The Vikings' quarterback seemed to regress in his second season. Did that tell us more about the player, or the Minnesota offensive scheme?
31 Mar 2005
Guest Column by Brian Hook
As TMQ often says, Denver's helmets should have a label that says "Insert running back, gain 1000 yards". So many Denver running backs have been accused of benefiting from "the Denver system" that we thought it would be a good idea to see what correlation might exist between the Denver "one-cut" zone-blocking scheme and overall running back success.
To gauge that we needed to trawl the archives for running backs that have gone from one system to the other. Unfortunately, this isn't as easy as it sounds.
Denver running backs seem to remain Denver running backs for life -- possibly because they have a nasty habit of suffering devastating injuries while a Bronco. Terrell Davis, Mike Anderson, Olandis Gary, Reuben Droughns, Quentin Griffin, and Tatum Bell have had almost all of their career carries as Broncos. The only player in recent memory that has gone from Denver to another team and achieved significant carries in both locations is Clinton Portis.
However, as luck would have it, some teams have recently transitioned to a zone-blocking scheme. Houston hired Joe Pendry as offensive line coach and installed a zone-blocking scheme for Domanick Davis between 2003 and 2004. Atlanta also saw a similar transition, this time by bringing in the godfather of the zone blocking scheme, Alex Gibbs, whom many a defensive player can thank for a career or season ending lower leg injury.
So this gives us four players to work with: Clinton Portis (Denver and Washington); TJ Duckett and Warrick Dunn (Atlanta); and Domanick Davis (Houston).
For those of you unfamiliar with the Denver offensive line scheme, they use a technique known as "zone blocking". In a "man" or "drive" blocking scheme the lineman is responsible for an individual, and the play is designed for a running back to hit a particular gap. The zone blocking scheme, on the other hand, has a lineman blocking an area instead of a designated defensive player. If multiple linemen are blocking an area than one can break off and block into the second level.
The offensive line typically moves as a unit laterally, and the result of their blocks should create some natural seams or gaps in the defensive formation. The running back is responsible for finding a hole, making a cut, and then running upfield. One of the key tenets of the Denver system is that the running back takes what he can get -- he should never dance around waiting for a hole to open. He needs to be agile, authoritative, and possess good instincts. Nothing fancy, just try to gain positive yardage.
A final element of the zone blocking scheme is the use of the much hated cut block to seal off backside pursuit. This means that any linemen on the backside of the play cut block defensive players in front of them, which drops the defensive players to the turf and, oddly enough, opens up holes for the running back. Note that the cut block is legal in this case, as long as the offensive lineman isn't hitting the defender from behind and as long as he doesn't roll up on his legs. But hitting him below the knees near the line of scrimmage is fair game, as much as the NFLPA doesn't want it to be.
Obviously getting defenders on the ground is one benefit of the cut block, but an intangible benefit is that defenders start worrying about their knees and ankles. They lose a bit of their aggression and speed since they're paranoid that some lineman is going to creep up on them and take out their legs. This has the benefit of slowing down the entire defense.
For the record, Joe Pendry claims that his zone blocking scheme doesn't rely on cut blocks. Given Davis's performance, he might want to reconsider that stance.
Before we start looking at the data, let's go ahead and establish up front that nothing we've discovered is conclusive. We're dealing with very limited sample data and a huge number of confounding variables. Portis had to deal with moving to a run-heavy offense with a passing game as threatening as a sleeping infant. In addition, Washington's starting right tackle (Jon Jansen) was lost at the start of the season. And, finally, Portis casually mentioned that he had been suffering from a shoulder injury he didn't want to "bother" the trainers about during the season.
While Atlanta managed to keep its personnel relatively intact, they turned over their entire coaching staff and installed brand new offensive and defensive schemes. In addition Dunn and Duckett flip flopped roles as primary ball carrier -- in 2003 Duckett was the workhorse but in 2004 that title moved to, well, Mike Vick, but for the sake of this article, we'll just note that Dunn had more carries than Duckett in 2004, a reversal from 2003.
One final note before we get to the analysis: we're not trying to analyze effectiveness (a la DPAR, DVOA, or Success Rate), we're instead trying to get a feel of the "nature" of their carries. Are they getting stuffed more often? Are they breaking off fewer or more long runs? Standard metrics such as yards/carry or standard deviation have a hard time telling us about the style of a runner, but an analysis of the distribution of their runs gives us a pretty good idea.
We took the running play information from 2003 and 2004 for Davis, Portis, Dunn, and Duckett and sorted the runs into buckets. These buckets were:
Discounting situational success, the first two buckets are "bad" runs and the second two buckets are "good" runs. We then graphed the results for each back comparing the frequency of runs within each category between zone and man blocking schemes. The results were inconsistent but still interesting.
Now, without looking at any real numbers, I think the graphs tell us a lot about each runner. Let's take a look at TJ Duckett first:
Duckett clearly improved with the zone running scheme. In fact, that he had more runs for 4-9 yards than for 1-3 yards is impressive. In 2004 he almost halved the number of stuffs he suffered, he had a far greater percentage of "good" runs, but his 10+ yard frequency stayed about the same.
In other words, he isn't a big play back, but in the zone scheme he was a lot more dependable for getting good, solid carries. Basically that distribution is exactly what you want from most running backs -- get me at least 4 yards over half the time. It is not a coincidence that he has the second best RB Success Rate of backs in 2004 (up from #20 in 2003). (RB Success Rate explained here.)
Of course, other factors are at play as well -- he wasn't asked to carry as much and it he was used very differently. In 2003 over 65% of his carries were on 1st down, whereas in 2004 less than 50% of his carries were on 1st down.
Then we have his teammate, Warrick Dunn, who went through his own changes:
The graph basically says "In 2004, Dunn was far more likely to get you positive yardage, but less likely to get you GOOD positive yardage".
Davis, on the other hand, isn't as cut and dry as expected:
The media have gone on and on about his lower yards/carry this year vs. last year (3.9 vs. 4.3). But the graph indicates that he was more consistent in the zone blocking scheme, i.e. a greater proportion of his runs were between the extremes of "stuffed" and "long run", which means he was stuffed less but also had fewer long runs. The numbers bear this out -- in the zone blocking scheme he had 73.3% of his runs between 1 and 9 yards, but in the man blocking scheme only 64% of his runs were in that range. Visually this is obvious, as is the fact that he had more stuffs and long runs.
It is possible that the lack of cut-blocking on the backside prevented Davis from breaking out past the 3 yard area and grabbing large chunks of ground.
Now compare this with Portis:
The graph there, again, clearly shows the trend difference -- there's a crossover between the "bad" region and the "good" region. In other words, in the man blocking scheme he tended to have more "bad" runs and fewer "good" runs than in the zone blocking scheme. He ran more often for more yards in the zone scheme, however it should be mentioned that Joe Bugel and Joe Gibbs recognized this near the end of the 2004 season and supposedly incorporated more zone blocking type plays, however looking at the overall game summaries I don't immediately see a difference.
With such limited data and with so many confounding variables we can't really draw any strong conclusions about man vs. zone blocking. But there are some interesting observations to be had.
First, the data seems to indicate that odds of getting stuffed drop with a zone scheme. This makes sense, since in a zone one-cut scheme the running back chooses the hole instead of sticking with the play's pre-selected gap.
Second, there seems to be no correlation between scheme and the ability to rip off long runs. Davis and Portis had more long runs with zone blocking; Duckett was the same; and Dunn had more in a man scheme.
Third, it is possible that Davis might have had a better year if Pendry had used cut-blocking to Davis's advantage -- his distribution of runs is similar between the two years with the exception of his lack of long runs.
The "Denver system" isn't a magical pill that a team can swallow to generate 1500 yard rushers with consistency, but obviously it has been successful for running backs in Denver. One reason it has not been widely adopted is time: it takes time to teach, time to master, and time to get the smaller, more agile offensive linemen that the system requires. If you take zone blocking and try to implement it with 340 pound behemoths, you will probably fail, and for better or for worse, 340 pound behemoths are what you'll find on a typical offensive line in the NFL.
Brian Hook is an avid football fan and occasional game programmer, having worked on games such as Quake 2 and Quake 3 in his career. He is currently hard at work on a multiplayer football management game. He can be reached at brianhook-at-hookatooka.com. If you have an idea for a guest column, something which takes a new look at the NFL, please email aaron-at-footballoutsiders.com.
5 comments, Last at 06 Feb 2007, 7:27am by Frank