Photo Credit: Keith Allison via Flickr Creative Commons
This is Part I of a three-part series.
Turn the clock back one year ago.
Imagine Browns fans were told that the team would draft a rookie quarterback who would start eight games and finish with a QB rating of 92.3, the third-highest mark by any Browns quarterback since the AFL-NFL merger in 1970 (min 150 attempts).
There would be euphoria. There might even be a parade.
Now come back to the present.
There is no euphoria. And while we almost had a parade, it certainly wasn’t to celebrate the season’s quarterback play.
Instead, we are engaged in our annual offseason tradition of sifting through draft prospects and potential trade targets at the quarterback position. The search for our quarterback-of-the-future continues.
And that Browns rookie who nearly posted a franchise-best QB rating since the merger?
Cody Kessler feels like a forgotten man.
Look, I get it. I’m certainly not a scout—something to always keep in mind when reading this blog—but Kessler’s deficiencies are rather obvious. The pop-gun arm. His short (6’1”) stature. Holding onto the ball too long. And, unfortunately, the concussions.
So what should we make of Cody Kessler at this point in his career? The sample size is still quite small—eight games started—but a comparison of Kessler’s 2016 season to other rookie quarterbacks in modern NFL history (since 1970) may provide some much-needed perspective about his future and that of the Cleveland Browns.
Since advanced metrics in the NFL are a somewhat new phenomenon, a historical analysis of rookie quarterbacks is dependent on the “basic” statistics accumulated in the modern era. For Part I of this series, this study will focus on five rate metrics available since 1970:
- Completion percentage (comp / att)
- Yards per completion (yards / comp)
- Touchdown percentage (td / att)
- Interception percentage (int / att)
- Sack percentage [sack / (sack+att)]
While the inclusion of four of these metrics should be obvious, note that yards per completion is used in order to approximate how much a quarterback is driving the ball down the field. In contrast, a common substitute—yards per attempt—is misleading since a quarterback can achieve the same level by either completing a high percentage of short passes or a lower percentage of deep and intermediate passes. Especially since arm strength is a lingering question with Kessler, yards per completion is a more relevant measure.
Unfortunately, a simple comparison of rookie quarterbacks’ raw numbers is misleading. For example, consider a comparison of Kessler’s rookie campaign to Bernie Kosar’s rookie season in 1985 as presented in Table 1 (below). On the surface, Kessler appears to have significantly outperformed Kosar, featuring a substantially higher completion percentage and lower interception rate offset by only minor downgrades in yards per completion, touchdown percentage and sack percentage.
But NFL passing offenses have changed drastically over the past 30 years, due to both rule changes and strategic innovations. As a result, Kosar’s rookie-season statistics look much more palatable when compared to the league average of QBs with 150+ attempts in 1985. Relative to their relative league averages, some of Kessler’s advantages seem to dissipate.
To put Kessler, Kosar and all NFL rookie on an equivalent playing field, this study calculates the z-score of each quarterback’s performance in these five metrics relative to the average performance of all quarterbacks with 150+ pass attempts for each season between 1970 and 2016. For example, Kosar’s z-score for completion percentage in 1985 is calculated by taking the difference between his rate (50.0 percent) and the league average of qualifying QBs in 1985 (54.84) and dividing the total by that season’s standard deviation (3.92). The resulting value, Z=-1.24, implies that Kosar was more than one standard deviation below the league average.
A revised comparison of Kessler and Kosar is presented in Table 2 (below). The results indicate that, since 1970, Kessler posted elite rookie values in completion percentage and interception rate relative to league averages. His interception rate was a full standard deviation better than the league average in 2016, with the fourth-best z-score in that category among rookies since 1970. Kosar’s interception score was comparable despite the differences in raw percentages; after accounting for a higher league-wide interception rate in 1985, Kosar’s z-score nearly matches Kessler, with the former finishing sixth among rookie QBs. While Kessler shines in two categories, he took sacks at a rate that was nearly two standard deviations worse than the league average; Kosar was slightly better than the league norm in 1985 at avoiding sacks as a rookie.
Kessler’s scores on completion percentage and interception rate put him in some rare company among NFL rookie QBs. Only four other QBs placed in the top ten of rookie rankings in both categories: Dan Marino (1983), Robert Griffin III (2012), Dak Prescott (2016) and Charlie Batch (1998). However, the similarities between Kessler and Marino end there, as the latter posted the second-best touchdown percentage (Z=1.62) and sack rate (Z=1.76) among rookies since 1970.
In other words, to evaluate Kessler’s rookie season from a historical perspective, one must incorporate all elements—including his negatives. To those ends, this study proposes using the simple sum of z-scores across all five categories. As demonstrated in Table 2, Kessler’s summed z-score equals -1.76, which establishes his 2016 season as the 33rd-best rookie campaign among the 96 qualifying rookie QBs since 1970. In contrast, Kosar’s score of -0.86 ranks 23rd. Context on these rankings will be offered in the next section.
(Before continuing, a few methodological footnotes. First, sack and interception rates are reverse scored to reflect a “good” score to be positive in value. Second, it is recognized that a simple sum of z-scores provides equal weight to all five categories, however the calculation of more appropriate weights is beyond the scope of this post. Finally, the identification of rookie quarterbacks is derived by comparing their draft year to the player’s season in question. This removes undrafted QBs from the analysis, as using the season of their first regular-season pass attempt would include some players—such as Kurt Warner—who were no longer rookies in any traditional sense.)
Using the summed z-score, Table 3 (below) provided a ranked list of the top 10 rookie seasons by quarterbacks with 150+ pass attempts in since 1970. Despite Dak Prescott’s stellar rookie campaign in 2016, Marino’s 1983 season still represents the most statistically dominant rookie year since the AFL-NFL merger. In addition to featuring some of the most accomplished quarterbacks since 1970, Table 3 should provide some perspective on the career trajectory of Robert Griffin III. The Opening Day starter for the Browns in 2016, Griffin posted the second-highest passing score among qualified rookies while with Washington in 2012. Griffin’s rookie-year performance is even more impressive given that the z-score metric ignores Griffin’s 815 yards and seven touchdowns as a rusher. Just outside the top 10 includes oft-cited Kessler comparable, Andy Dalton (12nd), as well as Peyton Manning (13th).
Table 4 (below) represents the worst 10 rookie seasons by quarterbacks since the NFL-AFL merger. The names on the list should provide alarm when evaluating Jared Goff’s rookie season, as only one name on the list—Alex Smith—went on to establish himself as a long-time starting QB in the NFL. That said, other successful quarterbacks also floundered as rookies and are near the bottom of this list, including Donovan McNabb (85th), Terry Bradshaw (84th) and Eli Manning (82nd). Nevertheless, the bottom third of this list is largely composed of quarterbacks with marginal career records.
Turning the focus back onto Cody Kessler, Table 5 (below) provides the z-scores and ranks for rookie quarterbacks who appeared in the NFL in 2016. While Prescott and Goff were already identified in the two previous tables, the results demonstrate that Kessler’s season was more impressive—from a statistical perspective, anyway—than Carson Wentz and also surpassed all rookies who failed to reach the 150-attempt threshold. Table 5 also reflects Kevin Hogan’s struggles as a passer for the Browns as a rookie, with a z-score of -14.64, far worse than his rookie counterparts (albeit in a small sample of attempts).
To put Kessler’s performance in the context of the Browns franchise, Table 6 (below) gives the summed z-scores of the seven rookie QBs who have had 150+ pass attempts while in a Cleveland uniform since 1970. While the Browns do not have a rookie in the top 20, Kosar (23rd), Brandon Weeden (31st) and Kessler (33rd) rank above the median among all NFL rookie quarterbacks. In contrast, Tim Couch (57th), Charlie Frye (68th), Colt McCoy (70th) and Eric Zeier (92nd) fall in the bottom half of the rankings.
Table 7 (below) expands the analysis to examine the five quarterbacks ranked above and below Kessler in terms of summed z-score. While none of the quarterbacks on this list would qualify as future hall-of-famers, Table 7 does include two QBs who started in a Super Bowl (David Woodley, Chris Chandler), a quarterback who just signed a massive contract extension (Ryan Tannehill) and another who looks primed to ascend to a starting role as a free agent this offseason (Mike Glennon). While it may be a bit too early to summarize the careers of Nick Foles, Zach Mettenberger and Weeden, Tom Owen’s rookie season in 1974 with the 49ers was the high-point in an eight-year career as a seldom-used backup. In sum, rookie quarterbacks with scores comparable to Kessler are a mixed bag, but many had solid, if unspectacular careers. (Note: Part II of this series will look more carefully at a list of player comparables.)
Weeden’s proximity to Kessler in the rankings was initially surprising given the former’s brutal rookie campaign for the Browns in 2012. As outlined in Table 8 (below), Kessler offered substantially better accuracy as a passer, holding mammoth advantages versus Weeden in the z-score for completion percentage and interception rate. In fact, Weeden was below league average in every category that involved the outcome of his passes. However, he did demonstrate one skill as a rookie—avoiding sacks—that offers a positive counterbalance. Even if Weeden’s method for sack avoidance was staggeringly awful, the substantial difference in sack rates between the two quarterbacks catapulted Weeden higher in the rankings.
Looking at sack rates among rookie quarterbacks, a first glance at the data suggests that it may have some predictive power of a QB’s future success; Part III of this series will investigate this more carefully. But whether it should be on equal footing with the other variables—especially in light of how it overshadowed Weeden’s poor numbers throwing the football—is at least an open question at this point. Simply removing it from the analysis and using a four-category sum of z-scores, Cody Kessler’s rookie season immediately vaults into near-elite status, as demonstrated in Table 9 (below).
Removing sack rates from the analysis, Kessler’s 2016 season ranks as the 13th-best rookie campaign among quarterbacks in the NFL since 1970. Further, only 15 of 96 rookie QBs generated a four-category z-score above zero. While not unanimous, most of the quarterbacks enjoyed stellar careers or appear destined for careers of equivalent success. Finally, note that the removal of sack percentage drops Weeden 20 spots in the ranking, from 31st to 51st.
Conclusions and Caveats
The goal of Part I of this series was to put Cody Kessler’s 2016 season in a historical context. Using z-scores to set rookie quarterbacks on a level playing field over time, the results of this analysis suggest that Kessler’s high completion percentage and low interception rate places him among the elite rookies in modern NFL history. In contrast, his below-average touchdown percentage and yards per completion rate possibly reflect the obvious concern among most Browns fans: arm strength. Nevertheless, as demonstrated in Table 9, Kessler’s overall performance on footballs that leave his hand rank him among the top 15 qualified rookie QBs since the NFL-AFL merger, a list that features numerous legendary signal-callers.
But, as discussed earlier in the study, Kessler’s penchant for taking sacks—regardless of who is at fault—tempers this enthusiasm. Kessler’s inability to avoid sacks ranked him 83rd of 96 rookie QBs when measured against league averages. While Part III of this study will investigate the predictive power of rookie QB statistics, a first glance at the data suggests that this may be a long-term worry. Overall, his sack avoidance issues temper a historical evaluation of Kessler’s rookie year; when it is added to the simple z-score calculations, Kessler’s rookie campaign drops from 13th to 33rd.
Demonstrating Kessler’s effectiveness throwing the ball isn’t new. But it is hoped that Part I of this study has offered some historical perspective on how we should evaluate the third-round quarterback moving forward. In Part II, I will develop a list of historically comparable quarterbacks to Kessler based on their rookie-season characteristics. Finally, in Part III, I will explore whether rookie-season statistics have any predictive power in determining long-run success in the National Football League.
In terms of caveats, the standard #notascout hashtag should apply here (and with every analysis on this blog). Further, and perhaps just as importantly, it is recognized that a quarterback’s performance is somewhat a function of team capabilities. It is possible that Kessler’s abhorrent sack avoidance rates would have substantially improved on a team with a better-performing center and healthy guards. Similarly, it is possible that Dak Prescott’s magical 2016 campaign would have been less stellar without a terrific offensive line and outstanding weapons at his disposal. While advanced statistical work may be able to tease out team effects of a quarterback’s success and failure since 1970, this analysis rests outside the scope of this study. Finally, while more advanced metrics of QB play are available for more recent seasons, this study relies on available “basic” statistics—and all their respective shortcomings—that allow for cross-era comparisons.