Pro Football Reference Blog

I was looking through the new leaderboards, including the yearly fumbles category, when I noticed that there sure seemed to be a lot of running backs from the late seventies and early eighties with lots of fumbles. Turns out there were 32 different running back seasons between 1978 and 1985 where a running back had 10 or more fumbles. Since 1990, there have been a grand total of three: Cleveland Gary (1990), Garrison Hearst (1995) and Travis Henry (2002).

So I decided to dig a little deeper, to see if this was really a trend, and if so, when it began. Thanks to the new database, I was able to pull all seasons since 1970 by a player who played running back or fullback during his career, with fumble info. Now, I cannot tell how many fumbles are due to rushing attempts, receptions, or kick returns or punt returns. Certainly, some of the fumbles by players listed as running backs are occurring on special teams plays, so the first thing I did was find the league-wide running back fumble rates for all touches, where a touch constitutes any play for which a back might have fumbled--rush attempts + receptions + punt returns + kickoff returns. Here is the yearly data for fumble rates on all running back touches, reported as Fumble Rate Per 100 Touches:

Year	Fumbles	Touches	Fum/100 Touch
1970	330	12982	2.54
1971	362	13362	2.71
1972	376	13790	2.73
1973	416	14591	2.85
1974	361	14770	2.44
1975	421	15596	2.70
1976	453	16648	2.72
1977	411	16580	2.48
1978	472	18394	2.57
1979	467	18156	2.57
1980	413	17497	2.36
1981	460	17695	2.60
1982	244	9463	2.58
1983	409	16917	2.42
1984	366	16344	2.24
1985	327	15981	2.05
1986	361	16160	2.23
1987	332	15302	2.17
1988	319	15455	2.06
1989	313	14878	2.10
1990	309	14120	2.19
1991	278	14167	1.96
1992	245	13816	1.77
1993	244	14297	1.71
1994	211	14245	1.48
1995	245	15243	1.61
1996	218	14912	1.46
1997	219	14930	1.47
1998	200	14704	1.36
1999	215	14503	1.48
2000	203	13969	1.45
2001	194	14122	1.37
2002	197	15011	1.31
2003	207	15393	1.34
2004	184	15088	1.22
2005	177	15276	1.16
2006	179	15322	1.17
2007	166	14986	1.11

So, it does appear that the league wide fumble rate for running backs has been dropping since the AFL-NFL merger, and was at an all-time low in 2007. Certainly, some of this must be due to changes in the usage of running backs on special teams, particularly star players. After all, most top running backs today are not involved in the return game, whereas in the 1970's, it was not uncommon for a player like Terry Metcalf or Greg Pruitt to be involved as a return man as well as get a significant amount of touches from scrimmage.

To test how much return touches affect the fumble rate, I divided the seasons into three groups: 1970-1981, 1982-1993, and 1994-present. Then, within each group, I sorted the players by the percentage of touches that were from scrimmage versus on returns. For example, here is the data for the 1994-present group, sorted by return touch percentage (ko & punt ret/total touches):

%Return	Fum   Touches  Fum/100 touches
100.0	71	2082	3.41
80-99.9	94	2854	3.29
60-79.9	64	2376	2.69
40-59.9	124	5870	2.11
20-39.9	154	8378	1.84
.1-19.9	585	42155	1.39
0.00	1723	143989	1.20

Just so you understand what you are looking at here, the 100.0% group represents all players designated as running backs who had all of their touches from special teams returns. The players in the 0.00% represents all running backs who had no special teams returns in a given season in that period. The fumble rate is higher on special teams, and the rates drop progressively depending on the percentage of touches attributable to special teams returns.

However, even the special teams to scrimmage fumble ratios is fairly uniform across eras. For the period 1970-1981, the fumble rate for running backs with zero special teams returns was 2.26 per 100 touches, and for running backs who got all their touches on special teams returns, a whopping 6.80 per 100 touches. The fumble rates on special teams plays appears to be almost three times greater than on scrimmage plays for running backs. The fumble rates for both special teams and scrimmage plays have been dropping steadily though, and the rate for each is about half of what they were in the 1970's.

Taking out special teams touches entirely, here are the fumble rates for only backs who derived all of their touches from scrimmage plays, for 1970-2007. To cut down on the length of tables, I divided the data into four-year groups (except for 2006-2007), but if anyone wants to see each year, let me know (the pattern is similar to the previous one above that includes special teams touches for all seasons). Here are the fumble rates per 100 scrimmage touches (rushes + receptions):

YEARS        FUM/100 SCR TOUCH
1970-1973	2.42
1974-1977	2.26
1978-1981	2.15
1982-1985	2.07
1986-1989	1.96
1990-1993	1.64
1994-1997	1.28
1998-2001	1.23
2002-2005	1.15
2006-2007	1.08

Even with punt and kick returns removed from consideration, we see that the fumble rates for running backs have been steadily declining over the last 38 years. Why is this? Clearly, it is not just because of changes in usage rates of primary running backs as special teams returners. Here are my thoughts:

1. Instant Replay
My first reaction when I saw the dramatic drop in the number of running backs with 10+ fumbles toward the latter half of the 1980's was that instant replay was a primary driving force. But after looking at the underlying data, I don't think instant replay is a major driving factor in what we are seeing. Instant replay was first introduced in the NFL for the 1986 season, and the original version was in effect through the 1991 season. The league stopped using replay for the 1992-1998 seasons, before it returned again prior to the 1999 season.

Looking at the data, there is no great decline in fumble rates after instant replay was adopted each time, or at least no greater decline than other periods when their was no change in the usage of replay. In fact, some of the biggest drops in fumble rates occur in the early 1990's, right after the league did away with replay the first time.

This is not to say instant replay is not a contributing factor. Most certainly, there are plays each year where a play that was called a fumble on the field is overturned, and the running back in 2007 does not get credited with a fumble when one in 1975 would have been. But let's assume that there were 32 running back fumbles overturned by replay and ruled down by contact in 2007. An additional 32 fumbles would have only increased the 2007 fumble rate to 1.30 per 100 touches, still well below the rate twenty years earlier.

I guess an argument could also be made that the usage of instant replay has a chilling effect on the officials, and they are far more likely to err on the side of caution, knowing they can be overturned if they rule a play a fumble instead of down by contact. Such an effect would build up over time as officials became more accustomed to being scrutinized by replay. Even if this were true, though, again I don't see it as a major factor.

Which leads me to a couple that I do think might play a role.

2. Changes in Team Climate Situations

When the AFL and NFL merged for the 1970 season, there were 26 teams. Of those 26, only one played in a dome (Houston), five others could be considered warm weather cities (Los Angeles, San Diego, Dallas, New Orleans, and Miami), and three might be considered moderate late season cities (Oakland, San Fransisco, Atlanta). The majority of the league's teams were in cold weather outdoor venues over the latter half of the season, and with the exception of games played in Houston, all were subject to precipation throughout the season.

Over the next thirty years, through expansion (Seattle playing in a dome, Tampa Bay, Jacksonville and Carolina), relocation (Cardinals to Arizona, Rams and Colts to play in a dome) and teams building domed stadiums (New Orleans, Atlanta, Detroit, Minnesota), the league shifted to far more games that were played indoors or in temperate climates later in the season. The changes were gradual, and it was not until the Oilers moved to Tennessee to play outdoors and Seattle moved into an outdoor stadium that there was any kind of shift the other direction.

I don't think climate fully explains the downward trend, but I do think the shift in percentage of games that are subject to bad weather conditions compared to the 1970's and 1980's is one of the driving forces behind the drop in fumble rates.

3. Passing Efficiency

Doug has occasionally used football topics to discuss things like poisson distribution and Benford's law in previous posts. Well, my background is in Biology, so in an effort to justify the money spent taking classes in mammalian cell genetics back in the day, I'm going to use this opportunity to discuss evolutionary biology, natural selection, and adaptation, and how similar concepts may relate to the fumble rate drop for running backs.

Humor me for just a second. Let's assume we have a group of grazing herbivores who live on an isolated island, but one large enough to support the herbivore population, and allow it space to move. On this island, there are no natural predators large enough to threaten the adult population, so it grows at its own pace and develops without any need to adapt to predators.

But then, a new predator is introduced to the island, and it is large enough to attack an average adult herbivore, and fast enough to catch an average adult herbivore. What's going to happen to the herbivore population over time?

It's going to change. Natural selection will result in certain traits being positively selected (and some negative traits being selected against) among the population. Maybe one of those traits is speed and quickness. The faster animals live to mate another day. Maybe it is size. The predator, given plenty of options, will tend to avoid a larger animal that could increase its risk of injury or cause it to expend additional energy. Larger size is then selected, and the smaller, slower animals are removed from the population and don't pass on their genetic traits. Maybe its something like vision, or hearing, to detect the predator and get an early jump. Maybe it is "off the chart" intangibles, so that the herbivores willing to devote hours to film study of predatory behavior are more prepared.

Okay, I'm getting silly, but I had to bring it back to football. Because I think that changes to passing efficiency over the last thirty years is having a profound effect on running back fumble rates. In 1978, the NFL introduced the "Mel Blount" rule regarding contact with a wide receiver outside of five yards from the line of scrimmage, and changed the rules for offensive linemen and pass blocking. There have been numerous other rule changes since then, many also having an effect on offense, usually promoting greater offense, particular passing.

Prior to the rule changes, passing offenses were more risky, featuring higher yards per completion, but lower completion percentages and higher interception percentages. The league wide completion percentages have been steadily climbing as the teams adapted to the rule changes, the West Coast offense became popularized, teams became more sophisticated and varied in their offensive approach, and young quarterbacks trained in these offenses entered the league.

So now picture our running backs and analogize back to the herbivores. We've had a shock to the system caused by the rule changes. Let's compare our fumbling running backs to animals that were smaller in size. Before this shock to the system, they could function just fine because they compensated in other areas. When passing offense was far more risky, teams could tolerate the occasional fumble if it came with an explosive back who could make big plays.

With these changes to the rules, though, the passing offenses become relatively safer and more efficient over time, while still having explosiveness resulting in more big point producing plays than the running game. The big play ability of running backs is still important, but less so when you can more readily get those big plays in the passing game. Thus, it becomes relatively more important that the running back not turn the ball over, to avoid taking the ball out of the hands of his team's increasingly efficient passing game. Much like the smaller animals, the fumblers are selected against with greater pressure than before. And just like a smaller animal better be at the far left of the speed curve to survive in the changed environment, the more fumble prone back better be at the far left of the explosiveness curve, moreso than before. He either adapts, or is so good in every other facet that teams are still willing to tolerate the flaw.

Maybe a Tony Dorsett would still survive because he was exceptional enough in other areas. (Dorsett had 8 or more fumbles in five different seasons). But what about a Wendell Tyler? Tyler had an amazing statistical season but fumbled 10 times in 1982--in a strike-shortened 9 games. And its not like that was an anomaly, as he had fumbled at least 9 times in both seasons where he had played a full schedule before then.

Basically, the tolerance for players like Tyler may have decreased over time, and either they adapted and learned to fumble less, or they got less playing time. The purpose of the running game has changed over time, now serving a greater role in lead preservation, but a smaller role in getting the lead in the first place compared to thirty years ago. Along with this change in focus has been a change in risk tolerance. In my opinion, this is yet another subtle effect of the changes in the passing rule.

Where have all the fumbling running backs gone? Like the Dodo Bird, Thylacine, and Quagga, they have been the victims of natural selection. They either changed their stripes, or they became extinct.