Tom Brady started 57 games in which his defense held the opponent to 10 points or less, the highest value among active quarterbacks. Would you be shocked to find out that the Patriots (and Tom Brady) - are 57-0 in those games?
As a quarterback, there clearly is a benefit in playing for a team that can hold the opponent to a low points total. Ben Roethlisberger is 52-1 when the opponent scored 10 points or less, Joe Flacco and Eli Manning are both 35-0, Carson Palmer is 31-0.
The reality is it's hard to lose a game in the NFL when the opponent is held to 10 points or less. Over the last four years, quarterbacks are 243-8 in such games for a staggering 0.968 percentage.
Yet if you were to read some of the headlines floating around after gameday - and if you were to give those headlines any credence - you could easily get the impression that it was the superstar QB all by himself who won those games.
The QB, whether you like it or not, is the only player on a football team who has a W/L record in his personal stats. But that doesn't mean the QB is actually winning games for his team. Teams can win games with a strong defense and a good ground game despite the QB, yet the QB will get the win on his record.
Which is why today I'm taking a closer look the W/L records of NFL QBs to figure out which NFL quarterbacks are winning games for their teams, and which QBs are just along for the ride. To do that, I'll look at 'wins over average' (WOA) to understand which offenses, and by extension their QBs, contributed to wins for their teams.
WOA was developed by Doug Drinen at Pro-football-reference.com (PFR) in a post on adjusting QB records. I've taken his approach and modified it slightly for this post. The underlying assumption for WOA is that it's a lot easier for a QB to win games when the defense holds the opponent to 10 points or less than it would be if the defense gave up 30 or more points.
WOA calculates a QB's winning percentage against a given range of points scored by the opponent (e.g., 0-10 points, 11-17 points etc.) and then compares that result against how many games an NFL-average QB would have been expected to win.
To calculate WOA, we first need a baseline to determine what 'NFL average performance' looks like. For our purposes here I'll use the four-year league average winning percentage in regular season games between 2013 and 2016. The table below summarizes the regular season records for all NFL teams for the last four years (excluding tied games), forms five clusters by points allowed, and shows the winning percentages for each cluster.
How to read the table: A QB (and by extension, the team) will almost always win the game when the defense holds the opposing team to 10 points or less (243-8 winning record). The NFL average winning percentage is .968 against the '0-10' points cluster. The more points a team allows, the lower its chances of winning the game. When your defense allows 32 points or more, your QB's chances of winning the game are very low (.0107 winning percentage, 38-318 record).
What separates the good from the merely average QBs is their ability to win games an average QB would not win. And that is exactly what WOA measures. We'll use Tony Romo's 125 career starts (in which had at least 10 pass attempts) to further explain the concept.
Tony Romo and Wins Over Average: Take Romo's record of 23-9 as a starter in 32 games when the defense allowed between 18 and 24 points.
The average NFL QB would be expected to win 16.7 of those games (32 games x .522 NFL average win percentage). Romo won 23 of those games, so he delivered +6.3 wins over average in this points cluster.
In the 11-17 points cluster, Romo has a 21-4 record in 25 games. An average NFL QB would be expected to win 19.5 games (25 games x .781 NFL win percentage). Romo won 21 of those games for a +1.5 WOA. Across all points allowed clusters, Romo has accumulated +11.0 WOA as per the table above.
Statistically, Romo accounted for eleven more wins for the Cowboys than an average NFL quarterback would have. At first glance, that doesn't seem like all that much, so in the next step I'll look at how number compares to other quarterbacks around the league.
Before we do that though, keep in mind that when I say "Tony Romo" or any other quarterback, I'm actually talking about that quarterback's entire offense. Quarterbacks don't win or lose games, teams do. But if quarterbacks are going to be measured by wins, let's at least make sure we understand the quality of those wins.
In the next table, I've listed the 29 projected starting quarterbacks and 12 backups with the most career regular season starts. I've calculated their Wins Over Average in the same way I calculated Romo's numbers above. Since every QB on the list has started in a different number of games, I've added an additional column in which I divided the WOA by the total number of regular season games started, which gives me WOA in %. That allows for a better comparison of the quarterbacks, regardless of how many games they started.
Note that only career starts with at least 10 pass attempts are included.
Career Wins Over Average by Quarterback, regular season games
The top QBs
No real surprises at the top, where Tom Brady, Aaron Rodgers, Drew Brees, and Matt Ryan outclass most other quarterbacks. A big surprise here, at least for me, is clearly the second-ranked Andrew Luck. In his five years in the league, Luck has averaged a mediocre 87.3 passer rating, which is far below Brady's (97.2) and Rodgers' (104.1) career averages, yet Luck (and his offense) has been winning a lot of games for the Colts. The combination of average passer rating and above average WOA suggests that the Colts have been winning more games than Luck's performance as a pure passer would warrant, and maybe that's part of his appeal: He's not really great passer but he and his team have consistently found ways to win like great QBs and their teams do.
Derek Carr and Kirk Cousins occupy a tier just below the top QBs. Carr struggled a little as a rookie, but has delivered two very strong seasons since, Cousins only became a starter in 2015. Using only the last two seasons to calculate each player's WOA percentage shows 19.4% for Derek Carr and 10.0% for Cousins. These two are poised to be the next superstar QBs in the league, and it's very gratifying to see how the dysfunctional Washington front office is ensuring that Cousins will join another team once the Redskins run out of franchise tags.
The caretaker QBs
The next group of QBs, with Ben Roethlisberger, Russell Wilson, Eli Manning, Philip Rivers, and Cam Newton take care of business at the QB spot, but are barely above replacement level. Ironically, this group has won five Super Bowls between them, which just goes to show how much of a team effort a Super Bowl win is.
Russell Wilson in some ways is the exact opposite of Andrew Luck. He has an above average five-year passer rating of 99.6, but that hasn't translated into Wilson (and the Seahawks offense) getting a lot of wins over average. The Seahawks win games primarily via their defense, but if they have to rely on their offense to pull out a game, WOA suggests they'll have to wait quite a while. And perhaps that's what's driving the rumored dissatisfaction in Seattle with Wilson. When it's time to deliver, the mailbox remains empty.
In any case, the caretaker QBs may not win their teams a lot of games, but at least they aren't losing too many either, at least as measured by WOA.
Just along for the ride
The next six QBs on the list (Carson Palmer, Andy Dalton, Matthew Stafford, Joe Flacco, Alex Smith, and Ryan Tannehill) have a WOA percentage that hovers around zero. By definition, these are average QBs. They win games they are expected to win, but don't deliver anything beyond that, their draft pedigree, monster contracts, or endorsement deals notwithstanding.
Sam Bradford, Josh McCown, and Blake Bortles, all have a significantly negative WOA percentage, meaning they (frequently) lose games an average NFL QB would win. All three are currently penciled in as starters for their teams because those teams don't have better alternatives, but all three teams will be looking for new QBs next season.
The young guns
The first group of 21 projected starters is followed by 10 more projected starters who 32 or fewer starts to their name. With stats like WOA, we quickly run into sample size issues with low career start totals.
But I wanted to include them anyway, because Dak Prescott is particularly noteworthy in terms of WOA. Consider that he has already accumulated half the WOA in 15 starts that Eli Manning needed 199 starts to achieve. But like the other (mostly young) QBs in this group, he'll have to show that he can win games consistently over a longer period, or he'll ultimately be relegated to the final group, the backups.
But if Prescott's rookie season is not a fluke, he'll join Carr and Cousins as the next generation of top QBs, while the other young guns will eventually settle into some of the lesser tiers described above.
This final group contains the journeymen of the NFL, most of which have already played for multiple teams and are clinging to their roster spot by a thread.
Obviously, there are many ways to evaluate a QB's performance. This one doesn't take any of the ubiquitous volume stats into consideration, and neither does it take any efficiency metrics into account. WOA is a proxy metric for whether a QB (and his offense) wins games for his teams. It doesn't matter if he does it ugly (e.g. a low passer rating), with a lot of mistakes (e.g. INTs), or if he awkwardly stumbles forward for a first down. WOA also doesn't care about the arm strength of a QB, his draft pedigree, how many rings he's won, or the number of endorsement deals he's signed.
The reason QBs are the only football players with a W/L record attached to their name is their inordinate impact on the result of a game. WOA is an attempt to measure just how good QBs are at winning games.
WOA is not a perfect metric by any measure, but taken with other valid metrics out there, I hope it helps solidify an overall picture of today's NFL quarterbacks, and perhaps raises a question or two.