A Blog by Jonathan Low


Jan 23, 2015

Deflating Outcome: The New England Patriots Fumble Prevention Record Is Statistically...Unusual

OK, all of you out there who don't care about sports in general and American football in particular, it's a week until the Super Bowl, so suck it up. 

The game itself is usually not very good and a substantial proportion of the television audience has probably lost interest by half time, due primarily to medico-economic factors like ratio of consumption to body mass divided by metabolism.

But, whatever. It's the event we really care about and given the combination of speed, money and violence that the game generally delivers, a huge audience can be expected, as well as a cornucopia of information about consumer behavior, rules, data, quaint folkways and strange recipes.

But this year, we actually have a controversy. And it's one that involves ethics, beliefs and pounds per square inch. Not with regard to the players' physiques but concerning the degree to which one of the competing teams, the New England Patriots, may or may not have deflated the balls they used in order to gain some sort of advantage in the game that got them to The Big Show. The Patriots aka the Pats, blew out their opponents that day 45-7. But it was discovered that 11 of 12 balls the Pats used were underinflated, ostensibly making it easier to grip the ball in cold, blustery conditions. Now given the score, some claim that nothing short of hog-tying their quarterback or giving the entire opposing team amphetamines (actually, let's not go there) would have made much of a difference. But still. Rules are rules.

Now, before continuing, let us stipulate that we will be rooting for the aforementioned Patriots. Let's not get into why: most people have trouble remembering who played from year to year anyway. But this is by way of saying we have nothing against the Pats, we just have a penchant for discerning meaning from often obscure points of data. And, given our interest in data science, the following article explains in great, even numbing detail that, in fact, some curious behavioral outcomes could, theoretically have resulted from ball manipulation. We will leave it to the reader to ascertain whether any of this matters. Reports say it had no affect whatsoever on the Las Vegas sports book odds. The real action anyway is on whether the opening coin toss will be heads or tails. JL

Warren Sharp reports in his blog, Sharp Football Analysis:

The 2014 Patriots were just the 3rd team in the last 25 years to never have lost a fumble at home!  The biggest difference between the Patriots and the other 2 teams who did it was that New England ran between 150 and 200 MORE plays this year than those teams

Yesterday I investigated whether or not the New England Patriots outperform expectations in bad weather.  I had several recommendations to look at home and road data, as opposed to just home data.  Mulling whether or not to undertake that further (time consuming) analysis, I watched this video:
I immediately noticed something that cannot be overlooked: the issue with ball security and fumbles.  Then I remembered this remarkable fact:
The 2014 Patriots were just the 3rd team in the last 25 years to never have lost a fumble at home!  The biggest difference between the Patriots and the other 2 teams who did it was that New England ran between 150 and 200 MORE plays this year than those teams did in the years they had zero home fumbles, making the Patriots stand alone in this unique statistic.
Based on the desire to incorporate full season data (not just home games, as a team theoretically bring “doctored footballs” with them on the road) I performed the following analysis:
I looked at the last 5 years of data (since 2010) and examined TOTAL FUMBLES in all games (as well as fumbles/game) but more importantly, TOTAL OFFENSIVE PLAYS RUN.  Thus, we can to determine average PLAYS per FUMBLE, a much more valuable statistic.  The results are displayed in the chart below.  Keep in mind, this is for all games since 2010, regardless of indoors, outdoors, weather, site, etc.  EVERYTHING.
(click to enlarge)
One can CLEARLY SEE the Patriots, visually, are off the chart.  There is no other team even close to being near to their rate of 187 offensive plays (passes+rushes+sacks) per fumble.  The league average is 105 plays/fumble.  Most teams are within 21 plays of that number.
I spoke with a data scientist who I know from work on the NFLproject.com website, and sent him the data.  He said:
Based on the assumption that fumbles per play follow a normal distribution, you’d expect to see, according to random fluctuation, the results that the Patriots have gotten over this period, once in 16,233.77 instances”.
Which in layman’s terms means that this result only being a coincidence, is like winning a raffle where you have a 0.0000616 probability to win. Which in other words, it’s very unlikely that it’s a coincidence.
I actually went back and researched 5 year periods for the entire NFL over the last 25 years. The Patriots ratio of 187 plays to 1 fumble is the BEST of ANY team in the NFL for ANY 5 year span of time over the last 25 years. Not was it just the best, it wasn’t close:
  1. 2010-2014 Patriots:  187 plays/fumble
  2. 2009-2013 Patriots:  156 plays/fumble
  3. 2006-2010 Colts:  156 plays/fumble
  4. 2005-2009 Colts:  153 plays/fumble
  5. 2007-2011 Patriots:  149 plays/fumble
  6. 2008-2012 Patriots:  148 plays/fumble
  7. 2010-2014 Texans:  140 plays/fumble
  8. 2004-2008 Colts:  139 plays/fumble
  9. 2006-2010 Jets:  135 plays/fumble
  10. 1999-2003 Chiefs:  134 plays/fumble
There are a few key takeaways.  First and foremost, the 187 plays/fumble dwarfs even the rest of the best seasons the last 25 years.  Second, the Patriots have been at the top of the NFL since 2007.
Ironically, as my study yesterday showed, the Patriots performance in wet weather home games mysteriously turned ridiculous starting in 2007.  In 2006, they went 0-2.  From 2007 onward, they went 14-1.
The next obvious question becomes, where were the Patriots in this statistic pre-2007?  Take a look:
(click to enlarge)
As you can see, the Patriots won their Super Bowls having a below average rate of fumbles lost given today’s average of 105 plays/game.  But in 2007, something happened to propel them to a much better rate (you’ll remember, that just so happened to be the same year they went 16-0 in the regular season).  But even looking at these numbers, its clear how insane the 187 number is:  they are almost running 100 MORE plays without a single fumble as compared to the 2002-2006 period when they won 2 of their 3 Super Bowls.
To further illustrate how these numbers are astonishing, the below graphics lay out clearly how far off the Patriots are from the rest of the league.  Its evident to the eye how far removed they are from the norm.  Whether we look at a histogram laying it out, where the Patriots and their 187 plays/fumble is far from the “bell shaped curve”:
(click to enlarge)
or the same chart as above, this time displaying color bands as we move away from the 105 plays/fumble average.  You can see the darker red band contains all teams but the bottom 3 and the top 3, and that the bottom 3 are very close to the darker red band.  Meanwhile, the Patriots are really in a league of their own:
(click to enlarge)
Could the Patriots be so good that they just defy the numbers?  As my friend theorized:  Perhaps they’ve invented a revolutionary in-house way to protect the ball, or perhaps they’ve intentionally stocked their skill positions with players who don’t have a propensity to fumble.  Or perhaps still, they call plays which intentionally result in a lower percentage of fumbles.  Or maybe its just that they play with deflated footballs on offense.  It could be any combination of the above.
But regardless of what, specifically, is causing these numbers, the fact remains:  this is an extremely abnormal occurrence and is NOT simply random fluctuation.
UPDATE: It was suggested that I look at ALL fumbles, not just fumbles lost.  With that said, let’s look there:
First, it should be noted (as the tables above show) that teams playing indoors fumble the ball less frequently.  Reasons are many, foremost the ball won’t be wet from precipitation, damp from late night condensation, and a variety of other reasons.  Which is why, if you look at the very first chart I posted above, you’ll see the teams who fumble the MOST/play are generally colder weather teams who play outdoors (PHI, DEN, BUF, PIT, WAS, NYG, KC, NYJ).  Whereas at the other end of the spectrum, aside from the Patriots in their own world, are HOU, ATL and NO, all dome teams.
The below graphic looks at ALL fumbles over 5 year periods the last 25 years.  I planned to cut this off at JUST the top 10 teams, but all we would have seen were the Patriots and dome teams.  Top 15 would have accomplished the same.  So I had to expand to the top 25 team periods.  As you can see, of the top 25 team-periods, 17 are dome teams, including 11 of the top 15.   First, let’s look at the chart, then we’ll look at comparisons to average:
(click to enlarge)
As is apparent, the Patriots are the only outdoor NFL team the last 25 years to average 70 plays/fumble or better, and they did it from 2007-2014 (four, five year periods).  Its simply uncanny, as the statistics above similarly showed.
  • Over the last 25 years, indoor teams averaged 43 plays/fumble (in all games they played that season, regardless of site, understanding that half their games would be played indoor sans-weather).
  • Since 2000, they improved to 46 plays/fumble.
  • Over the last 25 years, outdoor teams averaged 41 plays/fumble.
  • Since 2000, they improved to 43 plays/fumble.
The Patriots averaged 73 plays/fumble the past 5 years, almost 70% better than the 43 plays/fumble that outdoor teams averaged since 2000.
Next, lets look only at the current 5 year period:
The league average plays per fumble from 2010 thru 2014 was 50 plays/fumble.
  • For indoor teams, the average was 55 plays/fumble.
  • For outdoor teams, excluding the Patriots, the average was 46 plays/fumble (9 fewer).
The Patriots averaged 73 plays/fumble, almost 60% MORE than outdoor teams, and almost 50% MORE than the league average the past 5 years.
(click to enlarge)
Since we now can clearly in the data, both near term and long term, that dome-based teams (who play at least 8 games out of the elements) have an advantage in the fumble department, we can exclude them from comparisons to the Patriots.
If we do, I can produce a chart identical to the one at the very top which looked ONLY at fumbles lost.  This one looks at ALL fumbles, whether lost or recovered.  I think the point still remains:
(click to enlarge)
If this chart looks nearly identical, it should.  The Patriots are so “off the map” when it comes to either fumbles or only fumbles lost.  As mentioned earlier:  this is an extremely abnormal occurrence and is NOT simply random fluctuation.
Warren Sharp of sharpfootballanalysis.com is an industry pioneer at the forefront of incorporating advanced analytics and metrics into football handicapping after spending years constructing, testing, betting and perfecting computer models written to beat NFL and college football totals. A licensed Professional Engineer by trade, Warren now works as a quantitative analyst for multiple professional sports betting syndicates in Las Vegas and has parlayed a long-term winning record into selections for clients which move the Vegas line and beat the closing number with regularity.

The Patriots and Tom Brady Suspiciously Outperform Expectations in Wet Weather

According to the NFL, the New England Patriots were found to have introduced 11 under inflated footballs of the 12 they were required to provide during Sunday’s AFC Championship game vs the Colts.  The footballs were said to be underinflated by two pounds per square inch.  The incentive to having the Patriots offense play with underinflated footballs is that they are easier to grip, throw and catch as compared to properly inflated footballs.
Naturally, the immediate question arises:  “How long have the Patriots been playing with underinflated footballs?”  That’s impossible to know, but if the Patriots and Tom Brady believed it was to their advantage to underinflate the footballs for easier grip, presumably they would be doing it in wet weather, much like the weather in New England for the game vs the Colts.
I went through all NFL game books for the Patriots home games since the 2005 season.  What I found was, at a minimum, intriguing.  The game books list the “game weather”.  First, the data I share below assumes this weather report is at least somewhat accurate.  I did go back and cross reference every game vs historical daily weather reports for the area.  That said, its possible one or two additional games should be added in case the game book reported clear conditions when it was, in fact, wet.
Second, I presumed that Tom Brady, the quarterback, was the primary individual who would benefit most from underinflated footballs, so I excluded the 2008 season when he did not play due to injury suffered in week 1.  I won’t speculate as to whether the direction to underinflate the football was given by Brady or not, but I simply removed games he did not start.
The first table below shows the Patriots performance in wet weather in a game by game basis.  After the Patriots suffered losses in their only two home wet weather games in 2006, a strange phenomenon occurred:
The Patriots went 14-1 (93%) in Tom Brady’s home games played in wet weather since 2007.  Their only loss was to the San Francisco 49ers in 2012.  For some comparison, the Patriots went 51-9 (85%) in home games played in dry weather during that same period.  On average, in both wet and dry weather, the Patriots were favored by approximately 9 ppg.  In the NFL, 9 point favorites should win the game approximately 81% of the time.
(click to enlarge)
As you can see from the averages at the bottom, Tom Brady put up remarkably similar numbers in wet weather as he did in dry weather:  7.4 yds/att vs 7.6 yds/att, a 99 passer rtg vs 101 passer rtg, and a slightly better TD:INT ratio in wet weather as compared to dry weather.
The second table (below) looks more in depth at the results of those 15 wet weather home games since 2007 as compared to the Patriots performance in dry weather home games.  The compared results are strikingly different:
The Patriots went 31-29 ATS (52%) in dry weather home games, but 10-5 ATS (67%) in wet weather home games.  The oddsmakers on the games projected the Patriots would score an average of 28 ppg, whether the conditions were wet or dry.  But the Patriots scored 35 ppg in wet weather (+7) vs 31 ppg in dry weather.  They also held opponents to 5 ppg fewer in wet weather home games.
Thus, their average win improved from 30.7-19.6 to 34.6-14.3.  In other words, they went from winning games by 11 points to winning by over 20 points on average, despite being favored by 9 ppg in both scenarios.
(click to enlarge)
This analysis does not prove or disprove anything.  It certainly may fuel the fire of conspiracy theorists, but without any concrete evidence that the Patriots were underinflating footballs for Tom Brady in wet weather, we can only look at the game results.  While obviously suspicious, despite how strongly it appears “something” is helping the Patriots in wet weather, nothing can be proven by this study.
I always watched the Patriots and saw how they performed in wet weather.  Anecdotally it seemed like they “got it”: they knew the pass rush was slower and while other teams shifted AWAY from the pass and to more ground based games in the wet weather, it seemed the Patriots shifted the opposite direction, and passed the ball more frequently.  I always thought this was just “Bill Belichick and Tom Brady being smart and ahead of the game”.  Perhaps that is still the case.  Or perhaps they passed more because their offense played with underinflated footballs in wet weather while their opponent played with regulation footballs.  Its total speculation.  I am of the opinion that we will never know.  This whole allegation could be a lot of hot air.  I am not opining on what happened, whether other teams do it too, the level of advantage which is gained, etc.  I am simply providing actual data on game results and letting you use the data in conjunction with news/media reports to form your own conclusions.


Post a Comment