12 Mar 2012
Lewin Career Forecast 2012
by Aaron Schatz
Six years ago, Football Outsiders unveiled the college quarterback projection system known as the Lewin Career Forecast. Originally, the LCF projected the success of first- and second-round quarterbacks using just college games started and college completion percentage. Going back -- including when looking at quarterbacks from the years before the data set used to create it -- it had a strong record. After 2006, the record was not so strong. So last year, we debuted an updated version of the forecast, LCF v2.0.
The new version of the Lewin Career Forecast is built to apply only to quarterbacks chosen in the first three rounds of the draft. After that, quarterback success and failure becomes too difficult to predict. Part of the concept of the system is that scouts will do a good enough job identifying "system quarterbacks" so that those quarterbacks whose college stats are much better than their pro potential will naturally fall to the third day of the draft.
There are seven variables involved in LCF v2.0:
* Career college games started, with a minimum of 20 and a maximum of 48.
* Career completion rate; however, this is now a logrithmic variable. As a quarterback's completion percentage goes down, the penalty for low completion percentage gets gradually larger. As a result, the bonus for exceedingly accurate quarterbacks such as Tim Couch and Brian Brohm is smaller than the penalty for inaccurate quarterbacks such as Kyle Boller and Tarvaris Jackson.
* Difference between the quarterback's BMI and 28.0. This creates a small penalty for quarterbacks who don't exactly conform to the "ideal quarterback size."
* For quarterbacks who come out as seniors, the difference in NCAA passer rating between their junior and senior seasons. (For quarterbacks who come out as juniors or redshirt sophomores, this variable is always 5.0, which is the average increase for the seniors in our data set.)
* A binary variable that penalizes quarterbacks who don't play for a team in a BCS-qualifying conference.
* Run-pass ratio in the quarterback's final college season, with a maximum of 0.5.
* Total rushing yards in the quarterback's final college season, with a minimum of 0 and a maximum of 600.
These last two variables work together to penalize both quarterbacks who scramble too often and quarterbacks who take a lot of sacks (since sacks are negative runs in college), while pocket quarterbacks who are successful when they do run get a bonus.
The biggest question about LCF continues to be the importance of games started. This is still the most important variable in the equation. As I explained in last year's article, any quarterback projection system based on past performance is going to highly value collegiate games started. From 1990 to 2005, it was far and away the most important variable in determining the success of highly-drafted quarterbacks. However, there are questions about whether the rise of the spread offense is leading to number of quarterbacks who come into the NFL with a lot of collegiate experience yet still unprepared for the NFL-style game. Other quarterbacks have come into the NFL with less experience and done very well. The best example of this would be Cam Newton, who seems like the kind of guy who is built to break this system. He started only one year of Division I ball and looked like a huge risk, then put together one of the best rookie quarterback seasons in NFL history. Aaron Rodgers is another player who was underrated by the system; given the success of Newton and Rodgers, perhaps we need to consider adding junior college experience to the variable for collegiate games started.
Newton demonstrates where the system can go wrong, while Andy Dalton demonstrates where the system can go right. Dalton was the highest-rated prospect in last year's draft according to LCF and while his numbers (and his potential) don't match Newton's, his rookie performance surprised a number of observers who felt his arm wasn't strong enough to be a good NFL starting quarterback.
It's important to understand that LCF is meant to be a tool used alongside the scouting reports, not instead of the scouting reports. What matters is not which quarterback is ahead of which other quarterback by 100 points. Instead, what's important is who has an overall good or bad projection. Scouts still come first and foremost, but this method is valuable as a crosscheck device and should be part of the conversation about quarterback draft prospects.
With that in mind, let's look at the projections for this year's quarterbacks. These numbers represent an estimate for passing DYAR in years 3-5 of a player's career. The top prospects will be above 1,200 DYAR, and you should avoid quarterbacks below zero. Let's start with the top two guys, two of the highest-rated quarterbacks in LCF history who will also be the first two picks in the 2012 NFL Draft.
Robert Griffin, Baylor: 2,530 DYAR
Important stats: 40 games started, 67.0% completion rate, senior passer rating rose 45.3 points, 161 carries for 644 yards.
Andrew Luck, Stanford: 1,749 DYAR
Important stats: 37 games started, 66.4% completion rate, senior passer rating dropped -0.5 points, 47 carries for 150 yards.
Robert Griffin comes out with the strongest LCF projection of any quarterback we've measured. Here are the top ten quarterbacks by LCF projection since 1998:
Player Year LCF
Robert Griffin 2012 2530
Philip Rivers 2004 2476
Drew Brees 2001 2190
Colt McCoy 2010 2092
Carson Palmer 2003 1973
Peyton Manning 1998 1784
Andrew Luck 2012 1749
Chad Pennington 2000 1678
Brady Quinn 2007 1518
Jason Campbell 2005 1506
Griffin and Luck are basically LCF's dream candidates. They're both longtime starters with tons of college experience. Both have strong completion rates. Both get good yardage when scrambling. The biggest difference between the two according to LCF is what happened in their senior year. Luck, who was stellar as a junior, saw his passer rating stay constant. Griffin, on the other hand, improved significantly. The 45.3-point rise in his passer rating as a senior is largest senior improvement in our database (surpassing Jason Campbell, who rose 40.3 points) and the second-largest senior change in our database (behind only Rex Grossman, whose passer rating as a senior dropped 49.3 points). Statistically, Griffin's senior year was better than Luck's, his junior year not as good. This could indicate that Griffin is still improving, still learning, and still getting better, with more room to grow in the pros. Of course, it also could indicate that Griffin's 2011 season was a little fluky, and one of the arguments I've read against Griffin as a can't-miss prospect is that most scouts didn't have him as a first-round pick before his senior season. With all due respect to those scouts, it was pretty obvious within the first two or three games of the year that they were wrong. And even if Griffin's passer rating as a senior had stayed the same as his passer rating as a junior, Griffin would still have this year's highest LCF projection at 1,994.
Again, this little statistical exercise is not definitive proof that the Colts should draft Griffin over Luck. What's important here is that both quarterbacks come out as top prospects, and unlike with Colt McCoy, the scouting reports match the statistical projection.
One last note: The argument against "Luck and Griffin are about as close to can't miss as quarterback prospects can be" is not "well, people said the same thing about Peyton Manning and Ryan Leaf." We know more now than we did then. Leaf started only 24 games and completed just 55.4 percent of his passes in college. His LCF projection is at -407. If Football Outsiders had been around in 1998, we would have been arguing that Peyton Manning and Ryan Leaf weren't even in the same universe as prospects.
Nick Foles, Arizona: 1,391 DYAR
Important stats: 33 games started, 66.9% completion rate, 43 carries for -103 yards.
Foles is this year's version of Ricky Stanzi, a guy whose strong LCF forecast will probably end up moot because scouts will determine that he's not worth of a pick in the first three rounds. His film from 2011 has apparently dropped him on a lot of draft boards, and he had a poor combine performance. Greg Cosell calls him a "major projection" based on slow arm speed and an inability to drive intermediate-lenth passes. He also has an elongated delivery. I'm not a scouting expert by any means, and I haven't seen Foles play, but the scouting reports on Foles remind me a lot of the scouting reports on Andy Dalton, except that Dalton didn't have a problem with a slow delivery.
Kirk Cousins, Michigan State: 1,362 DYAR
Important stats: 38 games started, 64.6% completion rate.
One interesting note about Cousins is his fluctuating size. He's 6-foot-3 and weighed in at the combine at 214 pounds, which leads to a 26.7 BMI. That's lower than usual for quarterbacks, but not extremely low. However, he played the 2011 season at 202 pounds and 25.2 BMI. The data set used to create LCF v2.0 doesn't have a single quarterback listed below 205 pounds or 25.8 BMI. The team drafting Cousins needs to make sure he keeps up an intensive strength program so he's sturdy enough to take the hits he's going to take in the NFL. The ESPN Scouts Inc. profile of Cousins (note: Insider) lists him with "below average" durability.
Brandon Weeden, Oklahoma State: 1,011 DYAR
Important stats: 25 games started, 69.5% completion rate, 26.8 BMI
Ryan Tannehill, Texas A&M: 730 DYAR
Important stats: 19 games started, 62.3% completion rate, 55 carries for 296 yards.
Given the way they are discussed, you would never know that Ryan Tannehill had almost as many college starts as Brandon Weeden. Sure, Tannehill spent his first two years as a wide receiver before spending a year and a half as the Aggies' starting quarterback. But Weeden spent all those years as a baseball player, then was redshirted, and ended up only starting two full seasons. So while Tannehill may be a more raw talent than Weeden, and he wasn't as accurate as Weeden in college, he has far more potential. The LCF doesn't know that Weeden will turn 29 in the middle of his rookie season. When John Beck came into the league as an overaged prospect, he had 12 more games of starting experience than Weeden has and was three years younger.
Brock Osweiler, Arizona State: 248 DYAR
Important stats: 14 games started, 60.3% completion rate.
The LCF likes this year's quarterback prospects, with one exception: Brock Osweiler. Osweiler is built for LCF to hate. He has a low completion rate and only started one season in college (along with a single game in each of his first two seasons) before coming out for the draft early. LCF doesn't ding him for this, but there also need to be concerns about his height. He was listed at 6-foot-8 during the season, although he measured 6-foot-7 at the combine. Either way, he's taller than any quarterback in the LCF data set. The FO master database lists only two quarterbacks since 1992 who were at least 6-foot-7: Ryan Mallett and Dan McGwire. Dan McGwire also had very few games started (23) and a low completion rate (59.1 percent). You don't want to be compared to Dan McGwire.
The Asterisk
Russell Wilson, Wisconsin: 2,650 DYAR
Important stats: 48 games started, 60.7% completion rate, senior passer rating rose 64.1 points.
I would be remiss if I didn't at least mention the ridiculous projection that the Lewin Career Forecast spits out for Russell Wilson. Yes, that projection is even higher than the one for Robert Griffin. No, it doesn't particularly mean that Wilson is a sleeper prospect. There are a few things going on here that the LCF is just not designed to account for.
First and foremost, the change in Wilson's passer rating between his junior and senior years is insane. Remember that earlier I noted that Griffin had a larger senior year passer rating increase than any quarterback in our data set? Well, Wilson's senior year passer rating increase is 40 percent larger than Griffin's. But does it matter when the quarterback is playing in a completely different offense for a completely different school in his last year of college eligibility? At Wisconsin, Wilson got to pick apart defenses that were concentrating on stopping Montee Ball. At North Carolina State, I doubt opponents were quaking in their boots at the thought of Mustafa Greene and Dean Haynes. It goes without saying that there isn't another quarterback in the LCF data set who transferred between his junior and senior years.
There's also the issue of height, another data point where there's nobody in our data set that can be compared to Wilson. At first, it seems strange that LCF doesn't include a variable to discount short quarterbacks, but when you look at the data set that went into creating LCF the reasons are pretty clear. There's no penalty for being 5-foot-11, like Wilson is, because there are no quarterbacks in the data set who are shorter than 6-foot-0. There's no penalty for being only 6-foot-0 because the two quarterbacks who are 6-foot-0 are Drew Brees and Michael Vick.
Quarterbacks who are Wilson's height simply don't get drafted in the first three rounds of the draft, period. The FO master database only includes three quarterbacks who are below six feet tall: Seneca Wallace, Joe Hamilton, and Flutie. That's a fourth-round pick, a seventh-round pick, and an 11th round pick from 25 years ago. Even if we go all the way back to 1991, the only quarterbacks taken in the first six rounds at 6-foot-0 or shorter were Vick, Brees, Wallace, Joe Germaine (fourth round, 1999), and Troy Smith (fifth round, 2007).
Wilson too will probably be drafted on the third day of the draft, round four or later, which would render his absurdly high LCF moot.
