Tom Brady and His Incredible Luck in High-Scoring Games (Part I)

On Sunday afternoon, Tom Brady improved to 92-0 as a starter in games where his team scored at least 35 points, easily the best record in NFL history for a feat that leads to a win about 92% of the time since 2001. Peyton Manning (63-3), Drew Brees (56-5), Aaron Rodgers (42-2), and Ben Roethlisberger (35-4) not only all have multiple losses in games where they scored 35-plus, but they all lost a playoff game (two in Ben’s case) after scoring that much as well.

But never Brady. When the King of Kings puts up points, even if it involves two Matt Ryan pick-sixes as it did in Sunday’s 48-25 win over Atlanta, his teams win the game every time. On Sunday night in Baltimore, Patrick Mahomes just lost his third career game in 56 starts after scoring at least 35 points. He is 15-3 now. Thanks for nothing, Clyde.

I guess Brady’s just better than everyone else in high-scoring games then. End of post. No 5,000 words and graphs necessary.

But that’s just not my style. Instead, I am finally going to provide a deep dive into the data on quarterbacks and points allowed after years of citing these stats without much context. As it turns out, there is a sweet spot on the scoreboard where Brady does in fact outperform his top peers, but like with virtually every Brady stat that’s ever existed, he only outperforms them in team win percentage.

The Background

With roughly 18 years of experience in dealing with Tom Brady’s cult-like following on the internet, I know how his worshippers think better than probably anyone. While the casuals post pictures of his rings, and the weirdos post pictures of Gisele – as if her bank account isn’t the hottest thing about her – it’s easy to just ignore those people. Life is too short and you’re never going to find any substance there.

What better draws my attention is when they try to use any number besides the ring count to justify the case for their alleged GOAT.

Look, I get why they are so defensive and protective of his legacy. It’s really hard to prop up Brady as the greatest thing since sliced bread, which he doesn’t even eat. Would you feel comfortable in telling someone that the greatest football player of all time is someone who has been named first-team All-Pro, the best at their position, just one more time than Rich Gannon?

When someone brings up Michael Jordan or Wayne Gretzky as the GOAT in their leagues, they don’t need to rely on a ring count or team winning percentage to make their case. The individual resume speaks for itself. The record-setting dominance in both the regular season and postseason that still resonates decades after their retirements, as well as the eye test, it all speaks for itself. Just look:

When I come at the king, I don’t miss. Brady just pales in comparison to those two actual GOATs. But enough GOAT talk for today. The point is any pro-Brady numbers argument boils down into something very simple:

First, pick a split. Playoff games, division games, island games, December games, games after a bye week, 4QC/GWD opportunities, games against top 10 defenses, games against playoff/winning teams, games with 50+ pass attempts, games with 3+ interceptions, games taking three sacks, games in freezing temperatures, games with a passer rating under 80, etc.

When you compare Brady to his peers in the split, chances are he will have the best winning percentage, but he won’t have the best individual statistics. Pick any metric (PR, YPA, ANY/A, DVOA, DYAR, QBR, EPA, WPA, CPOE) that’s not the equivalent of an English teacher subjectively grading your Catcher in the Rye essay, and this will be true for just about anything you dig into.

This has always been the case too. He is eternally “Brady Just Wins” and that’s the way he was covered by media for six seasons in the Spygate era. You know, back when he shared a lunch table with Marc Bulger and Matt Hasselbeck when it came to quarterback stats, but the media forced him on us as the best in the game because his team won three Super Bowls by three points each. Jim Nantz would cream his pants weekly on CBS about the latest “Brady record” that was just a graphic of him (grinning pre-plastic surgery) with some crazy win percentage record the Patriots had back in 2001-06 or some “attempts without an interception streak” that Brady would then end with a pick in a playoff game he still went on to win.

But after Spygate came about in 2007, Brady developed into a better player and started playing with loaded offenses and the two most stat-inflating receivers (Randy Moss and Rob Gronkowski) of this era, if not of all time. He’s currently on a Tampa Bay offense that features four Hall of Fame-caliber receivers, three of which he had nothing to do with developing into studs. Keep in mind the 2018-19 Tampa Bay offenses, which did not have Gronk or Antonio Brown, and had error-prone Jameis Winston and Ryan Fitzpatrick at quarterback, threw for over 10,400 yards and 69 touchdowns in those two years. Yet here we are hyping up Brady for a nine-game winning streak with scoring 30 points where three of the games were against the Atlanta Falcons during a pandemic.

But the big-picture stuff can wait for another day as I’m getting off track again. This is about points allowed. Like Bruce Wayne seeing the Bat Signal, I was pointed to a tweet last week about Brady winning high-scoring games better than his peers.

I was asked if the graph was wrong, and initially I thought it was since the curves shouldn’t be that smooth. Then I realized the x-axis was when the “opponent scored at least x points” instead of “opponent scored exactly x points.” It reminded me of this old ESPN article that featured this paragraph:

“NFL teams have won just 23.6 percent of the time since 2001 when they gave up 21-plus points. Brady’s 66-50 record in these higher-scoring games equates to a .569 winning percentage — a whopping 2.4 times better than the league average. No other QB with 50 games played since 2001 has won even half of those tough-to-win games. Brady remains on top as the scoring bar climbs even higher. He is a league-best 20-29 (.408) when the other team scores 28-plus points, far above the .151 league average.”

The graph and this paragraph are both correct, but they are both very misleading. By using a minimum number of points allowed and not putting a cap on the other end, both are ignoring that Brady’s teams do a much better job of not having games where they allow 30+ and 40+ points. I’ve been on this for years.

While teams have won 27.2% of their games during Brady’s career where they allowed 21-plus points (it’s gone up since that 2016 article), that includes games where they allowed 38, 45, or 54 points too. Literally anything 21 or higher would count. However, when teams allow 35-plus points, they win just 7.9% of the time. If you look at games where teams allowed exactly 21 points since 2001, they win those 63.8% of the time. Why? That’s a below-average amount of points in this era. You should win those games, and if you’re a Hall of Fame quarterback, you should win those at a high rate.

In fact, I just tweeted something to this effect on September 3 about Brady, Jared Goff, and games where a QB’s team allowed 28-plus points. While 28 is the minimum, that ignores how far past 28 points the teams went. Sure enough, Brady’s average 28+ game had the lowest average (32.3) of points allowed while Goff’s had the highest (37.7) of the 46 QBs in the study. So, is Brady having the best win% really as impressive as it sounds when you learn that additional information? I don’t think so.

Confession: I too am guilty in the past of presenting these points allowed stats in misleading fashion or at least not with better context.

So, how do we improve on this?

The Key Ranges for Points Allowed

With NFL teams averaging a record-high 24.7 points per game in 2020, these are stats that will have to be adjusted in the future. It was harder to win a game in 2001, Brady’s first year as a starter, when a team allowed 21 points and the league average was 20.3 points scored. Now if you allow 21 points in a league where the average scored is 24.7 points, then you got an above-average game out of your defense in that regard. For reference, teams in 2021 are averaging exactly 24.0 points per game thru Week 2.

My thinking for over a decade on these points allowed stats is to bunch everything under 17 points together as a low-scoring game. Peyton Manning was 89-0 in games he finished when the team allowed fewer than 17 points, which is still the gold standard for that range. His only technical loss was in 2007 against the Titans (16-10) in a Week 17 game where he left very early for his playoff rest. I always figured anything under 17 just means you can score two touchdowns and a field goal and get the win. Should be no big deal for a Hall of Famer.

Then using 24-plus or “more than 24 points” are probably the way I’ve used these stats the most over the years on Twitter. Fans of Alex Smith and Carson Wentz can attest to that. From 2007 to 2019, the average points scored was 22.5 in the NFL. Since teams rarely land on exactly 22 or 23, I viewed 24+ as the minimum standard for a game with above-average scoring, especially since you likely need to land on 27-28 or 30+ to win that day.

My third favorite range of points was to use 27+ or 28+ points. This is basically saying “games where you had to score 30+ to win.” Anything higher and we’re probably stuck in a land of small sample sizes and terrible win percentages.

But with all the data available to us these days, we can do something better than this, right? What’s really the difference between allowing 22 and 24 points? So, I looked at every game for 2001-2020 (playoffs included) and got the winning percentage for each point allowed total. First, a bubble graph that shows the win percentage for each point total.

Obviously, as points allowed goes up, win percentage decreases. But you can see four little outliers with small samples of games (smaller bubbles) at 32, 39, 46, and 51 points allowed where winning percentage is higher than expected. The 46 is when the 2019 49ers won that huge 48-46 game in New Orleans to help them get the No. 1 seed. I think you know what 54-51 is all about.

Now here is the data grouped together in what I am going to consider the six key ranges for points allowed: 0-11, 12-18, 19-25, 26-32, 33-39, and 40+.

0-11: These are the easy games. With the exception of 9 points, these are all above 92% win rates. At 9 points, a few more teams are content with winning after kicking three field goals. Sounds like a Jets thing to me.

12-18: All still winning records above 60%, but we dip under 80% every time except for a bump at 14 points. Why are there bumps at 7, 14, 21, and 28? Those are the more common scores with touchdowns and extra points in this game. If a team scores exactly 14 points, chances are those two touchdowns were all they could get that day, but it wasn’t all they needed. If a team finishes on 13 or 15, they may have only needed a late field goal to win a 13-10 or 15-12 type of game.

19-25: Now we start to see losing records except for at 20 and 21. Why the bump there? Again, the more common scores. If a team scores 19, it could have just won a 19-9 or 19-16 game. You don’t see too many teams down 21-19 lose on a game-tying two-point conversion attempt. Nineteen is just what they needed that day. Similar thing at 23 points when a lot of games could end 23-16 or 23-20. But 23 is the lowest win% here at 36.3%.

26-32: Now we’re dropping under 26.5% except for the expected bump at 28 (35.8%) and the aforementioned outlier at 32 (42.9%) on just 63 games. Every other point total from 19 to 35 has at least 108 games in the sample.

33-39: This is miracle territory as only the outlier at 39 points is above 18%. What do we have to thank for that 6-22 record at 39 points? Thank Dan Quinn’s Falcons for not picking up that onside kick and losing to Dallas 40-39 last year. Thank Dwayne Rudd for taking his helmet off and helping the Browns lose to the Chiefs on opening day in 2002. Also a shoutout to the 2004 Seahawks for this inexplicable choke against the Cowboys on Monday night in a 43-39 loss. Two onside kick recoveries and Rudd keeping his helmet on from making that 3-25 (.107).

40+: I think it’s safe to group everything from 40 through 62 (the night the Saints destroyed the 2011 Colts) together for the last range. The only one with a winning percentage above 6.7% is that outlier at 46 due to the George Kittle play on fourth down in New Orleans in 2019.

Okay, now I have six ranges I feel better about using for this. Again, we’ll have to keep an eye on how things progress in the NFL with scoring. Maybe my days of doing “since 2001” stats can be dialed back once Brady (and Ben and Rodgers) finally retires. Then “since 2008” might become my new standard, since that was the year Ryan and Flacco changed the rookie quarterback expectations, and the 2007 Patriots had an influence on the league with the shotgun-spread offense. Team scoring shot up to 22.0 points per game in 2008 for the first time since 1965. But we’re not there yet, so it’s since 2001 for now.

Brady vs. His Peers in High-Scoring Games

How does Brady fare against his peers in these six different ranges? Unlike the original chart, I’m interested in comparing Brady to the only four quarterbacks who have sustained success with both winning games and accumulating stats on their way to locking up a first-ballot bust in Canton: Peyton Manning, Drew Brees, Ben Roethlisberger, and Aaron Rodgers. I’m sure the day will come when we can add Russell Wilson and Patrick Mahomes to this, but not today.

First, a very telling scatter plot of these five quarterbacks for each of the six key points allowed ranges. The x-axis is their win% as a starter with no games removed for injury (would affect Rodgers and Ben the most) or playoff rest (would affect Manning the most). The y-axis is their Adjusted Net Yards per Pass Attempt (ANY/A) in these games, which takes sacks into account, but does not include rushing touchdowns (would help Rodgers the most) or fumbles (would help Manning the most). This methodology is arguably most beneficial to Brady than the other four given his history of injury/playoff rest games/fumbles/rushing TDs. The original tweet’s graph used 1999-2020, so he also excluded Manning’s rookie year. I am excluding 1999-2000 as well, but that shouldn’t change much here. But for every other quarterback it is their full career minus two games in 2021 for Brady, Rodgers, and Roethlisberger, the last three standing.

Tale as old as time: Brady does not have the best statistics in any of the point ranges, but he has the best winning percentage (or at least a tie for it) in half of the six ranges. However, that’s not as dominant as one would expect from the way the original tweet by a guy named Tucker claimed it to be. You can also see that Brady’s lone sweet spot is the 26-32 range, but again, more on that below.

0-11: These QBs are a combined 248-2 in this range. Ben’s loss was a 9-0 game on MNF to the Jaguars in 2006, his first game after his motorcycle accident and an emergency appendectomy. He was just not healthy enough at that point. Likewise, Rodgers lost a 7-3 game to the 2010 Lions after leaving with a concussion after only 11 pass attempts. Brees had the most dominant stats (110.9 PR, 8.09 ANY/A). Brady (7.51) has the second-lowest ANY/A in this range.

12-18: Manning had the best record here (59-2), even including that playoff rest loss to the 2007 Titans. Brees (.909) and Rodgers (.902) are one game behind Brady (.919) in win%, but Brady has the lowest completion percentage, YPA, and second-lowest passer rating and ANY/A in this range.  The Tuck Rule is literally the difference here in Brady having the second-best win% and the fourth-best win% as he easily could have lost that game 13-10 without that rule and the greatest kick in NFL history. Meanwhile, Rodgers has the Fail Mary as one of his losses here (14-12 in Seattle) and a 2015 game against Detroit that would have been a 19-18 win if Mason Crosby made a 52-yard field goal. Remember, Brady has lost one game in his whole career after a missed clutch field goal (20-18 vs. 2012 Cardinals) and that comes in the next range.

19-25: In this range Brees does drop off statistically with the lowest win% (.629) and second-lowest ANY/A (6.92). Manning (.762) just edges out Brady (.753) for the best win percentage with Rodgers (.733) not far behind. In this range Brady moves up to third in passer rating (97.8) and ANY/A (7.24), but he certainly doesn’t dominate his peers or the stats Rodgers had (107.0 PR and 7.94 ANY/A). While I did mention the Stephen Gostkowski miss against Arizona costing Brady a win here, I raise you Adam Vinatieri missing a 29-yard field goal against the 2007 Chargers (23-21 loss) and Mike Vanderjagt missing a 45-yard field goal against the 2005 Steelers (21-18 loss) in this range for Manning. The point is Brady does not dominate his peers at winning games or how he performed in them at this range (or the first two ranges).

26-32: Every range is pretty tight, but this one has the most separation of them all. Despite Brady firmly ranking third in ANY/A (6.65) behind Manning (7.19) and Rodgers (6.90), he still has a winning record (37-31, .544) in this range while the others are all under 43%. Roethlisberger is especially bad here at 12-29 (.293) and 5.74 ANY/A. Why does Brady have such a winning advantage here despite not blowing away his toughest peers statistically? See Part II. I’ll give you a hint though: 28-3 is a game in this range.

33-39: Brady has the worst record (4-16, .200) and the lowest YPA (6.81), passer rating (85.7), and ANY/A (5.70) in this range. It includes two of his highest-scoring losses of his career in games against Manning: 38-34 in the 2006 AFC Championship Game and 35-34 on 4th-and-2 in 2009, a game that could have ended 34-28 if he didn’t turn into Alex Smith and threw past the sticks on that drive to deny Peyton the ball. But this is a bad looking range for Brady as his teams scored the fewest points (25.6) and allowed the fewest too (34.9). Manning had the best record (5-13) but not the best stats thanks to some of his most improbable comebacks. Not just the Patriots wins but also that 38-35 overtime win in Tampa Bay in 2003.

40+: While Roethlisberger had the lowest ANY/A in the first four ranges, he has the highest ANY/A in both the 33-39 and 40+ ranges. As I wrote in January after the 48-37 playoff loss to the Browns, he has a way of trying his best with huge passing days when the defense completely shits the bed to Baker Mayfield and Blake Bortles. So, this was amusing to see. Manning (1-11) definitely had some incredibly bad days when his team gave up a 40-burger, but it usually was a sign that the whole team imploded like the 41-0 playoff loss to the Jets and the 2013 Super Bowl against Seattle (43-8) that basically fell apart on the first snap over his head for a safety. Not only does Brady (1-6) have the lowest rate of 40+ games in his career, but his teams allowed the lowest average (40.9) above 40 points compared to Manning (43.2), Brees (44.0), Ben (44.6), and Rodgers (46.3). Brady got one win over Mahomes in 2018 (43-40) by virtue of getting the ball last that night. All Brees needed was one defensive stop on fourth down against Kittle in 2019 to beat the 49ers 46-45 and get the No. 1 seed, but he didn’t get it of course. Otherwise, he’d have the best record here at 3-13.

Again, where exactly does Brady shine over his peers? I removed the 26-32 range and summed up all the remaining games and put it in this chart:

Brady has the highest win% despite having the lowest YPA, completion percentage, and the second-lowest passer rating and ANY/A. His team averages the third-most points, but the key thing is his teams still allow the fewest points per game in these five ranges (23.8).

We should try to account for the different distributions of games into each range for the quarterbacks. For example, 23.8% of Brady’s starts have seen his team allow fewer than 12 points. That’s almost double that of Brees (12.2%) and Rodgers (13.8%). If Brees had 23.8% of his starts play out like that, he’d have over 72 such games instead of his actual number of 37. Remember, he was 37-0 when this happened. Likewise, Brady has the lowest percentage of starts with 33-39 points allowed (5.8%) and 40+ points allowed (2.0%).

In giving everyone the same distribution of points allowed as Brady, this is what the numbers look like for all games for 2001-2020:

Brady still has the best record without the best stats, but that’s always a given. Manning is right behind him now and Brees makes a huge leap with an extra 33 wins and raising his win rate nearly 11 percentage points. He’s definitely had the worst defensive support of these five quarterbacks.

Maybe the worst luck too, but when it comes to luck, no one holds a candle to Brady, the LOAT.

Check back for Part II where I find Brady’s voodoo magic in the 26-32 range.