Mixpanel Tackles the NFL
“Stats are for losers. The final score is for winners.”
– Bill Belichick
Before you ask: No, we’re not disavowing statistics.
We’re just a little antsy around here for football to start. So, being the data-driven geeks that we are, we decided to upload 15 years’ worth of historical NFL data into Mixpanel to see what we could learn from the past about the upcoming season. Not only has it been a ton of fun for everyone to play around with, but also it’s provided some really interesting data about why teams do what they do. (And we might have a few fantasy football players around the office.)
While everyone is putting the NFL data to good use in their own ways (don’t mess with our company fantasy teams this year), I decided to try to answer a question that’s been bugging me:
Why are the New England Patriots still favorites to win the Super Bowl, even though Tom Brady is suspended for the first four games of the season?
On its face, and as a Patriots fan, this doesn’t seem logical to me. No matter how good a team might be, missing a future hall of fame QB who’s at the top of his game seems pretty insurmountable.
I enlisted a fellow Mixpaneler (and avid Broncos fan) Brandon Skerda, one of our star Support Engineers, to help find the answers amid plenty of NFL data from the regular season and post-season from Armchair Analysis. Our goal? To attempt to prove out (or disprove) two things that we believe oddsmakers are assuming to be true:
- Assumption 1: The Patriots will win the championship despite Tom Brady missing the first four games of the season, because they’re a strong team in the regular season
- Assumption 2: The Patriots will win the championship because they’re strong all-around, and will have Tom Brady back for the playoffs
Are you ready for some data analytics?
For the purposes of this article, we’re going to look at some quantitative factors to the Patriots’ success in the Brady era (2000-2015) that have nothing directly to do with Tom Brady, to see if we can find some kind of metrics that point to why oddsmakers are still so bullish on Belichick’s Boys.
We took NFL data from more than 700,000 plays. We treated each play as an event, and each play had its own properties. Those properties included offensive team, defensive team, yards gained, special teams points, field surface, play type (rush/pass), points scored… and dozens more. Here’s a peek at the Live View of the project (specifically, Super Bowl 50):
Our customers put in millions of data points like these into Mixpanel every day, so this was a great way to see how flexible and fun data analytics can be!
After the NFL data points were in, we ran a variety of reports over the data to lift out and visualize some pertinent stats, including a lot of Segmentation reports and Formulas, too. This isn’t an exhaustive analysis, but it should provide a lot of data for discussion, whether you love the Tom Brady or love to hate him—or maybe you just love Mixpanel.
Pink stripes and NFL data
It goes without saying that the New England Patriots have been an amazingly successful team over the last 15 seasons. So, let’s consider the “worst case” scenario: the Patriots lose all four games without Tom Brady. No team in modern NFL history has ever overcome a 0-4 record to win the championship. Nor has a 0-3 team (although a few have managed to reach the playoffs). Therefore, it’s critical that the Patriots win at least two of their first four games to (historically and probability-wise) have a shot at winning it all.
Tom Brady has started for the New England Patriots consistently since he replaced an injured Drew Bledsoe in early 2001. The qualitative impact of Brady’s consistency as a starter is difficult to overstate, but it certainly poses a quantitative issue: there’s not a lot of Belichick-era, non-Brady Patriots data. The exception to this is 2008, which is the only non-Brady season since he got the starting job (thanks to a season-ending injury inflicted during the first regular season game by the Safety Who Shall Not Be Named).
So, let’s start with the most basic and obvious of Brady-related offensive stats to help us visualize what the Pats are missing: passing yards for each of the Patriots’ championship-winning seasons compared to the total passing yards in 2008, the only non-Brady season:
Patriots total passing yards,
Championship-winning seasons + 2008
From L to R: 2014, 2003, 2004, 2001, and 2008
Okay, that’s one heck of a spike there for their most recent championship season—5,069 total passing yards in 2014, meaning Vegas would be silly not to rely on Tom Brady being back for the playoffs. But a Patriots team giving up that kind of aerial firepower for four games doesn’t appear to bode well. However, the fact that the Patriots would have made the playoffs in 2008 (were it not for the 8-8 AFC West Champion San Diego Chargers, which I’m totally not bitter about) with a mere ~3,500 passing yards from a backup QB is encouraging.
It turns out, however, that the Patriots did better on offense in 2008 without Tom Brady than I thought they did:
Total points scored by NFL team, 2008
(regular & post-season)
New England was 5th overall for total points for the 2008 regular season
New England ranked 5th overall for points scored during their non-Brady season. That’s a little more than statistical noise! That year’s championship winner, the Pittsburgh Steelers, ranked 9th in total points scored. Their championship opponent, the Arizona Cardinals, were 1st. So while not an indicator (in the non-Brady year of 2008, at least) of ultimate success, both championship contenders were in the top 10. This looks promising for Assumptions 1 and 2.
But what’s the stat behind the stat? To me, it’s this: New England managed to land in a respectable spot for points scored in 2008 without their star QB; meaning, they weren’t some lame duck team without Brady.
But if offense were all it took to win it all, Arizona would’ve been the champs that year. Defense wins games, too. And it’s a great thing to look at when you’re assessing the strength of a team without their starting quarterback.
So, let’s see how New England’s D compares to the rest of the league:
Total yards allowed by defenses, 2000-2015
Most yards allowed in the league!! Get it together, New England!
Uh oh. Over the last 15 years, New England’s defense has allowed 96,302 yards—more than any other team in the league.
The Segmentation report above doesn’t show a lot of variation among the various teams (at least not among the top 12 for points allowed by defense), but that might be a good thing; consistently good defensive teams like New Orleans, Seattle, Indianapolis, and Green Bay still allow other teams’ offenses to gain a lot of yards. For comparison, Houston’s defense has allowed the fewest yards at 76,902. But they’ve only been in the league since 2002 and have not seen a lot of post-season play. Pittsburgh might be a better comparison to contrast with the Patriots; their defense allowed 81,352 yards over the same time period—almost 15,000 fewer yards than New England.
This makes me nervous. Maybe 2008’s winning record really was just a fluke? I had to look at the stat that counts to verify this question—defensive points allowed:
Most points allowed by team, 2000-2015
New England (not pictured): 11th fewest points allowed, 2000-2015
Whew! New England’s not even among the top 12 teams that allowed the most points! Oakland has the distinction of allowing the most points from their opponents, with 4,386. By comparison, the Patriots allowed a paltry 3,662.
Okay, that’s not paltry, but it’s not awful. As much as it pains me to say it, the data doesn’t lie: Baltimore has been the best, allowing only 2,858 points over the same timeframe:
Top 5 teams with most points allowed, 2000-2015
+ New England (21st) & Baltimore (last)
New England’s defense has allowed 3,662 points by opponents from 2000-2015
“Bend don’t break” defense is what New England is known for, and this set of NFL data bears that out. While the Patriots do allow the most yards, there are still 21 teams that allow more points than New England does. An important distinction.
This might be all that’s needed to prove out both Assumption 1 and Assumption 2: the Patriots have an historically solid offense AND defense and even if they don’t do well, Brady will be back. In the meantime, this consistency is nice insurance for their backup QB.
Speaking of whom (and speaking of insurance)… there’s not enough Jimmy Garoppolo data to determine how well he might fare during the first four games of the 2016 season, so I’m going to skip over his 31 total pass attempts and look instead at another offensive, point-scoring weapon: the kicker.
Given that the Patriots have won championships on the strength of their kicker, I want to peer into this position a little more closely to see if kicker points might be why Vegas gives the Pats such good odds.
Total successful FGs, 2000-2015
(reg. & post-season)
New England: 2nd in points from field goals, 2000-2015
New England is #2 on the list, with 488 successful field goals from 2000-2015. Only Baltimore has more with 504. That’s some crazy good kicking, boys. Here’s why:
The Patriots have had two very solid kickers since 2000: Adam Vinatieri, the Patriots’ kicker from 1996-2005, and then Stephen Gostkowski, who’s been with the Patriots since 2006. Both are top 20 all-time kickers. That’s some uniquely awesome kicking talent for one team.
With that in mind, we used Mixpanel’s Formulas feature here to compare the two kickers’ accuracy to see whether there’s a relevant correlation to winning the championship (the Patriots won in 2001, 2003, 2004, and 2014) that might be attributed to the kickers’ success.
(DATA NOTE: To do this, we made each dot represent one year, starting with 2000; Gostkowski’s career started in 2006, hence the sudden uptick.)
Gostkowski vs. Vinatieri, 2000-2015
Blue = Vinatieri, Green = Gostkowski (squares denote championship seasons)
Vinatieri won the 2001 championship with an 81% accuracy rate. He then saw an almost 20% dip in accuracy for the 2003 regular season from 2002, down to a mere 73%. Yet that Patriots team still won the championship. The following year saw a spike in his accuracy to a whopping 95%, culminating in another championship title. Vinatieri’s accuracy has fluctuated dramatically over the years, yet he notched three championships for the Pats, at least two of which were reached or won as the result of some crazy clutch kicks.
Gostkowski, on the other hand, has been remarkably consistent (he is, after all, the #3 all-time kicker in the league)—he has never dipped below 77% accuracy and has been trending steadily upward since 2012—but he’s netted just one championship title for the team. Gostkowski is statistically a better kicker, but Vinatieri has been more consistently clutch; he’s made the FGs when they mattered most.
Given Gostkowski’s accuracy and consistency, it’s likely that he is a factor in oddsmakers’ confidence in the 2016 Patriots team. But will that be enough to help win the first four games (Assumption 1), AND help them win it all (Assumption 2)? It’s more probable than not that the Pats are going to need a hell of a lot more offense.
To that end, let’s look at the most basic non-Brady offensive stat: points from the rush.
Top 12 teams with most rushing points, 2000-2015
Patriots are 1st in total rushing points from 2000-2015 (regular season & post-season)
I could probably wrap up this article right here. By the looks of this, all Jimmy G’s got to do is hand the ball off successfully. For the last 15 years, the Patriots have outscored all other teams in the NFL with the rush. New England has almost 2,000 total points from rush plays since 2000.
But wait a minute. The Patriots’ rush offense has been terrible for the past three seasons. If we look at just those seasons, it’s a much bleaker picture. The Patriots didn’t even make in the top 12 in 2014 or 2015. Also, a fair amount of those rushing scores have been Tom Brady’s classic QB sneaks (netting 159 points since 2000). Is this enough to disprove Assumptions 1 and 2?
To see if there’s still reason to be confident about a Brady-less Patriots team, I turned back to some team stats that speak more to the character of these teams as a whole, like how they perform under pressure.
What’s perhaps most impressive about Patriots teams under Bill Belichick is how well they do when they’re under duress—especially when they’re the away team:
Patriots regular season away wins, 2000-2011
Patriots won 6 of 8 away games in Brady-less 2008 season
Every team in the NFL plays eight away games every season. And the Patriots are often picked to win away games, even against rivals, even with hostile crowds, and even in 2008, when they won 6 of 8. Despite a lousy 2000 and a dip in 2009 (when they still finished 10-6), they’ve averaged around 5 away wins per regular season since 2000. Here’s how they look more recently:
Patriots regular season away wins, 2012-2015
Patriots’ most recent championship season highlighted
Heck, they won it all in 2014 with just five away wins—one fewer than their Brady-less 2008 season.
The reasons for this are many, and likely down to solid fundamentals and intangibles, as well as a zero-tolerance attitude towards poor performance (i.e., you mess up, you get benched). One of those key fundamentals that the Patriots are good at, and one that has been debated and dissected by far more sophisticated minds than mine, are turnovers.
You can win a game with only three points, but if you have “only” three turnovers, you’re likely to lose. So, let’s look at fumbles (those non-QB turnovers for a non-Brady look at the data):
Total # of fumbles: top 5 teams, 2000-2015
+ New England (26th) & Atlanta (last)
Since 2000, the Patriots have allowed the 6th fewest fumbles in the NFL
Arizona has had the most fumbles with 450, and Atlanta the fewest with 331. New England sits at #6 with 355. Not quite a conspiracy of efficiency, but most definitely impressive. When we look at how the Patriots stack up in the AFC East, however, the picture gets even more impressive:
AFC East: total # of fumbles, 2000-2015
L to R: Buffalo, 449 fumbles; Miami, 410 fumbles; New York Jets 375 fumbles;
New England, 355 fumbles – 2000-2015, including post-season play
Since 2000, the Patriots have had the fewest fumbles in their division (even when including the post-season). Given that New England plays the Bills, Jets, and Dolphins twice each, every regular season, it’s critical to hang onto the ball better than your divisional rivals. And given that two of those first four Brady-less games will be against the Dolphins and the Bills, making sure this positive trend continues is important.
Conclusion: do your job
Greatness takes many forms, both objective and subjective. And there are many, MANY more analyses we could run on NFL teams (and we likely will).
But if these few Mixpanel reports are any indication, the historical consistency of the Patriots appears to be what matters most to Vegas right now. That means that Assumption 1 (that the Patriots will win the “big game” because they can hold their own for four games without Tom Brady) is probably how oddsmakers are viewing even a 2-2 start. But don’t count out Tom Brady, who seems to be getting better as he ages; Assumption 2 (that the Patriots will win it all because Tom Brady will be back for the playoffs) is likely a big factor as well.
Fortunatly for the Patriots, there are many ways to win a football game, and ugly wins are still wins. That’s what Belichick’s getting at about stats being for losers; you can make stats say anything. What matters is that you win. Love him or despise him, you’ll never get a participation trophy from Bill Belichick. And you’ll never see a “Well, We Tried!” banner hanging from Gillette Stadium.
No caption needed.
No team—including the Patriots—is built around any one player. Like everything else, winning in the NFL requires the right combination of qualitative and quantitative factors. Our 10,000 ft. overview of this NFL data using Mixpanel only tells part of the story, but it’s a pretty intriguing case.
And if all all else fails, at least they’ve got Gronk.
Artwork courtesy Jack Kurzenknabe, and is in the public domain.