CFB Galaxy Methodology

How CFB Galaxy turns a season into a living team map.

Team Galaxy is an explorable visual model with two deliberately separate ideas: an orbital map for navigation and a statistical-neighbor model for comparison.

Open related toolBack to methodology hub
Transparent modelSource-awareReviewed July 2026

The short version

Plain-English explanation

Each conference becomes a solar system so the map is readable at a glance. Teams orbit their conference sun, while the CFB Gravity Core gives every conference system a shared visual reference point.

Orbit position is a navigation device, not a claim that two nearby planets are statistically identical. The team inspector calculates profile matches separately from normalized season features and labels those matches directly.

Inputs used

  • Conference membership and membership tenure for conference-system identity and deterministic team ordering.
  • AP rank where available, with a broader team-strength fallback for planet scale.
  • Conference gravity context built from loaded ranking, rating, winning percentage, scoring margin, offensive PPA, and defensive PPA signals.
  • Season-profile features covering offense, defense, style, explosiveness, finishing drives, havoc and turnovers, field position, pace, record, and scoring margin.
  • Season or weekly snapshot labels, data cutoffs, game results, and schedule context for time-aware stories.

What the model rewards

  • Conference systems with stronger loaded ranking, rating, record, margin, and efficiency profiles appear closer to the shared core reference.
  • Highly ranked teams receive more visual mass; strength supplies a documented fallback when poll data is absent.
  • Statistical-neighbor results reward similar multi-feature season profiles instead of one isolated ranking or record.
  • Weekly views preserve the selected cutoff so later results do not silently rewrite an earlier snapshot.

What the model does not claim

  • It does not claim physical orbit distance is a power ranking or statistical-similarity score.
  • It does not claim the conference gravity score predicts a game or reproduces a poll.
  • It does not claim a similarity match means the teams have equal quality, identical schemes, or interchangeable schedules.
  • It does not treat missing source fields as zero production.

Worked example

Example using Florida State

Selecting Florida State flies the camera to its ACC orbit and opens an inspector with the loaded snapshot, gravity context, profile matches, and schedule voyage. Those pieces answer different questions instead of collapsing everything into one unexplained position.

  • The ACC solar system explains conference membership and gravity context; it is not a league-table rank.
  • Florida State's closest profile matches come from the normalized feature model and remain independent of planet spacing.
  • A season voyage uses loaded schedule context and starts only when the user chooses it.

Known limitations

  • The current public experience is only as fresh as its loaded artifact and any live schedule or ranking enrichment shown on the page.
  • Advanced-stat coverage can vary by season; imputed feature medians are flagged during artifact generation rather than presented as observed values.
  • Conference realignment and incomplete weekly data can change the available systems, stories, and comparison context.
  • The 3D view may reduce effects or use an accessible 2D fallback when the device, motion preference, or launch controls require it.