What projects, and what doesn't · QBs · held out 2019-2025
How well one season predicts the next, measured on seasons the model never trained on. 1.00 would be perfectly predictable; 0.00 means last year told us nothing.
- Usage share (volume)0.46Beats guessing the QB average by 16%. n=2,421 QB seasons
- EPA per play (efficiency)0.05Not projectable — we do not forecast this. n=1,965 QB seasons
- Total EPA (value)0.51Beats guessing the QB average by 20%. n=1,965 QB seasons
Across all positions, usage carries year to year at 0.47 and efficiency at 0.07. So Keelon Russell's projection is a projection of opportunity — how much of the offense he runs through. How well he converts it is something this model does not claim to know a year in advance, and the number above is why.
2025 Production
Air yards
Where the ball goes, and how much of the gain is the throw rather than the run after it.
The ball travels 8.5 yards in the air on an average throw — about the median for the position. Another 2.4 comes after the catch.
Where he throws
attempts · completion rate · yardsDepth and direction come from CFBD’s passing detail, which is backfilled after the games: this covers weeks 1–2 only. Spikes, throwaways and intentional grounding are excluded before any average.
Performance Analysis · 2025 · vs QB peers
- Efficiency33
- Volume35
- Dual-Threat20
- Consistency50
- Clutch40
High-rep offense that leans on this QB's arm — value comes through cumulative production across a high number of plays.
- ▸Peak game: 1.00 EPA/play in Wk 2 vs UL Monroe (SP+ -22).
NIL Market Tier· 2025
On3 valuation ↗Meaningful starter. Local collective + position-group deals.
Tier is a model estimate based on position, school brand, performance rank, and usage — not a reported deal. NIL deals are private. For a real market valuation, see On3's NIL profile, which factors in social following and actual deal tracking.
Game Log · box score + EPA, week by week
Usage & Situational · Pro
Usage = share of team plays (CFBD has no true snap counts).