This week's DVOA commentary is all about worsts. Come find out where Washington stands among the worst special teams in DVOA history, whether San Diego has the biggest gap between offense and defense, and whether Baltimore or Jacksonville has the worst running game we've ever tracked.
10 Sep 2005
Guest column by Sean McCormick
At least one debate has raged on football websites this summer that has nothing to do with whether Tom Brady or Peyton Manning is better. Instead, the subject of the debate has been star Denver cornerback Champ Bailey. Going into last season, Bailey was widely considered the best cornerback in the league, a player capable of controlling one side of the field on his own with his cover skills, and the only thing up for debate was just how lopsided was the trade that brought Bailey and a second round pick to Denver in exchange for Clinton Portis.
Denver opened the year strong, going 5-1 and playing excellent defense, during which time Bailey's reputation remained intact. But something funny happened on the way to the Pro Bowl: Bailey played in two nationally televised games, first a Monday Night game on the road at Cincinnati, and then a Sunday Night game at home in the snow against Oakland, and in each game he was absolutely torched. Chad Johnson caught seven passes against Bailey for 149 yards and a touchdown. Jerry Porter did just as much damage, catching six balls for 135 yards and two touchdowns.
Those two games didn't destroy Bailey's reputation â€” at the end of the regular season he still received a Pro Bowl nomination for his play â€” but they did raise questions. And when Denver went into Indianapolis on wild card weekend and gave up 49 points, those questions became louder. Bailey couldn't really be blamed for the result; he spent most of the day lined up across from Marvin Harrison and watching helplessly while Reggie Wayne made Roc Alexander infamous throughout greater Colorado. Nevertheless, Bailey's presence in the lineup did nothing to help Denver defend the Colts' explosive pass attack, a fact that was made even more difficult to ignore when the Patriots smothered that same Colts attack the next week with a patchwork secondary.
So, let's just get the questions out in the open. Is Champ Bailey overrated as a player? Is he one of the best corners in the league or not? And did his presence on the Broncos make any difference at all last year?
Bailey's supporters â€” and there are many â€” tend to bring up the same few arguments to explain away the very public lapses in Bailey's game. They talk about his physical ability, usually by bringing in a quote from an NFL player or scout that declares Bailey to be an unparalleled cover man. They suggest that Bailey was asked to play on an island all year long, matched up one-on-one against the best receivers in the league. They play up other elements of his game, such as his tackling ability. And when in doubt, they fall back and insist that last year really wasn't as bad as it looked, and that if you watched the Denver defense play, you could see that Bailey was making a big difference. Sporting News' NFL columnist Dan Pompei recently wrote an article that enumerated every one of these points.
Pompei touches all the bases. He begins with a quote from Peyton Manning praising Bailey's â€œunbelievable cover skills.â€? He cites the STATS, Inc numbers showing that Bailey allowed 49 completions on the 83 balls thrown in his direction, a 59% completion percentage, but then turns around and insists that this number is not so bad considering the amount of man coverage Bailey was playing.
Pompei makes no attempt to provide any sort of context for this assertion. He doesn't provide the completion percentages of any other corners to compare with Bailey's, nor does he provide any evidence suggesting that corners playing man defense give up a higher completion percentage than those playing primarily zone coverage (the statistical evidence provided in Scientific Football 2005 suggests otherwise).
Pompei then pulls out his big guns, declaring that Bailey was assigned to shadow the opposing team's top receiving threat in 11 games last year, and that in 8 of the 11 games, Bailey held that receiver below their season average in yards. Only Chad Johnson, Jerry Porter, and Peerless Price (yes, that Peerless Price) had better-than-average games against Bailey. Pompei mentions Bailey's 84 tackles, fourth highest on the team. He explains away the gaffes against Cincinnati and Oakland, and declares, â€œCritics used those three plays to define Bailey's season, even though he was on the field for about 1,000 snaps.â€?
Then, curiously, Pompei hedges his bets at the end, noting that Bailey shouldn't be expected to handle Randy Moss one-on-one next year, that the rules currently favor the receiver over the defender so much that the days of the shutdown corner are over, and that Bailey is, considering the offensive environment, about as good as a corner can be.
Despite the high profile nature of playing cornerback in the NFL, and despite all the attention (and money) lavished upon the position by NFL teams, cornerback is arguably the most difficult position for fans to evaluate. There are a number of reasons for this. The vast majority of football fans watch the games on television, and ninety percent of cornerback play is invisible on television broadcasts. The corners disappear just after the snap, and unless the ball is thrown in their direction there is no way to know how well the corner was handling his assignment.
The problem is compounded by the lack of a significant statistical footprint for the position. There are benchmark numbers for most positions that a casual fan can look at to ascertain how a corner is playing. A quarterback might throw for 300 yards in a game or 3,000 yards in a season; a running back can average above 4.0 yards per carry; a linebacker can make 100 tackles in a year. Cornerback has no equivalent mark, and a brief glance at Bailey's conventional stat line gives very little indication as to what kind of year he had:
|Champ Bailey 2004 Traditional Statistics|
|Games||Tackles||Assists||Sacks||Fum. F.||Fum. R.||Int.||Pass Def.|
Tackles don't tell you much; it might mean that a player is solid in run support or that he is making tackles on receivers downfield instead of preventing receptions. Corners aren't likely to rack up big sack numbers, particularly if they don't play nickelback. Passes defensed and interceptions seem like the most relevant stats, but the best corners are likely to have a low number of each simply because teams don't throw in their direction all that much. What we need, in short, are better ways of tracking cornerback play.
Thankfully, several people are trying to address this need, but taking distinctly different routes to do so. Football Outsiders and Roland Beech of Two Minute Warning have built upon the successful play concept first pioneered by Pete Palmer, Bob Carroll, and John Thorn in The Hidden Game of Football and analyzed the conventional statistics in a way that sheds more light on how effective a cornerback's play was in the context of the defense as a whole. The defensive unit DVOA information (explained here) provided in Pro Football Prospectus 2005 provides a more sophisticated means for examining how successful a defense was at stopping each part of a passing attack. In his book Scientific Football 2005, KC Joyner has created his own set of statistics that fill in many of the gaps in conventional measurements.
The defensive playmaker statistics featured in both Roland Beech's Two Minute Warning site and in Pro Football Prospectus 2005 are probably the best place to start. The numbers do not show the opportunities a defensive player has to make plays, but they do show which players are most active on the field. They also provide welcome context, as they take Bailey's conventional statistical line and recast it to better demonstrate how Bailey's play contributed to Denver's defensive success.
For those unacquainted with the units of measurement, plays represent any play that the defender was involved in, stops refer to how many of those plays were successful at preventing offensive success (defined as 45% of the necessary yardage on first down, 60% on second down, and 100% on third or fourth down), defeats refer to the number of plays where the defender forced a change of possession, made a stop behind the line of scrimmage, or forced a fourth down, and stop rate percentage simply denotes the percentage of total plays that were stops.
Here is Bailey's PFP line, as compared with every other starting cornerback last year:
|Champ Bailey 2004 PFP Numbers|
|Plays||Stops||Defeats||Avg. Yards||Stop Rate|
|Rank||t. 9th||t. 3rd||t. 3rd||t. 40th||t. 6th|
By this set of measurements, Bailey was an elite corner in almost every respect. His stop number, his defeats, and his stop rate percentage were all tremendous. The Denver pass defense had a DVOA of â€“32% last year, second in the league, and that number was undoubtedly bolstered by having a cornerback who was involved in a stop on 48% of his plays, and a defeat on 26% of his plays.
The average yards per play jumps off the page, but all it really does is confirm what everyone already knew â€” Bailey had a difficult time handling some deep throws last year. When Bailey was beaten, he was beaten for big yardage, but the rest of the time he was active and very effective. Looking at this stat line, one would be tempted to conclude that the Bailey supporters are right and that's that.
But there are still a few things that are troubling. The play number is pretty high for an elite cornerback â€” by comparison, Chris McAlister was involved in 56 plays last year, Patrick Surtain in 70, Sam Madison in 59. Still, there were quite a few elite cornerbacks who had fantastic years while being involved in as many plays or more, including Nate Clements (94), Dunta Robinson (106), Sheldon Brown (109) and Ronde Barber (109). What's more surprising is that Bailey was involved in the same number of plays as Denver's other starter, Kelly Herndon, suggesting that teams did not shy away from throwing at Bailey, even with a less regarded player on the other side. But the most surprising thing of all is that you can see why teams were going after Bailey: because as good as his PFP numbers were, Kelly Herndon's were better in every respect.
|Kelly Herndon 2004 PFP Numbers|
|Plays||Stops||Defeats||Avg. Yards||Stop Rate|
According to the PFP numbers, Kelly Herndon was without question the most effective cornerback in the NFL last season. His stop rate was off the charts â€” the next closest corner was Fred Smoot at 51%, and Smoot benefited from being involved in 17 fewer plays. If you just glanced at these PFP numbers, and you happened to know that Denver ranked 6th in the league last year in pass defense, or that they ranked 4th in DVOA against the pass, you would think that the Broncos had the best set of corners in the league, and that they were probably trying to nail down a long-term extension with Herndon right now.
But of course, that's not what happened this off-season. The Broncos had a choice as to which free agent cornerback to place a first round tender on, and they decided to put it on Lenny Walls, a player who performed terribly in 2004. Teams were handing out ten million dollar signing bonuses to cornerbacks at a drop of a hat â€” Ken Lucas, the man Herndon was brought in to Seattle to replace, received a whopping $14 million bonus from Carolina. So how is it that a player that seemingly played so well at an impact position like cornerback was overlooked (only in the NFL can a five-year, $15 million dollar contract with a $4.5 million signing bonus be considered getting â€œoverlookedâ€?) while teams threw bouquets and cash at the likes of Anthony Henry?
There are two possible answers. The first is that the Seattle Seahawks just pulled off a major coup, the defensive back equivalent of the Chiefs' signing Priest Holmes away from Baltimore. If that's the case, we won't know it until this season when the Seahawks secondary improves radically and the team walks away with the division. The second possibility is that there is important information that is being missed by the PFP numbers. And until we can explain what is going on with Kelly â€œLockdownâ€? Herndon, we can't accept Bailey's PFP numbers as being a truly accurate reflection of his play.
We already know that the PFP numbers, like the conventional statistics they are based on, do not take into account the quality of opposition. We know that Bailey was matched up against the #1 receiver in most of Denver's games. That leaves the #2 receiver for Herndon, and a quick glance at Denver's schedule shows that the Broncos played a steady diet of teams without viable #2 receiving threats. Look at how the receivers rated for each of Denver's opponents:
|#1 and #2 Receivers on Denver's 2004 Schedule|
|Team||#1 Receiver||DPAR||#2 Receiver||DPAR|
|KC||Eddie Kennison||24.7||Johnnie Morton||19.3|
|JAC||Jimmy Smith||19.7||Reggie Williams||-7.7|
|SD||Eric Parker||16.2||Reche Caldwell||4.4|
|TB||Michael Clayton||39.8||Joey Galloway||13.0|
|CAR||Mushin Muhammed||41.5||Keary Colbert||7.2|
|OAK||Jerry Porter||5.3||Doug Gabriel||1.5|
|CIN||Chad Johnson||24.0||TJ Houshmandzadeh||31.0|
|ATL||Peerless Price||-8.3||Dez White||-1.3|
|HOU||Andre Johnson||18.4||Jabar Gaffney||16.7|
|NO||Joe Horn||41.6||Donte Stallworth||10.1|
|OAK||Jerry Porter||5.3||Doug Gabriel||1.5|
|SD||Eric Parker||16.2||Keenan McCardell*||1.2|
|MIA||Chris Chambers||5.0||Marty Booker||-3.2|
|KC||Eddie Kennison||24.7||Johnnie Morton||19.3|
|TEN||Derrick Mason||22.1||Drew Bennett||23.6|
|IND||Marvin Harrison||28.6||Reggie Wayne||44.0|
|IND||Marvin Harrison||28.6||Reggie Wayne||44.0|
These numbers are telling enough, but they actually require further qualification. Bailey was sometimes used to cover Tony Gonzalez (DPAR 43.1) and Antonio Gates (DPAR 35.3), who were in fact the top receivers on Kansas City and San Diego, respectively. In the first Indianapolis game, Jim Sorgi was at quarterback, while in the second game Herndon was matched up on Brandon Stokley, not Reggie Wayne. Bailey had a very tough slate of receivers to deal with, while Herndon was generally matched up against a receiver with no better than average ability.
While this list at least provides a degree of perspective, it still doesn't go nearly far enough in addressing the disparity between Herndon's outstanding PFP numbers and his perceived value around the league. An elite #2 corner is in its own way as valuable a commodity as a top #1 corner, as it allows you to leave him out on an island and roll your coverage elsewhere. In fact, most of the cornerbacks who switched teams in free agency were brought in to be second corners. It still seems likely that there was some element of Herndon's performance that isn't coming though. For that we turn to KC Joyner.
In the beginning of Scientific Football 2005, Joyner lays out the five areas where he felt current statistical tracking was insufficient. He wanted to know which defensive back was responsible for coverage on any given play, what the quality of the coverage was, what percentage of passes are completed against each defensive back, how many more completions the defensive back would have surrendered if not for drops or bad passes, and how much of a defensive back's yardage was surrendered while playing prevent defense. He then set about creating a database that would track exactly that sort of information. He watched a lot of game film, charted every passing play, and entered the results into his database. For more on Scientific Football 2005, see Jim Armstrong's review.
Sure enough, the holes in Herndon's game show up. Herndon was targeted for passes 96 times and gave up completions on 54.2% of those attempts. Those numbers are reasonable, but the yardage Herndon surrendered was not. He gave up 771 yards, 8.0 yards per pass attempt. The reason why the yardage total is so high is that Herndon struggled defending deep passes. He only allowed a 30.8% completion percentage on deep routes, but 323 of the yards â€” over 40% of his total â€” were given up on deep routes. Herndon played almost exclusively up on the line, which put him in position to provide the kind of run support that cut his yards per play in his PFP rankings down to nearly half of his yards per pass attempt. Unfortunately, that left him vulnerable to the deep ball.
Perhaps the most telling stat about Herndon is not the big plays he gave up but the big plays he could very easily have given up. Herndon had sixteen passes that could have been caught against him but weren't, and those passes would have resulted in an additional 303 yards for the offense. Admittedly, this is a bit of a subjective exercise on Joyner's part, but he applies the same standard to every defensive back, and by Joyner's reckoning only seven cornerbacks in the league were in danger of giving up more yardage than Herndon.
Let's get back to Bailey. In addition to his scouting profile, Joyner also puts out a fantasy football guide at the beginning of each season, and it is clear from his summary on Bailey in the 2004 preview that he was high on the corner and considered his acquisition a major coup for the Broncos. It is also clear that he was aware of a flaw in Bailey's game that was observable back in 2003, a flaw that would become more obvious to everyone as the 2004 season got underway:
â€œAs oxymoronic as this may sound, Bailey was a very good corner on every type of route except a go route. Bailey was beaten on go routes on many occasions and by many receivers. Bailey was beaten on 8 go routes last year. Not all of them were caught, but most of them were, and some of the ones that were incomplete were due to penalties on Bailey.
Bailey isn't a shutdown corner the way that Ty Law is, or even as a lot of other top level CBs are. He'll give up some big plays and get big penalties called on him. He was also beaten at times on shorter routes more often than you might expect. I'm not trying to say that Bailey isn't a top notch CB, or that he isn't a shutdown corner. He simply doesn't dominate the same way the top level CBs do.â€?
While Joiner breaks down each corner's performance into forty different categories divided by the type and length of route being run against them, there are clearly a few metrics which are more important than the rest: attempts, yards given up, touchdowns given up, total completion percentage, yards per attempt, tight/good coverage percentage, deep attempts and deep completion percentage.
Because Joyner breaks down game tape for every corner that had at least fifty passes thrown in his direction, he is able to effectively gauge a corner's relative performance in every aspect; for each of the forty categories, every corner is ranked. And when Bailey's numbers are put in context, they not only fail to look dominating, they look below average. Here is how Bailey fared in each of the crucial categories, along with how Kelly Herndon managed on the other side, and how the Denver secondary did as a whole:
|Scientific Football 2005 Cornerback Stats*|
|Denver Defense||Champ Bailey||Kelly Herndon|
|Att||N/A||N/A||90||t. 32nd||96||t. 22nd|
|TDs||19||t. 8th||7||t. 71st||3||t. 21st|
|Yds/Att||7.3||t. 20th||8.6||t. 65th||8.0||56th|
|Tight/Good%||17.2%||t. 7th||15.6%||t. 71st||33.3%||t. 3rd|
|Deep Att||N/A||N/A||24||t. 16th||26||t. 14th|
Bailey was targeted 90 times to Herndon's 96, an exceedingly high number for a player that is supposed to intimidate the offense. Generally, when there is an elite caliber corner on one side and a clearly less talented player on the other, offenses will target the second corner, a fact that comes out in the attempts distribution. Gary Baxter had 108 attempts, while on the other side Chris McAlister had only 65. Terrence McGee saw 126 passes thrown his way to Nate Clements's 88. Phillip Buchanon was targeted 76 times to Charles Woodson's 58. But in Denver, home to the supposedly best corner in the game, the pass distribution was almost even.
Clearly, offenses were not intimidated by the prospect of going after Bailey, and the rest of the numbers demonstrate why. Bailey gave up a high completion percentage on both deep patterns and short patterns (50% on the deep balls, 73.6% on the short stuff; his completion percentage on medium routes was a reasonable but by no means dominant 46.2%). His yardage total is very high, his yards per attempt is lower-tier, and the seven touchdowns Bailey surrendered made him one of the worst corners in the league in that department. Bailey gave up 12 receptions on deep patterns for 425 yards -- over half of his total yardage -- and four of his seven touchdowns came on those twelve receptions. It's exceedingly difficult to reconcile those numbers with Pompei's assertion that Bailey was being unfairly judged on the basis of three plays.
Does this mean we should discount Bailey's PFP numbers? Not necessarily. For starters, Joyner does not make any attempt to measure a defensive back's contribution to run defense. That's understandable, as a couple of tackles on running backs should not make up for getting toasted on deep balls. But Denver's rushing defense graded out as seventh in the league according to DVOA, and in part that was due to the tackling prowess of its starting corners. Joyner also makes no attempt to put a defensive back's performance on any given play into proper context. After all, giving up 10 yards on 3rd and 12 is a successful play for a defender, and Bailey's PFP numbers suggest that when he wasn't giving up the deep ball, he was very effective at preventing successful plays.
Then again, if Bailey was so effective, why doesn't it come through more in Denver's DVOA numbers? Yes, Denver was fourth in the league in passing DVOA, but a closer inspection of the numbers provided in Pro Football Prospectus 2005 shows that Denver's high ranking was almost entirely due to its tremendous ability to cover the tight end (-43.6%, 2nd) and running backs (-48.8%, 2nd). Their DVOA versus #1 receivers was only 7.4%, a substantial improvement over the year before Bailey's arrival, but still below average.
Denver's DVOA against #2 receivers actually got significantly worse in 2004, dropping from â€“9.1% to 14.4%, which both highlights the strangeness of Herndon's PFP numbers and throws into doubt the notion that Bailey's presence allowed the team to roll coverage effectively to the other side of the field. Both Herndon and Bailey played very little soft coverage, and it seems likely that both corners were in good position to help make plays against running backs coming out of the backfield.
Bailey's more active games show him bringing down backs over and over. In his week one game against Kansas City, for instance, seven of his eight plays were either run stops or tackling a back who had caught a pass out of the backfield. In the second game, against San Diego, Bailey made half of his six plays against Eric Parker, his primary coverage responsibility, and the other half against LaDanian Tomlinson. Against New Orleans, Bailey made four tackles on runs by Deuce McAllister, as well as another tackle on a McAllister reception.
The downside of Denver's playing its corners so close to the line was that both players put themselves in positions where they could and did get beaten deep. It's also highly likely that the reason Bailey's personal numbers look better than the DVOA against #1 receivers numbers is that he wasn't around to make tackles on plays that beat him deep, leaving a safety to clean up the mess and get tagged with a failed stop (Kennoy Kennedy posted just a 30% stop rate, and John Lynch a 38% stop rate, despite the fact that Kennedy played well and Lynch played very well last year).
No one doubts that Champ Bailey has a terrific skill set to play the cornerback position, but the first rule of playing cornerback is that you don't let yourself get beat deep, and for whatever reason Bailey has been reluctant to learn it. He had a problem with deep balls in 2003, and he had a bigger problem in 2004, when teams began to go after him in earnest.
In fact, Bailey's susceptibility to the deep ball ended up hurting other areas of his game as well; once teams started running deep posts against him on every play, Bailey's production against the run game went way down. Looking over Bailey's 2004 season as a whole, he clearly had a few bad games mixed in with a lot of good ones, and he should be given credit for having done well against a very tough slate of receivers. But an elite corner can't afford to be so vulnerable against the deep pass, and if Bailey doesn't correct his problem before Randy Moss comes to town, he may find himself with a lot fewer supporters this time next year.
67 comments, Last at 22 Jul 2007, 10:45pm by Chris L