During the last few weeks, I’ve had a lot of fun watching the various members of MHR, MSM and other fan sites quote statistics to support whatever point they are choosing to make. Most often, I have read a large number of posts regarding statistical comparisons between Jay Cutler and Kyle Orton.
I would like to affirm here that I am neither a Cutler detractor, nor a Cutler supporter. I also know next to nothing about Kyle Orton, so I’m not sure I can be called an Orton fan nor an Orton opponent. The reality is, over the years, I have learned to cheer for whomever happens to be under center, as I cheer on the entire offense/defense for a successful game.
I would also like to mention that I love to look at statistics, and I deeply appreciate all of the MHR writers who have taken the time to compile and analyze statistics for the rest of us. Having said that, however, I would like to mention that most statistical discussions of football players/teams tend to be incomplete (I prefer this term to ones like "biased" or "flawed"). I once had a college professor state that given the right set of statistics, he could prove just about anything.
Let me offer up an example. Using the statistics provided by pro-football-reference.com, I looked at the Denver Broncos’ coaches. I reviewed the coaches’ career records in 7 areas: Wins, Winning Percentage, Average Wins Per Season, Average League Offensive Rank (Points), Average League Offensive Rank (Yards), Average League Defensive Rank (Points), Average League Defensive Rank (Yards). Then I ranked the coaches according to those 7 areas to see which of Denver’s coaches should be consider "the best."
(NOTE: I included post season wins/losses in the calculations for each coach)
This is how the coaches rated:
Mike Shanahan 146 Jack Faulkner 9
Dan Reeves 117 Frank Filchock 7
Red Miller 42 Mac Speedie 6
John Ralston 34 Ray Malavasi 4
Lou Saban 20 Jerry Smith 2
Wade Phillips 16
Yet, this ranking is flawed because of the disparity between the coaches in the number of years they coached. Shanahan and Reeves coached for 14 and 12 years respectively, while Smith coached only 5 games.
So, let’s turn our attention to each coach’s winning percentage; that would seem to be a fair way to compare them, wouldn't it?
Career Winning Percentage
Miller .627 Malavasi .333
Shanahan .616 Saban .331
Reeves .596 Faulkner .291
Phillips .484 Filchock .268
Smith .400 Speedie .250
This look is also misleading, again due to the difference between the number of games coached. The more games a coach has, the greater the effect on his winning percentage. For example: Smith has a .400 winning percentage from a 2-3 record. Shanahan's percentage is based on a 146-91 record (or 47 times as many games).
What about average wins per season?
Career Average Wins Per Season
Miller 10.5 Malavasi 4.0
Shanahan 10.4 Filchock 3.5
Reeves 9.8 Faulkner 3.0
Phillips 8.0 Speedie 2.0
Ralston 6.8 Smith 2.0
This doesn't quite seem to do it either.
Maybe we should rank the coaches according to how effective they were in regards to their offensive and defensive schemes as measured by their average NFL offensive and defensive rankings.
Career Average Offensive Rank (Points)
Faulkner 6.0 Shanahan 9.5
Filchock 6.5 Saban 12.0
Phillips 6.5 Reeves 13.3
Speedie 7.0 Miller 15.3
Ralston 8.6 Smith 24.0
It’s interesting to note that Phillips – a supposed defensive guru – had a higher average offensive ranking in points than Shanahan – who is generally looked at as an offensive guru. But this still leaves us with questions. How about offensive yards?
Career Average Offensive Rank (Yards)
Phillips 5.5 Malavasi 9.0
Faulkner 6.3 Saban 10.2
Speedie 6.3 Smith 11.0
Filchock 6.5 Reeves 14.0
Shanahan 7.1 Miller 17.0
Again, note the interesting reversal of Phillips and Shanahan. What about defensive rankings?
Career Average Defensive Rank (Points)
Miller 6.5 Smith 11.0
Filchock 7.0 Reeves 12.0
Faulkner 7.0 Shanahan 13.9
Speedie 7.7 Ralston 16.4
Malavasi 8.0 Phillips 17.5
Once again, note the reversal of Shanahan (offensive-minded) and Phillips (defensive-minded).
Career Average Defensive Rank (Yards)
Malavasi 6.0 Smith 10.0
Faulkner 6.7 Shanahan 11.8
Speedie 6.7 Ralston 13.2
Filchock 7.0 Reeves 14.9
Saban 8.8 Phillips 23.5
If you go by winning percentage, or average wins per season, you can argue that Red Miller was the best Bronco coach to date.
If you go by the average offensive ranking of points scored, Frank Filchock becomes the best coach.
If you go by the average offensive ranking of yards amassed, Wade Phillips becomes the best coach.
If you go by the average defensive ranking of points allowed, Red Miller becomes the best coach.
If you go by the average defensive ranking of yards allowed, Ray Malavasi becomes the best coach.
If you go by the number of career wins, Mike Shanahan becomes the best coach.
. . . well, you get the point.
The issue with statistics is that each argument depends upon the statistics cited in support of the premise. The arguments arise when differing viewpoints choose to cite different statistics, or when said statistics are interpreted differently. Debate occurs when the different sides of the discussion refuse to agree on which statistics to use and/or how to interpret those statistics.
Football is, arguably, the most complex professional sport currently being played in the United States. There are 11 players on a side on the field for any given play. Each player has a set assignment to fulfill. The success or failure of any given play depends on each of the 11 doing their job effectively. A missed block by an offensive lineman can cause an offensive play to completely implode. A mistimed leap by a defensive back can result in a huge gain. Weather and playing surface can influence the outcome of a play. Refereeing can have an influence (need we cite Hochuli here?).
Individual statistics are great for arenas like fantasy football. But I would assert that individual statistics are not a particularly accurate indicator of the effectiveness of an individual player within the context of an NFL football game. Was the quarterback’s completion percentage a result of the passer’s competence, or were a lot of passes dropped, routes incorrectly run, balls tipped, etc. Could a running back have a great game without the other players throwing the correct blocks at the correct time?
I would humbly suggest that the only truly meaningful statistical discussion one can have regarding NFL players are those that include, in some fashion or another, all of the components that make up the plays. Anything less gives a skewed view of how good (or bad) an offense/defense is.
I’d love to hear other readers’ takes on the statistics issue.