Murray or Rosen: What Can We Quantify About Their Chances for Success

In the 2018 draft, Cardinals’ GM Steve Keim traded away the team’s 3rd and 5th round picks to move from No. 15 to No. 10 to select UCLA quarterback Josh Rosen. Presumably, this wasn’t a rash decision and came with the imprimatur of coaching and, more importantly, ownership.

Less than a year later, the NFL punditry is convinced Rosen is on his way out the door to make room for the Oklahoma Heisman Trophy-winner Kyler Murray. Such a quick about-face may be unprecedented, but this wouldn’t be a move without a rationale. Murray had one of the most efficient passing seasons in FBS history. A season even better than the prior year’s performance of Murray’s former teammate, fellow Heisman winner, and NFL rookie sensation Baker Mayfield.

While there are many in agreement that Murray has the profile of a future NFL star, opinions on Rosen range across the board. Some of the most connected NFL media have heard he’s only worth a third-round pick. While others have focused on the context surrounding Rosen: poor offensive line and poor coaching.

I’m going to dive strictly into Rosen from a statistical angle, one that not only incorporates his year-end numbers but adds additional layers to account for expectations based on draft position and sample size. I’m also going to compare Rosen to past rookies and see if he resembles the stars or busts.

The technique is Bayesian updating, which I’ve used in the past to give more heft to the collegiate statistical profiles of three-year starters Deshaun Watson and Patrick Mahomes versus one-year wonder Mitchell Trubisky, make the case that Brock Osweiler is who we think he is, and find the closest historical comps for the early-career quarterback performance of recently drafted quarterbacks.

Here is a formal definition for Bayesian inference:

Bayesian inference derives the posterior probability as a consequence of two antecedents: a prior probability and a “likelihood function” derived from a statistical model for the observed data.


When applying this to an estimate of true quarterback passing efficiency, the prior probability is the expectation based on draft position, the observed data are the NFL passing results, and the posterior probability is our updated expectation.

We don’t need a huge sample of passes for this analysis. We can update our estimate for a quarterback’s true, or go-forward passing efficiency after every throw. The more throws the quarterback has made, the more confident we can be in our updated estimate.

A look at the 2018 class

What you see above is the best estimates for the 2018 batch of rookies’ passing efficiency as measured by adjusted net yards per attempt (ANY/A).[efn_note]Adjusted net yards per attempt = (Net Yards + 20*Passing TDs – 45*Interceptions)/(Pass Attempts + Sacks)[/efn_note]

The starting point for each quarterback is his prior, i.e. our best estimate for future pass efficiency based on the historical average for quarterbacks at his draft position. Then we updated that prior based on the evidence, or the actual results each quarterback gives us from his play.

What should first jump out at you is how wide the spread is initially based only on draft position, with the No. 1 overall pick (Baker Mayfield) on average becoming an above-average QB, then a big gap before the No. 3 pick (Sam Darnold) starts slightly below league average, followed by another significant gap to No. 7 (Josh Allen) and No. 10 (Josh Rosen).

We don’t think of being No. 1 overall being that much different from somewhere else in the top-10 selections. But history shows it really is. We should have significantly less confidence in Rosen’s ability to succeed than Mayfield before they throw a single NFL pass.

If Murray is indeed the No. 1 overall talent in this year’s draft, he’ll have a big advantage in true passing efficiency versus Rosen. Our prior for Murray will be roughly 6.6 ANY/A, which is more than 0.6 ANY/A better than our updated estimate for Rosen, which is significant.

The best way to sum up the 2018 rookie results is:

  • Mayfield: Exceeded high expectations
  • Darnold: In line with the 3rd pick
  • Allen: Struggled early, but flattened out mid-season
  • Rosen: After brief success, steady descent
  • Jackson: Played well enough in limited action to improve his projection

Goff and successful rookies who struggled

When you look back at other first-round quarterbacks who struggled as rookies but ended up successful, you start to understand the importance of priors.

Jared Goff is the quarterback most often compared to Rosen and pointed to a reason Rosen can still be successful. But there are two material differences between the two: draft position (1 vs. 10) and the fact that Goff had such a limited sample as a rookie (205 pass attempts to Rosen’s 393). Goff was bad as a rookie, but he didn’t have enough pass attempts to make it as meaningful.

Matthew Stafford and Carson Wentz also weren’t great as rookies, but they were the first and second picks in their respective drafts. We tend to downplay draft position after a player has completed an NFL season. But it isn’t a coincidence that quarterbacks who turn things around after struggling were previously identified as the top players in their draft classes.

A rookie who struggled but didn’t turn it around

The graph above is full of hyped quarterbacks who never lived up to the hype. And the scary thing for Rosen’s prospects is that his end-of-year true efficiency projection is lower than all of them: Blake BortlesMark SanchezJake LockerBlaine GabbertChristian Ponder and E.J. Manuel.

Looking at where Rosen stands in contrast to these failed quarterbacks, and how far he is away from the successes above, should function as a bucket of cold water on anyone feeling good about his prospects. It’s not that there haven’t been quarterbacks who weren’t top selections, struggles as rookies and then became successes. It’s that we haven’t seen one in a long time.

If the Cardinals, or another team out there looking to trade for Rosen, want to play those odds, they can try. But you shouldn’t expect much. Most teams won’t get the chance to draft a quarterback worthy of the No. 1 or No. 2 pick, so taking a shot on Rosen can be a risk-reward play. But if the Cardinals see Murray as that guy, there really is no choice but to select him and not play the waiting-and-hoping game that Rosen can break a significant trend.

4 thoughts on “Murray or Rosen: What Can We Quantify About Their Chances for Success

  1. It makes sense that rookies would be below the league average. Imagine if they were above average. It would imply that the more experienced QBs were below average.
    It would,be interesting to see this analysis applied to other draft classes. Has there been a class when early-rounders have been in the upper quartile? What were the numbers on Ben, Matt Ryan, Eli, etc., in their rookie seasons?

  2. I probably should have been more explicit about what “estimated true efficiency” really measures. It’s the go-forward estimate for pass efficiency. It’s saying your expectation for a quarterback career if not drafted 1st or 2nd on day one should be below average NFL quarterback. Then the career estimate changes with evidence.
    I also make an adjustment for rookies that gives them more benefit than the average passer. This offsets the fact that rookies are normally worse than league average.
    Eli stunk as a rookie, Matt Ryan and Big Ben were excellent.

  3. How do we correct for the fact that draft position may reflect team need, or an unwillingness to trade? E.g., how does this change if we assume Darnold is the 2nd overall pick rather than the first (the market for draft picks is not perfectly efficient in transferring high picks to teams looking for QB upgrades).

  4. That’s a good point and something that I’ve thought about. Typically No. 2 picks are very close in efficiency to No. 1, because they’re usually drafted immediately behind another QB and would have been No. 1 in another class. Look at where Wentz is on the chart and that’s where Darnold would start as a No. 2 pick.

Leave a Reply

Your email address will not be published. Required fields are marked *