Elway thinks Tebow is not a good QB. Do you think Shanny might want him?

[War Paint]

I would love it if we got Tebow.

[rumplestilskin]

I realize your post is rhetorical, but I'll answer it anyway. The model has been more accurate NFL franchises picking in the first round, and its error rate has been going down as I refine it. Error rate is currently around 20%. My goal is to refine to near perfect reliability, and in that sense it has never left the drawing board.

Save your snotty comments. Since my model came out in 2005, a variety of media have come out with variations on the same theme. Some of those have been written about recently on ES. I suppose you find it impossible that someone not employed by the NFL, by a major media outlet, or by Football Outsiders can come up with a good model sooner, and a better model now. I run a multi-million dollar company and do this in my free time. If you have a substantial argument, I'll answer it in the future. Until then, this is my last response to you.

Here Here! ASF!

I am happy that there are people like you on this forum that can share their passion, knowledge and imagination and apply it to something like your model. I like the fact that the model is not biased. Unfortunately some people have preconcieved opinions based mostly on what they hear from media outlets and take it as gospel. I must admit I was not a Tebow fan last year before the draft and also I admit that I am ignorant to the college game. I did however see with my own eyes the past three games Tebow played in the NFL and he looked special at times and plays with a winners fire. I mean for a Rookie he could have done a heck of a lot worse.

[ArmchairRedskin]

The problem with Tebow is that accuracy is tied to mechanics. Since Tebow has poor mechanics, he has poor accuracy. In his three starts he's just under a 50% completion rate. He needs a lot of room to operate, has a long windup, and throws a lot of flutter balls. He's slow to reset on the run and has bad accuracy on long throws. His short to medium accuracy suffers if he's not able to step into his throws. Sure he can break one off for a long gain or even a TD, but that kind of success isn't sustainable over a long 16 or especially an 18 game season. The longer a season wears on, the less his mobility and running skills will help him. He'll take shots and he'll be injured. His legs won't be as fresh.I think all this will become readily apparent once he takes the helm full time.

[Geoff_K]

Since there is a minimum character limit i type this before i say ............ NO ! ;- P

[Atlanta Skins Fan]

The problem with Tebow is that accuracy is tied to mechanics. Since Tebow has poor mechanics, he has poor accuracy. In his three starts he's just under a 50% completion rate. He needs a lot of room to operate, has a long windup, and throws a lot of flutter balls. He's slow to reset on the run and has bad accuracy on long throws. His short to medium accuracy suffers if he's not able to step into his throws. Sure he can break one off for a long gain or even a TD, but that kind of success isn't sustainable over a long 16 or especially an 18 game season. The longer a season wears on, the less his mobility and running skills will help him. He'll take shots and he'll be injured. His legs won't be as fresh.I think all this will become readily apparent once he takes the helm full time.

To keep Tebow fresh and healthy, his rushes need to be limited to goal line situations, critical short yardage situations, and escaping sacks. I agree that health and long-term effectiveness depend on minimizing the number of hits he takes as a rusher.

As to his accuracy, he was very accurate in college. He's been a 50% passer so far in the NFL. You seem to have watched him and offer what seems to be a reasonable diagnosis and prescription for improving his accuracy. Let's assume for now that you are correct, and Tebow needs a good QB coach in the NFL. This may be what Elway was referring to in distinguishing Teblow as a great player vs great QB.

However, your comments completely fail to grasp why Tebow is effective. I missed this until recently, because previously I bought into the conventional wisdom that Tebow would never be a franchise NFL QB. Then I finally looked at his college stats when I redid my model to capture all QBs drafted in the first round in the past 20 years.

Tebow's college numbers blew me away. In college he had an 8.8% TD percentage throwing, 8.3% TD percentage rushing, and only 1.6% INT percentage. Put that together and you get 145 TDs vs 16 INTs, which has to be a record for ratio at that volume. He also threw for 66.4% completion percentage over 4 years, and no year was less than 64.4%. Very accurate, very consistent.

So far in the NFL, the TD vs INT trends are continuing, but not the accuracy. He's got 6% TD passing, 14% TD rushing, 3.7% INT, for 11 TDs and 3 INTs. Forgetting the rushing, his QB rating is higher than Bradford, McCoy and Clausen due to the passing TD/INT ratio, which is better than all the others, even while completion percentage lags all the others. By comparison:

• Bradford: 3.1% TD passing, 4.4% TD rushing, 2.5% INT. Overall: 19 TD, 15 INT.
  
• McCoy: 2.7% TD passing, 3.6% TD rushing, 4% INT. Overall: 7 TD, 9 INT.
  
• Clausen: 1% TD passing, 0% TD rushing, 3% INT. Overall: 3 TD, 9 INT.
  

So, Tebow with 11 TDs and 3 INTs is crushing the other guys in that department. Typically the downfall of young QBs is throwing INTs, and Tebow has kept those around par for the group, while dominating with much more TD production. This is why his QB rating is higher.

Looking back at his college stats, his INT% was only 1.6% and his accuracy was very high. It seems reasonable that he can get his INT% and accuracy closer to his college patterns, with a good QB coach. If he does, he has the opportunity to be lethal in the Steve Young mold.

NFL.com scouting report on Tebow says he "posesses very good arm strength" while criticizing mechanics.

College QBs with high completion percentage and low INT percentage, over multiple years, rarely fail in the NFL when they arrive with sufficient NFL physique and tools. In a nutshell, that is the model and that is Tebow. He's more of a project than you expect in a blue-chip prospect, but he's a winner and has tools that most blue-chip pocket passers will never have. His upside is at least as high as Bradford, but he does need a very strong QB coach and some patience for the next couple of years. People who doubt Tim Tebow will look very foolish in 5 years.

If the Redskins could get Tebow for a #1 or #2 pick, and sit him behind Grossman while letting the Shanahans develop him, they could have a dominant franchise QB for the next decade. Tebow blows away all other options available to the Skins at this time, and Grossman will be good enough (possibly very good) for as long as it takes to develop Tebow.

Skins should strike soon while the Broncos are in turmoil.

[TotalRecall]

Elway still wants Tebow. He said that he's a very, good football player, but we have to turn him into a very, good QB. Elway even said that if a coach came in during the interview process and said that he doesn't want Tebow, then Elway wouldn't hire him.

And no way to the speculation of getting Locker!!! Locker = Clausen.

[ArmchairRedskin]

Hey ASF, I'd love to see how a guy like Graham Harrell does in your model.

http://sports.espn.go.com/ncf/player/profile?playerId=160425

[Atlanta Skins Fan]

Locker = Clausen.

Clausen is an awful NFL QB, as I predicted. Locker projects worse. The only thing Locker does well vs other QB prospects is rush for TDs. Clausen beat him in all statistical QB categories, and by a wide margin for accuracy. Locker is one of the most obvious NFL busts I've ever seen discussed as a first-round pick.

Skins had better not draft him in any round. I've hoped that that level of QB stupidity left the building with Cerrato ... who drafted Druckenmiller and Ramsey, brought in Jeff George and dumped Brad Johnson, wanted to draft Harrington, Sanchez and Clausen, and helped cause the team to choose Campbell and Zorn over Collins, Saunders and Grilliams.

Mike Shanahan has a great record developing QBs, including Elway, but I don't know that he's nearly as good at picking QBs. (See McNabb.) Kyle Shanahan shows signs of being better at that, pulling Grossman and Beck off the scrap heap for free. I'd be happy to see Kyle Shanahan take a lead role in scouting and choosing any new QBs the team acquires. He values the right things.

[Command The 414]

so you called all of your NFL executive friends and checked on that, eh?

I would take Tebow over Locker in a heartbeat. I will eat crow if I am wrong, but Locker looks like garbage to me.

couldn't agree with you anymore...what exactly has Locker ever proven that warrants him a quality NFL QB? the guy hasn't and can't seem to stay healthy...and like it or not, Tim Tebow is a winner.....and his entangiables alone make him a better QB then Locker imho....I would take Tebow in a heartbeat here with the Skins

[Destino]

Tebow doesn't interest me at all. McDaniel's as an O coordinator though is extremely valuable IMO. The guy made Kyle Orton so much better and revived B Lloyd's career to the tune of leading the league in yards receiving. I don't see how Miami, San Fran, etc etc etc aren't leaping at the chance to hire him. As a GM and head coach he flopped but his work on the offensive side of the ball is very good.

[Atlanta Skins Fan]

Hey ASF, I'd love to see how a guy like Graham Harrell does in your model.

http://sports.espn.go.com/ncf/player/profile?playerId=160425

He does very well. However, the stats are tainted by association with the Mike Leach / Kliff Kingsbury lineage. Something about that system (spread offense) produces supersized stats that break models, much like June Jones at Hawaii. He went undrafted but is currently on the roster at Green Bay, put on the roster late in the year when Rodgers got injured. That means that he's valued by a team that understands QBs, but he has no chance to play behind Rodgers, sitting at #3.

He's a good example of why the model can't be taken alone, but only in tandem with a scouting report. Typically I apply the model as a filter on the group of QBs deemed by scouts to be top prospects, which avoids missing the obvious. I made an exception recently for McElroy, who has been deemed worth drafting in the lower rounds. Even that was something of a vote of confidence by scouts. For presumably good reasons, NFL scouts have decided that Harrell is not NFL caliber, and the model shouldn't be used to argue such basic assessments. It's best used to filter the group of QBs deemed NFL caliber, and it ranks these QBs very differently than scouts and draft position.

[ArmchairRedskin]

Well, If you apply NFL scouting reports to Tebow, then you have to take into account that, from your very own link,

Tebow really struggles with his accuracy. Release is far to slow to fit balls into spots against NFL defensive backs. Release point and mechanics (elongated, wind-mill delivery which comes out too low) likely need to be altered. Was not asked to run through pro-style progressions and struggled reading defenses, especially those with NFL concepts, in college.

Those are all huge knocks against him. Especially the accuracy and mechanics portion. You can learn to read defenses quicker and more accurately, but if your accuracy is affected by poor mechanics, then you are just not gonna make it. Tebow comes from the spread option that also had a lot of top shelf talent surrounding him. His stats were helped immensely by those factors.

I'm not sure how you can blindly plug in Tebow into your model and claim he'll be good without taking into account that all signs point to him struggling in an NFL offense. Sure he's got intangibles, but those are only good when they accompany the proper tangibles.

So far, in his limited starts, he's showing exactly the things that have been the knocks on him coming out. A low completion rate due to poor accuracy.

The longer he's the starter, the more teams will find and exploit his weaknesses. His throwing motion will never go away. Even though he can go out on practice field and show you a proper throw, when he's under duress he reverts right back to his bad habits.

I think your model is way off base on Tebow.

Time will tell, I reckon. All I know is that I don't want that time to be on the Redskins. Let Elway deal with him.

[Atlanta Skins Fan]

Well, If you apply NFL scouting reports to Tebow, then you have to take into account that, from your very own link,

Tebow really struggles with his accuracy. Release is far to slow to fit balls into spots against NFL defensive backs. Release point and mechanics (elongated, wind-mill delivery which comes out too low) likely need to be altered. Was not asked to run through pro-style progressions and struggled reading defenses, especially those with NFL concepts, in college.

Those are all huge knocks against him. Especially the accuracy and mechanics portion. You can learn to read defenses quicker and more accurately, but if your accuracy is affected by poor mechanics, then you are just not gonna make it. Tebow comes from the spread option that also had a lot of top shelf talent surrounding him. His stats were helped immensely by those factors.

I thought that the accuracy comment in that scouting report was interesting. If Tebow were not an accurate passer, that should show up in the stats for the year, or in the game logs. I've already mentioned the accuracy by year, so let's compare the game logs for Tebow and Bradford:

• Tim Tebow: college game logs
  
• Sam Bradford: college game logs
  

Sometimes QBs will have good accuracy stats for the year, but a closer look at game logs shows that the QB will have bad games against good teams, and then play very well against bad teams. But, Tebow has almost no bad games for accuracy in 4 years at Florida. He had only one game with under 50% completion rate, in 2007 (sophomore year). Bradford had 3 such games in 2 years + 3 games. So, I don't understand where the comment about accuracy comes from. Maybe he had some poor throws now and then, and maybe in those throws he had poor mechanics. But, overall, he was very accurate and consistently so from the game logs.

As far as top-flight talent goes, only one of his receivers was ever drafted in the first round (Percy Harvin). Harvin caught 133 balls over three years. Meanwhile Tebow completed 528 additional passes to receivers not drafted in the first round. He may have had very good talent around him at Florida, but you can't make the case that some blue-chip future NFL WRs carried him at Florida.

I've discussed the mechanics and don't dispute your assessment. However, Josh McDaniels, thin-skinned prick that he may be, was Tom Brady's coordinator and drafted Tebow. For subjective assessments about mechanics and what can be approved, I'll trust McDaniels over your opinion or mine. That's an area where pro QB coaches and pro offensive coordinators know better than we do, more often than not.

My main concern with Tebow, statistically, was that he had a limited number of ATTs per game. 23.5 ATTs per game, his last three years. That's low, and in my model it's a red flag that reduces confidence in the grade. Often low ATTs/game can signal a bust, but typically those busts also flashed late with only one year of great play. (The Cam Newton problem.) Tebow compensates by having played consistently over 4 years. My confidence would be higher if he had passed for 30+ ATTs/game.

Tebow isn't a sure-fire NFL franchise QB like Luck and Bradford. But, in my opinion, with his rushing TD tools and intangibles, his upside is in the league of Bradford, and much higher than QBs available in the 2011 draft. This is the context for my comments. As you say, time will tell.

[TheGoodBits]

I realize your post is rhetorical, but I'll answer it anyway. The model has been more accurate NFL franchises picking in the first round, and its error rate has been going down as I refine it. Error rate is currently around 20%. My goal is to refine to near perfect reliability, and in that sense it has never left the drawing board.

Save your snotty comments. Since my model came out in 2005, a variety of media have come out with variations on the same theme. Some of those have been written about recently on ES. I suppose you find it impossible that someone not employed by the NFL, by a major media outlet, or by Football Outsiders can come up with a good model sooner, and a better model now. I run a multi-million dollar company and do this in my free time. If you have a substantial argument, I'll answer it in the future. Until then, this is my last response to you.

Error rate is currently 20%, and the model applies only to QBs taken in the 1st round. I would venture to guess that the miss rate for QBs taken in the 1st isn't much worse than that in recent years.

Furthermore, you admit that there is a 20% error rate, yet when you post about QBs you act like it's the gospel. I wouldn't have a problem with it if you put a disclaimer at the end of every post saying *above opinion based on model with 20% error rate, but your posts make it seem infallible.

My doubt is not that you are not able to come up with a model for predicting QB success, my doubt is that ANYONE can create a formula for predicting QB success based simply on stats that is any more effective that traditional scouting. I would think your 20% error rate is pretty much as good as you're going to get it. Unfortunately, good scouts can achieve a success rate just as good or better by simply looking at film of the player and not subjecting his stats to any kind of formula.

[Mahons21]

Sometimes QBs will have good accuracy stats for the year, but a closer look at game logs shows that the QB will have bad games against good teams, and then play very well against bad teams. But, Tebow has almost no bad games for accuracy in 4 years at Florida. He had only one game with under 50% completion rate, in 2007 (sophomore year). Bradford had 3 such games in 2 years + 3 games. So, I don't understand where the comment about accuracy comes from. Maybe he had some poor throws now and then, and maybe in those throws he had poor mechanics. But, overall, he was very accurate and consistently so from the game logs.

But how many of Tim Tebow's passes were on options? The scouting report on Tim Tebow doesn't come from his stats, it comes from a scout watching him attempt to make actual NFL passes. I'd venture to say that majority of people who have a descent amount of knowledge of football, could tell you via the eye test that Tim Tebow isn't very accurate.

[Conn]

ASF, I think part of the problem here is that you're using completion percentage as a benchmark for accuracy. I don't think that's fair for a college player, because of the wide range of talent (or lack thereof) a college QB confronts in opposing defenses. Not to mention the wide range of talent that may or may not be present in his receiving core. If a QB's WR's are dominant (in comparison to the competition), or they are always open due to offensive scheme or opposing defenses not being as talented/well coached, you can have an artificially high completion percentage, propped up by "easy" throws. Mostly, you see this in players that are not forced to throw as many "NFL" throws in college.

Tebow may indeed have the arm strength to make all of those NFL throws (I think that he does), but does he have the accuracy? I think your use of completion percentage in college is a flawed approach...and I would argue that a superior means of figuring this out would be to study the prospect's games...and check out his ball placement. Whether his completion percentage is 60% or 70%, I believe that watching a QB's ability to place the ball on a throw is the most important thing...because whether he is elite in that category or not, his completion percentage can be artificially higher or lower than it should be, due to outside factors that are very hard to judge, and are even harder to quantify.

Is it subjective? Yes. Can it be part of a statistical analysis? No. And I think that's why you are over-looking it, purposefully or not. A flaw in your formula is that you must use information that is quantifiable. This is not. Yet I believe its the most (and maybe only) accurate way to discern a QB's....well, accuracy.

It's interesting that you say a few posts above that your formula should be paired with scouting reports to really avoid "false positives" (if that is the right term?). Yet, you seem to fail to recognize that accuracy is something that can more realistically be determined by the process of scouting, rather than by looking at the stats that effect the outcome of your formula.

[Glenn X]

ASF, I'm interested in your opinion of John Beck. Are you upbeat or downbeat about him?

[Atlanta Skins Fan]

Error rate is currently 20%, and the model applies only to QBs taken in the 1st round. I would venture to guess that the miss rate for QBs taken in the 1st isn't much worse than that in recent years.

Furthermore, you admit that there is a 20% error rate, yet when you post about QBs you act like it's the gospel. I wouldn't have a problem with it if you put a disclaimer at the end of every post saying *above opinion based on model with 20% error rate, but your posts make it seem infallible.

My doubt is not that you are not able to come up with a model for predicting QB success, my doubt is that ANYONE can create a formula for predicting QB success based simply on stats that is any more effective that traditional scouting. I would think your 20% error rate is pretty much as good as you're going to get it. Unfortunately, good scouts can achieve a success rate just as good or better by simply looking at film of the player and not subjecting his stats to any kind of formula.

Last time I checked (a few weeks ago), the model was 37-9 (80% accuracy) for QBs picked in the first 32 selections of the draft since 1991, while the NFL was 13-33 (28% accuracy) in the same period, with a few TBDs in the mix. Accuracy for the comparison was defined as model grade vs draft slot as better projecting the resulting performance.

I don't know what your definition of "recent years" is, so it's hard to evaluate your claim that the NFL has improved a great deal in accuracy. This is the same NFL that was all over JaMarcus Russell in 2007 and Sanchez in 2009, while drafting Clausen ahead of McCoy in 2010. The same NFL that let Schaub fall to the third round.

The NFL may be getting better at this, but they've got a long way to go when their accuracy rate over 20 years is 28.6%.

[ArmchairRedskin]

ASF, I don't understand how you use Scouting for a basis on whom to insert into your model while claiming your model takes the guesswork out of Scouting. Seems like a convenient way to explain away a guy like Harrell scoring rather highly (I assume) in your model. By the way, a lot of Scouts predicted 4th or 5th round for Harrell. He wasn't projected to go UDFA. Same with your boy McElroy. He's a mid to late rounder. Wasn't that your whole point? That McElroy is being overlooked by Scouts like Brady was? So I don't understand why a guy like Harrell wouldn't look like a sure fire star in the league. Sure his stats are inflated, by the scheme, but I don't see a plausible explanation as to why he's disqualified.

[Atlanta Skins Fan]

ASF, I think part of the problem here is that you're using completion percentage as a benchmark for accuracy.

OK, that's semantics, and I'll leave it alone. What I'm doing is using completion percentage as a core variable in projecting NFL success. This isn't really an arguable point, except for some mobile QBs who bust the model. Going back to my original model, I noticed that 62% was a magic threshhold for college stats. College QBs who passed for 62+% for 2 or more years, with a threshold number of attempts, had a high success ratio in the NFL. College QBs who failed that test had a high bust ratio. In various forms, you will see this general component in the better models that have come out in the past few years.

The seminal such model in the media is the Lewin Career Forecast, published in Pro Football Prospectus 2006 in August 2006. It's now cited as a bible by many sources, but actually came out 16 months after my own model.

At this point the completion percentage component is inarguable. I think I've got a better model than Lewin, and I also had it first, to my knowledge. However, I don't publish a Pro Football Prospectus.

---------- Post added January-9th-2011 at 09:19 PM ----------

ASF, I don't understand how you use Scouting for a basis on whom to insert into your model while claiming your model takes the guesswork out of Scouting.

The goal of the original model was very focused. It was a model to project future NFL franchise QBs, period. After watching an amazing number of teams blow first-round picks on QB busts, I became motivated to develop a filter by which to project highly touted QBs.

My idea was that if the model were proven accurate, then a team could skip the first round for drafting a QB unless a QB were projected to be an NFL franchise QB. For those so projected, a team could feel confident to go "all in" on the QB, trading up in the draft as high as necessary to get that QB.

The results of the first model were mixed. I concluded quickly that the model was better than the NFL, but the error rate was high enough that I couldn't hype the model as a silver bullet.

Gradually I've become more interested in trying to refine the model to make it more accurate. I expanded the data set to all QBs drafted in the first round in the past 50 years, while also including interesting QBs like Schaub (whom I touted). Again, the goals are narrow. I've not trying to replace scouting, but rather start with a premium blend of highly touted QBs, and filter those QBs. Now with about 50 data points (QBs) in the model, I'm able to test a lot of additional variables to see if they matter. So, for example, I concluded recently that YPA doesn't matter. The future franchise NFL QBs were almost exactly average for the group for college YPA. This makes YPA garbage data at best. However, some other stats have spiked very high in relevance. The current model factors some stats very strongly while ignoring others.

Very recently I've been looking at college winning percentage and sack percentage as factors, while also trying to determine how to factor number of ATTs per game instead of using a simple minimum threshold as I have before. These are tricky variables, and I don't have anything conclusive to report. They are examples of potential ways to improve accuracy above 80%. It's tough sledding, because additional variables can compress the spread of scores. Sometimes seeing more data has the effect of diluting the really important data.

I also have to be careful in looking at second-tier QBs like Dalton and Devlin, etc. Since I haven't done a control test for historical QBs drafted below the first round, it's hard to project upcoming QBs drafted below the first round with confidence. This is why I have been cautious with most of my comments about these QBs. I'm keeping my powder dry until closer to the draft, except when I see QBs like Luck or Locker who score so obviously as sure-fire franchise QBs / sure-fire busts, or McElroy, who struck me as a bargain if drafted in the 4th round, reminding me of Brady.

I'm posting in the Tebow thread because I missed him when he was drafted. I simply didn't bother to grade him at all at the time, due to all the dismissals of his performance, and also his low number of ATTs. When I did grade him finally (in the past month), I was stunned.

[worstSeat]

Tebow only if the Skins resign Shar to protect him.

[both are equally inane ideas, if we missed the humor.]

[Atlanta Skins Fan]

Very recently I've been looking at college winning percentage and sack percentage as factors, while also trying to determine how to factor number of ATTs per game instead of using a simple minimum threshold as I have before. These are tricky variables, and I don't have anything conclusive to report. They are examples of potential ways to improve accuracy above 80%.

Sacks (as % of ATTs) are a very interesting variable, but so far I can't test the variable historically more than a few years back. The reason is that I have not been able to find sacks as a college QB stat prior to 2003. Best resource I've found is rivals.yahoo.com, but those stats start at 2003. Does anyone know of a college stats resource that goes back a lot further? Ideally I'm looking for stats going back to 1991.

[nuposse87]

I wouldn't even trade a 4th rounder for tebow. Dude is going to be a bust and a half when it is all said and done. He has had three average games so far, and most of them meaningless games outside maybe the raiders in week 15 (it was meaningful for the Raiders). The best case scenario for this guy is a decent game manager. He's Jason Campbell lite. His mechanics make me want to cringe, I doubt he ever wins more than two playoff careers as a starter. I say this because once teams start REALLY game-planning for this guy, he is going to get leveled. He has great spirit, I can give him points for that. But I don't EVER want to see him in a skins uniform. Give me Locker 11 times out of 10. Although thats not the guy I'd pick there, but since we're talking strictly QBs. **** I'd take Newton/Pryor over Tebow, maybe im sippin on too much haterade

[Sekhmet728]

Shanahan has had the most success with gunslinger QBs in the past. I find it highly unlikely Tebow ends up in Washington.

[Atlanta Skins Fan]

My main concern with Tebow, statistically, was that he had a limited number of ATTs per game. 23.5 ATTs per game, his last three years. That's low, and in my model it's a red flag that reduces confidence in the grade. Often low ATTs/game can signal a bust, but typically those busts also flashed late with only one year of great play. (The Cam Newton problem.) Tebow compensates by having played consistently over 4 years. My confidence would be higher if he had passed for 30+ ATTs/game.

Tebow isn't a sure-fire NFL franchise QB like Luck and Bradford. But, in my opinion, with his rushing TD tools and intangibles, his upside is in the league of Bradford, and much higher than QBs available in the 2011 draft. This is the context for my comments. As you say, time will tell.

RETRACTION

I'm going to do something unusual here: call myself out for bad posts.

I don't have foundation to tout Tebow to the degree I have. It's true that I ran Tebow through the model and he projected very well, but that's only because I recently dropped the filter that would have excluded a QB with such few ATTs per game. I have not proven that the model is reliable on low number of ATTs, and in fact I ignored my prior conclusion that low number of ATTs made the model unreliable.

I compounded this mistake by making a classic error, which was observing a positive attribute and assuming that it was relevant as a projection factor. That's a Vinny-level error, and I'm annoyed to the point that I need to make a retraction. The specific error was observing Tebow's number of rushing TDs and assuming that this was a positive factor in projecting NFL success. This is similar to observing a QB's height, arm strength, GPA or charitable works and using such factors to puff up a prospect. It "feels right" that a prospect who is 6.6 ft tall, throws like a cannon, and has 4.0 GPU has positive attributes to make an NFL QB, but this is nonsense without proof of correlation. The 6.6 ft cannon may be a better baseball closer than QB, and the GPA and charitable works may be utterly irrelevant in projecting QBs. Similarly, rushing TDs are more likely correlated to RB success than QB success.

I have no basis whatsoever in the model to assume that Tebow's rushing TDs will help make him a better NFL QB. That was me talking, not the model, and I retract the comments.

I make this post after running some tests on 2003 NCAA QB data, and noticing immediately that high numbers of QB rushing TDs were inversely correlated to NFL success. In fact, this was a useful filter to combine with completion percentage. If I exclude the QBs with the top 10% of rushing TDs vs passing ATTs (guys like Tebow), then my top 10 list of QBs for the year is much more accurate in projecting to the NFL.

Tebow may have caused me to stumble into a very opposite point of view. It may be that QBs who score a lot of rushing TDs tend not to project well to the NFL. If so, Tebow's rushing TDs would be a negative factor and not a positive factor. I don't have proof of that opposite conclusion, but I do see enough alarming patterns to retract my comments in support of Tebow.

I'm not saying Tebow will bust, just retracting my comments. I misused the model to make an argument not supported by the model. Kudos to those who made good opposing arguments, which caused me to look more closely at the situation.