2012 Fumble Luck
by Rivers McCown
This weekend I was finishing up Bill Connelly's new book Study Hall: College Football, Its Stats and Its Stories, which -- quick plug! -- is available in our web store for $8.99. I came upon this passage in one of the chapters breaking down the basics of advanced statistics, from our dear leader:
"I often say I got into this to improve commentary, not to improve how well teams are run," says Aaron Schatz. "I think front offices in the NFL are smarter about things than fans give them credit for being. The trick for me in terms of regular fan acceptance is to get the color commentators and, particularly, the studio analysts used to it. We want the basic concepts to show up on broadcasts –- success rates, yards per down, fumbles luck, understanding the importance of opponent adjustments."
And it was then that it hit me that we hadn't run last year's fumble luck table on the site yet.
If you are new to the party: we don't claim that forcing fumbles is not a skill, nor do we claim that some running backs aren't very fumble-prone. (Bryce Brown, take a bow!) What we claim is that once the pointy ball hits the carpet, there is a lot of luck involved in which team claims the possession that comes with a recovery. In a 16-game season of close games, a lot of games can turn based on a few unlucky bounces on fumbles. These things tend to regress to the mean over the time. To name one example of how quickly a percentage can turn, in 2010, the Oakland Raiders kept a league-high 65 percent of their offensive fumbles, recovering 15 of 23 balls that hit the ground. In 2011, the Raiders fumbled 14 times, and recovered just seven of them.
So looking at the fumble luck stats, knowing that these things tend to regress towards the mean, we can glean a little bit about which teams are likely to be under or overvalued by conventional wisdom. Maybe the team will fumble as much as it did last year, but that doesn't mean they'll lose as many fumbles as they did last year. Those extra possessions can swing a few games.
Without further adieu, the 2012 fumble luck stats:
|2012 Fumble Luck
|OFF||FUM KEPT||FUM LOST||Lost%||X||DEF||FUM KEPT||FUM LOST||Lost%|
|OFF||FUM KEPT||FUM LOST||Lost%||X||DEF||FUM KEPT||FUM LOST||Lost%|
|OFF||FUM KEPT||FUM LOST||Lost%||X||DEF||FUM KEPT||FUM LOST||Lost%|
What jumps out in this table? Well, as Bill Barnwell pointed out last week, the Lions seem to have a pretty good case for a bounceback season based on their fumbles. The only problem is that the team everyone is looking up at in the NFC North, the Packers, were essentially just as unlucky as the Lions -- though they did have fewer fumbles on both offense and defense.
Another team that should be expecting a big regression in their favor this year -- and you'd know this if you read Football Outsiders Almanac 2013 -- is the Kansas City Chiefs. The Chiefs were not only among the five unluckiest teams on both offensive and defensive fumbles, but they also had a startling amount of unlucky interceptions. Alex Smith happens to be pretty good at not turning the ball over in the recent past, so you can certainly see that the Chiefs emphasized fixing this in the offseason.
(And yes, despite having the second-best offense by DVOA last season, the Broncos had the highest percentage of fumbles lost by any NFL offense. When you add that to the cauldron with Wes Welker and the camp rumblings of a hurry-up offense, there's solid theoretical ground to believe they could actually get better.)
On the other side of things, the Browns were lucky on both offensive and defensive recovery percentage last season. The offense has actually been outperforming the league average for a couple of years now -- Cleveland lost 29 percent of offensive fumbles in 2012 and a league-low 28 percent in 2011. If displaced Browns coach Pat Shurmur really did coach his way into a paper bag, as Mike Tanier hypothesized in the Cleveland essay of FOA 2013, perhaps he laid low on the turf and smacked fumbled balls in the direction of Browns offensive linemen. That would conveniently explain all the trip-ups in the running game, too.
Both Atlanta and New Orleans were also on the lucky side of things, which is one of the many little factors that went into us projecting the Carolina Panthers as NFC South champs in the Almanac. The idea that the Saints defense -- the worst in the league by DVOA -- was actually lucky on fumble events is part of the reason our projection system isn't forecasting a quick turnaround under Rob Ryan. Ditto the Falcons as they try to hold steady while replacing John Abraham and Dunta Robinson with Osi Umenyiora and rookies.
As further evidence that these things involve a lot of luck: the Falcons and Saints were among the five teams with the highest percentage of lost fumbles on offense in 2011, then they both dropped into the top 10 lowest in 2012. (Of course, it's not quite as impactful when you fumble as little as New Orleans and Atlanta do, but the point about luck remains.)
30 comments, Last at 22 Aug 2013, 4:31pm
#1 by Bobman // Aug 19, 2013 - 10:12am
Ouch. Colts fans knew we were lucky in various ways last year and were due to bounce back, as it were, but this in an unpleasantly specific reminder. Assuming luck evens out (and doesn't overcompensate) they will have to improve quite a bit just to stand pat. Happily this seems to be the case, but I could quote Obi-Wan Kenobi about bad feelings and all....
#4 by herewegobrowni… (not verified) // Aug 19, 2013 - 11:13am
If I am understanding the chart right, and reading it according to the FO context, you guys were lucky only on offense and not on defense, where as the Browns, Saints, Panthers**, etc. were lucky on both sides.
**Why is this supposed to help give the Panthers the leg up in the division, when regression to the mean will hurt them about the same as the Falcons/Saints?
#29 by Bobman // Aug 21, 2013 - 12:29pm
Thanks, you're absolutely right. Jeez, I'd have reversed those tables so the lucky are at the same end for both. Now I can sleep easy... for half the night.
#2 by Peregrine // Aug 19, 2013 - 10:47am
Are those numbers correct for the Patriots defense - 36 total fumbles forced? Guess that's the benefit of being in the AFC East. Only the Bears are in the neighborhood.
How persistent are the total number of fumbles by an offense from year to year? How much of that depends on the QB?
#9 by Travis // Aug 19, 2013 - 12:36pm
Officially, the Patriots forced 30 fumbles last year, which led the league.
#16 by RickD // Aug 19, 2013 - 3:44pm
Does that count the infamous Butt Fumble?
It's a stretch to say that the Patriots "forced" that one.
#21 by Travis // Aug 19, 2013 - 6:29pm
No, but the other four Jets fumbles in the game were.
#3 by Aaron Brooks G… // Aug 19, 2013 - 10:54am
What's interesting about that Oakland example is that despite their fumble luck decreasing from 65% to 50%, their total lost fumbles also decreased, from 8 to 7.
How do year-to-year fumble amounts vary?
#5 by Spleen (not verified) // Aug 19, 2013 - 11:38am
I wonder - how many of those NE forced fumbles came from That Jets Game?
#8 by Karl Cuba // Aug 19, 2013 - 12:14pm
I think they will have quite a few from the 49ers game, though not all 'forced' as the niners had a great deal of trouble executing the snap in that game.
#11 by Scott Kacsmar // Aug 19, 2013 - 1:20pm
Correct. Kaepernick was charged with 4 fumbled snaps, meaning 4 carries for 0 yards that night. SF did not lose any of them.
#12 by herewegobrowni… (not verified) // Aug 19, 2013 - 1:32pm
Likewise, a pretty large fraction of both the PIT offense and CLE defense in this table are concentrated in the first game between them.
#6 by wiesengrund // Aug 19, 2013 - 11:50am
So for ATL, you're saying the offense regression from their incredible 44% luck towards 50% will not only be stronger and more relevant than the defense regression from their handed-on-a-silver-plate 69%, but that it will also somehow lead the even more lucky Panthers, in defense and offense, right by them?
#7 by wiesengrund // Aug 19, 2013 - 11:55am
Ah, might have read the chart wrong, is Def sorted the other way, with "LOST" meaning, the Defense recovered? If so, the diff between ATL and CAR is 0.4 percentage points in total, which I guess could be construed as a "little piece" indeed. Apologies for the misread, the Def regression does seem palpable that way.
#10 by herewegobrowni… (not verified) // Aug 19, 2013 - 1:14pm
Exactly what I was getting at a few posts up--I was confused as to the reading defense backwards, and how the NFC South would be affected as it was predicted to.
Will be very interested to see the crew's North divisions analysis in light of this and other data (particularly since the North divisions play each other this year.)
#13 by Nick Bradley (not verified) // Aug 19, 2013 - 2:30pm
The Chiefs were 2 sig below the league average...wow.
Team Recoveries Non-Recoveries Combined Recovery Rate
NO 18 8 69.23%
CLE 24 12 66.67%
WAS 25 13 65.79%
SF 23 12 65.71%
ATL 16 9 64.00%
CAR 21 12 63.64%
HOU 19 12 61.29%
TB 14 10 58.33%
CHI 25 18 58.14%
JAC 16 12 57.14%
BAL 16 12 57.14%
MIN 20 16 55.56%
NYJ 22 18 55.00%
NE 26 22 54.17%
OAK 14 13 51.85%
CIN 19 18 51.35%
SD 21 20 51.22%
NYG 14 14 50.00%
PIT 23 23 50.00%
IND 14 14 50.00%
SEA 15 16 48.39%
ARI 14 15 48.28%
DAL 13 15 46.43%
STL 14 17 45.16%
MIA 15 20 42.86%
PHI 19 31 38.00%
TEN 12 22 35.29%
BUF 14 28 33.33%
GB 8 16 33.33%
DEN 10 24 29.41%
DET 9 24 27.27%
KC 9 26 25.71%
#14 by Mike White (not verified) // Aug 19, 2013 - 2:39pm
Just curious: is there also such a thing as interception luck? Not from the perspective of whether an INT occurs, but rather, where an INT occurs. Put differently, how much control do defenses have over where their INTs will occur. Is picking a ball off at an extremely opportune location from a points perspective (i.e. deep in your territory or deep in your opponent's) a repeatable skill or just largely the result of luck and randomness?
#15 by Karl Cuba // Aug 19, 2013 - 3:07pm
That's a very good question. Intuitively you'd expect a zone heavy team to capitalise on the reduced space as the defense backs up to its own goal line. I'd love to see some stats on this.
#18 by Bowl Game Anomaly // Aug 19, 2013 - 4:28pm
While it makes intuitive sense that interceptions near either endzone are more valuable than interceptions near midfield (because they are more likely to lead directly to points scored or taken away), all the research I've seen has been unable to verify it.
#17 by RickD // Aug 19, 2013 - 3:57pm
A table like this would give us a chance to test the randomness of fumble recovery. If, say, you look at the defense column alone, you have 32 independent tests of the hypothesis that "half of all fumbles are recovered by the defense." Normalize these puppies, look at the sum of the squares, and compare that number to the Chi-Square distribution, yes? That should give you a p-value.
#19 by RedDog (not verified) // Aug 19, 2013 - 5:53pm
Why do these numbers differ so vastly from the official NFL scoring?
When I look at the NFL page (linked from Travis' posting No 9) for Def
Patriots forced 30 fumbles and recovered 21. FO: forced 36, kept 16
Chicago: forced 26, kept 20. FO: 27/11
Bills: 21/9 vs. 20/12
(Actually, I also have a hard time reading the tables above. One time, "more luck" is top, other time "more luck" is bottom)
#20 by Dennis // Aug 19, 2013 - 6:23pm
Is the overall percentage of fumbles recovered by the offense consistent from year to year?
#22 by Bowl Game Anomaly // Aug 19, 2013 - 8:31pm
Yes, in fact even different types of fumble have consistent rates. League wide fumble loss rate across all types of fumbles tends to be 50%.
According to that, fumbles are lost 55% of the time on runs, 47% of the time on completed passes, and also 47% of the time on sacks. (Disclaimer: that was based on just half of one season.) I remember reading somewhere that aborted snaps have a low loss rate, something like 20%.
#23 by RickD // Aug 20, 2013 - 3:34pm
That article linked to only refers to aggregate rates. It doesn't show that the rates are consistent from year to year.
#24 by Karl Cuba // Aug 20, 2013 - 4:45pm
It's still a great article, I'd love too see FO include those numbers in their work.
#25 by Vincent Verhei // Aug 20, 2013 - 5:35pm
RE: The disparity between fumble numbers here and on other sites: It's amazing in 2013, but accurate data on total fumbles is still hard to find sometimes. Fumbles LOST, sure, but total fumbles aren't always available, especially when you're looking for total fumbles by a team's opponents. Sometimes you can find fumbles FORCED -- a defender hits an offensive player and knocks the ball loose -- but A) as noted, those numbers will be all over the map, and B) It doesn't include things like botched snaps and the Buttfumble.
#26 by Aaron Brooks G… // Aug 20, 2013 - 7:26pm
How do you guys handle those end of game kickoff situations, where you'll have like 7 fumbled laterals on one play?
Or the incomplete-swing-pass-that's-actually-backwards-but-goes-out-of-bounds? And who is that attributed to -- the QB?
#27 by Vincent Verhei // Aug 20, 2013 - 9:01pm
In the first instance, it's not actually a fumble unless it hits the ground. Even then, I'm guessing, Aaron filters them out.
On the second, in the official PBP, it's a run for negative yards and a fumble on the quarterback. At the end of the year, we designate those as "incomplete laterals" and either mark them as passes or ignore them entirely, I'm not sure which.
#28 by Travis // Aug 20, 2013 - 9:34pm
You can always get official totals for fumbles against (which includes aborted plays, muffs, special teams fumbles, and unforced fumbles) on NFL.com, but they're making that site more impractical every year. (On some pages, the link for team stats is under "Schedule"; on others, it's under "Stats." And the individual stat pages are a disaster.)
#30 by adamcarley (not verified) // Aug 22, 2013 - 4:31pm
It would be nice to see a scatter plot of # fumbles Vs % lost.