Welcome to my first official simulation of Round 1 of the 2021 NFL Draft! My analytics-based simulation is based solely on a contextual, data-driven model that aims to do one thing: maximize each team’s potential to win as many games as possible in 2021. So before reading any further, take note :
I am NOT trying to predict or guess what teams will REALLY DO on draft day.
For this particular file, the model looked at current rosters, the overall potential free agent market, and the prospects for the 2021 draft. How exactly? Well, this is how my drill works …
I use my lead draft model, explained at the top of this article, to create a numeric value for each player. These qualifications can be compared between years. I then use my NFL model, which considers the potential free agent market at each position, to create projected earnings contribution metrics by player, position group, and side of the ball. These are added together to predict win totals for the season. (Here is an example of these metrics for WR.) The results quantify the strengths and weaknesses of the current NFL rosters. My model also takes into account as many known elements of training philosophies (from current staff) as possible, and each team’s opponents in 2021. My model then “selects” the prospect from the draft that would produce the highest win total. for each team in the next season.
Here’s the bonus: I have projections and results for every team and draft prospect they’ve selected over the past 15 seasons. I examine the results on the field for each season, objectively analyzing what happened and identifying the trends and strategies that led to success or failure. I also ask coaches, front office executives, and even players to help me understand why the results came about. These subjective inputs help shape the results, which means that the model gets “smarter” every season.
There are many real-life efficiencies that could be achieved through draft pick operations. I can’t help but notice them in certain cases. Still, for the sake of this particular drill, I did not allow exchanges. If you were working for an individual team, an analysis like this could help create a strategy to identify potential business partners, as well as vulnerabilities where other teams are more likely to get particular players, especially if free agency is in place.
Finally, another change to this year’s simulation is a real refinement of how computer vision in the field weighs on predictions. This is typically a huge factor, as a prospect’s most recent season’s field performance is the most valuable ingredient. This college season was strange (to say the least); the number of games played, when they were played and even the situations / contexts faced were quite different from previous years. So I had to look for factors over longer periods of time (and their trajectories) and normalize all past measurements (like we would get from the combine) for the 15 seasons using computer vision to make sure every comparison was as equally possible.