Scoring Methodology

The Philosophy

Every player on every FBS roster receives a score based on what they've actually done on the field, not recruiting stars, not potential, not hype. Our system evaluates five measurable dimensions of a player's resume, then aggregates those into team-level rankings that reflect true roster strength heading into the 2026 season.

The result is a transparent, data-driven picture of who has the most proven talent. A five-star recruit who hasn't played a snap scores lower than a three-star senior who has started 40 games. A transfer portal pickup doesn't get credit for his new school's brand. He carries the resume he's actually built.

Player Scores: Five Components

Position Weights

The five components are summed and then multiplied by a position-specific weight to produce the final player score.

From Players to Teams: The Starter Model

Team scores aren't a simple average of every player on the roster. Instead, we identify the top 25 projected starters using the same position counts for every team. This ensures every team is evaluated on the same terms; no team gets an advantage from having a 90-man roster versus an 85-man roster.

Starters are selected by prioritizing players with a projected starter designation, then by career starts, then by total score. Within each position group, the top player's score is weighted more heavily than the depth behind him. The weights vary by group size: two-starter groups (RB) use 60/40, three-starter groups (WR, LB) use 50/30/20, four-starter groups (DL, DB) use 40/30/20/10, and the five-starter offensive line uses 35/25/20/12/8. This reflects that your best player at each position matters more than your depth.

The quarterback's score counts 1.25x in the team average, reflecting the outsized importance of the position. Kickers and punters count at just 0.2x, preventing weak specialists from disproportionately dragging a team down.

Roster Turnover Penalty

Teams that rely heavily on transfer portal additions face a small penalty. If more than 10 of a team's 25 projected starters are transfers, the team loses 0.5 points for each transfer starter beyond that threshold. A team with 15 transfer starters, for example, would lose 2.5 points. This reflects the reality that rosters assembled largely through the portal need time to build chemistry and cohesion. A team that retains its core and supplements through the portal isn't penalized; only rosters that are majority-transfer face a modest adjustment.

Position Grades

Position grades compare each team's starter group against every other team at the same position. We calculate a weighted average of each team's starters within a position group, then assign a percentile rank from 1 to 99. This means a position grade of 85 indicates that team's group is better than 85% of all FBS teams at that position. The overall team grade is a weighted average of all ten position grades, with quarterback counting slightly more and specialists counting less.

Win-Loss Predictions

Season records are predicted using a position-group matchup model that evaluates each game as a head-to-head contest between units, not a single-number comparison. For every scheduled game, we compute how each team's offense matches up against the opponent's defense, and vice versa, convert that into a calibrated win probability, then aggregate the per-game probabilities into a most-likely season record.

Each team's starters are grouped into three units:

• Offense: weighted average of QB, RB, WR, TE, and OL starter scores (using the same position weights: QB 1.25x, OL 1.70x, etc.)
• Defense: average of DL, LB, and DB starter scores
• Special Teams: average of K and P starter scores

The matchup advantage for each game is:

The equal weighting between offense-vs-defense and defense-vs-offense ensures the head-to-head math is zero-sum: if Team A is the favorite over Team B, Team B is the underdog over Team A by exactly the same margin (before home field). The home field boost of ±3.5 is calibrated to approximate the standard 3-point home advantage used in Vegas point spreads, scaled to our scoring system.

From advantage to win probability. The raw advantage is then converted into a per-game win probability using a logistic curve:

The constant 22 was chosen by minimizing prediction error across 1,625 historical FBS games from the 2024 and 2025 seasons. Lower values made the model systematically overconfident (predictions in the 90%+ band only won 84% of the time in reality); higher values made the model too cautious. At the calibrated value, an advantage of 0 reads as a 50/50 coin flip, an advantage of 10 reads as a 62% favorite, and an advantage of 30 reads as an 80% favorite. The slope is intentionally gentler than a Vegas closing line because our prior carries more residual noise than a sharp closing market.

Each game also shows an approximate point spread (e.g., −7.5 favored or +3.0 underdog) alongside the win probability, converted from advantage units at a ratio of roughly 0.75×. This gives a familiar reference point for how lopsided each matchup is, though these are roster-based projections, not betting lines.

From per-game probabilities to a season record. Once every game has a win probability, two natural questions follow: how many wins should the team average, and what single record is most likely? The first is just the sum of per-game probabilities, shown as Stat W-L on the predictions page and team cards. The second is the peak of the probability distribution over possible final records, shown as W-L and computed via a Poisson binomial roll-up of all per-game probabilities. The two diverge whenever a team is favored in many narrow games: twelve games at 60% each averages 7.2 wins, but the single most likely actual record is 7-5, with substantial probability of finishing 6-6 or 8-4 and only about a 0.2% chance of running the table 12-0. The predictions page surfaces both numbers along with the probability of going undefeated and the probability of finishing with exactly one loss.

This model captures matchup dynamics that a simple roster average cannot. A team with an elite offense but poor defense will lose to a balanced team, even if the unbalanced team has a slightly higher overall rating. Games against non-tracked opponents (FCS, etc.) are counted as guaranteed wins in the season-record roll-up.

On each team's page, every scheduled game shows its win-probability badge directly in the schedule; you can also click the chart icon next to any game to see a full position-group matchup breakdown showing how each offensive unit stacks up against the opposing defense, and vice versa.

Head Coach Factor

Coaching matters. A team's head coach influences game preparation, in-game adjustments, player development, and overall program culture. To capture this, every game prediction includes a head coach factor based on the head coach's career body of work, not just one or two recent seasons.

We use a mechanical Coach Power Rating (CPR) on a 0–100 scale, where 60 represents a true .500 coach. CPR is built from every FBS head-coaching stint in a coach's career, weighted by season-by-season performance relative to roster strength, with bonuses for sustained excellence, conference titles, and national championships. Year-1 stints at a new program carry a small transition tax that fades by year two. The result is a single number that travels with the coach across schools.

This produces a factor ranging from −3.0 to +4.0, with 0.0 for an average coach. The asymmetric clip reflects that the mechanical CPR model systematically pushes first-year head coaches and established sub-.500 coaches to the bottom of the distribution; raising the floor to −3.0 keeps elite separation at the top while compressing the noisy bottom. In each game, the coaching adjustment is the difference between the two coaches' factors. A matchup between a +3.0 coach and a −1.0 coach shifts the advantage by 4.0 points in favor of the better coach, comparable to home-field advantage.

First-year FBS head coaches with no prior head-coaching record receive a neutral starting factor near 0.0, then update as their CPR resolves over their first season. New hires with prior FBS head-coaching experience carry their existing CPR into the new program, minus the year-1 transition tax.

Examples (2026):

• Kirby Smart (Georgia), Ryan Day (Ohio State), Dan Lanning (Oregon): elite tier, +3.0 or higher
• Kalen DeBoer (Alabama Y3), Steve Sarkisian (Texas): high tier, around +2.0
• An average career .500 coach: 0.0
• A first-year HC with no FBS record: starts near 0.0

The coaching factor is calibrated so roster talent remains the primary driver of predictions, but coaching provides a meaningful edge in close matchups where the rosters are evenly matched.

Gauntlet Factor

Strength of schedule is already captured through SOS in the rankings, but consecutive high-difficulty games carry an additional toll. The gauntlet factor reflects the cumulative fatigue and preparation cost of facing top-tier opponents in back-to-back weeks without a BYE to recover.

A “strong” opponent is any team currently ranked in the top 40 by roster average. The first strong-opponent week in a streak carries no penalty. The second consecutive week subtracts 1.0 point from the team's advantage. Each additional consecutive strong week subtracts another 1.0, capped at 3.0 total. A BYE week, or any game against a non-top-40 opponent, resets the count: rest restores readiness.

The penalty is asymmetric: only the team in the gauntlet absorbs it. Their opponent that week gets no compensating bonus and faces a normal matchup.

Example: Florida's 2026 schedule runs Auburn, Ole Miss, Missouri, South Carolina, Texas in five straight weeks (no BYE). The Auburn game carries no penalty, Ole Miss is −1.0, Missouri −2.0, South Carolina and Texas each −3.0. The Week 8 BYE then resets the count before Georgia and Oklahoma in Weeks 9 and 10.

Display Scores

The raw team averages are rescaled to a 1–99 display range for readability. Individual player scores are shown as their raw values, preserving the natural separation between elite and average players.