Maximizing Your Fumble Luck

Submitted by The Mathlete on August 2nd, 2013 at 3:10 PM

If nowhere else in the world, it is a well-established tenant on MGoBlog that forcing fumbles are mostly random and recovering them is almost more random. Historically, when I am evaluating an offense or a defense I exclude all fumble plays. They are huge swings in value for a game or player with very little correlation to the overall quality of the player or team. Some of what I found has caused to me reconsider the inclusion of certain fumble plays.

The Study

To look at how different types of plays contributed to fumble totals, I broke down the rate of lost fumbles (blue bar, left axis) and the odds of the defense recovering a fumble (yellow line, right axis) for six types of plays. Looking at over 700,000 plays from the last ten years, here is what I found.

image

Excluding sacks and punt returns, all other play types generate a lost fumble about 1% of the time. Punt returns are lost a little over 1 in 50 plays and sacks are the big defensive opportunity with 1 in 9 sacks causing a fumble and about 1 in 17 resulting in a turnover for the defense. What else is interesting is that the further you get away from the line of scrimmage, the more likely the defense is to recover a fumble. Completions and positive rushing plays are at the low end for lost fumbles but at the high end for defensive recoveries. As a defense, if you can generate fumbles down the field, there is a good chance for a turnover.

Based on this data, going forward I will be including plays where the defense recovers a fumble after a sack in my evaluations for team defense and offenses. Because of their increased incidence I felt like the generation of the swing play through a quality defensive play like a sack didn’t warrant the exclusion. All other plays where a fumble is lost will continue to be excluded as more luck than skill.

 

Keeping the Ball in the Offense’s Hands

For offenses crossing the line of scrimmage is about the best thing you can do to reduce your odds of losing the ball. In the above chart completions and rushes for positive yardage both generated the lowest total rate of lost fumbles. I broke those rushing plays down to see which position was the biggest culprit.

image

After crossing the line of scrimmage, quarterbacks are the most likely to lose a fumble on a running play. Fullbacks and running backs both fumble the ball on 1% of their positive rushes but fullbacks are specially trained fumble recovery machines as there is more than a 10 percentage point gap between the defense’s ability to recover a running back’s fumble as opposed to a fullback’s fumble.

The offset of this data is even more striking. Here is what fumble rates look like for running plays (not sacks) that never make it back to the line of scrimmage.

image

The data here is clearly overrun with bad snaps and failed handoffs. Quarterbacks lose a fumble on 1 in 15 non-sack rushing plays attributed to them behind the line of scrimmage. Running back and wide receiver rates are also much higher than on other plays.

A 7% fumble rate but a 35% defensive recovery means that quarterbacks are given responsibility for a fumble on more than 1 in 5 plays that don’t cross the line of scrimmage. Wide receiver rushes also have a better than average chance of recovering their own fumbles behind the line of scrimmage.

I looked at these behind the line of scrimmage numbers and ran them against whether a defensive player was credited for the forced fumble. On runs crossing the line of scrimmage a defensive player is credited with a forced fumble 80% of the time and that number is relatively consistent across all positions.

If you adjust the behind the line of scrimmage to say try and account for the non-forced fumbles (take the forced fumble numbers and assume they represent 80% of the total) the loss rates behind the line of scrimmage drop considerably. The rates are still higher than post line of scrimmage plays. QBs fall to 2.6%, WRs to 2.2% and RBs at 1.4%. When the fumble is forced by the defense behind the line of scrimmage, defenses recover nearly 70% of fumbles by backs and receivers but barely 50% of quarterbacks. In fact, of all forced fumbles on scrimmage plays, wide receivers and running backs lose the fumble 68% of the time, where quarterbacks only lose the ball 61% of the time. Not sure exactly what that is from but it’s a pretty stark difference and with thousands of plays for both running backs and quarterbacks, one relatively immune from sample size concerns.

Up All Night to Get Flucky*

Based on these numbers a couple things stood out to me:

  • Sacks produce fumble at an obscene rate compared to any other play
  • Don’t skimp on the fundamentals, poor snaps and hand-offs are a major source of fumbles
  • Positive plays are good for the offense, getting past the line of scrimmage greatly reduces the chance that a fumble occurs, but increases the defense’s chances at a recovery if one is forced
  • Hitting ball carriers behind the line of scrimmage is a good way for a defense to generate fumbles
  • HOLD ON TO THE DAMN BALL! Punt returns are the most likely play to result in a lost fumble.
  • Not all fumbles are created equally, defenses recover nearly 70% of fumbles that are forced and only 45% when they are not.
  • Quarterbacks are fumble prone but their teams are better at recovering them than other players’ fumbles

*Sorry, my kid has been singing this song for weeks now, I had to work it in to this article somehow.

Comments

Sextus Empiricus

August 2nd, 2013 at 12:17 PM ^

Seminal work... auspicious on news of Denard's supposed decreased role in his first year.

The picture at the top should be excluded though right? That is not a quality defensive play ... but rather a freak of Rees. Will your ongoing analysis count that or not? It seems like it would from what you are saying. What is the criteria distinguishing these freak fumbles from ass kicking blind side fumble causing take downs?

GotBlueOnMyMind

August 2nd, 2013 at 1:35 PM ^

I do have one point to dispute. Although I agree that fumble recovery rates are generally random and return to the mean, forcing fumbles is most certainly not random. Just look at a guy like Charles Tillman who has perfected the art of stripping the ball-carrier. The rate of forced fumbles is definitely something that can be improved upon through teaching proper technique and such.

Tacopants

August 2nd, 2013 at 9:39 PM ^

While it's probably true that certain players are fumble prone, and it is possible that some guys who work on trying to get fumbles are slightly better, its unlikely that that group of players really has a unique skillset.

That's the beauty of stats, you can run the numbers all you want, there will always be outliers because that's how life works. Is Charles Tillman better at forcing fumbles? Maybe. Or maybe he's the 1 in 1000 defensive players that experiences a multi year run of luck where a larger percentage of his hits cause fumbles. Maybe he is luckily always matched up vs notable fumble machines. It's arguable that when we see that we would think its skill. The only way we will ever know for sure is to replicate tackles and attempted strips across the nfl, which clearly will not happen.

I guess my argument boils down to this: my belief is more that fumbles happen due to lack of execution (think clowney vs smith or any fumbled snap) or a stupid decision by the offense (stretching the ball, holding it in the wrong hand, shifting hands) than it is caused by the defense. Defenders are taught to securely tackle first, which is hard enough.

Any guy who says he's been working extra hard trying to cause fumbles is probably wasting their time, time probably spent more profitably learning how to do something else better that's in their control.

And there are other things to think about as well. If Tillman has a better training method that lets say doubles ff% from 1% to 2%. If people buy into it, then paradoxically you have now proven that its a skill, but have also commoditized that skill so that the expectation is that everyone should be able to learn it and force fumbles at the average rate, now 2%, and it's a random process again!

TL;dr fumbles are random!

Shop Smart Sho…

August 2nd, 2013 at 2:10 PM ^

I know you're reading this, since you posted about it at TBL.  You should probably go back and re-read that pargraph.  You missed the emphasis on "down the field" in contrast to sacks resulting in turnovers.

GoBlueInNYC

August 2nd, 2013 at 3:20 PM ^

Regarding the "defense is more likely to recover the fumble if it happens downfield" finding, is that most likely a function of player traffic? If a fumble happens at or near the line of scrimmage, you've got your QB, RB, possible FB, and the entire OL all clammoring around to get it. If it happens downfield, after a completion let's say, then you most likely have the player who fumbled and maybe one other offensive player in the area. Defensively, you'll have the player who forced the fumble, plus any other players in the area or in transit to assist, such as safeties flowing towards the ball to assist.

89Grad

August 2nd, 2013 at 3:46 PM ^

Very well done.  It would be interesting to compare the fumble rates to decades earlier.  I guarantee that the fumble rate was lower in the pre-1990 timeframe.  Seems like that was when the art of stripping the ball became mainstream. 

hfhmilkman

August 2nd, 2013 at 3:55 PM ^

Here is one possible explaination.  When a fumble is unforced, for example a bad handoff, a flubbed punt, or just a case of the oops, the offensive players are in a position to make an attempt to recover the fumble.  They also know before anyone else that they fumbled.

In the case of a player being sacked or tackled they are often not in a position to make a recovery attempt because they are being tackled or worse have been knocked silly.  The only exception is if the ball pops right back to them. 

So an interesting question is what hapens to the recovery rates for the offense when the fumbler DOES NOT recover the fumble.  Can we make the statement that the person who just fumbled the ball is in the best position to recover since they are aware before anyone else that they did fumble and two they know where they fumbled to?

The next question is can defensive recovery rates be improved if players are coached to wipe out the fumbler under the premise they are most prepared to make the recovery?  Or are you still better off just having everyone pile on and hope your player gets it?  What would be interesting is look at the percentage of fumbles recovered by the fumbler.

Not that it is very big but where do punts and FG attempts fit in.  My guess is they probably have very high rates of offense recovery along with kickoffs.  But as expected with fewer defenisve players in the vicinity the recovery rate is going to be higher.

 

 

bronxblue

August 2nd, 2013 at 4:04 PM ^

Great stuff, as usual.

Maybe I missed it, but did you track who on the offense or defense recovered and/or caused the turnover. Might be interesting to see who tends to be the most responsible for deciding those plays.

Proclus

August 2nd, 2013 at 4:32 PM ^

Fascinating analysis. A quick point of orthography: the first sentence should probably begin "It is a well-established tenet . . .". A well-established tenant would be like that sixth-year graduate student who's lived in the same house on Catherine Street since his sophomore year while roommates come and go around him.

jdon

August 2nd, 2013 at 5:16 PM ^

One point of contention.  I see a definite correlation between teams who sack the quarterback and teams who recover fumbles.

It would seem to me that if a team that sacks the quarterback more often stands to recover more fumbles then teams that consistently get to the quarterback would be more likely to recover fumbles. 

yes?

jdon

August 2nd, 2013 at 8:00 PM ^

my contention is that fumbles are not completely random and that there is a correlation to the quality of a defensive line or that of the offensive line...    the next step would be to see if teams who produce low amount of sacks recover fumbles at a rate different than teams that produce a high amount of sacks  Against the average overall...

jdon

 

Michael Scarn

August 2nd, 2013 at 6:09 PM ^

Only pointing it out because I have made the same mistake more than a few times - tenet, not tenant.  Then again the fumble debate occurs so frequently on here that it probably has started paying rent, so maybe that's actually what you meant.

MichiganG

August 2nd, 2013 at 7:24 PM ^

As you pointed out, fumbles are recovered more often by QBs since there aqre many instances where they can drop on the ball when there's a bad snap or what not.  But I was wondering why that may be the case for forced fumbles, too.  My main hypothesis is that it has to do with being surrounded by thir linemen and potentially a RB (since most forced QB fumbles are related to sacks).  And, chances are, there's a lineman who is watching the QB at that moment since it means someone got by him.  

MGlobules

August 2nd, 2013 at 9:40 PM ^

more. 

Not all fumbles are created equally, defenses recover nearly 70% of fumbles that are forced and only 45% when they are not.

Might this be because if you're fumbling on your own you know it and have that fraction of a second more to pursue the fumble?

  • Interesting stuff. 

"tenet," fwiw. 

elm

August 3rd, 2013 at 12:28 PM ^

Is it that getting past the line of scrimmage reduces your chances of fumbling or is it that fumbling reduces your chances of getting past the line of scrimmage?

I'm having trouble figuring out a causal story for why a player would get a better grip on the ball once they're into positive territory.  The causal story for the other possibility is more obvious: if you fumble in the backfield, you can't get positive yardage on the play.

JTrain

August 3rd, 2013 at 9:33 PM ^

HATS TO THE BALL.

Great pursuit + unbelievable effort yields recovered turnovers.

How many times have you heard Mattison talk about not loafing. Going 100%. "Cupping" / containing.

A good bounce goes a long way....but we can't control that. The rest of the aforementioned, we can.