Team Strength

Opponent-adjusted, from the play-by-play, with garbage time thrown out.

What the columns mean

Strength: Points per game better than an average Power 4 team, offense and defense solved together for every team at once. It takes no notice of the record: a good team that has lost two close games is still a good team.
SOS: Strength of schedule, ranked. Not the average opponent - the number of games a top-20 team would be expected to win against these fixtures, with home and away counted.
SOR: Strength of record, ranked, and Record odds is the number behind it: the chance a top-20 team would have done at least this well with this schedule. A low percentage is a better resume.
vs top 25 / vs the rest: Points per game better or worse than this team's own rating expected, split by the calibre of opponent. This is the column that separates a contender from a team with a tidy record: won-or-lost cannot tell beating a poor side by three from beating them by thirty.

How accurate is it

Tested the only way that means anything: rate the season up to a given week, predict the week that has not happened, and score it. Across 799 games from 2023 to 2025 that the model had never seen, it missed the final margin by 12.65 points on average. The closing betting line, over the same games, missed by 11.91. Being a little behind the market is the honest result - the market has injuries, weather and everyone's opinion in it - and anything claiming to beat it on a page like this is fitting its own homework.

Home field is worth 2.5 points, measured against games the model had not seen rather than taken from the fit. Fitting it gives nearly six, which is twice what anyone has ever measured: Power 4 teams host weak non-conference opponents far more often than they visit them, and the home term quietly collects the difference.