When a backup quality, retread quarterback comes in and has a few good games as a starter, the conventional wisdom is that it won’t last. Whatever flash or flourish that may come from a QB that didn’t make it initially is most likely a temporary aberration from the norm, and a return to the mediocrity that made him a backup is inevitable.
And so it seems to be with Case Keenum’s new-found success as starting quarterback for the Minnesota Vikings. It won’t last. Pretty soon he’ll revert to his statistical mean, and he’ll once again be carrying a clipboard on the sideline. And so it has been at times with quarterbacks trying to star in their own second act.
The Problem of Using Statistics to Predict Sports
In the world of using statistics and regression analysis to predict the future, it is assumed that results tend to a normal curve, with the most frequent results occurring near the statistical mean, and fewer further away from it.
This works well, for example, in determining how precise an automated machine calibrated to cut a 6-foot length of pipe is likely to be in the future, based on a large sample of previous results. A well designed and calibrated machine will likely cut most pipe at exactly 6 feet, with perhaps a significant number of cuts within a millimeter shorter or longer, due to some rattle or loose parts, for example, and a relative few further off due to an occasional jam or malfunction. This distribution of results conform to a normal, or bell shaped curve on a graph.
But for guys playing football in an uncontrolled and changing environment, it is much less certain that individual results tend to that same normal curve- which is essential for accurate predictions- and why so many predictions about football- and sports in general- are inaccurate.
The best that can be said is that collectively, a team or player performance results may conform to a normal distribution curve. For example, NFL team win/loss records will coalesce around a .500 winning percentage in any given year in a normal distribution, or individual metrics such as QB passer rating, may also conform to a normal distribution for all quarterbacks collectively over a given season. But this doesn’t give any insight into how a particular quarterback will perform in the future.
As you can see from these graphs, QB X & Z had these passer ratings over the first 36 games of their career. The dotted line represents the trendline, and in both cases is fairly flat, one around a passer rating of 85, the other just below 80.
One is of a quarterback that went on to be a career journeyman- fairly average- and the other will be a future first ballot Hall of Famer.
But based on these two graphs, can you tell which one is which? Probably not.
And the reason is pretty clear. There is no real trend in either chart, and based on these charts alone the only reasonable conclusion would be that both QBs would likely be roughly mediocre QBs going forward, based on their performance over their first 36 games. Perhaps the one with the higher trend line around which the data points are distributed would have the better career going forward.
But such was not the case. QB X was in fact Brad Johnson’s early career passer ratings, while QB Z is Drew Brees’ early career passer ratings. In his subsequent career, Johnson continued the trendline established early on in his career, while Brees quickly improved to a 100 passer rating trendline for the rest of his career-to-date.
Do a QB’s Early Years Predict Future Performance?
Considering the example and charts above, the answer is no. QBs from Brees and Peyton Manning to Brett Favre to Terry Bradshaw have performed significantly better later in their careers than they did in their early years.
And yet there are plenty of cases where early performance did predict the future- as was the case with Brad Johnson. Aaron Rodgers, Russell Wilson, Dan Marino, John Elway, Roger Staubach and to a lesser extent Fran Tarkenton’s future performance could have been predicted from the results of their early years. In the case of Rodgers and Wilson, they had a relatively flat trendline at a high (100) level early on. Dan Marino’s early trendline was slightly downward from a very high level early on- a trend that continued the rest of his career. Elway had a slightly upward trendline from a low level that continued slowly over his long career. Staubach and Tarkenton had gently rising trendlines early in their careers, which continued for Staubach but eventually flattened out for Tarkenton.
In any case, it is not unusual for a QB to improve over his first few dozen games, as he learns the NFL game, etc. Sometimes talent limits that improvement (Brad Johnson), and sometimes QBs never really improve (Christian Ponder). But even Hall of Fame quarterbacks can start out slow.
Drew Brees had a 74.0 passer rating over his first 25 starts, completed 59.5% of his passes, had 27 TDs and 29 INTs, and an ANY/A (adjusted net yards per attempt) of 4.92.
In his subsequent 218 starts, he had a 99.0 passer rating, completed 67.58% of his passes, had 454 TDs, 196 INTs, and an ANY/A of 7.48. Apparently he did not revert to his “statistical mean” established over his first 25 games.
Terry Bradshaw, over his first 26 games, had a 46.6 passer rating, completed 46.6% of his passes, with 19 TDs and 46 INTs, leading to an ANY/A of just 3.35. This was during the 1970s, when passing wasn’t as prolific as today, but nevertheless well below average in his day.
In his subsequent 140 games, he had a 75.0 passer rating, completed 52.51% of his passes, including 191 TDs, 146 INTs, and a ANY/A of 6.18. Not earth-shattering, but nevertheless substantially better than his first 26 games.
Brett Favre, after 26 games, had a 79.9 passer rating, completed 63.07% of his passes, including 33 TDs and 31 INTs, and an ANY/A of 5.34.
He went on, in his subsequent 273 games, to have a passer rating of 86.7, completed 61.87% of his passes, with 475 TDs and 303 INTs, and an ANY/A of 6.33.
Similar to these QBs, Case Keenum, after 26 games, had a 78.4 passer rating, completed 58.4% of his passes, including 24 TDs and 20 INTs, en route to an ANY/A of 5.35 - fairly similar to both Brees and Favre.
And since those first 26 games, Keenum has improved to a 98.6 passer rating overall, completed 67.5% of his passes, including 16 TDs and 5 INTs, and an ANY/A of 7.35 - remarkably similar to Drew Brees (TD/INT ratio, not raw numbers obviously).
So, what does this tell us about Keenum’s future performance? Comparing him with Hall of Fame QBs is premature to be sure, but there is a discernible trendline as you look at Keenum’s performance over his first 36 games.
Let’s start by looking at Keenum’s passer ratings by game over his short career. Keenum has established an upward trendline to a high level, which is also a little less erratic late in the graph. Notice also the pattern of higher low points in the second half of the chart. His average per game passer rating prior to joining the Vikings was 80.26. His average per game passer rating since joining the Vikings is 99.06. He also has more 100+ passer rating games in 11 games with the Vikings than he had in his entire pre-Vikings career span of 26 games. He also has a current streak of four 100+ passer rating games. Pre-Vikings, Keenum had only one streak of two 100+ passer rating games- the first two of his career in 2013.
Next, let’s look at completion percentage. Here is an even more pronounced upward trendline, again with higher low points over the course of time, and improving to a relatively high level for all QBs. It should be noted as well that Keenum’s average yards per pass increased, rather than decreased, generally as completion percentage improved.
Lastly, below is a QBR chart for Keenum. QBR attempts to give more weight to critical plays (completing a 2 yard pass on 3rd and 1 is better than completing an 8 yard pass on third and 10), while discounting less successful plays and stats (i.e. garbage time) that do not have a big impact on the outcome of a game. It also attempts to assign value to a play (i.e. a QB doesn’t get as much credit for a bad throw when the WR makes a great catch for a completion). This is perhaps the most significant upward trendline, going from a low to high level relative to all QBs. Keenum’s QBR rating has gone from just under 40 overall in each of the last three seasons to over 75 this year - nearly doubling it.
Obviously the upward trendlines here for Keenum look auspicious for his future performance. Of course any number of things could impact that trend, such as injury or some negative external factor, but absent that, I can’t think of a quarterback having established a significant upward trendline to a high level early in his career, who reversed course immediately following that. In some cases the trend may reach a plateau (as it did for Tarkenton) or really represent a bridge from one level to another (like with Brees). But without some external change or injury, it seems less likely for an upward trend to abruptly reverse, once established and to a high level.
So, what then explains Keenum’s Improvement?
Keenum himself gave some important insight into his improvement when asked about it during an interview after the Atlanta game with Peter King of si.com/MMQB, who gave Keenum one of his offensive player of the week awards in his weekly column:
OFFENSIVE PLAYERS OF THE WEEK
Case Keenum, quarterback, Minnesota. This is a weekly award, because this ridiculous Keenum had a day, again, in Atlanta: 25 of 30, for a season-high 83 percent, with two touchdowns and no picks, in the 14-9 win over Atlanta. But it is more than that... Keenum, in four games, is an 80-percent passer, with nine TDs and two picks, for a passer rating of 124.1. Never in his pedestrian five-year career has Keenum had this kind of run, and he’s done it with the hot breath of Bridgewater on his neck.
“I don’t think those things are correlated,” Keenum told me after Sunday’s win. “In my mind I’ve done a good job of keeping my blinders on. I’ve learned to compartmentalize things. Honestly, I am doing the same things I always do, and just trying to be better. I may have gotten better through experience and very good coaching here. Plus, this is by far the best receiver group I’ve ever worked with.”
Beyond that, it was clear early on that Keenum had a good chemistry with the rest of the players. Kyle Rudolph commented on that in an interview with Chris Tomasson yesterday as well:
“Since the first play he walked into the huddle in Pittsburgh, he has had charisma and confidence about him,” Rudolph said after the Vikings’ 14-9 over over Atlanta on Sunday at Mercedes-Benz Stadium. “Like, ‘Look this offense isn’t going to take a step backward just because I’m stepping in.’ ”
You could see that confidence in the way the other players behaved toward him during the game in the huddle, and again in the comments they had about him after the games. And as the wins have piled up, so too has the confidence all around.
“Any time you’re successful, I think that breeds confidence,” Keenum said. “I think guys have confidence in each other. I think the receivers have confidence in me. I have confidence in the receivers. The offensive line is trusting each other. We trust the defense to make a stop every time that they’re out there. And I think the defense trusts us to go down there and win ballgames and put drives together when we need them.”
Keenum is the first quarterback to win the NFC Offensive Player of the Month (for November) since Brett Favre in 2009, after going 4-0 for the month and putting up numbers only Tom Brady could match. Keenum had this to say about it:
“It meant a lot, it really did,” Keenum said. “Those things are very cool. I think even though it is an individual honor, I think it says a lot about our offense. Our offensive line has been great protecting, guys down the field are making plays, catching the football and running, and even out of the backfield, running backs, too. I think it’s a team award.”
But perhaps the biggest thing that has led to Keenum having a breakout season is his seemingly newly-developed, and uncanny short-area elusiveness when it comes to avoiding sacks when under pressure. Keenum is far and away the leading matador in the NFL in that respect. It’s not like Russell Wilson or Fran Tarkenton wheeling and bootlegging it out of the pocket, or even Aaron Rodgers kinda side-stepping and rolling outside- it is a unique combination of pocket presence, short-area quickness, balance and agility while in the pocket- all while being able to keep his eyes down the field.
This piece in Inside the Pylon does an excellent job of illustrating how well Keenum has done in avoiding sacks, while also keeping his eyes downfield and making plays. It’s these kind of plays that has led Keenum to the highest QBR among active quarterbacks in the league this year.
Pressure to sack %:— Ryan Smith (@PFF_Smith) November 23, 2017
Case Keenum: 4.8%
NFL average: 17.6% pic.twitter.com/2kcNhN2t1n
You can see in this play that while the Vikings offensive linemen have given up pressures from multiple defenders, they remained engaged with the defender, allowing Keenum to judge the leverage and movement his linemen have, and move accordingly to avoid a sack. That makes it easier for Keenum than if there were simply free rushers, as having to juke multiple guys to avoid a sack is a much more difficult task. But Keenum has also been successful at times avoiding a sack by at least one free rusher- see the Inside the Pylon clips- and still made a big play as well.
Seriously, anyone acting like Case Keenum this year is no different to years past, it's just the situation...— Sam Monson (@PFF_Sam) December 3, 2017
Does not make this play in years past. pic.twitter.com/1XdBsHMmcf
Here again in this clip above against Atlanta, Keenum is able to move outside the pocket this time, and with his peripheral vision move just enough away from defenders- while still focusing his eyes downfield and making a key completion to Diggs for a first down and extending a critical drive in the third quarter.
Keenum’s Improvement Is Real
Perhaps more than statistics and trendlines, these clips- along with many others- show how Keenum has upped his game as a play-maker. He may not have the cannon to fire a rope 50 yards down the field to a receiver at the back of the endzone for a touchdown, as Favre did memorably in 2009 against the 49ers, but he does have a certain saavy in avoiding sacks, extending plays, and still delivering an accurate throw- all while not making many mistakes with the football. And he’s had some of his best games against the toughest opponents- and on the road. All that suggests not a quarterback playing over his head, but a quarterback coming into his own.