Methodology Deep Dive

How CFBTrack calculates Dominance Score.

Dominance Score is built to explain why one season looks stronger than another without reducing the answer to only wins, points, or poll rank.

Open related toolBack to methodology hub

Plain-English explanation

The score starts with season-level team performance and turns several different signals into comparable percentile-style components.

Those components are blended into one 0-to-100 style score so users can compare modern seasons without pretending a single raw stat tells the whole story.

Inputs used

  • Offensive quality from source-provided CFBD PPA, success rate, and explosiveness where available.
  • Defensive quality from source-provided opponent PPA, success rate allowed, and explosiveness allowed, where lower allowed values are better.
  • Schedule-aware strength and advanced-quality signals from SRS, SP-style overall strength, and unit-level ratings when available.
  • Average margin, win percentage, total wins, undefeated status, final rank, national-title-season status, and playoff status.

What the model rewards

  • Teams that are strong on both sides of the ball instead of only elite in one phase.
  • Teams that beat good schedules and create separation on the scoreboard.
  • Teams whose postseason and final ranking support the underlying efficiency profile.
  • Seasons that remain strong after the model compares them against the rest of the available 2003-and-later sample.

What the model does not claim

  • It does not claim to be a betting projection.
  • It does not claim an older season outside the modeled data is weaker because it is not present.
  • It does not claim the top score is the only defensible answer when the gap is narrow.

Example using Alabama

For an Alabama season, the model looks beyond the win total. A team with elite offensive efficiency, a top defensive component, a strong schedule-aware profile, a large average margin, and a title signal will rise because the same season checks several boxes at once.

  • If another Alabama season has a similar score, the best-teams page flags the race as debatable.
  • If a season wins a title but has weaker efficiency or schedule context, it still gets credit but does not automatically win the model.

Known limitations

  • The modern best-teams model is scoped to seasons with enough comparable 2003-and-later data.
  • Advanced-quality inputs are only as complete as the loaded source data for that season.
  • CFBTrack does not recreate CFBD's play-level PPA model; it audits local mapping, scaling, and route calculations around the source-provided fields.
  • Poll and postseason treatment can still leave close calls that fans should inspect directly.