College football transfer portal board
Every quarterback, running back, wide receiver and tight end in the transfer portal, with what each did last season: his share of the team’s work, his opponent-adjusted EPA per play, and his value over replacement (how much better he was than his level’s backups, times how often he had the ball). Pick a program to see its offense in plain words, how much of last season’s work at each position is leaving, and an experimental fit list. Team pages: all FBS teams, team ratings.
How to read it
Value over replacement compares each player with what his team would get from its next man up: the average opponent-adjusted EPA per play of backups at his position and competition level that season, weighted by how often they played. A positive number means he was better than a typical backup at his level; a big one usually means he was good and played a lot.
Market tiers are rough price levels from public information only. Sparvi doesn’t know what anyone is paid. The rules:
Projected yards are experimental. For any entrant and any program, the projection estimates his yards there next season (QB passing, RB rushing, WR and TE receiving): his projected share of the program’s work at the position, competing with its returners, committed transfers and signees, times the offense’s volume, times his efficiency, with an 80% range. A program’s room changes how many yards an entrant would get there, not how entrants compare with each other, so the program list’s order is nearly the same everywhere; best landing spots on each player show where his yards would be highest. Projections are shown only for players who reached last season’s backup minimum: below it there’s too little to go on, and in past portal windows veterans with little playing time were projected well above what they got. Quarterbacks and tight ends moving down a level (say, from a Power conference school to a Group of Five one) are still projected somewhat above what they’ve gotten in past windows. Its status: experimental: its QB, RB and WR projections beat both simple baselines (last season’s yards; last season’s share) in a 2021-2025 backtest, but those seasons were also used to find and fix bugs, so the clean test is the 2026 season (scored in January 2027); its TE projections tied the baselines. It doesn’t know about health, makeup, or money, and it treats a player who hasn’t committed as if he could land anywhere.
The fit list is experimental. For a program, it ranks entrants by a model of how likely each one is to win the job and play above his level’s starters within seasons there, from his production, his playing style, the program’s offense, and the jump in competition. Its status: experimental: its pre-registered test (2021-2024 portal windows, 311 successful transfers) tied every baseline: recall at 10 0.148 vs. transfer rating 0.142, raw production 0.153, profile only 0.151 (random 0.045); edges -0.005 to +0.005, each 90% interval spanning zero. The system and fit terms add nothing detectable over ranking players by quality alone. So treat it as a starting point for a conversation, not a recommendation. From the January portal window on, every program’s list and the simple rankings beside it are frozen each day in Sparvi’s public forecast record, so how the fit list actually does will be on the record.
Method, data, and what this can’t tell you
Data. Portal entries, rosters, recruiting classes and play-by-play from CollegeFootballData.com. Each entrant is matched to his previous team’s roster by name and school; entrants who can’t be matched aren’t shown. Statistics are from FBS and FCS games outside garbage time. “Opponent-adjusted” means each play is credited net of what that defense allowed in the same stat that season.
Definitions. Share is of the team’s whole season. Strengths are his highest-percentile performance components among those he had enough plays for. Similar players are the nearest player-seasons at his position over the last seasons by playing style, size and efficiency. “Old offense vs. this one” compares his last team’s offense with the program’s on run/pass balance and how passes are spread between receivers, tight ends and backs.
The projection. CFB player projection v1.2.0: team volume from the program’s last two seasons and coaching continuity; each player’s share from his last season, experience, level and recruiting profile against everyone at the position; efficiency regressed toward his position and the program’s last season. It is trained only on seasons before the one it projects, through 2025 (the 2026 season is held back as a sealed test).
The fit model. CFB system fit v1.1.8 (experimental: its pre-registered test (2021-2024 portal windows, 311 successful transfers) tied every baseline: recall at 10 0.148 vs. transfer rating 0.142, raw production 0.153, profile only 0.151 (random 0.045); edges -0.005 to +0.005, each 90% interval spanning zero. The system and fit terms add nothing detectable over ranking players by quality alone). It is trained only on transfers whose outcome was complete before the season the board looks ahead to (classes through ).
What this can’t tell you. Health and medicals, character and makeup, NIL cost, academics, and anything only film shows (route running, blocking, contact balance). Offensive linemen, defenders and specialists aren’t on the board yet. Stars and ratings from recruiting services are used only inside the market tier.
Data provided by CollegeFootballData.com. Sparvi is not affiliated with the NCAA, any conference, or any school.