This morning Browns Head Coach Hue Jackson named Brock Osweiler the starting quarterback for the team’s first preseason game against the New Orleans Saints on Thursday. The news was at least somewhat surprising, as the trade by which the Browns acquired Osweiler was largely a salary dump by his former team, where the Browns acquired a second round pick as compensation for taking Osweiler off of the Houston Texans’ hands.
While Osweiler is being paid starter money, the consensus opinion going into training camp was that the Browns would start the season with one of the two quarterbacks getting reps with the first-team offense, second-year signal-caller Cody Kessler or second-round rookie DeShone Kizer.
Osweiler was bad for the Texans last year, really bad. Osweiler’s 5.8 yards per attempt ranked dead last among 30 NFL passers with at least 300 attempts last season, with a material gap between Osweiler and 29th ranked Carson Wentz (6.2 yards per attempt).
Multiple observers have noted that Osweiler has been the most impressive quarterback during OTAs and training camp so far, and he performed better than Kessler or Kizer at the Browns’ intra-team scrimmage last Friday night. It makes sense that Jackson would want to give the first starting shot to the quarterback who looks better at this point.
But is it the right move for coaches to base their decisions on a handful of weeks of OTAs and training camp? Those of us in the “analytics” business can probably boil most of our insights into a few broad themes, the dangers of small samples being near the top. But before why write off Osweiler’s relatively strong training camp, we need to look at his historical performance and see what it tells us about how he’s likely to perform moving forward.
What should we expect from Osweiler moving forward
During draft season, I showed how you can use Bayesian statistics to quantify the differing levels of confidence we should have in Deshaun Watson‘s and Mitchell Trubisky’s nearly identical pass efficiency. That analysis illustrated why we can be much more confident in Watson’s based on his larger sample of pass attempts (1207 versus 572), and that Trubisky was more of an unknown following only one starting season. We can use a similar analysis to find a realistic range of outcomes for Osweiler’s pass efficiency moving forward based on his 881 career attempts.
Bayesian statistics is a way of using smaller samples to estimate the true values, in this case the Osweiler’s true passing efficiency. If before Osweiler throws a pass we assume he’s a league-average quarterback, then we can use his actual performance over time to update or assumption. The greater the sample of throws, the more we know about Osweiler and the more narrow our expected range of outcomes.
I had NFL play-by-play data to use here where I didn’t at the CFB level for the Watson-Trubisky analysis, so we can not only see the end result, but also track how our expectations for Osweiler’s future pass efficiency have changed over his career.
Osweiler hadn’t always looked like the troubling passer we saw last season. In fact, our expectation for how Osweiler should perform were solidly in line with your league-average quarterback through around 370 pass attempts, or roughly a full low-volume NFL season. Having a league-average quarterback isn’t something that causes fan bases to cheer, but they’re still a rare commodity and perhaps even worth more than the $16 million (roughly 10% of the salary cap) that Osweiler will make this season.
What happened to Osweiler in the second half of his career pass attempts is the real story. It coincides with the Texans’ Week 3 matchup at the New England Patriots last season. Those who saw the game that Thursday night witnessed a catastrophically bad offense performance from the Texans. Osweiler threw for fewer than 200 yards on over 40 attempts, and the Texans left Gillette stadium with zero points on the scoreboard.
Osweiler didn’t perform much better the rest of 2016, and his median projection for pass efficiency going forward now stand at roughly 6.2 yards per attempt, which is worse than all other passers than Osweiler in 2016. Arguably the more troubling part is that Osweiler’s upside, or the threshold of the top 10% of outcomes, now has narrowed to 6.9 yards per attempt, which is only slightly better than the NFL average of 6.8.
What should the Browns do?
It’s understandable that Jackson and the rest of the Browns staff would want to get a longer look at Osweiler if they view him as the most impressive quarterback of the offseason. But there are times when decisions should be made based on a broader context. Is it possible that Osweiler will vastly improve his performance from what we saw in 2016? Of course there’s a chance. But by using analytical analysis we can estimate that the chance that Osweiler exceeds league-average passing is only 10%.
There are many factor that going into a quarterback’s passing efficiency that is out of his control: luck, offensive line play, receiver skill, the effectiveness of the running, and the defenses ability to prevent poor game script. But there isn’t anything about the 2016 Texans that jumps out as putting Osweiler at a disadvantage. What’s more likely is that the inefficient Brock Osweiler we’ve seen in NFL action is who we will see if he starts in 2017, not the impressive offseason player.
If I were making the quarterback decision for the Browns, I’d eventually give the starting nod to Kizer, as long as he doesn’t perform so poorly that it looks like malpractice to do so. Kizer is an unknown, which is a good thing in comparison to what we know about Osweiler. While Osweiler is a decent athlete, Kizer also likely brings more to the table in terms of rushing value, which can be undervalued in a setting like training camp where the focus is on how the passing game is being executed.
It’s also important that the Browns avoid a worst-case scenario: Starting Osweiler and having him and the rest of the team perform just well enough to prevent Kizer getting a serious look this season. It’s not likely that this will happen, but if I were Browns management I’d want to know as much as possible about what I have in Kizer before having to make decisions about how the team should approach the 2018 draft, where they have enough capital to make a run at a top quarterback prospect.