Tom Brady fumbled just twice at Gillette Stadium last year.
The Patriots rarely fumble, particularly at home. It's a fact that kept hitting me on the head as I tried to make sense of the full Deflategate implications. Only a charter member of the tinfoil hat brigade would comb over Brady's passing statistics and conclude that he would turn into Mark Sanchez if the ball was slightly puffier. But if a softer football resulted in three or four fewer fumbles per year…well, that sort of thing can affect the standings. Or a playoff game. Or a legacy.
A statistical study with some alarming conclusions appeared on SharpFootballAnalysis.com soon after the Deflategate story broke. It found that the Patriots have been avoiding fumbles at a mathematically improbable rate since 2007. That's one year after road teams began supplying their own footballs, thanks in part to lobbying by Brady and some other top quarterbacks.
The study caused a stir in January, but as deflated-ball suspicions moved away from Bill Belichick and toward Brady and equipment-room underlings, stories of the Patriots gaining an overall competitive advantage (other than a happy quarterback with a better grip on the football) were gradually sidelined.
Now that both Brady and the Patriots were penalized and the NFL universe has taken sides and gone to the mattresses, those fumble statistics loom large again. Could we possibly be looking at a competitive advantage that allowed the Patriots to hold on to the football, through good weather and bad, big hits and small? Or is the NFL punishing the Patriots because of the major cover-up of a minor crime, because of the sin of arrogance or for some other reason?
The January study looks damning: The Patriots' fumble rates appear to be an extreme statistical outlier. But I am fussy about where I get my data, so I performed my own study using the statistics available from, among other sources, the Football Outsiders database and the media-only site NFLGSIS.com. My goal was neither to implicate nor exonerate Brady; nothing found on a spreadsheet could do that. But with radio hosts and karate-dojo dads telling me that soft footballs fumble-proofed the Patriots offense, it was important for me to find out whether or not there's any real evidence.
Fumble Study Findings
The table shows the NFL teams with the lowest fumble rates, expressed in fumbles per 100 plays, from 2012 through 2014, as well as the NFL average for those three years:
|NFL fumble rates (2012-2014)|
|Team||Fumbles per 100 plays|
"Fumbles" for the context of this study includes offensive fumbles lost and recovered, but not special teams fumbles or the rare plays where a defensive player fumbled after a turnover—because we are interested in Patriots offensive fumbles, after all. Botched snaps are included to account for the remote chance that an underinflated football might make the center exchange a little easier.
As you can see, the Patriots have had a low fumble rate over the last three seasons—but not the lowest in the NFL, let alone some kind of shockingly aberrant rate.
Perhaps Brady alone is reaping the benefits of an easier-to-clutch football. The next table isolates quarterback fumbles.
|Quarterback fumble rates (2012-14)|
|Team (QB)||QB fumbles per 100 plays|
|Falcons (Matt Ryan)||0.42|
|Bengals (Andy Dalton)||0.45|
|Broncos (Peyton Manning)||0.48|
|Saints (Drew Brees)||0.49|
These are fumbles that come as a result of sacks, sneaks, scrambles or botched snaps.
Again, Brady and the Patriots rank among the NFL's best at avoiding fumbles, but there is nothing statistically damning in his data. If anything, he is just hanging out among his few peers.
A funky home/road fumble split could also implicate Brady and the deflators. The Patriots can bring their own footballs to road games, but the Wells Report (warning: NSFW language) suggests that it would be harder for minions to fiddle with footballs on the road than in Foxborough, where they know the locations of the most secluded lavatories.
But there is no such split from 2012 through 2014. The Patriots fumbled 21 times at home in the last three seasons, 20 times on the road.
Teams generally fumble slightly more often at home than on the road because they get to run more plays at home than on the road. As for the cold weather in Foxborough playing a role, the following sample of home and road dome and bad-weather teams shows little rhyme or reason for home-road splits.
|Home-road fumble splits for selected teams (2012-2014)|
|Team||Conditions||Home Fumbles||Road Fumbles|
That brings us back to the Patriots' 2014 home fumble total. They fumbled just three times in Foxborough in the regular season, once more in the playoffs. It's a very low fumble total, but not an unprecedented one. The Saints and Panthers each fumbled only three times at home in 2013. Five other teams have fumbled just four times at home in the past three years, including the cold-weather Ravens and Giants in 2012.
We could push back further in history to search for a Patriots advantage. The Patriots fumbled just 11 times in 2011, losing six. The Packers fumbled only nine times that year, losing five. The Saints lost five of seven fumbles that year. The Patriots had an exceptional fumble-avoidance season in 2010: six fumbles, four lost. The 2010 data truly stands out, as the Patriots had a net turnover differential of plus-28 that year.
But if you have to go back five years to find a smoking gun, you face a major problem: The gun ain't smoking anymore. Pile the remarkable 2010 season atop very good low-fumble seasons in 2011 and 2014, and a trend begins to emerge. The Patriots excel at avoiding fumbles, year in and year out, whereas most teams' fumble totals fluctuate much more wildly. If you are looking for reasons to be suspicious about tampered, easy-to-grip footballs, the consistently low fumble rates will make you suspicious.
But those suspicions may not hold up under scrutiny.
The Trouble with Fumbles
Before we get our hands too gritty with the data, this old math teacher must offer a few notes of caution.
Note One: Fumble totals are small.
The average offense fumbles 16.55 times per season, or just over once per game. Small numbers are statistically finicky in a number of ways. They are heavily distortion-prone, for example. The difference between the fourth-least fumble-prone team over the last three seasons (the Browns, with 41 fumbles) and the 20th (the Bears, with 50) is nine fumbles: an average of three per season. Imagine all of the factors that can affect a three-fumble-per-year swing—a third-string quarterback in the huddle, one or two extreme-weather games, a tough division, etc.—and you realize that you must treat the data gingerly.
Small numbers don't break down into smaller numbers very helpfully. If we try to separate fumbles by weather conditions, for example, we quickly end up with over-granulated data: a game here and a game there, a fumble here and a fumble there—a data set that combines a few Patriots games with an end-of-season Browns game from three years ago and that weird Bears-Ravens game when a monsoon swept in near halftime but it was sunny in the fourth quarter (November 17, 2013). That's the kind of data that's easy to bruise and abuse.
Fragile data must be sliced or combined with care. One thing we don't want to do is magnify potential distortions by, say, expressing fumble data in a plays-per-fumble format. Most teams run about 1,000 plays per year. Divide them by a low figure like 10 fumbles and you get 100 plays per fumble. Divide by a league-average 16.55 fumbles and you get 60.44 plays per fumble. We just took a six-event swing and made it look like a 40-play difference: great for making differences appear dramatic. But the devil is in the denominator. If you want to make small numbers look vastly more significant, divide by them. The tables above are expressed in fumbles per 100 plays: It's a subtle difference, but it shows that fumble-prone and fumble-stingy teams are separated by increments, not dozens.
Note Two: The Patriots are good.
If you harbor the belief that the Patriots' 14 years of dominance are entirely the result of spying and squishy footballs, you might as well click away now. The Patriots are a great team and have been a uniquely great team for over a decade.
The data shows that the Patriots are significantly above the NFL average at avoiding fumbles and have been for a period of years. The Patriots are also well above NFL average in completion rates, yards per pass, yards per play, various defensive stats and just about any metric that can be used to measure quality over a period of years. Just about any NFL chart you create—whether you start in 2001, 2007 or 2012—will show the Patriots hovering near the top far more than you would expect an "average" team to do, because the Patriots are not average.
The Patriots have also sustained their excellence for a historically unusual period of time. Clump together even four Patriots seasons with four seasons by just about any other franchise, and you end up comparing four outstanding seasons to some combination of another team's ups and downs. Create a 2011-14 data set and you are comparing four years of Brady-Belichick to a combination of Peyton Manning and Tim Tebow, the powerhouse Seahawks and a team that went 7-9 with Tarvaris Jackson at quarterback, the Ravens of Super Bowl glory and salary-cap purgatory, the Saints through a Bountygate seesaw ride and on and on.
So Patriots data arrives with a thumb on the scale, and the more years you lump together, the heavier the thumb gets. The Patriots have a 100-28 regular-season record since 2007; their statistics are not going to play nicely with the other statistics. If the fact that the Patriots are in the top three of multiple categories over multiple years is considered evidence of guilt, then the Patriots are always going to look guilty to you.
Fumbles, Sacks and Backups
The Patriots' fumble rates are low, though not extremely or historically low, over the last three years. Factor in the 2010 and 2011 seasons, and the rates get even lower. Is there any reason to believe that deflated footballs may contribute to these low rates?
As mentioned in the last section, the generally high quality of the Patriots' play certainly contributes to their low fumble rate. But that's not a very satisfactory answer, because it contains some circular reasoning: The Patriots don't fumble because they are awesome, and they are awesome because they don't fumble. Are there any other statistical forces that might be tugging the Patriots' fumble rates so far below the league average?
There are two easy-to-tease-out statistical indicators that do just that.
Sacks and scrambles accounted for 164 of the 523 offensive fumbles in the NFL last season. On a per-play basis, a sack is far more likely to result in a fumble than any other offensive play. A team that avoids sacks is likely to have a low fumble rate. As the table shows, the Patriots have been one of the league's best teams at avoiding sacks for the past three seasons.
|Fewest sacks allowed (2012-2014)|
There has been a relatively strong correlation between team sack rates and team fumble rates since 2012. (For stat junkies, the r-squared correlation coefficient of 0.144 is pretty darn good for data this messy.) On average, teams give up one fumble for every 2.3 sacks allowed. Isolating for just quarterback fumbles, the average quarterback fumbles once for every 5.1 sacks.
Now, if either Brady or the Patriots got an extra benefit from a soft football, it would likely show up in these sack-fumble ratios. Imagine Terrell Suggs or Von Miller trying to strip-sack Brady, only to find that his fingers have sunk deep into the fluffy ball; aggregated over several seasons, such an advantage would show up in the stats.
No such advantage exists. The Patriots average one fumble per 2.27 sacks, essentially the NFL average. Patriots quarterbacks surrender just one fumble for every 5.81 sacks, which is stingier than the NFL average. Remember, if Brady can squeeze the football like it's Charmin, he should be able to endure many sacks without fumbling. But like so many Patriots fumble statistics, their sacks-per-fumble rate falls in the "very good" category, not the "historic and suspicious" category. See for yourself in the table:
|Sacks per QB fumble (2012-2014)|
|Team||Sacks per QB fumble|
The data suggests that the Patriots are not necessarily a team that rarely fumbles. They are a team that rarely allows sacks, which therefore depresses their fumble rates. Decreased sack totals also keep the quarterback healthy, which feeds directly into the next force that has kept Patriots fumble totals low.
Whether looking at one, three or six Patriots seasons, you are basically looking at Tom Brady data, with backups only appearing in mop-up roles where they rarely throw a pass in duress. The same can only be said of a handful of other teams. That's a big deal, because established starting quarterbacks have lower sack rates and fumble rates than backups.
In 2014, 47 of the 523 offensive fumbles in the NFL (just below 9 percent) were the direct results of sacks or of aborted snaps by quarterbacks who were not Opening Day starters. Colt McCoy, Mark Sanchez, Austin Davis, Zach Mettenberger and other backups often have a huge impact on fumble totals in a small amount of playing time.
Teams with high quarterback continuity tend to have lower fumble totals over multiple seasons; that goes not just for the Brady-Peyton-Brees triumvirate, but for the Tony Romo Cowboys, the Andy Dalton Bengals, the Joe Flacco Ravens and the Cam Newton Panthers. If you think Flacco is fumble-prone because he stands still and suffers strip sacks, it's probably nothing compared to what his backup would do.
To get a sense of how all of these forces interact, compare the 2012 to 2014 Patriots to the 2012 to 2014 Bengals. The Patriots have had eight fewer fumbles than the Bengals in three seasons; the Bengals surrender 1.54 fumbles per 100 plays, near the middle of the pack and .35 fewer fumbles per 100 plays than the Patriots. Andy Dalton has started all 48 games in three years, though backups Bruce Gradkowski and Jason Campbell managed to squeeze in two fumbles in extremely limited playing time. Dalton is no Brady, but he holds on to the football. Bengals quarterbacks only suffered one fumble per seven sacks. The Bengals also protect Dalton well. He and his subs have suffered just 98 sacks in three seasons. Bengals quarterbacks only fumble 0.45 times per 100 plays, a lower rate than Brady and the Patriots.
The Bengals are a perennial playoff team that plays in a somewhat-cold-weather city in a cold-weather division. Their fumble rates are similar to the Patriots' rates because their quarterback is usually healthy and well protected. The eight-fumble difference in three years starts to shrink when you look at who is doing the fumbling.
The same thing happens when the Patriots are compared to the Ravens, who have an even lower fumble rate. It happens when the Patriots are compared to the Saints and Falcons, except that weather can be cited as a greater factor. It even happens in comparison to the Packers and Panthers, a pair of teams that have had to rely on backup quarterbacks in recent years but still have fumble rates comparable to the Patriots'. The Patriots' fumble rate is no outlier. It's part of a cluster of quality teams with stable quarterback situations. The Patriots just have more quality and stability than every other team of our generation.
Let's face it: If deflating the football a few tenths of a PSI below 12.5 inoculates teams against fumble-itis, the NFL should deflate footballs to a few tenths of a PSI below 12.5. Coaches should order the footballs inflated to exactly 12.5 PSI, the minimum NFL requirement, to get the most anti-fumble benefit permissible by the rulebook. If a quarterback liked highly inflated, easy-to-fumble footballs, it should affect his draft status. If there was a tangible benefit here, it wouldn't be something Belichick and Brady discovered and were able to sit on for eight years. Josh McDaniels, Bill O'Brien, Matt Cassel and others would have spread the secret around the NFL by now.
The January study looks damning: It shows the Patriots as a far-fringe statistical outlier. But the Sharp study has several problems. One of them has already been hinted at: Using "plays per fumble" as a metric mathematically magnifies tiny changes in very volatile data. But there is a much simpler problem: The Sharp study eliminated dome teams before collecting the data!
Eliminating dome teams compensates for any impact weather might have on fumble tendencies. It also eliminates the Saints, probably the most Patriots-like team of the last decade in every way except climate. The Saints have been contenders with stable, high-quality coaching and quarterback play for most of the last six seasons. Not coincidentally, their fumble rates have been lower than the Patriots' rates over the last three years.
The Falcons, another team with stable quarterback play and coaching for several recent seasons, are also chopped off the back of any study that eliminates dome teams. Assuming Lucas Oil Stadium is considered a dome because of its retractable roof, some signature Peyton Manning seasons were also excluded from the mix.
Such omissions would make sense if domed stadiums had a major impact on fumble totals. That is not the case, as one of the earlier tables illustrated. The Saints, notorious for massive home-road splits, fumbled 17 times at home in the last three years, 20 times on the road. The Falcons fumbled 24 times at home, 14 times on the road. The Lions, who mix dome home games with annual trips to places like Chicago and Green Bay, fumbled 31 times at home in the last three years, 28 times on the road. The Rams, who are likely to deal with extreme heat, damp conditions or nerve-racking crowd noise in NFC West road games, fumbled 30 times at home, 27 times on the road in the last three seasons.
Weather has, at most, a tiny impact on fumble totals. Sack rates and the presence of a backup quarterback in the lineup have much greater impact. Any attempt to isolate Tom Brady's fumble tendencies that excludes Drew Brees, Matt Ryan and several seasons of Peyton Manning is going to include some massive built-in distortions.
Like most sane people living west of Interstate 91, I believe Tom Brady strongly persuaded some equipment guys to underinflate footballs because that's the way he likes to grip and throw them. I am also certain that Tom Brady would be a Hall of Famer and multi-Super Bowl champion given any regulation football to throw.
If the Patriots received any additional advantage from using deflated footballs, like a significant decrease in fumbles, I might be persuaded to become a hard-liner on the team's punishment: Brady's four-game suspension and the team's lost draft picks would almost seem lenient, and the "tarnished legacy" crowd would have my sympathetic ear.
But if Deflategate is all about a veteran quarterback bending some rules to get a little customization for his footballs, then I am as eager as ever to get the appeals over with, welcome any reduced sentences and forget this whole thing five minutes after Brady returns to the field.
The research suggests that the Patriots got no significant fumble advantage from Brady's choice of footballs in the last three years. The Patriots are good at not fumbling because the Patriots are good at avoiding the many things that lead to fumbles.
That doesn't exonerate Brady, but it does prove that previous statistical studies failed to implicate him. And it also underlines, once again, what trivial events these dire offenses sprung from.
Mike Tanier covers the NFL for Bleacher Report.