just one element of the model.

Probably the best possible argument against the model is what Slimm said, that it's descriptive, not predictive. But isn't that the eternal dilemma with science? Just because the Sun has risen millions of times doesn't mean it will rise tomorrow. And yet I'm betting it will.

But speaking of that, the model's been around for several years. It existed when Russell Wilson was drafted. You saw how highly the model rated him. It's a funny story, because the owner of the site was pretty embarrassed by how highly Wilson was coming out, and even put an asterisk on him. Because he might not even go in the top 100. And yet, we all know what happened. It's most outrageous prediction came through, so that's pretty big.

What I really like about statistical models is that they don't presume to know anything or explain why anything happens. They just pick up on things that are actually happening and will likely happen and reap the rewards. So you can be in complete ignorance of the subject or the subject might itself be completely unknowable, yet you can make solid predictions. In any area of life, not just football. It's awesome.

 

[ToddPhin]

Well, completion percentage has gone up by a huge amount in the NFL and I imagine it has, too, in college. But it's very likely they adjust their data by era, like they do for their other stats, like DVOA. Not to mention completion% is just one element of the model. Look, these people are not simpletons. They know what they're doing. Probably the best possible argument against the model is what Slimm said, that it's descriptive, not predictive. But isn't that the eternal dilemma with science? Just because the Sun has risen millions of times doesn't mean it will rise tomorrow. And yet I'm betting it will.

What I really like about statistical models is that they don't presume to know anything or explain why anything happens. They just pick up on things that are actually happening and will likely happen and reap the rewards. So you can be in complete ignorance of the subject or the subject might itself be completely unknowable, yet you can make solid predictions. In any area of life, not just football. It's awesome

Not suggesting that they're simpletons. Let's address it a different way: if Dan Marino's college stats were input into their model, would he be a pass or a fail? I'm guessing he'd be a fail. This applies to Josh Allen b/c their statistical model pertains to this current era which is heavily influenced by the Air Raid, Spread, and Pistol offenses.... even though Allen's OWN OFFENSE applies more to what their prior model would've been about. Now, are they adjusting their data to properly fit Allen's offense, or are they cramming him as a square peg into a round hole?

 

[where's th'fish]

Sorry about the simpletons bit, that came out a bit short. I edited out that part, but you were already replying.

Anyway, I'm pretty sure they adjust their model by era, just like with the pros. I don't know about Marino in college, but his pro stats (DVOA) are adjusted by era so that they compare well to today's players, unlike his conventional stats.

Second, along with completion% the model also incorporates many other factors, including ypa and offensive efficiency. That would cancel out the impact of whatever particular offense the QB is running, so long as it works. Just like in the pros a great QB will never be in a mediocre offense, I assume something similar happens in college. If Allen's offense is more vertical, his comp% will suffer, but his ypa should be better and his offense should be very good anyway. And if he were being brought down by his teammates, the model incorporates strength of supporting cast, as well as strength of schedule.

 

[Pachyderm_Wave]

Sorry about the simpletons bit, that came out a bit short. I edited out that part, but you were already replying.

Anyway, I'm pretty sure they adjust their model by era, just like with the pros. I don't know about Marino in college, but his pro stats (DVOA) are adjusted by era so that they compare well to today's players, unlike his conventional stats.

Second, along with completion% the model also incorporates many other factors, including ypa and offensive efficiency. That would cancel out the impact of whatever particular offense the QB is running, so long as it works. Just like in the pros a great QB will never be in a mediocre offense, I assume something similar happens in college. If Allen's offense is more vertical, his comp% will suffer, but his ypa should be better and his offense should be very good anyway. And if he were being brought down by his teammates, the model incorporates strength of supporting cast, as well as strength of schedule.

Well, Allen's YPA for his career is nearly 8 YPA. I'm not sure I understand where the red flag is here.

It seems to be taking statistical snapshots instead of incorporating the entire picture.

Secondly, ask the model what Wyoming's record was the 4 years before Josh Allen got there.

 

[ToddPhin]

Sorry about the simpletons bit, that came out a bit short. I edited out that part, but you were already replying.

No worries man..... but thanks for the correction though my friend.

Anyway, I'm pretty sure they adjust their model by era, just like with the pros. I don't know about Marino in college, but his pro stats (DVOA) are adjusted by era so that they compare well to today's players, unlike his conventional stats.

but we're not talking about adjusting for era, as that defeats the purpose here. We're talking about taking a guy from a prostyle era and inputting his stats into this current statistical model, seeing as that fairly parallels Allen's situation in Wyoming.

Second, along with completion% the model also incorporates many other factors, including ypa and offensive efficiency. That would cancel out the impact of whatever particular offense the QB is running, so long as it works.

Respectfully, can you (or the models) tell me from Wyoming's 6.6 y/a, Washington State's 6.7 y/a, Michigan State's 6.6 y/a, and Bowling Green's 6.8 y/a, what type of offense their respective QBs run? ... and by looking only at the stats, can you tell me whether an inefficiency is caused by the QB, the offense around him, or both? {BTW, Wyoming runs an antiquated prostyle, while Washington St is Mike Leach's spread.}

Just like in the pros a great QB will never be in a mediocre offense, I assume something similar happens in college.

College QBs simply cannot do it by themselves though, as there's clearly a massive difference between an NFL QB with 7+ years of experience to go with an experienced surrounding cast, compared to a 2nd year college QB surrounded by a slew of freshman college kids and inexperienced 1st-year starters.

What happened to one of the most efficient college passers of all time, Baker Mayfield, in that 2nd half vs UGA when he faced the same type of heavy pressure, the small throwing windows, and all the 2nd/3rd & Longs that Josh Allen faced ALL YEAR LONG?? He completed just 9 of 24 drop backs.... for 7 total pts..... while his offense gained just 37 total yards on its first 5 possessions. Again, a QB can't block for himself, can't catch his own passes, and can't will his ground game out of the basement all by his lonesome. If there's heavy pressure in Allen's face every other snap, and his receivers are struggling to gain separation, and his ground game can't keep him out of obvious passing situations, then HONESTLY, what the heck is he supposed to do to make the passing offense look like it's running efficiently, within the confines of a prostyle system no less?????

If Allen's offense is more vertical, his comp% will suffer, but his ypa should be better and his offense should be very good anyway.

What does the offense being "more vertical" have to do with anything when you're pressured on 41% of your attempts and can't hold the ball longer than 3 seconds without defenses getting in your face? How is it that Allen's YPA "should" be better when he has no time to throw it deep, and has an inexperienced group of pass-catchers who aren't well versed at uncovering when plays break down and Allen is scrambling to feed them the ball?

And if he were being brought down by his teammates, the model incorporates strength of supporting cast, as well as strength of schedule.

I'm respectfully gonna call BS on this one. Between the 2 of us, it took CK and I a few hours independently to isolate everything that was going on in Wyoming's offense. Probably took me an hour alone just to nail down everything that was transpiring with Allen's offensive line. That said, I HIGHLY DOUBT that the people behind these models spend HOURS on every player in college football, let alone hours on Josh Allen, to ensure complete accuracy of their model as it pertains to Allen.

 

[ToddPhin]

....and just to touch on the myth that college QBs should be able to overcome any adversity just b/c great NFL QB's will allegedly never be in a mediocre offense:

the main caveat being though- that in the pros, a great QB is significantly more seasoned than in college, and a great pro QB is typically great b/c he has the mental ability to play the chess game at the LOS. They've been in the league long enough to recognize everything, check into a favorable matchup, and they have guys around them with infinitely greater football experience to facilitate it all. It takes them YEARS to reach that point....... whereas Josh Allen had only 2 years in his college Wyoming offense, and 6 of the starters around him were either starting true freshman or 1st-year-starters.

Does anyone honestly believe these QBs come into the league immediately ready to take over games and turn their offenses around regardless of circumstance? What did Wentz & Goff do as rookies?.... but according to this day and now, they're great QBs. It took Matt Ryan 8 years before he could actually take over a game. It took Brady 5 years to surpass an 87 rating. It took Brees 8 years to finally surpass 4,500 yards. It took Peyton Manning 4 years to reach 63% completions..... and 6 years to finally get his INT% under 2.0.

In 2013 when Brady's targets were injured and his O-line was troubled, New England's passing offense ranked 14th in scoring, 15th in rating, 16th in completion percentage, 18th in Y/A, and 19th in TD%. I think that's pretty much mediocre.

What happened to Andrew Luck in 2015 when his offensive line was horrendous and he had no ground game to support him? The best prospect since Elway, the smartest young QB in the game, and he completed just 55% of his passes en route to a 74.9 rating.

What happened to Peyton Manning vs Seattle in the Super Bowl, despite him playing some of the best ball the QB position has ever seen? Was Denver's offense not less than mediocre that day? Was New England's offense not mediocre during those games vs Rex Ryan's NY defense and Baltimore's D when they were in Brady's face all game?

Prior to the heavy rule changes of 1978 (created to steal some of the thunder away from defenses dominating football), the scoring avg was just 17.2 pts in 1977..... with 562 INTs to 388 TDs.... 51.3% completions....... and a Passer Rating of 57.8.
Is that b/c there were no great QBs in the league? -nope. Bradshaw, Griese, Staubach, Fouts, Stabler, and Tarkenton are all in the Hall of Fame. Only 1 QB that year cracked 59% completions. Go look at Dan Fouts' stats before and after the '78 rule change. Looks like a completely different guy.
Put great QBs in crappy situations and they WILL NOT make chicken salad out of chicken s**t.

 

[ToddPhin]

I still vividly remember those 2 stretches of games for Brady & Manning when people were jumping off a cliff thinking they were washed up...... when Brady in 2013 had a passer rating of 74.9 through the first 8 games, throwing just 9 TD and completing a mere 55% of his passes...... and then in 2010 when Manning had that 5 game stretch where he went 1-4, throwing a whopping 13 INT and posting a miserable 74.2 rating when most of his receiving corps was out.

 

[where's th'fish]

It'll be fun to follow Allen and Mayfield's career. You guys obviously feel pretty strongly about them!

 

[ckparrothead]

I like them both.

They don't all have to look the same to be successful.

 

[nick1]

Mayfield will be better then Allen who will be a bust. I feel pretty confident in that. Allen is far too raw, you guys are falling in love with his tools but playing QB is by and large a mental game

 

[ToddPhin]

I like them both.

They don't all have to look the same to be successful.

Hey Chris, I just rewatched Allen vs Iowa and Mayfield vs UGA last night back-to-back and came away thinking Allen's tape was easily more impressive. I wasn't expecting that. Notably more NFL caliber throws by Allen. Better & cooler under pressure is Allen, escapes the pressure better, and keeps his eyes downfield better under duress.......... And even though Allen's offense is outmatched most of the game and his team never stands a chance, it's not until the end of the game when TRAILING BY 3 SCORES that we actually see Josh making bad decisions, panicking, and forcing poor throws, unlike Baker who seems to do so in this game anytime he doesn't have an easy throw or a clean pocket, regardless of the scoreboard.

Contrary to Allen, Baker gets everything from the shotgun, with the field SPREAD OUT, loaded with premium weapons, and never having to turn his back to the defense. He's afforded about a dozen completions that shouldn't even count toward completion% (unlike Allen).... and in the 1st half Baker has ALL DAY to throw and receivers that are running WIDE OPEN (again unlike Allen), yet it's still mostly easy completions for Baker, far far easier than the stuff Allen is attempting behind a weaker pocket and into tighter windows. It's like a good college diver competing against an Olympian- the college diver looks equally impressive on the surface, but it's only b/c his degree of difficulty is significantly less. In the 2nd half, Baker misses most of his NFL caliber throws or chucks INTs on them, now that he's actually being pressured and facing tighter windows in ways that Allen did ALL YEAR..... or Baker gets bailed out by receivers who actually MAKE PLAYS with the ball in the air (while Allen's don't).

Josh Allen sticks THIRTEEN opposite-field throws from the far hash to the opposite numbers and beyond in this game alone..... out of 16 attempts I think. The 3 incompletions were: a drop after the receiver keeps his route too far upfield out of his break (so the ball arrives a little low); a drop on a throw that Allen puts low & away from the corner where only his guy can get it; a drop b/c Allen puts it a bit high & inside. All 3 catchable though, with 1 being a justifiable drop. We're left with Allen accurately hitting 15 of 16 cross-field boundary throws. Accuracy issues my a**. He makes these throws look so easy, yet they're balls that all travel 30+ yards of air distance that he makes look like 15-20..... meanwhile most of Baker's passes travel about 15 yards and under vs Georgia.

It's hilarious though- I broke down every good throw Allen made vs Iowa to the guys at CollegeFootballMetrics (who hate Allen b/c he doesn't fit their statistical models, but LOVE Baker b/c he does), and these guys actually treated these 15+ passes as "MEH, 50 other guys can make those throws, why should I be impressed?!". They're somehow convinced that he only made 2 good throws in this game, and they're infinitely more concerned about criticizing Allen for not attempting any passes directly in front of him (or between the hashes), you know- the EASIEST STUFF for a QB to make, meanwhile they completely give Baker a pass for not attempting any perimeter throws even though they're significantly tougher on QBs, especially against NFL corners. I remember when everyone thought Tannehill was farrr too raw to start as a rookie, but he already excelled at nailing those NFL caliber perimeter throws in college, and so that aspect transitioned easily into Philbin/Sherman's offense. I imagine Tannehill's ability to stick those throws from day 1 helped keep corners more honest, less able to sit on in-breaking routes on short-medium drops and sit on the downfield stuff on deeper drops...... but yeah- who cares about Allen's ability to beast those throws in ways that very few people on the planet can. Totally insignificant and irrelevant amIright. :sigh:

Mayfield OTOH attempts just 1 out throw all game, and it's a short play-side toss from the near hash (clip below). But yeah- somehow all the long opposite-field throws by Allen while working from under center aren't NFL level stuff. :sigh again:

Disclaimer: I know that CK likes Allen, so this post isn't intended to suggest he doesn't.

 

[ToddPhin]

Mayfield will be better then Allen who will be a bust. I feel pretty confident in that. Allen is far too raw, you guys are falling in love with his tools but playing QB is by and large a mental game

Wait up, so the guy with the highest Wonderlic score of the bunch who's impressed in meetings with his great memory recall ability is suddenly being criticized for not having the "mental tools to succeed"?

Allen has the brains.... the size.... the athleticism.... the escapability.... the mobility.... the ability to create with his legs AND arm on the run..... the arm (in every regard)..... and already possesses the ability to stick the tougher NFL caliber throws out of a pro style offense..... but he's "too raw"?..... let alone more raw than the kid from the Mike Leach influenced spread?..... more raw than the USC kid with 21 INTs in his past 22 games who doesn't even know how to use his checkdown?..... or more raw than the kid from Louisville whose program is concerned they might not be able to transition right back into their desired pro style offense this year b/c Lamar Jackson forced them to deviate too far away from it?

 

[Pachyderm_Wave]

Disclaimer: I know that CK likes Allen, so this post isn't intended to suggest he doesn't.

I don't. I think he's going to be one of the most overdrafted quarterbacks I've ever seen if he indeed gets selected in the top 5 or 10 picks. This is the one people are going to feel stupid about 5 years from now in my opinion.

I'd take him at a certain point, but he's always been a 2nd rounder to me. There's 4 quarterbacks I take over him in this draft.

 

[Pachyderm_Wave]

As for Lamar Jackson, I can't ever recall a quarterback being selected in the 1st round after he went out and threw 4 interceptions in his last game. I know he wouldn't have a Heisman Trophy if he played an SEC schedule. He's 1-3 with seven interceptions in his last 4 games against mediocre SEC teams.