Model read: Justin Herbert projects 9 rush yds (median), 11.6 under the 20.5 line — the model leans under. Its yards projection is lifted most by carry share (+10.0 yds) and pulled down by rush defense (-0.1 yds). Research signal — not a bet.
| Week | Opp | Yds | Tgt | Result |
|---|---|---|---|---|
| 25 wk17 | HOU | 37 | — | OVER |
| 25 wk16 | DAL | 42 | — | OVER |
| 25 wk15 | KC | 0 | — | UNDER |
| 25 wk14 | PHI | 66 | — | OVER |
| 25 wk13 | LV | 8 | — | UNDER |
| 25 wk11 | JAX | 21 | — | OVER |
| 25 wk10 | PIT | 19 | — | UNDER |
| 25 wk9 | TEN | 57 | — | OVER |
| 25 wk8 | MIN | 62 | — | OVER |
| 25 wk7 | IND | 31 | — | OVER |
Graded against this week's 20.5 line, not the line each game was actually posted at — so it answers “how often would this number have cleared”, which is the question a bettor is asking today.
Research tool. ProprScore is an ABSOLUTE 35–89 scale (never higher): it measures how far our projection sits from the line relative to the outcome's own spread — NOT a betting edge or confidence. A quiet slate tops out low, and that is information. Testing found no betting edge on the main line.