PFF QB decision-making analysis (and Dak Prescott, of course!)

Say that you'd like to take a stab at furthering the progress of football analytics - what to do? Here's one standard approach: identify an area of study that you feel raw stats fall short in, ponder how to go about measuring more useful results, execute the process, and then hold up the results to our consensus, intuitive expectations. If the results generally hold to those overall expectations (exceptions are fine, and in fact it's a red flag to lack them), you might be on to something.

That is what some members of PFF recently did regarding Quarterback "decision making", as outlined in this article (click here for the link). It's interesting stuff, and you should check it out; the article lays out the theory, the process, and the results. If you don't want to read the full article, the following will provide a quick summary, before then jumping into analysis of individual QB results.

Summary Of Theory, Approach, And Overall Results

The general idea was simple enough: take every route and model a predicted Expected Points Added (EPA) figure for it, based on a number of factors (route, separation, ability of closest DB, etc) at the time the QB chose to throw the ball. This could then be compared to actual EPA results for the actual throws taken. While the prediction might miss for any given single throw or decision, over a large enough set of plays one could expect the kinks to iron out. Note that all data was taken from 2019-2021 in order to involve a big enough sample size.

PFF first thought to chart the frequency that a QB chose the best available route on a throw to the QB's overall final EPA, to see whether value was generated by how often a QB found the best option out there. Unsurprisingly, the results were a useless cloud; this makes sense, given that (for example) a QB might target a high expected EPA route (a good decision) even with a teeny-tiny better EPA route available at the same time (technically a "miss" by this evaluation).

The team then charted the projected EPA for the sum of all throws for each QB versus the QB's actual end EPA. These results were much better:
As you can see, there is a steady rise in actual EPA produced as QBs made better overall decisions, and just as importantly the association is fairly tight - a correlation value of 0.72 is darn good when it comes to sports analysis of this sort! Without knowing the identities of QBs to further confirm the validity of this approach, these results lent confidence in the accuracy of the predicted EPA analysis.

By the numbers, this analysis holds promise! The next step is to look at the actual results, both to confirm that the overall results conform to general consensus as well as to then parse what those results say about various players.

Ranking By Decision-Making

The first results simply laid out the overall quality of QB "decision-making" based on the projected EPA of all a QB's throws from 2019-2021. Below is the graph laying out various passers, followed by grouping results into various tiers with some of my own analysis of each group. Without further adieu:
The Goat - Tom Brady - it's an excellent sign that this analysis not only lists Brady as the top decision-making passer, but even places him all by himself at the top of the mountain.

The Elite - Aaron Rodgers, Lamar Jackson, Matthew Stafford - being a very experienced QB who hasn't suffered a major step back in physical ability seems to help in this measure, and speaks to Rodgers and Stafford being just short of the top. The presence of Jackson makes sense too, as the running threat of his own legs and teammates have been thought to open up opportunities for Baltimore's low-volume, run-first offense.

The Very Good - Ryan Tannehill, Joe Burrow, Justin Herbert, Dak Prescott, Jimmy Garoppolo, Jalen Hurts, Patrick Mahomes - the next tier could be seen as the "very good" one, and it has an interesting mix of names. Tannehill, Prescott, and Garoppolo generally will have their reads talked up, and this seems to confirm that notion. Hurts is probably the closest thing to Jackson in terms of the run threat and reduced volumes helping a more careful throw selection; if he can keep these numbers up as the offense looks more to his arm, that would be a great sign for his future. The presence of young guns Burrow and Herbert helps explain why they're already knocking on the door of the league's elite QBs, and while Mahomes takes up the rear of this group he's no slouch himself.

Above Average - Josh Allen, Mac Jones, Baker Mayfield, Cam Newton, Taylor Heinicke, Gardner Minshew - we're starting to see names we presume to be less effective, meaning "slightly above average" decision making is no longer enough to carry a QB to higher levels of performance. Allen being here suggests that he makes up for being a bit weaker with his decisions with execution (more on that in the next section), and the trio of Jones/Heinicke/Minshew could all be read as guys who outperform their physical arm ability. The presence of Newton might surprise, but like some of the other mobile QBs here his game has long been about buying time to find higher-yield throws; at his best, Newton carried a weak set of receiving weapons in an MVP season by working to find better pass targets.

Average - Russell Wilson, Kyler Murray, Teddy Bridgewater, Trevor Lawrence - if I'm a Jaguars fan, I take it as a very good sign that the rookie Lawrence was able to hit a solid average in decision-making despite his inexperience and the tire-fire around him. Murray is developing but has work to do (and if he's boosted by mobility, this might help explain his good but still unsatisfying perceived performance), Bridgewater should be expected to do well enough in this area, and Wilson might be making up for a terrible OL, limiting his ability to survey the field unmolested.

Below Average - Derek Carr, Kirk Cousins, Philip Rivers, Jared Goff, Matt Ryan, Mitchell Trubisky - seeing this set of names together starts to explain how being below-average in typical decision-making for throws can leave fans with a sour taste, high levels of productivity or not. Carr, Cousins, and Goff have all had their ample production in recent years, but all seem to leave a bit wanting. Rivers and Ryan are both perceived to have lost something from their arm, which as we'll see further down appears to have an impact on a passer's overall decision-making (theory of what is behind that still to come!). Trubisky gets a shrug from over here.

Poor - Drew Lock, Daniel Jones, Tua Tagovailoa, Drew Brees, Deshaun Watson, Carson Wentz, Sam Darnold, Ben Roethlisberger, Ryan Fitzpatrick, Kyle Allen - we have quite a lot of inexperience here (Lock, Jones, Tagovailoa, Darnold, Allen), which likely speaks to why they're here. Wentz has always been something of a forcer of throws, so seeing him this low is no surprise. But Brees, Roethlisberger, and Fitzpatrick? I think I know why they're so low: all three have in recent years been left with very poor arm strength, meaning that some higher-yield routes have likely been off the table for them. It seems a lack of an arm caps the on-paper value of the "decisions" a QB can make, meaning their type (including Rivers, Ryan, and even Bridgewater) have to be judged by a different standard than the rest. It can be expected that a guy with a "pop gun" arm needs to make up for it with his well he pulls off the throws he makes...which sets up the next section very nicely! That also goes for the yet-ignored Watson.

Rock Bottom - Jameis Winston, Andy Dalton, Jacoby Brissett, Nick Foles - give Dalton the "pop gun" asterisk, though I'm not sure that any of the other names here fully deserve that despite fairly limited arm talent. Winston (the anti-Brady) being ranked this low almost seems to be the cherry on top of confirming the validity of this analysis; he and Brady are the Manny Ramirez and Ichiro Suzuki, respectively, of this analysis (the two former MLB legends being the end-point calibration for fielding metrics for the worst and best performances).

Ranking By Execution

The projected EPA of all throws ("decision making") is one thing, but the passer still needs to then make the throw. In theory, the passers who most outperform actual EPA versus projected are the ones who are either throwing guys open or are doing something with their throws to create more production for a given set of passes taken than others passers would:
The Gold Jacket - Drew Brees - Brady might be the ultimate decision-making passer in today's game, but Brees one way or the other has been getting the job done even under physical decline. Perhaps this involved utilizing Michael Thomas's hands, Alvin Kamara's RAC, ball placement on throws...but the numbers say that his throws continued to work even at the end of his career.

The Elite - Patrick Mahomes, Deshaun Watson, Aaron Rodgers, Kirk Cousins - Mahomes and Watson are both regarded as having some of the best arms in the business, and that seems to translate into throwing targets into better results than typically expected. Cousins probably executes in a lesser-but-similar manner as Brees, while it has long been regarded that what makes Rodgers great is a ridiculous combination of strong reads and strong execution.

The Very Good - Justin Herbert, Matthew Stafford, Dak Prescott, Ryan Fitzpatrick, Ryan Tannehill, Jimmy Garoppolo, Philip Rivers, Lamar Jackson, Derek Carr, Jameis Winston - there's an interesting diversity here, with some terrific pure throwing arms (Herbert, Stafford, Winston), passers who mix solid arms and execution together (Prescott, Tannehill, Garoppolo, Carr), and a couple more guys with weaker arms who know how to make up for it (Fitzpatrick, Rivers). If Jackson could retain his positive showings in each category in this piece while becoming higher volume and seeing defenses key more on his team's aerial attack, he would truly cement his doubters as wrong.

Above Average - Joe Burrow, Russell Wilson, Josh Allen, Teddy Bridgewater - there is no denying that Wilson has not been his peak self more recently, lending the question of how much these categories are affected by supporting cast (or whether he really has taken a step back). Burrow won't need to execute much if he continues to make good decisions with his throws. Allen would be the biggest surprise in this piece, until we factor in both that this includes his pre-breakout 2019 season and that the value he creates with his legs aren't directly measured here either.

Average - Jacoby Brissett, Tom Brady, Mac Jones, Kyle Allen, Kyler Murray, Matt Ryan, Ben Roethlisberger, Tua Tagovailoa - this group contains plenty of more limited arms, illustrating the value of making strong-enough pass decisions that an execution boost isn't needed. Murray is the notable outlier here; his presence suggests issues with consistency and/or making full use of his strong arm.

Below Average - Andy Dalton, Carson Wentz, Jalen Hurts, Jared Goff, Maker Mayfield, Mitchell Trubisky, Gardner Minshew - Wentz and Goff don't produce enough baseline value from their decision-making to get away with relatively weak execution, especially given their natural arm talent, so their placement here further explains their downfall. Mayfield's performance has been generally understood to be reliant on an effective play-action attack; he has only been a little above average at finding valuable targets despite that context advantage, and to this point he's even been a drag compared to his peers once he's put the ball into the air. Improvement in this area from Hurts could cement him into Tier 2 overall performance, and that would spell trouble for the Cowboys.

Poor - Drew Lock, Nick Foles, Daniel Jones, Taylor Heinicke - not much of a surprise here! The names speak for themselves.

Barf - Sam Darold, Trevor Lawrence, Cam Newton - Lawrence can be excused given that he was only a rookie and in an atrocious situation. Darnold might be permanently affected by the tire fire he was developed in, and Newton's placement here makes complete sense given how hard he's fallen off; his shoulder issues have rendered his arm weaker and prone to major inaccuracy.

Overall Player Conclusions

PFF was also kind enough to lay out both on a single graph:
The easiest way to interpret this is to imagine a straight line running from the upper left to the lower right - landing somewhere on that line would render a QB roughly average no matter whether by pure decision-making, pass execution, or partial quality at each. The farther a name is above the line, the better the passer overall, while the reverse is true for those below it.

By that accounting, Rodgers takes the overall crown from the past three NFL seasons, which lines up given his across-the-board strengths. Brady and Mahomes are right behind, though each gets there is almost opposite ways, and we can see that Stafford truly has ascended while Jackson has been a unique case. Inspecting tier by tier from there, the overall results certainly appear to align with general expectations; PFF is definitely on to something here!

The Dak Prescott Verdict

No league-wide analysis like this on BTB is complete without taking a particular look at the relevant Dallas Cowboys, and these results seem to further reinforce the general conclusions held about the man behind center for the team.

Prescott overall is both well-rated and yet a clear notch behind the best of the best. How the next stretch of his career unfolds will largely be determined by whether his present placement indicates that he is knocking on the elite door - ready to kick it in with a little more polish - or the ceiling he has and at best will continue to hit when circumstances aren't a major drag. While it's premature to slot Burrow and Herbert clearly ahead of Prescott in the present only given their similar performance to this point, the fact that the two young passers have only just gotten started suggests that they are more likely to improve, and improve to greater degrees, than Prescott is in the years to come. It would be pretty crazy to not prefer each to Prescott for the years to come.

Perhaps the most useful bit Cowboys fans can take from this exercise with regards to their own is the fact that Prescott places well both with decision-making and with execution. Some continue to insist that Prescott can't read the field well, but NFL experts have tended to firmly disagree and this hard data further points to the likely truth: he knows what he's doing with his reads and progressions. The quality execution numbers for the passer might come as a bigger surprise; we can see from these results that high-end arm talent can be the source of an execution boost, but on the flip side the likes of Brees and Cousins demonstrate that there are multiple paths to getting the best results per pass. Prescott certainly isn't being carried by pure arm talent, and we know he will suffer demerits from the occasional off-target shorter pass, so it's interesting to ponder how he's making up for those hits. Is it from his typically reliable deep throws? His ball placement on longer sideline passes? Those are inquiries for another time.

Note: the PFF article includes a section that we too-often aren't given - a layout of what the authors hope to do next to further improve this analysis! That's juicy, as this specific analysis had some unavoidable limitations (e.g. available sample size of data) while the approach in general still strikes me as being reasonably suscepible to extremes (good or bad) from QB supporting cast. It's just a start, but given how nicely the results generally conform to expectations the start is an excellent one.

Another user-created commentary provided by a BTB reader.