Two more blowouts this weekend have us asking: has this been the worst set of playoff games we've ever seen? Plus, DVOA and DYAR ratings from Sunday.
25 Aug 2004
Accurately measuring defensive performance poses a great challenge to all statistical sports research. Football is no different than baseball and basketball in that regard. Football researchers do have a leg up on their counterparts due to the number of traditional defensive statistics found in game recaps. With a little effort, we can find numbers such as: tackles (assisted and unassisted), tackles for loss, interceptions, pass deflections, sacks, forced fumbles, and fumbles recovered. And while it's not a complete or perfect list, it can be fairly useful in gauging individual defensive performance.
Cornerback may be the most important defensive position in the NFL. Football talking heads stress how important the running game is, but in the NFL pass offense and defense are the keys to winning. A secondary that sports two solid cornerbacks has a ready-made advantage in professional football. But what makes a good cornerback?
Deion Sanders was often criticized for his inability and/or reluctance to make tackles in run support. Was the criticism fair? Probably not. Sanders, like all great cover corners, wasn't paid to help stop opposing running backs. His job was simply to prevent his man from receiving the ball. If they did throw to his receiver, he often picked it off or at least deflected it away. Those are the traits of a great NFL cornerback.
Most tackles made by a cornerback come in the passing game after their wide receiver has caught the ball. Unless playing in a deep zone coverage to prevent a first down or touch down, this cornerback has already failed. Which leads to an important assumption for this article: Tackles are a negative statistic for cornerbacks. Interceptions are, of course, a positive statistic. Pass deflections are positive too. And it would be unjust to penalize corners for sacks or tackles for loss. With these assumptions in mind, we can reasonably ascertain what type of statistical profile a good cover corner would possess.
Personally, I prefer to keep such statistical measurements simple. Here's the formula for my Cornerback Coverage Rating:
Interceptions are obviously worth more than pass deflections, but they tend to vary wildly year from year, whereas pass deflections are more consistent. So I arbitrarily decided to value them equally. Good corners should post lower ratings than their less talented colleagues. Let's see how it works.
It's hard to find a better testing ground than Hawaii. Last year's Pro Bowl, supposedly, featured six of the league's best cornerbacks.
|Player||Cornerback Coverage Rating (CCR)|
To paraphrase the old Sesame Street bit, "Can you see which two don't belong?" I will cut Troy Vincent some slack because of his previous exemplary service and Father Time (Vincent's average CCR from 1994 to 2001 was a healthy 2.75). But Champ Bailey, by CCR standards, did not have a good year. Looking at Bailey's career, however, 2003 looks anomalous. Champ Bailey's career CCR numbers:
|Year||CCR for Bailey|
What went wrong in 2003? Champ posted his highest tackle total since entering the league (71). And his pass deflections and interceptions total took a big drop (from 26 in 2002 to 11 in 2003). Perhaps the Redskins were content to play more zone coverage last season. Such a huge rise in CCR isn't unprecedented.
Pro Bowl CB Chris McAllister had a similar season in 2001.
|Year||CCR for McAllister|
Chris McAllister was able to regain most of his previous CCR form. It will be interesting to see if Bailey can do the same.
There are some obvious limitations to this metric. Corners toiling in zone packages will always be undervalued. Also, team tackle totals tend to vary wildly from squad to squad (in the colleges this disparity is magnified). In 2003, the Dallas Cowboys recorded over 300 fewer total tackles than the New York Jets. Depending upon whether the team recorded more or fewer tackles than would be expected given their total number of defensive plays, we can adjust a cornerback's total tackle number. For example, the 2003 Philadelphia Eagles recorded 13.15% fewer tackles than you would expect from their team defensive plays total (885 tackles on 1019 defensive plays). Troy Vincent and Lito Sheppard are both penalized by adding 13.15% to their respective tackle totals. The Indianapolis Colts were credited with 4.08% more tackles than expected, so Nick Harper and Walt Harris are rewarded by subtracting from their tackle numbers. This new metric, let's call it Adjusted Cornerback Coverage Rating (ACCR), is Byzantine and doesn't lead to any great rankings shakeups, but in some instances it produces smoother results.
Upon initial research, it seems elite corners post ratings below a 3.00 almost annually CCR and ACCR are not the ultimate tools for evaluating cornerbacks. I'm sure as football analysis evolves, they will become obsolete. But I feel they at least give us something to start with.
Click here for a list of CCR ratings from 2003 for all cornerbacks registering at least 30 tackles.
Editor's notes: Thanks to Michael for submitting this interesting new statistic. There's a lot of potential here, particularly when our database is detailed enough that we can separate tackles in the passing game from tackles in the running game to remove a penalty from cornerbacks who play behind really bad front sevens.
I'm a bit mistified at Michael's findings given my article back in April regarding how different teams performed against the opposing offenses' best receivers. This analysis seemed to indicate that Champ Bailey was far and away the best member of the Washington secondary. One problem with Bailey's rating may be that pass deflections are a bit of a nonstandard statistic. As Len Pasquarelli writes here (in section on Tedy Bruschi): "The 'passes defensed' category is one of those esoteric numbers, with the statistic not yet officially recognized by the league, one in which the criteria varies from team to team." Bailey's "passes defensed" dropped from 19 to 7 last season, but Fred Smoot's also dropped from 10 to 5. Perhaps there was a change in how the number was counted for the Redskins.
Michael Knight is a graduate student at Western Kentucky University studying history. He is a fan of any NFL team the "Swami" picks to lose against the spread.