TCU has played much better in the second half of games this year. What other schools have seen dramatic shifts of play after halftime?
18 May 2007
In Pro Football Prospectus 2006, we ran an article by David Lewin that introduced a new system for projecting rookie quarterbacks based on college statistics. As far as we know, this research was the first of its kind. We've talked a lot about the college quarterback projection system since then. We feel it is one of the signature pieces of Football Outsiders research, on the same level as DVOA ratings, the third-down rebound theory, and the inconsistency of fumble recovery.
As some of you may know, three weeks ago Gil Brandt wrote an article on NFL.com previewing quarterbacks for the draft. This article contained a section about projecting rookie quarterbacks based on college games started and completion percentage that resembled David Lewin's research.
Since this article ran on April 26, FO has not commented on the Brandt article publicly. However, it is now three weeks since this article ran. Although our editors at ESPN The Magazine are still trying to settle this matter with NFL.com, we feel that we need to say something before our research is permanently identified with another writer.
We hope that two article excerpts will help people understand why we feel so strongly about this issue, and why there is more to it than coincidence or similarity in concept.
First, these are the first two paragraphs in an article written by Aaron Schatz for the ESPN NFL Draft 2007 magazine. You can find it on page 31, at the start of the section on quarterbacks:
In 1998, the Indianapolis Colts faced one of the most important decisions in franchise history. With the No. 1 overall pick, would they take Manning or Leaf? Peyton proponents argued that he was more mature and accomplished. Leaf backers argued he had the stronger arm and more growth potential. But all the Colts really needed to know was four numbers: Manning started 45 college games, completing 63% of his passes. Leaf started just 24 games, completing 54% of his passes. The future couldn't be more clear.
It's true: College statistics really can predict NFL performance. For our 2006 Pro Football Prospectus, we studied 10 years' worth of drafts and discovered that the greatest indicator of NFL success for QBs taken in the first two rounds is the number of college games they started. Philip Rivers, for one, started 51 games at NC State. Donovan McNabb started 49 college games and Carson Palmer started 45. On the flip side, busts-to-be Joey Harrington (28), Jim Druckenmiller (24), and Akili Smith (19) had relatively little starting experience.
Second, this is the fourth paragraph in an article written by Gil Brandt on NFL.com, posted April 26:
There seem to be two important predictors for success when drafting quarterbacks. One is games started. Peyton Manning started 45 games and completed 63 percent of his passes. Ryan Leaf started 24 games and completed 54 percent of his passes. Philip Rivers started 51, Donovan McNabb 49 and Carson Palmer 45; all three have quarterbacked teams to playoff games. Joey Harrington started 28 games, Jim Druckenmiller 24 and Akili Smith 19. All were first-round picks in the past 10 drafts, two are no longer in the league and Harrington will be playing for his third team in three years.
Readers can come to their own conclusion about what is going on here, but we would like to point out two oddities:
285 comments, Last at 22 Apr 2009, 5:55pm by andy3887