QBASE 2016

by Andrew Healy

(Ed. Note: Surprise! Our free content version of the QBASE projection article was all ready to go this morning... and then a huge trade between Los Angeles and Tennessee superseded the first few paragraphs. The Cleveland Browns will still get to select a quarterback with their top pick, but it's the Rams who will get to choose between Carson Wentz and Jared Goff. And if rumors are correct, the Rams' choice at No. 1 will likely leave the Browns with the quarterback we thought they should take in the first place. Read on for more details... -- Aaron Schatz)

Draft the wrong quarterback. Fire the general manager. Draft the wrong quarterback. Fire the general manager. The Browns have been stuck in that cycle since coming back into the league as an expansion team in 1999. During that time, the Browns have drafted eight quarterbacks, four in the first round. They have also changed general managers nine times over that stretch, more than any other team. Whether Cleveland's latest GM, Sashi Brown, avoids the early exit of his predecessors depends on his ability to succeed where the others failed. Whatever else he does may not matter if the Browns yet again choose the wrong quarterback.

Fortunately, our Quarterback-Adjusted-Stats-and-Experience (QBASE) projection system is here to help. If it had been around for all the previous drafts since 1999, it would have told the Browns to stay away from Tim Couch, Johnny Manziel, and their other first-round flops. And it has a clear preference for the name the Browns should turn in on draft day this year.

To come up with NFL projections for this year's top quarterback prospects, QBASE looks at college performance, experience, and expected draft position. The last of these is included to account for the scouting information that college stats miss. To allow some time for development, QBASE projects a quarterback's passing efficiency in the third, fourth, and fifth year of his career according to our measure, Defense-adjusted Yards Above Replacement (DYAR). Note that rushing value is not part of this projection. 50,000 simulations produce a range of potential outcomes for each prospect.

QBASE favors quarterbacks expected to go high in the draft who also have a relatively long resume of college success according to the stats. Those stats include completion percentage, yards per attempt, and team passing efficiency. Most importantly, all those stats are adjusted both for the quality of the defenses that a prospect had to face as well as the quality of his offensive teammates.

The projections may seem pretty pessimistic, but remember that the Browns' experience with drafting quarterbacks is just a particularly extreme case of a more general condition: most prospects fail. This year, QBASE only projects one quarterback to be a good proposition to succeed in the NFL, and it's not the one most mock drafts have going to Cleveland.

Changes From Last Year's Model

This year, the model behind QBASE changed little from last year, although I did tweak the formula just a bit in terms of how it weights experience. One thing that bothered me in the original model was that it seemed to over-penalize players who started only one or two college seasons. I felt this was likely overfitting past data. In other words, putting a big weight on years started helped the model come up with more accurate numbers for previous prospects such as Russell Wilson, but some of that is likely a coincidence about the past rather than a useful indicator for predicting the future. If you do the thought experiment of what Wilson's NFL career would look like if he had come out a year earlier, it seems unlikely that he would have been a dramatically worse player.

The updated formula basically adjusts for that idea. Rather than getting the same bonus for each year as a college starter, a prospect now gets a bigger jump going from one year to two than he gets for going from two to three, which is in turn bigger than the bonus going from three years as a starter to four. In the resulting projections, the top four-year quarterbacks (mainly Philip Rivers and Carson Palmer) are clustered closer to the other top prospects than they were previously.

Some corrections to our past data also leave just two quarterbacks in the data who only started one full college season. Last year's model incorrectly listed both Michael Vick and Ryan Tannehill with one year as a starter. With the corrections, only Mark Sanchez and Brock Osweiler qualify to be listed with just one starting season. The projections for Sanchez and Osweiler may be too harsh because of that, but still, no adjustment is going to make QBASE think Sanchez made sense as a top-ten pick.

The other big change to this year's model is the inclusion of FCS quarterbacks who played during the last ten years and were drafted in the top 100: Joe Flacco, Jimmy Garoppolo, Tarvaris Jackson, and soon (presumably), Carson Wentz. Unlike many earlier small-school prospects, such as 2000 49ers third-rounder Giovanni Carmazzi, these four players have easily available college statistics. The QBASE projection for Josh McCown has also moved up because our data now incorporates his senior year as a transfer to FCS Sam Houston State.

2016 QBASE Projections

Carson Wentz (North Dakota State)

Projecting college quarterbacks to the NFL is inexact enough without adding the complication of a collegiate schedule built around FCS schools. Here, we are presenting our projection for Wentz that makes the most generous assumptions we think make sense for the quality of opposing defenses and the quality of his teammates. The projection essentially calls Wentz's overall situation at North Dakota State -- which includes some good surrounding talent that helped win the school a fifth consecutive FCS title -- relatively, but not extremely, favorable for posting good stats. (In building the new model, other FCS quarterbacks in our database are treated similarly.)

Here are Wentz and the quarterbacks since 1996 who have both similar QBASE projections and similar opponent adjustments in the model. The projection sees Wentz as similar to some major-conference quarterbacks, such as Drew Stanton and Akili Smith, who faced relatively weak competition.

The hope here is obviously that Wentz will resemble Joe Flacco, a small-school quarterback whom QBASE would have underrated. But while we should be extra cautious with the projections for players from unique situations such as Wentz and Flacco, QBASE is not generally biased against small-school quarterbacks. In addition to Jimmy Garoppolo, it also gives considerably higher rankings to quarterbacks such as Ben Roethlisberger (1,227 DYAR) and Chad Pennington (1,113 DYAR), even though they have even larger corrections for playing weak mid-major schedules.

Our estimates for Wentz's chances of being a bust, 62 percent, are certainly higher than one would hope to see in a No. 2 pick. His statistical profile (62.5 percent completion rate as a senior with 7.9 yards per attempt, 17 touchdowns, five interceptions) fits with other highly drafted quarterbacks who struggled in the NFL. But we want to be even more cautious than usual with our estimates since Wentz comes from a situation that's hard to put into the proper context. Our model has its misses -- and Flacco, who played in a similar situation at the University of Delaware, is one of the biggest.

Nevertheless, while Flacco is the hope if the Browns draft Wentz, he is also the exception to a broader rule. Successful NFL quarterbacks usually show in college not only throws that leap off the film, but also clear and consistent evidence of efficiency, whether they have played at Miami of Ohio or Miami of Florida. Wentz's projection reflects the kind of thin statistical resume that has often predicted first-round busts in the past.

Jared Goff (California)

While Goff is not a sure thing, his estimated chances of succeeding in the NFL are only a little bit lower than Marcus Mariota's, QBASE's favorite prospect from the 2015 draft. Goff had the kind of numbers in his final collegiate season (64.5 percent completion rate, 8.9 YPA, 43 TD, 13 INT) that successful NFL quarterbacks usually have. He posted those numbers against a good, but not great, set of defenses (ranked No. 30 in college football by our estimates). He had little help in terms of NFL-caliber talent at receiver or on the offensive line. The only teammate at those positions who projects to be drafted in the first four rounds this year or next is potential third-round receiver Kenny Lawler.

Goff becomes QBASE's No. 9 quarterback prospect of the last 21 years. Mariota occupied that same spot last year, but has moved up in the reshuffling since another year of data has been added to the model.

Players from earlier years are part of the model that creates the projections, so the takeaway here is not that Goff is likely to be as good as his fellow Cal alum just below him on the list. Goff placing so highly instead says that his resume resembles those of other prospects who succeeded as NFL quarterbacks. Our projection gives Goff the best shot this year, almost 40 percent, of being the kind of upper-tier player who solves a team's long-term quarterback dilemma.

Paxton Lynch (Memphis)

At first blush, Lynch's 2015 numbers (66.8 percent completion rate, 8.5 YPA, 28 TD, 4 INT) look promising. The bloom comes off the rose, however, when you correct for Lynch running up his numbers against the No. 86 slate of opposing defenses. While QBASE finds enough in the raw numbers for prospects such as Ben Roethlisberger to produce a high projection despite weak competition, Lynch's stats do not quite rise to that bar.

Even with QBASE's pessimism about Lynch's outlook, the projection gives him about a 33 percent chance of being at least an adequate starter. Quarterbacks with those odds of success have historically gone in the second round, so the projection would view it as a reach if, as some mock drafts currently suggest, the Broncos selected him with their first-round pick.

Connor Cook (Michigan State)

Cook projects as a below-replacement-level NFL quarterback. His basic stats in 2015 (56.1 percent completion rate, 7.7 YPA, 24 TD, 7 INT), in an era when college quarterbacks routinely complete more than 60 percent of their passes, recall other NFL flops such as Jake Locker.

Cook did not even face a particularly strong set of opposing defenses. The defenses Michigan State faced ranked No. 51 in our estimates. Moreover, Cook had the benefit of playing with the most NFL-caliber teammates of any prospect in this year's class. Those teammates include All-America left tackle Jack Conklin, likely a first-round pick in the draft this year. That Cook failed to do better in college despite favorable circumstances makes him a long shot to succeed in the NFL despite going in the second round in many mock drafts.

Christian Hackenberg (Penn State)

Compared to Cook, Hackenberg played against a slightly harder college schedule and had less support in NFL-caliber offensive teammates. But Hackenberg projects further below replacement level because his expected draft position is lower and his 2015 stats (53.5 percent completion rate, 7.0 YPA, 16 TD, 6 INT) are even worse.

No quarterback in QBASE's database (top-100 picks since 1996) has succeeded with remotely similar stats. The list of previous top-100 picks with completion rates under 55 percent in their last college seasons -- Brock Huard, Dave Ragone, Kyle Boller, Marques Tuiasosopo, and Quincy Carter -- augurs poorly for Hackenberg's NFL prospects.

(Ed. Note: In the original article on ESPN, we listed Josh McCown as a quarterback with a completion rate under 55 percent in his last season. McCown had a completion rate of just 51 percent in his final year on the FBS level, but a completion rate of 60 percent as a senior transfer at FCS Sam Houston State.)

Dak Prescott

The projection rates Prescott far ahead of Lynch and Cook, prospects currently expected to go earlier in the draft. Often compared to Tim Tebow earlier in his college career, Prescott did not have as much NFL-caliber talent surrounding him, even with projected 2017 first-round wide receiver Fred Ross. Prescott's projection here accounts for that, as well as his performance against the toughest set of opposing defenses (No. 14 in FBS last season) of any quarterback prospect in this year's draft.

QBASE does not predict Prescott to be a likely NFL success. But his 45 percent chance of being at least an adequate starter gives him enough upside to make him very much worth a Day 2 draft pick.

Here are the full historical QBASE projections using the updated model, which also incorporates an additional year of data. Note that some of the system's mistakes, including Matt Ryan, Brian Griese, and John Beck, are discussed in depth in last year's article introducing the system.

(Ed. Note: a condensed version of this article originally appeared on ESPN Insider.)