Connor Dwyer and Andrew Gullotta
Super Bowl LV has all the quarterback storylines: the greatest of all-time vs. the greatest right now, the six-time Super Bowl winner in Tom Brady vs. the defending Super Bowl champion in Patrick Mahomes. But, what about the other 30 teams in the NFL?
ESPN reporter and insider Adam Schefter predicts that upwards of 18 franchises will turn to a new quarterback under center to start the 2021 season, with the New Orleans Saints, Indianapolis Colts, and New England Patriots headlining the expansive list.
Quarterback shake-ups aren’t new (Tom Brady left New England for Tampa Bay just last season), but 2021 has an added variable that makes shopping for a new signal-caller even harder: a 10% decrease in the salary cap. This feature will force teams to spend more prudently when filling holes on their roster, especially at quarterback. In a year when getting a “bang for your buck” is of utmost importance, teams must look to maximize their return on investment at the quarterback position.
With that being said, Fordham Sports Analytics Society Vice President Connor Dwyer and I set out to create a model to predict the monetary value of a quarterback based on a number of statistical variables that balances individual value with team performance.
Franchises evaluate signal-callers through many different lenses, from individual stats to the team’s overall performance. Implementing stats such as DVOA, Effective Yards, Pro Football Focus (PFF) grades, Approximate Value, and Passer Rating convey a quarterback’s individual value to the team. Meanwhile, accomplishments like winning percentage, playoff wins, and Super Bowl victories express overall team value. For example, Nick Foles would have never received a four-year, $88 million deal if it weren’t for his .778 winning percentage, four playoff wins, and Super Bowl MVP when starting for the Philadelphia Eagles from 2017 to 2018.
After looking at a plethora of signal-callers, we decided to build our model from only quarterbacks who received a contract within the last three years by using their statistics leading up to the contract. We employed weighted averages based on the number of games played each season to determine each statistic needed for the model (as seen below). To determine the correlation between a quarterback's stats and his percentage of cap, each statistic was considered an independent variable, while the percent of cap was considered the dependent variable.
To account for inflation over the last five years, each contract was defined as a percentage of salary cap instead of a monetary value. This feature ensures that there are no discrepancies between players caused by inflation.
Once the data set included 30 quarterbacks, we used the Data Analysis Toolpak function of Excel to generate a multi-variable regression model. By applying this function, we determined that there was an “r value” of 0.947, and an “r squared value” of 0.897, signaling a strong correlation between the statistics and the percentage of the cap for the quarterbacks.
We then used each variable’s coefficient to determine an equation that predicts the percentage of the cap based on the various quarterback statistics. For example, one could enter Cam Newton’s stats for the 2020 season to predict his next contract’s average salary.
To ensure the equation is effective and the data is normal, we ran the model using the statistics of the 30 quarterbacks used to create it. From there, we graphed the residuals vs. the actual percent of the cap, as seen below. No apparent patterns or direct relationships between the two suggest that the data is normal, which confirms that a linear model is a good choice for this sample. If in fact a pattern emerged in the residual graph, one may look towards non-linear methods of analysis.
Once we determined the regression model and accompanying predictive equation, we took our best shot to answer the initial question: How much are this year's free-agent quarterbacks worth on the open market?
We looked at four of the more prominent free-agent quarterbacks heading into the offseason: Dak Prescott, Andy Dalton, Mitchell Trubisky, and Ryan Fitzpatrick. To predict what contract value each signal-caller should receive this offseason, the statistics for each player's last two years were collected and are shown below.
Armed with the weighted averages for each of the four quarterbacks, we inserted the statistics into the predictive equation, resulting in the percentage of cap that each player should take up. Because of the uncertainty in the 2021 salary cap, we made projections based on two potential salary cap thresholds.
Starting off with the 2016 Offensive Rookie of the Year, Prescott is positioned to become one of the highest-paid quarterbacks in the NFL. Based on individual value and team impact, the two-time Pro Bowler should garner 14.89% of a team’s salary cap. With next season’s threshold predicted to be anywhere between $175 and $198.2 million, Prescott is projected to earn an average annual value (AAV) between $26.06 and $29.52 million based on our predictive model.
Moving over to the former second overall pick, Trubisky has had a disappointing start to his career, but still offers value at the quarterback position. Over the past two seasons, he won more than 58% of his starts and accumulated more effective yards than two quarterbacks in our study who both garnered significant long term contracts (Nick Foles and Jimmy Garoppolo). With that being said, Trubisky is projected to take up 8.58% of a team’s cap, which when converted results in a contract with an AAV between $15.02 and $17.01 million.
Looking at a few veteran quarterbacks, both Ryan Fitzpatrick and Andy Dalton hit the free-agent market in 2021 and have plenty of experience commanding an NFL offense. These two veteran gunslingers combined for 41 starts over the past two seasons and collected 9 wins in 17 games in 2020.
Fitzpatrick is an admiral backup with starting potential, as he posted a 76 PFF Offensive grade, which places him in the top half of our sample size for that statistic. However, the gunslinger finished in the bottom ten quarterbacks for both Approximate Value per Game and Passer Rating. With that being said, our model projects him to take up 5.73% of a team’s salary cap next season. This prediction sets him up for an AAV between $10.02 and $11.35 million next season.
As for the former Bengal, Dalton’s play with the Cowboys proved that he is a formidable backup who can keep his team afloat during the season. However, his production over the past few seasons does not suggest he deserves a full-time starting job. For example, The Red Rifle placed in the bottom ten for every metric used in the study, including DVOA, Passer Rating, and TD%. Our model projects that he will consume 1.38% of the cap, which converts to a predicted salary between $2.41 and $2.73 million.
While Tom Brady and Patrick Mahomes duel for the Lombardi Trophy in Tampa Bay, more than half of the NFL may face major turnover at the league’s most important position. In a unique offseason, teams will have to pin-point matches between player and monetary value at an unprecedented rate. Front offices and coaches will undoubtedly use predictive models similar to this one to acquire their next franchise signal-caller, so they can be the next team duking it out on Super Bowl Sunday.
Sources:
https://www.footballoutsiders.com/info/glossary#:~:text=DVOA%20stands%20for%20Defense%2Dadjusted,this%20stat%20is%20called%20VOA. https://www.pff.com/grades https://nextgenstats.nfl.com/glossary https://www.pro-football-reference.com/about/glossary.htm