Results
The End of Season Ranking
After all rankings for each position group were created, they were organized into a heat map by their grade and their rank compared to their peers. Below are the heat maps by grade and by rank. Green indicates that the position is a strength, red indicates a weakness, and yellow represents that the position is neither a strength nor a weakness.
The table of rankings (left) and grades (right) are shown. Green indicates a strength, red red means the position is a weakness, and yellow is neutral.
To determine the positional priority for each team, the ranking table on the left was used as the primary source. Visually, the color difference is a lot more prominent on the ranking table than the grade table, due to the presence of outliers and similarity between the grades in the grade table. On the right hand side, the quarterback and running back contribute to the skewness the most because they are only defined by one player, meaning their volatility between observations are much higher. Since the other positions are averaged across all players within their groups, their means are more likely to be centered close to the average. Therefore, the ranking table was determined to be the primary source, while the grade table was used to break ties or delve deeper into each team’s situation. After that, the lowest three position groups were identified for each team and put into another table.
The top three needs for each team.
Since the goal of the model is to improve the draft model accuracy from the previous model, the top three position groups were recorded to remain consistent with the 2022 model. However, the beauty of grading by individual players and positions this time around is that this model can specify the position within the group that needs the most attention. Take a look at the Carolina Panthers for instance. One of their top three weaknesses was the offensive line, which was ranked 27th among all teams. It could suffice to say that they can just acquire an offensive linemen to satisfy the weakness. However, take a look at the individual players.
From the table, it is clear that despite the unit performing horribly, the offensive tackles actually performed well. The issues were on the interior offensive line, and this was caused by not just poor play, but also several injuries to the position group. Therefore, if somebody wanted a specific recommendation, the model would prescribe drafting a guard or a center as opposed to a tackle. The appendix will have a data table with specific positions for all phases of the offseason, but for simplicity the model will just refer to the position groups as a whole.
The Free Agency Period
One of the biggest flaws of the 2022 model was that it did not account for free agency. This is the period where returning NFL players that do not have contracts are free to sign with any team, whether its their own or elsewhere. The player movement caused by free agency has an impact on the positional needs of a team. For example, the intial 2023 model showed that the Baltimore Ravens needed to address the running back position. They signed former Titans running back Derrick Henry, who ranked in the top 10 last year. As a result, the Ravens skyrocketed up the running back rankings. However, their highly ranked offensive line saw three of its players leave the team to sign elsewhere. Therfore, the Ravens offensive line ranking plummeted to the bottom. This shows that a position group can be improved in free agency, while another unit can lose players and turn into a draft need. Therefore, a free agent spreadsheet was created to track player movement across the league.
After the intial model was built, potential free agents were identified from each team, and placed into a list sorted by their player grade. Once free agency began, the list was consistently updated to record the status of each player. There were also quite a few players that were traded to other teams, and since they were technically not free agents, their movement was recorded off to the side. Once free agency settled down, every teams's position groups were recomputed and resorted to reflect the additions and subtractions of the team. Below is the new heat map for the rankings and grades for each team.
The table of rankings (left) and grades (right) are shown. Green indicates a strength, red red means the position is a weakness, and yellow is neutral.
Something else that happened during the re-ranking was the inclusion of players that were "filtered out" from the k-means clustering. For position groups such as offensive and defensive line, there were some players that played a vast majority of the season, but were not included in the initial ranking because they either did satisfy the minimum snap count or don't produce a high number of statistics to have been included the first time. To account for this, a player was included in this round of the rankings if they played at least half the season. To determine the grades for those players, their statistics were scaled to reflect a full 17 game season, and then a close substitute was identified based off of raw statistics. The player was then given the same grade. As a result, some of the positional unit grades could have been inflated. Therefore, it is important to look at a table that shows the change in rankings and change in grades. Below are tables that show the improvement or regression of a team based off of rankings and grades.
These tables show the change in rankings and grades from the end of season to the end of free agency. A green cell indicates improvement while a red cell shows regression. A gray cell reflects a team that neither improved nor got worse. The right hand side column in each table shows the total change in team quality.
These tables really show how a position group can affect the draft prioritization for teams. A solid green cell shows that a team improved the position group significantly in the free agency period, while a solid red cell means that a team lost key players at those positions. Gray cells indicate that either a team did not address the position or the acquisitions did not move the needle for that team. The model can also show which teams overall improved the most and regressed the hardest. For example, the model would seem to indicate that the Carolina Panthers are the most improved team while the Dallas Cowboys suffered the most. However, it is important to consider the volatility of the quarterback and running back. Since those position groups are determined by one player, losing a player and failing to replace him can skew a team's true offseason impact. Look at the Dallas Cowboys. Although their ranking change was -46 points, over half of that was because they lost their starting running back and failed to replace him yet. While this is a downside in the model, at least it shows a very obvious need for the Cowboys in the draft. The entire row for each team needs to be assessed to gauge the true impact of a team's offseason during free agency.
Finally, a new draft prioritization was created and produced in the table below. A green cell indicates here that the position group is a new team need that was not in the initial draft prioritization. Out of the 32 NFL teams, 26 of them had at least one new positional need. This is a big deal, since the 2022 model failed to capture these changes shown in the 2023 model. There is a real potential here to improve the accuracy of the model, and it will be interesting to compare the initial 2023 model to the updated 2023 model.
The updated draft prioritization for each team. A green cell highlights a new position of need for each team.