How the Rankings Work
These rankings are designed to measure how well teams have actually performed on the field. The model evaluates offensive and defensive efficiency, adjusts every performance for opponent strength and game location, and recalculates the entire FBS network until the ratings stabilize.
From game performance to national ranking
Measure Performance
Each team-game is evaluated using five offensive and defensive efficiency metrics.
Adjust for Opponents
The quality of the opponent is incorporated into every offensive and defensive performance.
Neutralize Location
Home-field advantage is removed so performances can be compared on a neutral basis.
Solve the Ratings
The model iterates through the full schedule until team ratings converge.
Team Performance Rating
The Team Performance Rating, or TPR, is the primary score used to rank teams. It represents opponent-adjusted performance across every usable game through the selected week.
Every team receives separate offensive and defensive ratings. The overall rating is the average of those two components.
Because the FBS ratings are recentered during the solving process, an average FBS team sits near zero. Positive ratings represent above-average performance, while negative ratings represent below-average performance.
Five metrics power the model
Each team-game performance is built from five efficiency measures. Before they are combined, each metric is standardized against the FBS population so unlike statistics can be compared on a common scale.
Offensive Rating
Offensive performance rewards higher PPA, higher success rate, greater explosiveness, and better finishing.
Havoc works in the opposite direction for offense. The model rewards teams that allow less defensive disruption.
Defensive Rating
Defensive performance evaluates what the opponent was allowed to produce. Lower opponent PPA, success rate, explosiveness, and finishing are better.
Defenses are also rewarded for producing havoc themselves, so disruptive defensive play contributes positively to the rating.
Every performance is evaluated in context
A team's raw game performance is not its final rating. The model adjusts offensive performance using the opposing defense's rating and defensive performance using the opposing offense's rating.
This creates an interconnected system. A team's rating depends partly on its opponents, while those opponents depend on their opponents. The model repeatedly recalculates those relationships until the ratings converge.
Expected Neutral Margin
TPR can also be translated into an estimated point-margin scale. The model multiplies overall rating by its calibrated points-per-rating-unit value.
This produces an expected neutral-field performance margin relative to an average FBS team.
Strength of Schedule
Strength of Schedule is the average location-adjusted strength of the opponents a team has actually faced through the selected week.
Positive SOS values represent schedules harder than the FBS average. Negative values represent easier schedules. SOS Rank #1 is the most difficult schedule.
SOS is displayed separately from TPR
Opponent quality already affects TPR through the opponent-adjusted rating process. The displayed Strength of Schedule value is a descriptive summary of schedule difficulty and does not currently act as an additional independent multiplier or bonus to TPR.
Under the hood
The following details describe the current Performance TPR v1 implementation.
1. FBS statistical baselines
Metric means and standard deviations are calculated using FBS-vs-FBS statistical observations through the selected week.
2. Z-score standardization
Each available game metric is converted to a z-score relative to the FBS baseline. This puts metrics measured on different scales onto a common standardized scale.
3. Weighted performance composites
The five standardized metrics are combined using the current provisional weights: PPA 1.00, Success Rate 1.50, Explosiveness 0.75, Finishing 0.50, and Havoc 1.25.
4. Home-field neutralization
Non-neutral games are adjusted for home-field advantage before the observations enter the opponent-adjustment solver.
5. FCS treatment
FBS teams remain eligible for ranking when they play FCS opponents. Those opponents are represented using fixed FCS offensive and defensive priors rather than being solved as FBS teams.
6. Early-season stabilization
The model includes a one-game prior toward average. This dampens extreme ratings when a team has only a small number of games while allowing the actual data to increasingly dominate as more games are played.
7. Iterative opponent adjustment
Offensive and defensive ratings are repeatedly recalculated from each team's game observations and the current ratings of its opponents. A damping factor is used to stabilize the process.
8. Recenter around average FBS
After each iteration, FBS offense and defense are recentered separately so the average FBS offense and defense both equal zero.
9. Convergence
The solver continues until the largest offensive or defensive rating change falls below the convergence threshold, or until the maximum number of iterations is reached.
10. Weekly snapshots
Each ranking run includes only completed games through the requested week, producing a separate historical ranking snapshot for that point in the season.
Current Model Constants
Rank teams by performance, not reputation.
The model does not attempt to reproduce a human poll. It is built to evaluate what teams have demonstrated on the field through the games played so far.