Matthew Stafford and the Lewin Projection System

In Football Outsiders latest Four Downs feature, Bill Barnwell uses the Lewin Projection System as criteria to evaluate Matthew Stafford.

If they select Stafford, he'll become another player in the long line of Lions draft fiascos from this decade.

How can we be so sure? The answer has to do with the research first presented at FootballOutsiders.com by David Lewin -- namely, that the only two stats that matter for a college quarterback are games started and completion percentage. For quarterbacks taken in the first two rounds, the numbers are inextricably linked to NFL success.

In his three seasons at Georgia, Stafford started 33 games and completed 57.1 percent of the passes he threw. That puts him in some pretty uninspiring company.

QBs similar to Stafford
Quarterback	Starts	Comp Pct.
Patrick Ramsey	38	58.9
Jake Plummer	40	55.4
Shaun King	39	55.5
J.P. Losman	27	57.8
Matt Ryan	32	59.9
Matt Stafford	33	57.1

Yes, reigning rookie of the year Matt Ryan appears on that list, but with a bit of an asterisk. Ryan played for a Boston College team that had no running game and threw 654 passes in his final season, skewing his completion percentage some. Even if Ryan doesn't come with an asterisk, a one-in-five shot of picking a quarterback of Ryan's caliber shouldn't encourage Stafford's selection in the slightest.

The Lewin Projection System, for the unfamiliar, is a system devised to project a college quarterback's pro performance based on just two college stats: games started and completion percentage.The system, though built on a somewhat shaky foundation, has been reasonably successful since it was introduced in 2006. Because it doesn't move from causation to correlation, but correlation to causation, the reasons for its effectiveness are a bit speculative.

On paper this paints a bleak portrait of Stafford, but Barnwell makes a significant mistake. In 2007, Lewin adjusted his projection system to weight quarterbacks selected in the first sixteen picks more for games started, and quarterbacks taken in picks 17-64 more strongly for completion percentage. Ramsey, Plummer, King and Losman were selected 32nd, 42nd, 50th and 22nd. Assuming Stafford is taken in the top ten, and few dispute it, he's only true comparable on the list is Matt Ryan.

Barnwell attempts to explain away Ryan's presence by mentioning Boston College's poor run game and Ryan's very high attempts in his senior season. It's not a very strong argument. Attempts are dismissed by the Lewin Projection System. Ryan's abnormally high number of attempts should provide a truer portrait of Ryan's completion percentage, not a less accurate portrait, and predictably enough, Ryan's completion percentage in 2007 (59.3%) is very similar to his completion percentage in his first three seasons (60.1%) and his career completion percentage (59.9%).

He then begs the question:

Stafford was directly preceded at Georgia by the recently retired David Greene; both spent their entire college careers under head coach Mark Richt in similar offensive systems. Stafford's college numbers are actually worse than Greene's, with the latter completing 59 percent of his passes and averaging 8.01 yards per attempt to Stafford's 7.83. If Stafford was really a star in the making, wouldn't he have put up better numbers, in the same system, than a guy who washed out of the NFL without taking a professional snap? If it was our $25 million guaranteed, the answer would need to be yes.

As if Greene and Stafford played on perfectly similar teams against perfectly similar opponents. Despite that, it seems like a tough argument to refute if only because it's baffled by so enough qualifiers as to eliminate most comparisons. We need two pro prospects, both who played for the same coach, in the same offensive system and probably in the last ten to twenty years. One pairing springs readily to mind though: Carson Palmer and Matt Leinart. At USC, Leinart completed 64.8% of his passes for 8.59 yards per attempt. Directly preceding Leinart at USC, Palmer completed 59.1% of his passes for 7.53 yards per attempt. Palmer first started under Paul Hackett, and it was his worst season, but even under only Pete Carroll, his numbers are inferior to Leinart's: 61.3% completion percentage and 7.75 yards per attempt. One could then say, if Palmer was really a star in the making, wouldn't he put up better numbers, in the same system, than a guy stuck at backup after three seasons in the league? Well, you could, but it wouldn't be very compelling.