FCS board — the FCS model, fitted on its own scale (zero is the average rated FCS team). These ratings are not comparable to the FBS board: the two are fitted separately because the level of play differs. MPR only — no SP+ win probabilities, no CFP odds. Coverage is 2021–2026; the Expected column starts in 2022 (2021 is the first season fitted, so it has no prior season for the replay to calibrate against).
Luck — what didn’t repeat
Both are results that outran the plays behind them, so both tend to come back to zero: each sums to ~0 across FBS, and a team at the top of one list this season is usually near the middle of it next season (year-over-year r ≈ −0.18 for both). A positive number is not a knock on the team and a negative one is not a promise — it is the part of the record that did not come from anything repeatable.
What explains this team
Each value is the team's deviation from the league average on that factor × the factor's fitted weight, in points of the model's projected margin. Both sides are oriented the same way: positive = better than league average at that factor. The "pct" figure is the raw opponent-adjusted standing in the season, so a factor the model gives little independent weight to (the per-play efficiency stats overlap heavily) shows its true rank instead of a bare 0.00. The ten nets add to the Offense and Defense chips, which decompose the projected margin — not the MPR.
Schedule & results
Record odds — odds of each record at each point of the season
First column: the selected model's win probability for that game. Every following column: the chance you have exactly that many wins after that game (regular season, workbook methodology — Φ(margin / 17)). Actuals mode pins played games to their real result and projects the rest of the season with the selected model.