CFB Schedule college football
2026 season

Where these numbers come from

Every source, every constant, and the things this site will not do with them.

Sources

Data sources
WhatWhere it comes from How often
Schedules, scores, venues, TVESPN's public college football APIHourly, and every five minutes while a game is in progress
Betting linesESPN's odds record, DraftKings, taken before kickoff and frozen thereWith the schedule
Player statistics, game logs, splitsESPN, asked one athlete at a time from each team's rosterAfter a team's game goes final
Depth chartsOurlads, honest user agent, 12-hour reuseDaily
Efficiency ratingsThe CFB King, a sister site, 2025 final numbersOnce a season until 2026 games are played
Kickoff weatherOpen-Meteo, at the venue's coordinatesHourly, and only inside 16 days

The constants

Four numbers decide a great deal of this site, so they are written down rather than left in the code.
Model constants
ConstantValue What it does
Home field2.4 pts Added to the home side before two ratings are compared. Not used on a neutral field, and the preview pages say so where it applies.
Margin spread16 pts The standard deviation used to turn a points difference into a win probability: a normal curve, so a 16-point edge is about 84%.
Win credit5 pts Rating points added per win above what the schedule implied, when the projected bracket orders teams.
Ratings spreadabout 48 pts The full range of the efficiency file, best to worst. It is why a win percentage cannot be added to a rating directly.

What is derived, and how

Efficiency margin is points scored per 100 plays minus points allowed per 100 plays, adjusted for the quality of the opponents faced. The adjustment is the sister site's; this site consumes the published number and does not recompute it. Everything below is arithmetic on that number and can be reproduced by hand from what the pages show.
Schedule strength is the mean efficiency margin of a team's opponents. A team whose schedule includes a program with no rating is averaged over the rated ones only, and its team page says how many that is.
The ratings differential on a preview is one team's rating minus the other's, plus 2.4 for the home side. That is the whole calculation. It is not a projection, a value, an edge or a lean, and nothing here compares it with the betting line to say one is better than the other.
The default pick in Predict mode is the team with the higher rating, plus home field. Where a betting line exists the line is used instead, because a market that has taken money knows more than a rating from last season.
The projected bracket orders teams by rating plus five points for every win above what their own schedule implied, where each game's expectation comes from the two ratings and home field through the normal curve above. It is not win percentage: a perfect record scored on a 0-100 scale would outweigh a rating range of 48 and seat unbeaten teams from weak schedules above everybody.
Against-the-spread records are accumulated, not reconstructed: each line is stored while a game is upcoming, frozen at kickoff, and graded afterwards. Every rate is printed with the number of games behind it.

What this site will not do

It does not publish a projection dressed as a pick, and it does not compare its own numbers with the betting line to tell you one is better. It does not invent a value it does not have: a game with no line says so, a player with no statistics has no page, and a rank is withheld until the population it is taken over is complete. Where a number could be derived two ways it is derived once, and where two sources disagree the page says which one it is reading.
When something here has been wrong, the page that carried it says so rather than quietly changing — the playoff format page carries one such note.

Freshness

The stamp in the footer of most pages is when that page was built. A player page is stamped when his numbers were last read, which is the only date that means anything there. The live board carries its own clock and says when it last asked.