Introduction
In the highly competitive landscape of American football, the success of a team often depends on the performance of its position groups. From the precision of the quarterback to the tenacity of the defensive line, each position plays a vital role in shaping the outcomes of games and ultimately the entire season. As football strategies evolve and teams seek a competitive edge, the need for a comprehensive position group ranking model becomes increasingly evident. The complexity of football, with its multifaceted nature and dynamic interactions between players, demands a new approach to evaluate team dynamics. Historically, the unpredictability of player performance, coupled with the evolving dynamics of the game, makes it a hard task for football franchises to consistently identify and secure valuable talent through the draft. Teams face the challenge of navigating through a multifaceted landscape where the promise of high draft picks comes with no guarantee of success. The challenge of securing top-tier talent is weakened by the reality that player performance at the professional level is influenced by many factors, many of which are difficult to quantify and predict.
Moreover, the financial implication of high draft picks are substantial. Teams invest not only in talent itself but also in the anticipation that these selections will translate into on-field success. The gap between the perceived value of a draft pick and the actual impact on a team’s performance is exemplified by instances such as the Chicago Bears’ selection of quarterback Justin Fields with a high pick. Despite the significant investment, the anticipated increase in on-field success remained the same. This underscores the critical need for a comprehensive and data-driven approach to evaluating and prioritizing offensive and defensive position groups. This study seeks to go beyond the surface of football discussions, diving into the need for teams to go beyond the realm of chance and instead make informed, value driven decisions in the draft. By making a comprehensive ranking model, this research aims to equip teams with the tools needed to boost their offensive and defensive capabilities and, in turn, enhance their prospects for success.
This research serves as the second edition of this model and updated to reflect the 2024 NFL offseason, where the ultimate goal is to accurately predict the results of the 2024 NFL Draft. Part of the motivation behind repeating this model had to do with the results of the previous model. Although the methodology reflected a credible and reliable way to rank teams by position groups, the results did not meet expectations. Some of the reasons included correlations between position groups and ignoring free agency. The methodologies remain the same, but there are some slight modifications based off of the findings from the previous edition. The table below presents some of the changes made to the model from the previous version.
Model Changes
The first set of model changes have to do with two relationships between position groups, discovered in the previous iteration. The first one has to do with the linear relationship between the quarterback and wide receivers. The initial research showed that the ranking of a quarterback correlated with the ranking of the corresponding receiver corps. This makes sense, since the quarterback throws the ball to the receivers, while the receivers catch the ball from the quarterback. To solve this, the model now ranks individual players as opposed to the team statistics. This is a more powerful way to assess not just the quality of players, but also the depth of a position group. It could help explain a teams situation at any position, or it could help break a tie in a prediction.
There was also an inverse relationship between the defensive line and linebackers. For example, a team with a top defensive line often had a bottom tier set of linebackers. This can be explained by defensive schemes. Below are images and explanations of the two main defensive formations and their effects on the rankings.
A 4-3 Defensive Formation
A 3-4 Defensive Formation
Notice how the 4-3 defense has more DT and DE’s than LB’s, and vice versa for the 3-4 formation. This caused the inverse relationship between those two position groups. To solve this issue, the defensive ends in a 4-3 and the outside linebackers in a 3-4 were combined into a new position group, called the Edge Rushers (EDGE). This made sense since the 4-3 defensive ends and the 3-4 outside linebackers are in similar positions on the field and have similar responsibilities. As explained in the Sources and Variables tab, the edge rushers will be a part of the defensive line rankings.
The new edge rusher position group is not the only change made like this. Since the model now grades individual players, all positions under a position group have been split up. The other two instances of this are on the offensive line (Tackles vs Interior Offensive Linemen) and the defensive backs (Cornerback vs Safety). This was also a good approach for the new model, since the draft prioritization can now be identified for a specific position as opposed to a generalization. An example would be the Carolina Panthers, where their offensive line was very poor this year. When the grades came out for the players on that line, it was shown that the tackles actually did just fine, but the interior offensive linemen were very poor. Therefore, a recommendation was made after the season for the Panthers to improve the interior offensive line and not the tackles.
The final major change had to do with updating player movement during free agency. The previous model did not account for free agency, where players that do not have a contract have a chance to sign one with another team. As a result, the previous model was making recommendations based on the end of the previous season, when a need could’ve already been taken care of during the free agency period. Updating team needs could be the big change that improve the draft prediction accuracy.
Finally, there are some smaller changes for aesthetic, such as using min-max scaling to design grades like in a video game. The range is from a 99, representing the best player in a position group, to 65, which is either the worst player that qualified for a grade or no player qualified for a grade.