Predicting which rookie QBs will not be starting by 2020

The analytics behind drafting and retaining NFL Quarterbacks

Image: Erik Drost

Imagine you’re a top FBS college quarterback, and after years of impossibly-hard work, you’ve finally landed your dream job — you were just drafted by an NFL franchise. The world’s richest sports league has deemed you one of this year’s top rookie quarterback prospects nationwide. You are very likely to be a multi-millionaire. You will be more famous than you were in college; many more people will recognize you now. It’s a great start. I bet you’re ecstatic. Unfortunately, unless you’re one of the elite few, the sad truth is that your joy will be short-lived. Sincerely, enjoy it while you can.

About to start? Don’t blink.

Most NFL players have short careers; this is widely understood. What is perhaps less known is how much shorter a player’s starting career can be. The average QB career lasts 4.4 years in the NFL, but the average QB is a starter for only 1.6 years during his career. In fact, 71% of all quarterbacks drafted from 1966 to 2000 have not started a single full year, and a full 47% of all quarterbacks drafted during this time have not started even a single game as of 2018. The numbers are especially astounding when you consider the millions of dollars that even later-round players can earn over the length of their rookie contracts.

I chose the interval from 1966–2000 because 1966 is the start of the Super Bowl era, and 2000 is when the oldest active QB (Tom Brady) was drafted. This interval provides a view into the full career of quarterbacks. The yellow bars are self-explanatory. The orange line depicts the cumulative percentage that the yellow bars make up from left to right (e.g. 13 1st-rounders with 0 years starting experience is 21% of all 1st-rounders, while another 13 QBs (also 21%) have started for 1–3 years, making the cumulative percentage 42%.

Who will win the Kentucky Derby — Dream Maker, Dark Prince, or Baffin Bay?

So how effective is the draft process, you may ask? Well, the earlier a quarterback is drafted, the more likely he is to get (and retain) the starting job. 79% of first-round quarterbacks and 63% of second-rounders started for at least one year for their respective teams. By contrast, only 14% of quarterbacks drafted in the sixth-round or later started for at least one year over their entire career. Of course, players considered “starter material” are drafted earlier, and are generally given the chance to prove themselves earlier. But these early starters have historically done better and lasted longer than their later-round counterparts who have received equivalent opportunities, suggesting the drafting process is not as broken as some may have you believe.

* Average total years QBs were starters. This list only includes QBs that were given the opportunity to start for at least one year — thus controlling for any possible early-round-bias.

Yes, there have been many first round busts, and the G.O.A.T. himself was famously a sixth-rounder, but on the whole, teams have done the best they can with the information available to them. Scouts evaluate hundreds of players based on small samples of performance statistics, interviews, coaches’ recommendations, and isolated film clips; asking them to consistently predict things like longevity, leadership qualities, and toughness from these disparate sets of data is like asking you and me to pick the winner of the Kentucky Derby based on the names of the horses.

The NFL Combine matters, kind of…

The NFL Scouting Combine has been around in some form since 1982, and the data shows it has been a key input into the draft decision-making process for early-round quarterbacks. I looked at how QBs have done at four different Combine tests — 40-yard dash, 3-cone drill, vertical jump, and broad jump — from 2000 to 2018, and plotted the performance against their draft position.

The chart shows results for four tests for all QBs from 2000 to 2018. The results are organized by draft round.

For first and second-rounders, performance at the Combine actually matches the draft position. But that’s where the correlation ends. Passers drafted in the third round were sixth in overall Combine performance, and quarterbacks that eventually went undrafted placed fifth during the Combine.

Quarterback draft round compared to performance at Scouting Combines from 2000 to 2018. Combine rank is calculated through an average of averages — I averaged the performance of each draft round class for each event and generated corresponding ranks in each event. I then averaged the ranks for each draft round class to come up with the overall Combine rank.

Mystery solved, partly…

Clearly, when evaluating quarterbacks, athleticism is somewhat important but is not the deciding factor. Passers league-wide have gotten faster and more scramble-happy, but the lumbering greats like Brady, Roethlisberger, and Brees continue to stand in the pocket and deliver, time after time. The Combine measures athleticism, but not passing skill. So perhaps we can modify the model above to take college performance into account. I did this by calculating each quarterback’s equivalent fantasy football points per game, and their quarterback rating, and put it together with Combine performance.

College and Combine ranks are calculated with the following weight: equivalent college fantasy point = 60%, college quarterback rating = 20%, combine performance = 20%. College QB rating and fantasy points are highly correlated, obviously; by giving QB rating some weight, I am placing slightly more emphasis on the passing game.

This leads to three interesting results: undrafted players fall into a more predictable eighth place; third-round players inexplicably fall even further — into seventh place, and first and second-rounders stay absolutely put. The rankings stay relatively the same even if the weighting between college statistics and the Combine are varied. The data proves that first and second-rounders are noticeably ahead of their peers, both athletically and in passing skill. Undrafted players can perform well in the Combine, but their college stats are just not at the level needed for the NFL. As for the players in the middle, the data is unclear, but there are some likely explanations — these players have likely gotten better (or worse) over time, having better (or worse) seasons immediately before the draft, while others have had good high-profile rivalry or bowl games, or some QBs have gotten recommendations from their college coaches for being good leaders or hard workers. Injuries and off-field issues can have a material effect as well — we have seen many players’ draft stock drop due to non-football issues.

Putting it all together

How do quarterbacks actually perform once they get drafted in the league? We looked at starter longevity earlier, but now let’s look at passing and rushing performance by draft round. One way to evaluate this is to look at fantasy points per game. I used standard NFL fantasy football scoring (four points for passing touchdowns, six points for rushing touchdowns, 1 point for every 25 passing yards, 1 point for every 10 rushing yards, and -2 points for each turnover), and looked at historical statistics to calculate fantasy football points. The chart suggests how adept NFL scouting is — first rounders outperform their cohorts sizeably, and performance dwindles thereafter by draft round.

Data for all quarterbacks drafted in the Super Bowl era.

Finally, let’s take a look at how quarterbacks drafted in each decade have performed in the NFL. Once more, some clear trends emerge.

Quarterback performance, measured as equivalent fantasy football points calculated from historical statistics. The data is organized by the decade that the QB was drafted in.

Quarterback scoring has gone up as time has passed (no surprises there). More interestingly, the data shows that QB careers have lengthened as the decades have passed. QBs from the 1960s to 1980s that started 11 or more seasons actually have lower average career points than quarterbacks that play less — evidence of the drop-off in performance as they aged. But the trend is gradually reversing as time passes, and passers drafted in the 2000s who played 11 seasons or more have done better than their counterparts who have shorter careers. As the golden age of quarterbacks dawns, the trend can only get stronger.

The 2018 Draft Class Extraordinaire

So by putting our newly-found knowledge (thanks data science!) to use, we can draw some conclusions for how much longer each first-rounder of the 2018 QB draft class will be starting for. These durations are statistical predictions based on each QB’s performance so far in the NFL season. The R-Squared (coefficient of determination) between fantasy points for an NFL starter’s first 16 games and those for his overall career is a very healthy 0.72, which enables us to make some reasonable predictions here:

Pick #10 Josh Rosen (Arizona Cardinals), T minus 2 years: As someone who holds two degrees from UCLA, it hurts me to tell you that the data points to a quick exodus for The Chosen One. Rosen broke many UCLA passing records, and had a decent showing in the Combine, but he is averaging less than 10 fantasy football points per game. There is a disappointing lack of talent around him, and unless the Cardinals can add some offensive support for Rosen next year, he will unfortunately be the first casualty of the 2018 QB group.

Pick #3 Sam Darnold (New York Jets), T minus 5 years: Darnold has been dealing with a foot injury, and like Rosen, has a dearth of receiving options. As the season has progressed, he has shown moments of brilliance amid a flurry of turnovers, and unless he can clean up his act, Darnold has less than 5 years left as a starter.

Pick #7 Josh Allen (Buffalo Bills), T minus 5 years: Allen is averaging over 14 fantasy football points per game, which normally would put him on track for a longer career. But a disproportionate percentage of his points are coming from rushes, while his passing stats are mediocre, all of which brings up major questions regarding sustainability in the long-term.

Pick #1 Baker Mayfield (Cleveland Browns), T minus 12 years: I am going to jump with two feet (and a cartoon anvil anchor) on the Baker Mayfield bandwagon. He definitely has his rookie moments, but he is averaging nearly 17 points per game on the season. After drafting Johnny Manziel, Brandon Weeden, Brady Quinn, and Tim Couch, it looks like the Browns may finally have gotten one right, and barring injury, Mayfield looks set to be the franchise quarterback the team has been searching for all these years.

Pick #32 Lamar Jackson (Baltimore Ravens): Lamar Jackson is now officially the Ravens’ starter, but there is not enough data yet on Jackson to make a prediction, so I will steer clear of this one for now.

All data in this article is taken from one of the best football statistic websites of all time: www.pro-football-reference.com. The data is current to 2018 Week 13. All charts and analysis presented here are the sole opinion of the author of this article.