College football
College QB rating
Top ten Heisman board
How this works
- Rating
- Scaled like passer rating: 100 is an average FBS starter, and 15 points is one standard deviation.
- 80% range
- Where a full season would likely land, 80% of the time. Wide early in the year.
- Heisman factors
- Ranks likely winners, not the best passer: quarterbacks, receivers, tight ends and backs scored together on production, team record and rank.
How the rating is built
The engine is opponent-adjusted efficiency. Where the NFL rating leans on EPA, the college version is built on ANY/A (adjusted net yards per attempt) because that's the deepest signal available across all ten seasons. Two college-specific fixes do the heavy lifting: an iterative national strength-of-schedule that stops a G5 quarterback feasting on bad defenses from outranking an SEC starter, and a garbage-time discountthat down-weights snaps once a game is decided so a 49–point blowout doesn't inflate the line.
What matters
- Who you played: the single most important college-only adjustment. Raw stats lie when half the schedule is overmatched; the national SoS is what separates real production from inflation.
- Efficiency per dropback (adj. ANY/A): yards, touchdowns, and a heavy interception penalty, per attempt, against the defenses you actually faced. The engine of the whole thing.
- Avoiding sacks: part processing, part protection; it survives as a real, repeatable negative.
- Rushing: credited, because the modern college QB runs. It's the difference between rating Jayden Daniels and rating a stationary passer with his arm stats.
What doesn't
- Garbage time: yards piled up after the game is decided. Left in, they reward blowouts; we down-weight them to a quarter of a leveraged snap.
- Raw counting stats: total yards and touchdowns are schedule- and tempo-dependent. A big number against a bad defense is a small number once you adjust.
- Team record: it's a quarterback rating, not a team rating. The two diverge constantly, and that divergence is exactly where the Heisman gets interesting.
- Narrative: not in the rating. It is in the Heisman model through record and rank.
The formula
Rating = 0.55·z(opp-adj ANY/A) + 0.15·z(comp%) − 0.15·z(sack rate) + 0.12·z(rush yds/game)Heisman factors: the best quarterback vs the likely winner
The Heisman view answers a different question than the rating: who will the voters actually crown? The skill rating is pure passing efficiency; the Heisman model adds what voters reward: team record, team rank, conference, and raw production.
Since October 2026 the model is position-aware. Quarterbacks, wide receivers, tight ends and running backs sit in one candidate pool. A quarterback is credited for his rating; everyone is credited for yards and touchdowns per game against others at his position. A non-quarterback carries a fixed discount, about a quarter of a quarterback's odds on the same profile, because that is how voters have behaved. It is fit to the vote shares of every Heisman finalist, 2016–2025.
Walked forward through 2019–2025, each season scored by a fit on the seasons before it, the eventual winner's median model rank is 2 and 18 of the 27 finalists in the pool landed in the model's top five. The old quarterback-only model could not rank DeVonta Smith, Travis Hunter or Ashton Jeanty at all. The position-aware model had Smith 2nd in 2020. It still misses 2024: Hunter won on two-way play, and the model only sees his receiving. Defensive players are not scored.
During the season the board leads with the Kalshi market and lets the model re-rank the players it scores. A player Kalshi prices but the model cannot score keeps his market price. The 2016–2025 seasons on the board still show the earlier quarterback-only model.
Sources
Observational research on cfbfastR play-by-play, player stats and schedules, 2016–2025 (FBS only). The 2026 season-to-date rows are rebuilt from ESPN box scores and play-by-play with the same components and garbage-time discount, scored on the 2016–2025 scale. Strength of schedule is an iterative national rating; the playoff proxy is each team's end-of-season Elo rank from the schedule data (no public poll dataset spans the full period); the 2026 board uses this site's power rank. The Heisman model is fit to finalist vote shares from 2016–2025 and validated walk-forward 2019–2025; wide receiver, tight end and running back production comes from the college WR/TE/RB rating's play data. Records are regular season plus conference championships.
The NFL version is the NFL QB rating, which uses EPA and completion percentage over expected. The Heisman view also scores the running backs, receivers and tight ends.