NFL Draft: The Stats that Correlate to Quarterback Success

The quarterback is the most important position in professional football. He touches the ball on essentially every play and is one of the only players on the team that plays every offensive snap. Quarterbacks often determine NFL success. Teams with great quarterbacks, like the Chiefs with Patrick Mahomes, often have high playoff aspirations and Super Bowl hopes, while those without a solution to the most important position in football (like the Chicago Bears) struggle to win lots of games. Just one of many examples is how the Tampa Bay Buccaneers switched their quarterback from interception-prone Jameis Winston to three-time NFL MVP Tom Brady, then won the Super Bowl shortly after.

The most popular avenue of finding a franchise quarterback comes through the NFL Draft. Each year, the worst teams from the previous season obtain the best picks and the opportunity to take their future quarterback. However, a highly picked quarterback is no guarantee to be a great player. For example, none of the quarterbacks picked from 2009 to 2016 are on the team that drafted them. While these players include good quarterbacks in bad situations, namely Matthew Stafford, it mostly consists of disappointing players, such as Mark Sanchez, Blake Bortles, and Jared Goff. In order to find which quarterbacks have a better chance to succeed in the NFL, we can take a look at the stats of previous quarterbacks that were drafted.

Tom Brady’s move from the Patriots to the Buccaneers demonstrates the importance of quarterbacks in the NFL

Analyzing College Quarterback Stats

When looking at the college stats of the quarterbacks in the draft, many people take into account all the stats equally. They will view a high number of yards in the same way they view a high completion percentage. However, certain college qualities carry over to the NFL, while others do not. To find which college quarterback stats are most consistent from college football to the NFL, we can look at the correlation between the college and NFL values for the past drafted NFL quarterbacks for various stats. The NFL passer rating statistic, one which is used by many to determine the quality of NFL quarterbacks, uses four main statistics as its inputs: completion percentage, touchdown percentage, interception percentage, and yards per attempt. While passer rating has its flaws, the inputs are useful for determining which quarterbacks have had NFL success and which have not.

For each of the graphs and models that are shown in the remainder of the article, the sample of players consisted of all quarterbacks drafted from 2000 to 2020 with at least 20 games started. No FCS quarterbacks (e.g. Joe Flacco and Carson Wentz) were included as their college statistics would have been skewed from playing with such inferior competition. This produced a sample of 78 quarterbacks, each with a sizable sample of pass attempts as they started several games.


The most consistent quality from college to the NFL is accuracy. The most common statistic to determine accuracy is completion percentage. Of course, it does not completely account for accuracy because the simple calculation of completion percentage cannot account for the fact that certain quarterbacks attempt more difficult throws than others. While some sites attempt to consider the difficulty of pass attempts using a completion percentage over expected metric, the simple completion percentage stat is more accessible and simpler to use and understand.

Correlation: r = 0.368

The quarterbacks that were very accurate throwing the ball in college often had good accuracy in the NFL. The correlation coefficient (r) of 0.368 indicates a moderate correlation, although touchdown percentage, interception percentage, and yards per attempt all had lower values. There are numerous examples of players that were accurate in both college and the NFL, including Deshaun Watson, Teddy Bridgewater, and Russell Wilson. On other other hand, players like Derek Anderson and Michael Vick were not very accurate in college nor the NFL. There were only a few major outliers: Brandon Weeden and JaMarcus Russell were two players that were accurate in college but not in the NFL, while Drew Brees and Kirk Cousins were among the few that had average accuracy in college but superb accuracy in the NFL. For the 2021 quarterback prospects, this finding bodes well for Mac Jones, Zach Wilson, and Justin Fields, who had the three best completion percentages of the prospects for this year. Conversely, Sam Ehlinger, Jamie Newman, and Kellen Mond had the three worst completion percentages of this year’s prospects, meaning they could struggle to be accurate in their NFL futures.

Teddy Bridgewater had a high completion percentage in college and in the NFL


After accuracy, the second most consistent passing quality from college to the NFL is turnover rate. Interception percentage can be used to measure how turnover-prone a quarterback is in both college and the NFL. Again, interception percentage is a simple stat using only the number of interceptions and the number of attempts, so the pass difficulty was not accounted for.

Correlation: r = 0.358

The correlation of interception percentage from college to the NFL was very close to the correlation of completion percentage. This is likely because both stats measure a form of accuracy, as quarterbacks who complete passes at a higher rate are also going to avoid throwing the ball to the other team at a high rate. The least turnover-prone quarterbacks in college, such as Marcus Mariota, Teddy Bridgewater, and Dak Prescott, did not throw many interceptions in the NFL. One notable exception includes Kevin Kolb, who had just 4 interceptions in 432 pass attempts (0.9%) during his final year at Houston, but 25 interceptions in 775 attempts (3.3%) in the NFL. Additionally, players that throw lots of interceptions in college also throw many interceptions in the NFL, on average. For example, it is not surprising that Jameis Winston, notorious for his interceptions, has thrown so many picks during his NFL career (3.4%) considering that the threw interceptions at a high rate in college (3.9%) as well. Among this year’s prospects, interceptions are a red flag for Jamie Newman, Justin Fields, and Kyle Trask, but a lack of interceptions is a positive for Trey Lance, Zach Wilson, and Mac Jones. When evaluating prospective NFL quarterbacks, two of the most important stats to investigate are completion percentage and interception percentage.

Jameis Winston’s turnover tendency with the Buccaneers should not have been surprising given his college stats

Yards Per Attempt

Yards per attempt primarily measures the explosiveness and efficiency of a quarterback. The consistency of yards per attempt from college to the NFL is not as high as that of completion percentage or interception percentage, but still is somewhat predictive.

r = 0.251

Unlike completion percentage and interception percentage, which are both measures of accuracy, yards per attempt measures how explosive a quarterback is. Quarterbacks with a high yards per attempt often throw the ball very far don’t throw a lot of checkdowns. While some quarterbacks with high yards per attempt in the NFL had a high yards per attempt in college (like Russell Wilson and Philip Rivers), most had only average values in college. While there is a slight correlation between college and NFL yards per attempt, the NFL results for yards per attempt are mostly random.

Touchdown Percentage

The least correlative stat from college to the NFL was touchdown percentage. Touchdown percentage measures how often a quarterback scores by throwing the ball.

r = 0.081

With a correlation so close to zero, there is essentially no relationship between college touchdown percentage and NFL touchdown percentage. Lamar Jackson, Patrick Mahomes, and Aaron Rodgers, the quarterbacks with the three highest touchdown percentages in the sample, all had just about an average touchdown percentage in college. Only one player had a phenomenal touchdown percentage in both college and the NFL, which was Russell Wilson. Having a low touchdown percentage in college does not matter very much: for example, Drew Brees had a college touchdown percentage in the 21st percentile, but now has the 2nd most touchdown passes in NFL history. Of the 2021 quarterback prospects, Mac Jones, Kyle Trask, and Zach Wilson had the best touchdown percentages. However, this does not mean they will have success scoring touchdowns in the NFL.

Analyzing Physical Traits

Whenever NFL Draft analysts and scouts talk about which quarterbacks they think will succeed in the NFL, they almost always mention a quarterback’s height or weight. Draft experts often warn people that a certain player won’t be able to play as well in the NFL because of a small height or a small frame. However, short quarterbacks like Kyler Murray (5’10”) and Baker Mayfield (6’1″) have been drafted as the first overall pick despite their height being a perceived weakness. Does the height or weight of a quarterback really matter?


First, we can analyze the importance of weight on NFL performance. Draft scouts generally warn against quarterbacks that are too light or too heavy. This makes sense as those who are too light would be very likely to take sacks and get injured, while those who are too heavy would be unable to navigate the pocket very well. However, the sample of 78 quarterbacks drafted over the last 20 years shows a different trend.

When comparing weight (at the time of the player’s drafts) and NFL passer rating, we can see that weight has an effect on the future performance of quarterbacks. In this sample, the quarterbacks that weighed less generally had a better passer rating in the NFL than quarterbacks that weighed more. Most of the good quarterbacks in the NFL today had a below average weight during their draft. Meanwhile, several of the more high-profile quarterback busts of the last 20 years had a much larger weight than average, including JaMarcus Russell, Josh Freeman, and Brock Osweiler. It is important, though, that the sample used in the graph above does not include all the quarterbacks who did not start at least 20 games in college. Therefore, it is likely that weight is at least somewhat predictive of future NFL performance, but it is not definite.

JaMarcus Russell, who had the largest weight of the sample of QBs, was a bust after being the number 1 pick in 2007


When draft scouts discuss the height of a quarterback, they often say that players that are smaller would have a harder time throwing over their offensive linemen and seeing the field. However, just like weight, we can evaluate the truth of that claim. By examining how impactful height is on NFL passer rating, we can detect the true importance of height.

As seen by the graph above, the height of a quarterback actually may matter, but in the opposite fashion that expected. In the sample of 78 quarterbacks that I used, those that were smaller actually had better passer ratings than those that were taller. The average passer rating of players under 6’2″ (74 in) was 85.4, which was about 3.5 points greater than those with an average height, which is about 6’3″ (75 in) or 6’4″ (76 in). While there were no significant differences due to a small sample size, the results do show that being shorter does not change the NFL outlook of a quarterback.

Other Factors

Draft Round

Obviously, the round that a quarterback is drafted matters. However, the sample that I used consists of quarterbacks that already started 20 games in the NFL. Therefore, the quarterbacks included were already good enough to have started many games. In the sample of quarterbacks that I used, the draft round actually did not have a large impact on NFL passer rating.

Surprisingly, the quarterbacks drafted in later rounds had the same average passer rating than the quarterbacks drafted in early rounds, such as round 1 or 2. However, the sample sizes for each round except round 1 included less than 10 players. This means that very good outliers, like Russell Wilson in round 3 or Dak Prescott in round 4, had a large effect on the mean value. Meanwhile, the sample of round 1 quarterbacks included 37 players, with a balanced mix of good players and bad players.


The last major factor in determining the future success of drafted quarterbacks is their age at the time of the draft. The age of a quarterback can matter to a large extent since very young quarterbacks have more time to develop before entering their primes, while older quarterbacks have to be effective right away to make the first round investment worthwhile.

Of the quarterbacks in the sample, all but 2 were between ages 21 and 24 when they were drafted. The two outliers were Chris Weinke (age 29) and Brandon Weeden (age 28). Both of these quarterbacks did not perform well in the NFL, with passer ratings of 62.2 and 76.0, respectively. Of the quarterbacks who were between 21 and 24, those who were 21 on the day of their draft performed the best, on average. Therefore, it is reasonable to assume that younger quarterbacks are better to draft than older quarterbacks on the day of the draft.

After being drafted at 29 years old, Chris Weinke was unable to find success in the NFL despite winning a Heisman trophy at Florida State

Multiple Linear Regression

Now that we have explored the effect of several stats and traits on the future of a quarterback prospect, we can combine them to get an overall picture of how successful a quarterback may be. Since the sample size of quarterbacks was very low, I did not want to include too many variables in order to avoid overfitting. Eventually, I settled on four inputs for the regression. The first input was completion percentage, which was found to be predictive of future NFL completion percentage, one of the inputs into the NFL passer rating formula. Secondly, I included college interception percentage as it shows how turnover prone a quarterback is. Unfortunately, this variable is not very significant in the regression due to its collinearity with college completion percentage. After that, I included the weight of the quarterbacks since it was shown that quarterback prospects that weigh less generally perform better. The last variable I included was draft age, since younger quarterbacks traditionally have a better NFL outlook than older ones. The variables, coefficients, and significance levels are shown below.

DrAge = Draft Age, CmpP.col = college completion %, PassTOP.col = college interception %, Wt = Weight


One way to examine the accuracy of the model that I constructed is to look at a residual plot.

The players located towards the right of the graph above were predicted to be good passers by the model, while those located towards the left were predicted to be bad passers by the model. Players located towards the top exceeded expectations, while those towards the bottom disappointed.

The NFL Passer Rating model had several successes and failures within the sample. The most well-liked players by the model were Teddy Bridgewater, Russell Wilson, and Kyler Murray. Bridgewater was worse than his expected 95.3 passer rating, but Wilson met and exceeded his expectations. The model also successfully predicted that Chris Weinke would not be good in the NFL because of his age. On the other hand, some of the misses included Matt Ryan, who was better than expected, and Geno Smith, who was worse than expected.

Wilson was one quarterback that the model expected to do well in the NFL

Predictions for 2021 Quarterbacks

Now that I have made a model that can help to predict how good a quarterback’s passer rating will be in the NFL, I can apply the model to the quarterback prospects for 2021. The predicted NFL passer rating for 9 quarterback prospects in the 2021 NFL Draft are shown below.

Proj_Rnd = Projected Round, yhat = predicted NFL passer rating, all percentages are shown as decimals

Of this year’s quarterback prospects, my passer rating model likes Zach Wilson the most. Wilson has the second highest completion percentage, the second lowest interception percentage, and the lowest weight of all the quarterbacks in the draft class. Behind him is Mac Jones, who had an outstanding completion percentage that was over 77% to go with a low interception percentage and low weight. The reason Jones is below Wilson is that Jones is a year older. However, both Wilson and Jones had advantages, which were playing against below average competition and playing with phenomenal players, respectively. After that is Trey Lance, who had 0 interceptions in 2019. Lance’s 2019 stats were used since he only played in 1 game in 2020, which was a showcase game where he threw 2 interceptions, interestingly. Additionally, the accuracy of Lance’s projection is unknown since he played for North Dakota State, an FCS school. No FCS quarterbacks were included in the sample which was used to make the regression model.

Zach Wilson has the highest expected NFL passer rating of the 2021 quarterback prospects

Trevor Lawrence, the presumptive number 1 pick, ranks fourth in these rankings. Even though the model predicts that he will have a very high passer rating in the NFL, it has him below Wilson, Jones, and Lance since they each have a better completion percentage or interception percentage. The lowest ranked quarterback that is expected to be picked in the first round in Justin Fields. This is largely because of his high interception percentage, above average weight, and age of 22 years. It is important to note, though, that the regression only accounts for passing predictions, not overall quarterback play. Therefore, the regression does not fully capture the ability of Justin Fields (and Trey Lance) to pick up yardage even after the pocket collapses.

Of the quarterbacks that are not expected to be picked in the first round, Kellen Mond ranks the highest as he has a low interception percentage and weight to make up for his low completion percentage. Trask and Newman are expected to be the worst quarterbacks of this class. Both are old compared to the rest of the quarterbacks of this year, and they both have several red flags. Trask weighed the most of all the quarterbacks shown above, and he has a high interception percentage. Jamie Newman had the lowest completion percentage and highest interception percentage of this year’s prospects, meaning he will likely have trouble with accuracy in the NFL.

The rankings produced by the model are not completely accurate. For example, almost all draft scouts have Trevor Lawrence ranked first by a wide margin and have Justin Fields ranked well above Kellen Mond. There are several things the model could not take into account. It only used one season’s worth of data, so it is possible that a quarterback had a very good year which was not counted. Additionally, due to the nature of the sample the model could not consider the projected draft round of a quarterback, which is usually a good predictor of success. Additionally, the model could not include aspects of a quarterback like his ability to go through reads and to make plays when the pocket collapses, things which NFL scouts took into deeply. Therefore, the rankings should be used as a general tool in conjunction of draft scouts’ observations for whether a quarterback is worth being picked higher or lower than expected based on the consensus.

Kyle Trask is not expected to be very good in the NFL despite being a Heisman contender during his final year at Florida


While the futures of drafted quarterbacks are difficult to predict, some stats can be used to better capture the likelihood of a player having success. Completion percentage and interception percentage correlate the best from college to the NFL, meaning that accuracy is the most likely attribute to carry over from college. Physical traits can also be used to predict the success of quarterbacks as lighter quarterbacks have a better NFL passer rating, on average, than heavier quarterbacks. Since the sample I used only consisted of quarterbacks that were good enough to have started 20 games, the draft round did not significantly predict the NFL passer rating of a quarterback. However, age was a major factor since older quarterbacks were predicted to have less NFL success than younger quarterbacks. Finally, using a multiple linear regression, the quarterbacks with the three highest projected NFL passer ratings in this year’s draft are Zach Wilson, Mac Jones, and Trey Lance.

One thought on “NFL Draft: The Stats that Correlate to Quarterback Success

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