AI Football Play Calling: What's Coming
This feature is not live yet. The play recommender is being built now, and when it ships it will be part of the paid plan. The play designer, playbook and play sheet PDF stay free. This page explains how the recommender will work, so you can decide whether it is worth waiting for.
The short version: you will tell Atlas Chalk what defense you expect to see, and it will rank the plays in your playbook by how well they have worked for your team against that look, by down and distance.
The problem
Most game plans are built from memory. You remember the counter that broke for 40 yards against a 5-3 last year. You forget the six times it went for one yard. The big play sticks and the ordinary results fade, so the call sheet ends up weighted toward the plays you remember best, not the plays that work most often.
Charting every snap fixes part of that, but a stack of charts does not tell you what to call on 2nd and 6 against Cover 3. Someone still has to sort it, count it and compare it, usually on a Sunday night. The recommender will do that sorting for you.
How recommendations work
The recommender will use the plays you chart in Atlas Chalk. Each snap you enter records the play you ran, the defensive look, the down, the distance and the yards gained. Over a season that becomes a record of what your offense does against each front and coverage.
When you prepare for an opponent, you will enter the fronts and coverages you expect to see. The recommender will then rank your plays by situation against those looks, and show the score, the number of snaps behind it, and a one-line reason for each.
It will only use what your designer shows your program. A youth program does not assign personnel groups to plays, so its recommendations will be by formation and concept, not by personnel.
It will not use anything about your players. No roster, no names, no ratings of individual kids. The inputs are plays, defensive looks and results.
Like every AI suggestion in Atlas Chalk, a recommendation will be a suggestion. You decide what goes on the call sheet.
Success metric
Each charted snap will be graded three ways:
- Positive yardage: the play gained at least one yard.
- On schedule: the play gained at least half the distance on 1st down, 70% of it on 2nd, and all of it on 3rd or 4th. This is the standard success-rate definition used in college analytics.
- Explosive: a big gain. The starting cutoffs depend on your level: 10-yard runs everywhere, 15-yard passes for youth and middle school, 20-yard passes for high school and college.
A play's score will blend those three rates, weighted 30% positive yardage, 40% on schedule and 30% explosive. Those weights and the explosive cutoffs are starting points, not findings. You will be able to change them for your program in Settings.
Sample size matters as much as the rate. A play that worked 2 times out of 2 should not outrank one that worked 30 times out of 50. So a play with only a few snaps will be pulled toward your team's average until it has enough snaps to stand on its own. By default that pull is as strong as 10 snaps of average results, and you will be able to change it or turn it off.
Every score will show how many snaps it is based on. Anything under 8 snaps will be greyed out and marked low sample, so you know which numbers to trust and which to treat as a hunch.
Join the waitlist
Leave your email and we will tell you when the recommender is ready.
FAQ
Is the AI play caller available now?
No. It is still being built and will come as part of the paid plan. The play designer, playbook and play sheet PDF are free today.
What data will the recommender use?
Only the games you chart in Atlas Chalk: the play you ran, the defensive look, the down, the distance and the yards gained. It will not use anything about your players.