27 Sep 2007
61 catches for 911 yards and seven touchdowns.
51 catches for 775 yards.
22 catches for 403 yards and five touchdowns -- in just three games.
One of these stat lines is not like the others.
OK, so Randy Moss almost certainly won't continue to produce at this level over a full 16-game season. That would result in the NFL's best ever receiving season by any and all measures, and even put LaDanian Tomlinson's touchdown record within reach. Nonetheless, it's fairly apparent that he's going to meet and exceed his KUBIAK projection (the first of the three statlines listed) with some ease. The second prediction comes from Bill Barnwell's June 28th article on Wes Welker, who has also confounded FO's prediction systems.
So, what happened? There are two parts to this story: Moss is not who we thought he was, and the Patriots aren't either. For Moss, it's a case where statistics don't tell the whole story. DPAR and similarity scores don't have a variable for "this player clearly couldn't be bothered in Oakland." When it came to forecast his 2007 production, all we really knew is that he was an aging receiver who had once been dominant but relied on speed he might not have any more. A more interesting dilemma is figuring out what what this year tells us about his future. Has his half-hearted play in Oakland given him another year or two of peak performance elsewhere? No idea.
Fortunately, the changes in New England's offense are much easier to quantify. Bill Barnwell's article showed a target percentage by role for the New England offense in recent seasons, with a 2007 projection. Let's revisit that table, adding in standard deviation from 2003-2006 as well as the stats for the three games so far this year:
|Patriots Target Percentage By Role, 2003-2006|
Now, it's unlikely that the Patriots will stick with this precise distribution of passes -- if nothing else, one player or another is bound to pick up an injury eventually which will see somebody else getting their pass targets. If Moss continues tearing up the league, his share of the passes may go down as teams increasingly sell out to stop him. Nonetheless, both Moss and Welker are currently getting nearly 4.5 percent more targets than any recent New England receiver. Unsurprisingly, they're outperforming FO predictions, which were based in part on recent Patriots offensive trends. It's true at this stage in the season we're not dealing with all that many targets -- the difference between Moss and Welker in 2007 and Deion Branch in 2004-2005 is less than 1.5 passes per game. However, the variation is in excess of four standard deviations above the mean for the primary receiver, and over three and a half higher for the second wideout. This makes it more likely that we are seeing a shift in the Patriots' offensive philosophy and not just an artifact of a small sample size. The extra targets for the primary receivers seem to have come at the expense of the tight ends, who are less involved in the passing game than they have been, and "other receivers." The Patriots have only thrown to four wide receivers this year, and those four players have all been healthy. New England rarely uses an empty backfield, and when they do the formation usually involves a tight end (and the play is often a Tom Brady sneak on third or fourth-and-1).
It's probably not a coincidence that, after spending a summer loading up on wideouts, the Patriots are throwing the ball more to imported wideouts. Previous Patriots teams had regarded wide receivers as fungible, and it will be interesting to see if this apparent shift in offensive game planning is reflected in New England's front office decision-making going forward. After all, when the Patriots restructured Randy Moss' contract a few months ago, they also made him a free agent in 2008.
62 comments, Last at 07 Oct 2007, 4:37pm by lolicheck
The college football season returns in a big way with South Carolina-Texas A&M, Wisconsin-LSU, and Clemson-Georgia headlining opening weekend. Seventh Day Adventure is back with advanced stats, analysis, and F/+ predictions.