Looking at the fumble numbers posted in
this analysis &
this analysis I believe there is an issue with their calculations. I think they are including
incomplete passes as part of the total plays from scrimmage numbers. Obviously you can't fumble on an incomplete pass and therefore, teams that pass more are going to have a higher number of plays per fumbles. This data would also need to be normalized because the schedule is unbalanced. The Patriots play almost 40% of their games against the AFC East and from what I can tell the Jets have been at or near the bottom of the league in forced fumbles the last few years. I didn't have time to go through the Patriots schedule the last 5 years and correlate that with the defensive forced fumble stats.
Based on my read of these links -
http://www.footballperspective.com/the-definitive-analysis-of-offensive-fumbles/
http://www.footballperspective.com/2013-fumble-recovery-data-has-jets-cowboys-at-extremes/
I think its safe to say that the rate of fumbles (not just fumbles lost) is different for different types of plays. Unfortunately I couldn't find total fumbles (nfl.com just has fumbles lost) for receiving/rushing plays. A runner out of the back field faces a greater opportunity to be gang tackled and it would make sense that its harder to hold on to the ball when two or three defensive players are grabbing/punching at it. So I also think the analysis would need to be corrected for run/pass balance.
Perhaps the statistical outlier becomes much less of an outlier when the data is corrected for incomplete passes, run/pass balance, and schedule or perhaps not. I do know that there is a reason that scientists have their analysis peer reviewed before publishing it. IMO it looks like the authors both made the same mistake and stopped asking questions once the information supported the theory they wanted to prove. To make things worse there is a certain segment of the viewing public that will accept this analysis as gospel.
The nice thing is that NFL.com does have both forced fumbles and fumbles lost for all offensive plays. Really wished they had the same information for rushing attempts and receptions. Here is the data for the Patriots for 2010 through 2014.
Year Scrm Plys Rush_Att Pass_Att Comp Inc Plays (w/o inc.) FUM Lost Scrm Plys/Fumble Plays/Fumble
2010 986 454 507 331 176 810 9 5 109.6 90.0
2011 1082 438 612 402 210 872 13 5 83.2 67.1
2012 1191 523 641 402 239 952 14 7 85.1 68.0
2013 1138 470 628 380 248 890 24 9 47.4 37.1
2014 1073 438 609 392 217 856 13 4 82.5 65.8
5470 2323 2997 1907 1090 4380 73 30 74.9 60.0
So the numbers above do not match up perfectly with the Sharp Football & the other analyses. I didn't include playoff snaps so this could explain some of the variation. Neither one of those studies talks about what exactly Total Offensive Plays includes but it sure looks like it includes incomplete passes. The scrimmage play/fumble numbers are very close to the numbers put forth in the two analyses conducted. The last column plays/fumble is the number of scrimmage plays minus incompletions. These numbers could be further refined as I don't know if the scrimmage plays includes special teams (believe it does).
Finally here's a couple of interesting reads on
Ben Jarvis Green-Ellis. The WSJ articles talks about a drill he use to run in college to help develop his technique and the Bleacher Report talks about the fact that he fumbled three times in two games and has some interesting points on his state of mind. Sounds like the first fumble may have gotten inside his head.