How we’re using player evaluations, clearly defined position requirements, and a large language model (LLM)—such as ChatGPT, Claude, or Gemini—to create a more consistent starting point for assigning offensive and defensive positions.
Every season begins about the same for us.
We spend a couple of practices evaluating our players.
The goal isn’t simply to see who is the fastest or the strongest. It’s to collect measurable data that helps us understand each player’s athletic traits and gives us a starting point for building our team.
Like many youth football programs, we evaluate things such as speed, change of direction, strength, aggressiveness, age, and weight. Those evaluations are combined into an overall score that helps us identify our better football players.
Once the evaluations are complete, we typically take that overall score and begin assigning players based on our offensive and defensive position priorities. Our highest-rated player is considered for our highest-priority position, the next highest-rated player for the next priority, and so on.
That process has worked reasonably well for us over the years.
But this season we started asking ourselves a different question.
Were we fully using the information we had already collected?
Looking Beyond the Total Score
A Total Score is useful because it identifies your best overall football players.
But it doesn’t necessarily identify the best fit for a particular position.
Two players may have nearly identical overall scores, but very different athletic profiles.
One player may have exceptional speed and change of direction.
Another may be stronger, more aggressive, and more physical.
Depending on the position, either player could be the better choice.
That led us to a simple question.
Could a large language model (LLM) help us dig deeper into our evaluation data and give us a better starting point for assigning positions?
Notice the question wasn’t:
“Can an LLM build our roster?”
The question was:
“Can it help us better interpret the evaluation data we’ve already collected?”
The Cost of Guessing
One of the things we’ve noticed over the years is that, without a structured evaluation and position assignment process, many youth coaches spend a good portion of the season moving players from position to position.
A player starts at running back.
Two weeks later he’s playing tight end.
A week after that he’s playing defensive end.
Every position change means learning new assignments, new techniques, and building confidence all over again.
Sometimes those changes are necessary.
But sometimes they’re simply the result of not identifying the player’s best fit early enough.
If we could improve that process before the season began, we believed it could benefit both the player and the team.
We Quickly Learned the Evaluation Spreadsheet Wasn’t Enough
Our first attempt taught us something almost immediately.
The evaluation spreadsheet by itself wasn’t enough.
The LLM had no way of knowing what we were looking for at each position.
It didn’t know what made a good offensive guard.
It didn’t know what traits we valued in a linebacker.
It didn’t know which positions required speed, which required physicality, and which depended on characteristics that aren’t measured during evaluations.
That wasn’t a problem with the LLM.
It was a gap in our own documentation.
Most of that information existed only in our heads.
Documenting the Methodology
That realization turned into a much bigger project than we expected.
We started documenting:
- What each evaluation drill actually measures.
- What responsibilities each offensive and defensive position has.
- Which measurable traits matter most at each position.
- Which important characteristics aren’t measured and still require coaching observation.
- Our position priorities.
- Our roster construction rules.
The more we worked on it, the more we realized we weren’t just writing better instructions for an LLM.
We were documenting our Position Assignment Methodology.
Player Evaluation and Position Assignment Are Different
One of the biggest lessons we learned is that player evaluation and position assignment are two different processes.
Player evaluations tell us what athletic traits a player possesses.
Position assignment determines where those traits create the greatest value for the team.
Those are not the same thing.
A player with the highest Total Score isn’t automatically the best choice for every position.
Some positions place a premium on speed.
Others reward strength and aggressiveness.
Some require exceptional change of direction.
Others depend heavily on traits our evaluations don’t directly measure, such as leadership, football instincts, reliable snapping, ball security, or tackling ability.
Those are still coaching decisions.
The evaluations simply help us make better-informed ones.
Where the LLM Fits
Once the methodology was documented, the LLM became much more useful.
Instead of asking it to assign positions, we asked it to apply our methodology.
We upload the completed evaluation spreadsheet.
It analyzes each player’s measurable traits.
It compares those traits to the position requirements we’ve documented.
It considers our position priorities and roster constraints.
Then it creates a first draft of offensive and defensive position assignments, identifies position battles, explains its reasoning, and highlights areas where coaching observation is still needed.
That first draft isn’t our final roster.
It’s a much better starting point for our coaching staff.
The Biggest Surprise
Going into this project, we thought the biggest benefit would be getting better position assignments.
Looking back, we think the biggest benefit was something else.
The process forced us to clearly define what we’re looking for at every position.
It also forced us to examine whether our evaluation drills actually measure the traits we value most.
Those conversations have already made us better coaches.
The LLM simply helps us apply that methodology more consistently.
Final Thoughts
We’re still learning.
We’ll continue making adjustments as we use this process throughout the season.
Some recommendations will be right.
Some won’t.
That’s expected.
The coaches still make the final decisions.
But if we can consistently place more players in positions that match their measurable traits before the season begins, we believe we’ll spend less time moving players during the season and more time helping them develop where they’re most likely to succeed.
Whether you ever use a large language model or not, we’d encourage you to spend some time documenting how you assign positions.
For us, that turned out to be the most valuable part of the entire project.
