Edge Labs EDGE LABS Market intelligence
MLB CFB NFL
Board Research Tools The Lab Record Vault Sign in Get full access
Edge Labs · College football · Regression watch

Who the record says is due to move.

One finding drives this whole page · records drift back toward the talent on the roster. We take each team's prior-season record and set it against the talent it actually signed. The teams that won far more than their talent says are flagged to give some back. The teams that won far less are flagged to bounce. This is a tendency and a watchlist · the direction the base rate leans, not a score, not a pick, not a bet.

The base rate, stated straight: across 2020 to 2025, point-in-time, teams that outran their recruiting talent gave back about 15 win-percentage points the next season, and the underachievers rose about as much. n is more than 150 on each side. It is a tendency, not a guarantee · any single team can buck it.

~15 pts
avg win-% swing · next year · n>150/side

Filter

watchlist
Reset
The gap is how far a team's prior-season record sat above or below its talent rank, in percentile points · bigger gap, stronger the pull. Only FBS teams with at least eight completed games are in the universe · n=0.
Regression-down candidates Outran their talent
n=0 · shown
Won far more than the roster's talent in the prior season. The base rate says these tend to give some back. A lean, not a lock.
No regression-down candidates in All FBS for this season.
Bounce-back candidates Underran their talent
n=0 · shown
Won far less than the roster's talent in the prior season. The base rate says these tend to climb back. A lean, not a lock.
No bounce-back candidates in All FBS for this season.

A tendency, not a forecast. Every team here is flagged by one thing only · how far its prior-season record sat from its recruiting talent. That gap pulls back toward the roster on average · about 15 win-percentage points · but it is a base rate across 0 teams, not a promise about any one of them. Injuries, a new coach, a quarterback leap · any of it can override the lean. We do not turn this into a pick or a bet, because the regression signal does not beat the spread and we tested that. Watch this season · prior record vs talent · n=0 · research, not advice · 21+.
See the full national talent power ranking →