Changing The Way We Evaluate NFL QB Prospects

There is no more exciting time in the NFL, in my opinion, than the NFL Draft. Studies have shown that successful rosters are built through the draft, and it makes sense; the top players tend to be kept by their respective teams, and thus don’t often reach free agency. The surplus value of getting a cost-controlled player for four years, in addition to a fifth-year option, is huge, and whereas free agency fills short-term roles, the draft is where a team builds the foundation of their roster.

This is specifically true when it comes to quarterbacks. Outside of a 43-year-old Tom Brady, how many productive quarterbacks generally reach free agency? Additionally, even if they do, the surplus value of signing one isn’t the same as drafting one. The Vikings, for instance, will pay Kirk Cousins over $30 million this season, which forces them to be very strict when it comes to spending on other areas of their roster. Teams like the Chargers, Cardinals, Bills, Browns, and Ravens have had the luxury to support their quarterbacks with a strong roster, since they’re on rookie contracts, while teams like the Eagles and Rams (Carson Wentz and Jared Goff) have been able to elevate their quarterbacks into Super Bowl-worthy production with the talent they have been able to put around them. There is something to be said about the overall flexibility of having a quarterback on a rookie contract, which is why studies have shown that more successful teams generally have these coveted commodities.

How coveted are successful young, cheap quarterbacks? The 49ers just traded two-first round picks and a third-round pick to move up nine spots to the third overall pick to get the third-picked quarterback in this draft! Many had them projected as a playoff team as is, but the pay-off and benefits of being able to fortify their roster around a rookie quarterback made the investment worth it for them.

With there being less than a month before the draft and San Francisco’s trade up generating a lot of discussion about quarterback evaluation, now would be a great time to not only look at this quarterback class, but the process of quarterback evaluation in general. Today, we’ll be looking at the stability of metrics from college to pro, as well as what has correlated to success at the pro level, looking to discover a pattern to help us identify who the most successful quarterbacks will be.

Unlike other positions, quarterback evaluation is essentially 100% about skills and production. Whereas the evaluation other positions, such as receiver or cornerback, is heavily affected by athletic testing measures, this isn’t the case at the quarterback position, and for obvious reason; the days of hand size impact evaluation are in the past. Thus, production from college to pro should be more stable than other, less “skilled” positions. Is this actually the case. Let’s take a closer look, splitting production into three main categories: style of play, overall output, and time of production.

Note: All metrics courtesy of Pro Football Focus


How a quarterback produces at the college level is brought about a lot during the evaluation process, as it can theoretically affect how they adjust to various NFL offensive schemes. One would think this starts from the ability to make quick decisions, but by simply looking at time to throw metrics, this isn’t actually the case:

That’s right; having a low time to throw at the college level has actually had a negative correlation with NFL success. Now, by no means would I want to suggest that making quick decisions is a BAD trait, but there is a lot to suggest that time to throw is mainly a product of the offensive system a quarterback is in, rather than their overall abilities; the stability of time to throw from college to pro is quite low. Deshaun Watson, for instance, went from having one of the lowest time to throws in college to one of the longest. Was this him changing? Probably not. Rather, it was a product of the offensive environment he was in changing, and, generally, a quicker time to throw would suggest more “schemed” rhythm passes that make a quarterback’s life easier. Plus, there’s always the opportunity cost with getting rid of the ball quicker, based on NFL quarterback data from the past three years:

How do you produce more air yards, thus being a less replaceable quarterback in general and being able to elevate the talent/coaching staff around you? Well, you’d have to throw passes further down the football field, which involves holding onto the ball longer. Thus, the overall stability of average depth of target and its correlation to future NFL success should not be too surprising:

With it being a style statistic, you wouldn’t expect a higher average depth of target to have an extreme relationship with quarterback success; your intentions don’t make you productive. However, it is clear that having a higher aDOT is much more about the quarterback than their time to throw. In other words, quarterbacks who throw to the more valuable areas of the football field will likely continue to do so at the NFL level, which makes them more likely to succeed.

What about the ability to not covert pressure into sacks? Sacks can obviously be a killer for an offense and have an extremely negative effect on their expected points added (EPA), so being able to identify this skill would be critical. Unfortunately, that isn’t the case:

It’s easy to be frustrated with a quarterback’s propensity to take sacks, but it is very difficult to identify which ones are able to help their offensive lines by not converting pressures to sacks. Thus, this isn’t a strength or weakness that should factor much, if at all, in the quarterback evaluation process.

Overall Production

So, quarterbacks who throw down the football field appear the be the ones to covet, but none of this matters if they haven’t been able to establish a track record of production:

Thus, discovering the key statistical indicators are future success are key.

Reading blitz packages is often something seen in quarterbacks described as being “pro ready”, but does it actually provide much predictive significance? The numbers wouldn’t seem to suggest that:

Given that the time that a quarterback is blitzed is such a small sample, it makes sense that we cannot identify who the top quarterbacks against the blitz are in a small sample. Perhaps, additionally, this is also something that can be more easily developed over time, though I’d lean on it being more of a product of the statistic not having enough time to stabilize.

Speaking of small sample size performance, be careful of taking too much from a quarterback’s skills under pressure. Performance in a clean pocket has three times the predictive power of future NFL success, based on the coefficient of determination, than performance under pressure, while there was a positive relationship between production with no play action as opposed to production without play action.

Furthermore, we know that quarterbacks who throw the ball down the field tend to have a greater margin for error, so it isn’t a surprise that the performance in the short passing game isn’t very predictive of future NFL success:

By contrast, performance on 10-19 yard throws is around three times more predictive, while performance on 20+ yards is around five times more predictive. Once again, NFL teams should be looking for less replaceable quarterbacks capable of producing a lot of air yards, and that doesn’t come through short passes.

Speaking of making big-time plays, it’s also clear that quarterbacks who are capable of producing them should be coveted more than those who limit negatives, based on the data:

The old saying goes that coaches can “coach out the bad” in quarterbacks, and although that isn’t exactly correct, you’d certainly rather take your chances getting rid of the negatives than the high-end production of a quarterback, which is much more difficult to improve on. Limiting turnovers can sometimes be as simple as being ultra-conservative and taking check downs at low depths of target, and since these are the types of quarterbacks than NFL teams continue to move away from, limiting turnovers should continue to be valued less than producing big-time plays.

Another way to produce big-time plays is by being a factor in the running game. As we’ve looked at in the past, having a mobile quarterback often leads to a more efficient rushing attack, as the more options defenses have to take into account, the more unlikely they are to have an idea as to what you’re running. Additionally, mobility is becoming seen as a necessity at the quarterback position, and that certainly is backed up by the success of recent quarterback prospects:

Rusher rating is essentially an adjusted version of PFF’s rushing grade, but using their rushing yards per game as the main variable. Obviously, Lamar Jackson’s college production is an outlier, but the relationship actually only becomes stronger if you remove him! Besides Jackson, Dak Prescott, Deshaun Watson, and Kyler Murray stand out as quarterbacks who produced a lot as a runner in college, while Josh Rosen and Dwayne Haskins stand out on the opposite end of the spectrum. When we speak to having a “higher floor” in terms of projected outcomes, we often think of overall polish, but really, it’s by being an effective rusher; your margin for error is much higher because even in a game in which you struggle to pass the ball, your effectiveness as a runner at least stresses out the opposing defenses and allows for a certain baseline of production. This is something to keep a very close eye on when it comes to projecting this current quarterback class.

Finally, it shouldn’t shock you in the slightest to find out that quarterbacks who are more accurate tend to be more successful at the pro level. However, I do want to caution that accuracy numbers should be adjusted with what depths of target they are throwing to:

We’ll get to a current quarterback draft prospect that has completely bucked this trend, but it is simply not enough to deem a quarterback accurate. Rather, their aDOT matters tremendously; being accurate throwing 10+ yards on average down the field is much more impressive than being accurate in the short passing game.

Time of Production

Recency bias can be quite the drug, as quarterbacks are generally evaluated based on what they last did. Struggle slightly in your final year? It doesn’t matter what you did in the past, while those coming off a breakout year jump up draft boards significantly. However, this probably shouldn’t be the case:

In the end, taking the whole body of work for a quarterback reigns as the supreme method of evaluation, as opposed to simply looking at their final season. The “improvement” narrative is one that plagued Justin Herbert last offseason, for instance, while Josh Allen was also much worse in his final college season. On the other end of the spectrum, quarterbacks like Drew Lock and Paxton Lynch didn’t exactly “progress” in a way that helped them be successful at the NFL level. It’s often wonder how to weight what you just saw from what a quarterback has done in the past, but, as all projections should be, it’s about looking at the entire body of work- statistics take a great deal of time to stabilize.

What about quarterbacks with one year of production? The sample size of those prototypes who also were drafted high enough to play at the NFL level is limited, but it is worth noting that of the quarterbacks in this study, the three who quality for this description – Dwayne Haskins, Mitch Trubisky, Kyler Murray – all rank in the top-ten in terms of the downgrade of their college production score to their PFF grade. Obviously, Murray is a great prospect, but it would have been silly to expect his elite 94.9 PFF grade to translate to the next level smoothly, and he still would have rated tremendously with a small sample downtick. For Haskins and Trubisky, it’s clear that if you play for one season, perhaps the bar should be much higher.


Using this information, I have been able to construct a model to project this year’s quarterback class. With a moderately strong correlation (r=.45) that is double of the simple correlation of raw PFF grade from college to pro, it certainly appears to have some predictive power, especially when considering the overall variance of prospects:

Top Five:

  1. Baker Mayfield
  2. Kyler Murray
  3. Marcus Mariota
  4. Deshaun Watson
  5. Joe Burrow

HM: Lamar Jackson

Bottom Five:

  1. Trevor Semien
  2. Jeff Driskel
  3. Daniel Jones
  4. Drew Lock
  5. CJ Beathard

HM: Brandon Allen

Now, there will be misses, as there is with all quarterback evaluation processes (it was certainly too low with Jones), but when you add in the adjustment of scheme and supporting cast, the predictive power only grows. Thus, my model’s projections for this year’s quarterback class are certainly eye popping:

  1. Justin Fields
  2. Trevor Lawrence
  3. Zach Wilson
  4. Trey Lance
  5. Mac Jones

To be fair, Fields, Lawrence, and Wilson are all rated very similarly to each other, but, in the end, it favors Fields. Why?

  • Fields is the most accurate of all of them
  • This is despite having the highest average depth of target, which should make him less accurate
  • Fields is the best overall rusher of the three
  • Fields has the best big-time throw/turnover-worthy play split
  • Fields has been the best in the “money areas” of the fields

Most importantly, though, most of the knocks against Fields (high time to throw, taking sacks) are statistically insignificant data points compared to what he does well. He is a tremendously well-rounded quarterback prospect, and if the 49ers are able to secure him with the third overall pick, they should be ecstatic. That isn’t a slight on Lawrence, who has the strongest overall track record, or Wilson, the best deep-ball thrower, but if I had to bet on one player being an elite quarterback in this class, it’d be Fields, who ranks as my models’ top QB prospect ever, though Lawrence (3rd) and Wilson (6th) rank as players certainly worth being #1 overall picks in practically any year.

With one year of data at a FCS school (North Dakota), Lance is an impossible projection, especially since he isn’t very accurate and didn’t produce a lot of big-time throws. However, he still graded out well from PFF, has a very high average depth of target, and is the best overall rusher in this class- his floor may be higher than you think because of that ability. I’d certainly want to gamble on him than Jones, who isn’t mobile and has the lowest average depth of target of these five, though all five rate as tremendous quarterback prospects worth selecting in the top ten.

There are certainly flaws in any way of quarterback evaluation, but my hope is to find a “less flawed” and objective method. Finding out which data points are most predictive is important, and when we do that, it’s clear that Justin Fields has become a very underrated and over criticized quarterback prospect. It is ludicrous to suggest that he’d be not one of the top three quarterbacks taken, and whichever team selects him will be quite pleased. Should it be the 49ers, I’d certainly look into their Super Bowl odds for not only this year, but the years after as well. In the end, situation and development ultimately determine the outlook of these prospects, yet in terms of the identification process, there are some inefficiencies that may be leading to certain players being overlooked. It will be quite interesting to see how this quarterback class shapes out, and how the evaluation of this position continues to evolve over time!

One Comment Add yours

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s