Same player. Two settings.
Measured Combine drill and NFL OUT route.
0 s = detected turn · 100% = entry speed · shared yard scale
- Combine speed retained
- NFL speed retained
- NFL separation at throw
No reliable scouting advantage yet
This result is not strong enough for a positive competition submission.
Predicting average separation
| Model | Error ↓ |
|---|
Error compares each player’s average separation with their predicted average. It is not the miss on each individual route.
The tempting broader result
Years and player groups
2024 was previously explored. These are exploratory checks, not untouched confirmation. No new 2025 scoring.
What this can tell us
What is actually measured?
Speed after the detected turn divided by speed before it. A broad turn can preserve speed without showing a sharp plant or recovery.
What you see
- The same receiver’s real tracking in both panels.
- The complete drill and NFL route from snap to throw, aligned at the turn.
- Only the receiver. Opponents and the quarterback are not invented.
- After one recording ends, its last observed point holds and is labeled.
Measurement and example selection
The first sharp-cut detector rejected most rounded speed-outs. A wider window was registered before prediction scores. Its results remain exploratory. The apparent turn is a coordinate-based proxy, not a verified foot plant.
Missing separation is unknown. Untargeted routes with a recorded value are included. No future NFL coordinates are used as prospect prediction inputs.
NFL / Kaggle 2027 · derived non-commercial research plots · no game film or raw tracking downloads