Tomorrow is game of the century in the NFL. The Denver Broncos travel to the Detroit to take on the Lions. Will the Lions go to 6-2? Will John Elway make a comeback and save Denver’s season?
Oh wait, that’s only the game of the century in the Berri house, where we have a husband who was born in Detroit and a wife born in Colorado. For the rest of the world, the game of the century is the Patriots vs. the Colts. And given the perceived importance of this game, I thought I would add to the hype and devote my weekly QB Score column to this game.
Tom Brady and Peyton Manning
Just how good is Tom Brady? To answer this question we need a measure of performance. Analysts note yards, completion percentage, touchdowns, and interceptions in discussion a quarterback. The NFL’s QB Rating attempts to summarize these statistics in one number.
But as Dr. Z of Sports Illustrated notes (and The Wages of Wins confirms), the QB Rating has its problems. So we need a new metric.
And given that you are reading The Wages of Wins Journal, it should not be surprising that I am going to turn to QB Score. As noted, QB Score is calculated as follows:
QB Score = All Yards – 3*All Plays – 30*All Turnovers
So this model incorporates passing yards, rushing yards, yards lost from sacks, pass attempts, rushing attempts, sacks, interceptions, and fumbles lost.
After eight games, here is how Brady looks via QB Score.
QB Score = 1,401
QB Score per game = 175.1
QB Score per play = 4.88
I have data back to 1994 on every quarterback who played. Across this sample (of more than 1,000 quarterbacks) we see only 71 quarterbacks who have posted a QB Score for an entire season in excess of 1,400. And again, those 71 got to play the entire season. Brady has only played half a year.
The best QB Score posted was 2,559 by Peyton Manning in 2004. Manning in 2004 also posted the best per game mark (159.9) and per play mark (4.78). At this point in the 2007 campaign, Brady is on pace to best all these marks.
What is not getting as much acclaim this year are the numbers posted by Manning. Here is how he looks via QB Score in 2007.
QB Score =967
QB Score per game = 138.1
QB Score per play = 3.93
Here are all the quarterbacks who have posted a per-game mark of at least 138.1 since 1994.
Peyton Manning (2004): 159.9
Daunte Culpepper (2004): 149.3
Kurt Warner (2000): 146.0
Peyton Manning (2006): 141.6
Jeff Garcia (2000): 139.8
That’s it. Five quarterbacks in the past 13 years have posted better numbers than Manning this year. And two of these quarterbacks were Manning.
So the game tomorrow is truly going to feature the two best quarterbacks in the game. But we should note that Manning has had seasons like this before. For Brady, though, appearing at the top of the numbers chart is new. Here is what Brady has done in his career:
2001: QB Score per game (50.0), QB Score per play (1.53)
2002: QB Score per game (70.1), QB Score per play (1.67)
2003: QB Score per game (71.9), QB Score per play (1.92)
2004: QB Score per game (84.9), QB Score per play (2.50)
2005: QB Score per game (109.5), QB Score per play (3.01)
2006: QB Score per game (75.2), QB Score per play (2.04)
Brady has been above average each of the past four years (average QB Score per play is 1.85). But the numbers from this year are simply not consistent with anything we have seen before. Obviously something has changed, and if I had to guess, that something is named Randy Moss.
Consistently Talking About Inconsistency
The impact a motivated Moss has had on Brady’s performance brings us to the second story I wish to tell today. It certainly looks like Brady’s numbers have changed because his teammates have changed. And this leads us to conclude that the statistics we see in football – unlike the numbers we see in baseball and basketball – are not just about the player. A quarterback’s numbers not only reflect his ability, but also the ability of his teammates, the quality of his coaches, and the quality of defenses he faces. The impact of all these factors can cause a quarterback’s numbers to change dramatically from season-to-season and from game-to-game.
To see the game-to-game pattern, consider the Week Eight QB Score rankings. So far this season 25 quarterbacks have attempted at least 14 passes in at least five games. Of these 25 quarterbacks, 18 have had at least one game with a QB Score per play in the top 10 for a week, and one game with a QB Score per play ranked below 20 for the week. In other words, 72% of quarterbacks have already had one “great game” and one “bad game.”
And that highlights the nature of football and the numbers we have to evaluate performance. Last week Brett Favre had an amazing game against the Broncos, posting a QB Score of 7.57 (the eight best performance by a signal caller this season). But in Week Six his QB Score was -0.21 while in Week One his performance was -0.72. Such inconsistency is not only seen in Favre’s statistics. Again, 18 quarterbacks have been at both ends of the rankings. Only Tom Brady, Jeff Garcia, Peyton Manning, Ben Roethlisberger, Carson Palmer, Jay Cutler, and David Garrard have yet to offer a performance that would be thought of as “bad.”
Of course we are only in Week Eight. As the weeks go by, we can expect each of these seven quarterbacks to post a bad game at some point. In fact, we should not be surprised if Brady and/or Manning lose their perfect marks tomorrow.
And the Running Backs…
Way at the top of this post is the overall running back rankings. All I will say about these is that running backs are not entirely consistent from year to year either. At least, Shaun Alexander, Clinton Portis, and Larry Johnson are not posting numbers in 2007 consistent with what we saw in the past. Maybe I should devote an entire post to this topic sometime in the future.
– DJ
For more on QB Score, RB Score and what these metrics mean see
Consistent Inconsistency in Football
Football Outsiders and QB Score
The Value of Player Statistics in the NFL
Wages of Wins Journal 
demcavs
November 3, 2007
Not trying to quibble, but defining “great” as being in the Top 10 might not necessarily be accurate, if for instance only 6 (or any number less than 10) quarterbacks played well in a given week.
Great blog.
dberri
November 3, 2007
I agree. How about “very good”?
T.G. Randini
November 4, 2007
1. Are your all time rankings for QB score normalized? Any ranking must be normalized to the CONTEXT for each year because the ‘game’ changes all the time. Normalization, of course, is Sabermetrics 101… which is beginning freshman year college basic analysis. Data that is not normalized is grade school level analysis.
2. If a QB scores a TD on a 1 yd sneak… he gets –2 yards in your system… and –2 yards/play? Wow… better throw the ball for a TD. Wait… he gets –2 yards for that offense also! Better hand the ball off to someone else to get the system punishment!
3. That’s the problem when you construct a system where the regression is based on simple macro analysis… where Macro = best fit (macro) INSTEAD of Macro = sum (micro 1, micro 2,… micro t).
Kent
November 4, 2007
Is Jon Kitna the best QB in recent Lions’ history?
dberri
November 4, 2007
Kent,
Heading into this week he was only average. And a few quarterbacks managed to get past average (Charlie Batch and Scott Mitchell to name a couple). We will see where he ranks after today. If he keeps playing well he might finish with one of the best seasons in recent history. Which tells us how bad quarterbacks have played for the Lions in recent history.
Kent
November 5, 2007
Kitna is a really good prognosticator. Before the season started he said Detroit would win at least 10 games!