Draft Results and Discussion

Pre Free Agency:

Accuracy: 46%

When compared to the 2022 model, the pre free agency prioritization accuracy essentially matched the 2022 model accuracy. The pre free agency draft prioritization only beat the 2022 model by 2 picks, or 2%. This makes sense since that these two prioritizations were taken within the same context of each other. However, the updated draft prioritization u beat the pre free agency model by 8% and the 2022 model by 10%. Thus, updating for free agency leads to more accurate draft predictions.

The post free agency draft prioritization performed better across the league compared to the pre free agency version. This can be seen on the correct column in each table. The post free agency model only had one team that had zero correct picks, while the other model had 4. Likewise, the post free agency model went three out of three for two teams, while the other model had none.

Although it was an improvement from last year, the model still left a lot to be desired. It still feels like a low accuracy compared to personal expectations. However, there were some trends from observation of the NFL Draft that could explain the reasoning.

First, there were a bunch of teams that drafted the same position of need more than one time. This tended to happen particularly in the later rounds of the NFL Draft. This makes sense since teams with positions of need were willing to draft the position multiple times to make sure that weakness is truly addressed. Likewise, teams also might have just needed more than one player at that position. As a result, it is possible that the scoring system is too strict. Maybe in the future, duplicate selections should be counted, or a position group should count if it is addressed at any point in the draft.

The other thing that could have happened is that this model is ran off of pure numbers, with no personal bias taken into account. Thus, there may be certain position groups recommended to teams that would make no sense in real life context. For example, the Carolina Panthers and the Indianapolis Colts had quarterback in their draft prioritization. However, they had drafted a quarterback in the previous draft with a very high pick. It would make no sense for those two teams to do that in this draft. Thus, this model should be of assistance to teams, not a direct recommendation. In the future, personal opinion should be factored into the model to identify potential false positives.