Analysis of the Enemy: Opposing Quarterbacks
In The Wages of Wins, Southern Utah University professor David Berri unveiled a new system for measuring quarterback play that distilled a quarterback's primary duty-- maintaining control of the football-- into a tidy number using a formula he developed. The system, entitled QB Score, rewards a quarterback for simply gaining yards, through air or ground, and heavily punishes a quarterback for turning the ball over.
QB Score= Total Yards - (3 x Plays) - (50 x Turnovers)
Following the release of his book, which offers alternative statistical systems for various sports, Berri received flak for some pretty hollow criticism of NBA stat guru John Hollinger, but his methodology for ranking quarterbacks works well as a companion to other measures.
Unlike passer rating, QB Score rewards quarterbacks for their rushing contributions, and it lacks the inflated value touchdowns add to the equation. Unfortunately, developing a useful formula is only a small part of the battle in college football analysis-- with 120 FBS teams, acquiring the data for any statistical system is an often arduous task. Due to the efforts of SB Nation's California Golden Blog, however, the QB Score battle has already been fought and won. The Golden Blog compiled QB Score data for every quarterback in every FBS conference, and I've used those results to break Ohio State's 2011 quarterback opponents into tiers.
Tier One: The Positive Producers
| Player | QB Score | Completion % | Y/A | Interception % | Passing TDs |
|---|---|---|---|---|---|
| Denard Robinson (Michigan) | 981 | 62.50% | 8.8 | 3.80% | 18 |
| Taylor Martinez (Nebraska) | 472 | 59.20% | 8.3 | 3.60% | 10 |
| Nathan Scheelhaase (Illinois) | 146 | 58.70% | 6.9 | 3.00% | 17 |
| Kirk Cousins (Michigan State) | 55 | 66.90% | 8.4 | 3.00% | 20 |
| Terrance Owens (Toledo) | 34 | 59.50% | 7.6 | 3.10% | 13 |
According to QB Score, the Big Ten was a QB strong league last season, placing nine starters in the positive rankings. Northwestern's Dan Persa led the way (1,102), but Ohio State does not play Northwestern this season, leaving Denard Robinson as the top returning quarterback. Robinson was the Michigan offense last season, accounting for 67% of the Wolverine's yards and scoring 32 touchdowns.
In most cases, a returning player with Robinson's productivity and exposure would be an early candidate for Heisman chatter, but new Michigan offensive coordinator Al Borges abandoned Rich Rodriguez's spread-option scheme in favor of a multiple offense with West Coat roots, and it's unclear how Robinson will respond. To his credit, he passed well in the spread-option, although his interception rate is the highest of the quarterbacks listed with a positive QB Score.
Interception rate, in fact, seems to separate the really successful pure passers from the passers who do well enough but are more valuable for their athleticism. Like Robinson, Taylor Martinez operated in an option offense (although not nearly as proficient of one) and passed well enough to overcome some weak peripheral numbers. His completion percentage, 59.20%, and his interception rate, 3.6%, are not good, but Martinez overcomes them by piling up yards. An 8.3 yards per attempt figure makes him an opportunistic passer, even if his mechanics prevent him from becoming a proficient one.
Clearly, Kirk Cousins is the best pro-style passer in the Big Ten, and he's a bit penalized by QB Score because of it. Dual-threat quarterbacks essentially receive double the opportunities to score under the formula-- a limit of QB Score-- so his number is a bit artificially low. Cousins peripheral numbers are excellent; a high completion percentage, yards per attempt figure, and a low interception rate makes him one of the premier passers in all of college football.
For what it's worth, Terrelle Pryor had a QB Score of 502, an excellent figure that would have placed him third in the Big Ten for returning starters, behind only Persa and Robinson.
Tier Two: The Mediocre to Bad
| Player | QB Score | Completion % | Y/A | Interception % | Passing TDs |
|---|---|---|---|---|---|
| Russell Wilson (Wisconsin) | -13 | 58.60% | 6.8 | 2.67% | 28 |
| Robert Bolden (Penn State) | -37 | 58.00% | 7 | 3.60% | 5 |
| Matt McGloin (Penn State) | -48 | 54.90% | 7.2 | 4.20% | 14 |
| Rob Henry (Purdue) | -305 | 53.10% | 6.1 | 4.30% | 8 |
| Tyler Hansen (Colorado) | -402 | 68.30% | 6.7 | 3.65% | 6 |
| Austin Dantin (Toledo) | -539 | 66.10% | 6.5 | 4.20% | 7 |
| Stephen Morris (Miami) | -570 | 53.60% | 8.1 | 5.90% | 7 |
Could the Russell Wilson hype be off? Wisconsin's new quarterback transferred from NC State, choosing the Badgers over Auburn and wizard offensive coordinator Gus Malzahn. Wilson brings a level of mobility that Wisconsin usually lacks at quarterback, and his acclimation to the West Coast-ish Badger offense will be interesting to watch in the early part of the season.
For all his suitors, however, Wilson's numbers are remarkably unremarkable. A mediocre completion percentage and low yards per attempt figure indicates an inefficient passer who produces only through plentiful opportunity (526 passing attempts in 2010).
Penn State's quarterback battle could end in a tie, but I don't see why it should. Not only did Bolden outperform McGloin in key categories-- completion %, interception rate, debilitating pick-sixes-- but Bolden's a true sophomore with great natural ability and room to grow. McGloin's a redshirt senior walk-on whose best attribute seems to be some vague "moxy" that endears him to the coaching staff. In this case, a tie isn't really a tie, since the player with the long-term potential is the clear best choice for the program.
Toledo has a quarterback competition of its own, but there's more statistical grey area in their battle. Junior Austin Dantin began 2010 as a starter, but a midseason injury opened the door for sophomore Terrance Owens, who outperformed Dantin in yards per attempt and interception rate while providing a mobile threat with his legs. Dantin's completion percentage was about 7% higher than Owens, though, making the quarterback decision difficult for third year head coach Tim Beckman.
The offense as a whole performed better under Owens than Dantin, according to Bill Connelly's S&P+ measure, perhaps due to the Rockets' focus on rushing the ball. Toledo ran more on both standard and passing downs than the national average, and if Beckman wants to continue with the rushing emphasis, he may choose the dual-threat Owens.
Tier Three: The Truly Terrible
| Player | QB Score | Completion % | Y/A | Interception % | Passing TDs |
|---|---|---|---|---|---|
| Jacory Harris (Miami) | -1,327 | 54.80% | 6.6 | 5.50% | 14 |
| Patrick Nicely (Akron) | -1,513 | 49.40% | 5.2 | 3.80% | 10 |
Avert thine eyes! One quarterback on this list holds multiple excuses for scoring so poorly, the other doesn't. Sophomore Patrick Nicely survived a turbulent season under Akron's first year head coach Rob Ianello, transitioning to a pro-style system after spending his freshman season and entire high school career in a spread attack. Nicely led a team drained of talent to a 1-11 finish, staving off a winless season only by defeating Buffalo in the final game of the year, which coincided with Nicely's best performance.
Nicely may not start at Akron this season-- he's competing with transfer Clayton Moore and true freshman Zach D'Orazio-- but his circumstances can at least explain the poor 2010. Jacory Harris can't really explain anything about his poor season. From four interceptions against Ohio State to a benching in favor of a true freshman, Harris may have performed the worst of any preseason Heisman contender ever. With the lowest QB Score of any BCS quarterback, Jacory Harris' days as rising star are ignominiously over.
9 comments
|
0 recs |
Do you like this story?
Comments
Dan Persa Did NOT Graduate!!
He was injured for the last 3 games last year, and is @ 80% now, but expected to play opening day. You need to re-work all your numbers – If Persa is 100%, he has to be #1, or close to #1 as a returning QB.
Persa is indeed back
and per last year’s numbers, he’d easily be number one.
However, if he’s still openly walking with a limp less than two weeks from the first game (per Pat Fitzgerald himself), I doubt he’ll perform like he did last year.
I’m surprised to see Wilson so much lower than, say, Scheelhaase. He has comparable completion percentage and Yds/Att, and a better interception rate. I guess since the formula doesn’t take scoring into account (for better or for worse, I guess), Wilson’s numbers suffer.
His supporting cast will be infinitely better at Wisconsin than it was at NC State, so I wonder how much better his number will be this season.
He also gets to play a much lesser schedule at Wisconsin
'11: Minimum Goal: Win 10 games again
'10: 7th in offense, 41st in defense. Division Champions. 10-4. (6-3)
'09: 3rd in offense, 107th in defense. 7-6 (4-4)
Tomahawk Nation Nole-Holds-Barred Analysis of FSU Sports!
Follow Tomahawk Nation's Twitter feed!
by Bud Elliott on Aug 24, 2011 10:47 AM EDT up reply actions
Tyler, what are your thoughs on taking YAC away from numbers that evaluate QBs?
On the one hand, ball placement can increase a WR’s probability of making some YAC. On the other, if you throw to Calvin Johnson and he drags a tackler for 6 yards, that doesn’t seem like it should be a credit to the QB.
Then, you factor in constraint plays like the draw, which only work because of the threat of the pass but never get credited to the QB. And various types of screens that are probably run to keep defenders from cheating against specific passes and/or runs—these could be credited/not credited to the QB when it isn’t/is due.
So does elimination of YAC from advanced stats make sense? Or does it make more sense just to keep it in there? Or should we just accept the limitations of statistics in football?
"Statistics are like bikinis. What they reveal is suggestive, but what they conceal is vital." - Aaron Levenstein
I’m not a statistician, so there are better people out there to answer that question, maybe even on this site, but my feelings on football statistics in general is, they better be specific if they’re not measuring new data in a new way. For example, QB Score measures quarterback production in a very specific way— how well do you personally move the ball while controlling it— which allows us to account for its limitations, like undervaluing Kirk Cousins.
Sabermetrics work because baseball is the perfect numbers game— discrete battles between batter and pitcher break the data into clear categories, making it both easier to gather and apply. Basketball and football don’t have that advantage, but Dean Oliver and Ken Pomeroy have distilled the essence of winning team basketball into four categories that generally reveal good teams over bad ones.
Basketball lacks the depth of statistical analysis that baseball has, although there’s a lot of high-level stuff going on in basketball that we don’t see— like the Houston Rockets’ proprietary measures— because it requires time and resources that fan websites can’t devote to acquiring the data. I’m sure certain baseball teams do the same, but the divide isn’t as sharp in baseball as basketball. There are many gritty, detailed conversations fans can have in baseball— like debating the exact value of a utility infielder— that do not happen in either basketball or football with the same level of precision.
Football suffers from the basketball problem to a great degree. Bill Connelly’s work at Football Outsiders and Football Study Hall is easily the most interesting advanced football stuff existing today, but there’s a data accessibility problem that makes incorporating it into team breakdowns fairly difficult.
So, that was a circuitous way of saying, accessible data is more important than formulas. Yes, I think that we need a statistic that doesn’t give YAC credit to quarterbacks, but we could also use one that does, and then compare the difference. But what we need more than anything is someone who will measure the YAC-less QB yardage of 120 FBS teams and make that information freely accessible.

by 














