Spaces:
Running on Zero
Running on Zero
| title: HAWK ONE | |
| emoji: 🏈 | |
| colorFrom: blue | |
| colorTo: red | |
| sdk: gradio | |
| sdk_version: "4.44.1" | |
| python_version: "3.11" | |
| app_file: app.py | |
| pinned: false | |
| # HAWK ONE | |
| A football film room built on one simulation engine with two rule profiles (NFL and college). | |
| Watch a play run, click a player, redraw his path or change his reaction time, and see what would have happened. | |
| Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference | |
| ## What it does | |
| - **Run play / Next play.** Offensive and defensive coaches call the snap ("Coach decides") or you pick the calls. Next play carries the drive forward (down, distance, field position, hash spotting). The play auto-plays. | |
| - **Level switch.** NFL or college reruns the play under that level's rules and hash width (catch, down by contact, pass interference, ineligible linemen downfield, hash marks). | |
| - **Player card.** Click any player for height, weight, 40 time, top speed, separation from the nearest opponent over time, and speed over time. | |
| - **What-if.** Drag the points on a player's path (add, remove, retime), or change a defender's or the QB's reaction time. Press **Run what-if**: the original and your version are replayed hundreds of times on the same random seeds, with a verdict, the numbers, a histogram, reasons, and a check of whether his body can run what you drew. Several edits at once are supported, with the credit split between them. | |
| - **Path Lab.** The engine tries variations of one player's path and ranks them. | |
| - **Recorded plays.** Load a tracking CSV (format in `tracking.py`) to inspect and annotate a real play. "Save this play as CSV" writes a sample. | |
| ## Run locally | |
| pip install -r requirements.txt | |
| python app.py | |
| ## Files | |
| | File | Role | | |
| |---|---| | |
| | `app.py` | Gradio app (single screen) | | |
| | `film.py` | Film viewer (HTML/SVG/JS served in an iframe) | | |
| | `render.py` | Result panels (verdict, numbers, credit split, Path Lab) | | |
| | `whatif.py` | Paired Monte Carlo, feasibility, path suggestions | | |
| | `engine.py` | Physics, collisions, blocking, coverage, rule hooks | | |
| | `coach.py` | Play calling and the drive loop | | |
| | `attrs.py`, `playbook.py`, `formations.py` | Player attributes, plays and calls, alignments | | |
| | `rules.py` | NFL and NCAA rule profiles and rule functions | | |
| | `tracking.py` | Tracking CSV import and export | | |
| ## Status and caveats | |
| - The simulation is plausible but not fitted to league data. Use the direction and the reasons more than the exact yardage. | |
| - What-if runs on simulated plays. Recorded plays can be inspected and annotated; running them through the engine needs an adapter that starts from recorded alignments. | |
| - Rules marked "verify" in `rules.py` (game clock, pass interference flat vs capped, overtime, kickoffs, targeting) need confirmation against the current NFL and NCAA rulebooks. | |
| - Engine, what-if maths, result panels and the viewer (including the iframe-to-page bridge) were run headlessly. The Gradio app itself still needs a first run in your environment. |