Methodology Deep Dive

How CFBTrack compares teams.

Team comparison pages are built for debate: pick teams, choose a metric, and read the result with season, sample, and metric direction in view.

Open related toolBack to methodology hub

Plain-English explanation

The comparison tools use compact season-level records so the page can show multiple teams without shipping unnecessary raw data to the browser.

Scatter views pair two metrics at a time, fit a simple trendline when the sample supports it, and summarize outliers for readers who do not hover every point.

Inputs used

  • Selected teams, season, conference filters, and metric choice.
  • Season-level raw and advanced stats such as efficiency, volume, record, scoring, and opponent-adjusted values.
  • Metric direction labels that keep lower-is-better and higher-is-better fields readable.
  • Fallback season context when a newer season does not yet have usable coverage.

What the model rewards

  • Teams that stand out after the selected filters and metric direction are applied.
  • Outliers that sit above or below a trendline in a way users can inspect.
  • Comparisons where teams share enough season and stat context to make the chart meaningful.

What the model does not claim

  • It does not claim two teams are comparable in every era or schedule environment.
  • It does not turn correlation in a scatter plot into causation.
  • It does not rank teams globally unless the selected view is explicitly a ranking surface.

Example using Michigan

A Michigan comparison can place one season beside Ohio State, Alabama, Georgia, and Texas for a chosen metric. The chart shows where Michigan sits in that selected frame, while links back to team and season pages preserve the underlying context.

  • Changing from a volume metric to an efficiency metric can change the read.
  • A team can be an outlier because the selected metric combination is unusual, not because the model is making a complete team-quality verdict.

Known limitations

  • The tools are strongest when selected teams share enough season and data coverage.
  • Charts simplify complex football context into selected axes or bars.
  • Fallbacks can keep a page useful while a newer season is still incomplete, but the fallback should remain visible.