Last updated: 2026-07-27 · Season: 2026 · Scoring: half-PPR · Licence: CC BY 4.0

Why this exists

Most college fantasy projections are either paywalled, limited to the biggest conferences, or published without saying how they were produced. This dataset is none of those. It covers every FBS program, the method is written out below including the parts that did not work, and the CSV is free.

Coverage

All 138 FBS programs. For comparison, Yahoo Sports' 2026 college fantasy product covers the Power 4 conferences plus Notre Dame — 68 programs — leaving 70 FBS schools outside it.

Conference	Programs	In Yahoo's 2026 product
Big Ten	18	Yes
ACC	17	Yes
SEC	16	Yes
Big 12	16	Yes
American Athletic	14	No
Mid-American	13	No
Sun Belt	13	No
Conference USA	11	No
Mountain West	10	No
Pac-12	8	No
FBS Independents	2	Notre Dame only

Methodology

Every input is restricted to what is knowable before a season starts: prior-season production, games played, and position. No in-season hindsight.

projected_points = ppg × ( w · games_last_year + (1 − w) · expected_games )

expected_games = 12          a healthy college regular season
w              = 1.0         durability weight

At w = 1.0 this reduces to prior-season total points, which is deliberate — see below.

What we tried that did not work

These are published because negative results are the useful part. Candidates were compared by Spearman rank correlation and top-N hit rate, predicting 2025 from 2024:

Raw prior-season total points orders the broad field best, because volume tracks a secure role — but it has the worst hit rate on the actual top tier (34%), since it buries players who missed games.
Per-game rate (ppg) has the best hit rate on the top tier (46%) but ignores durability and role entirely.
A ppg-lean durability blend (w = 0.35) looked flat on broad rank correlation and was a severe regression in practice: against a realistic opponent field in the championship simulator it won 0.3% of titles versus 11.0% at w = 1.0. The mechanism is that the blend credits partial-season players for games they never played — six real games projected as roughly ten.
Empirical-Bayes shrinkage toward the positional mean and class/age multipliers both reduced rank correlation when only one prior season of evidence was available, so neither is used.

Fields

Column	Meaning
overall_rank	Rank across all positions by projected points
name, position, school, conference	Player identity and program
position_rank	Rank within the player's position
projected_points	Projected season fantasy points, half-PPR
projected_adp	Projected average draft position
per_game_points	Projected points per game
floor, ceiling	Low and high outcome bands
is_newcomer	True for incoming players with no prior FBS season

How to cite

Use freely, including commercially, with attribution:

Saturday Night Showdown, "College Fantasy Football Projections 2026",
https://saturdaynightshowdown.com/data/ (accessed 2026-07-27). CC BY 4.0.

Corrections

If a projection looks wrong, it may well be — tell us and we will look. Contact.

COLLEGE FANTASY FOOTBALL PROJECTIONS 2026