Methodology
The Davis Rankings turns a season of results into one score per team using a fixed formula. There are no ballots, no committee, and no preseason assumptions: every FBS team starts each season with a blank slate, and its score is recomputed from scratch every week using only games that have already been played. The table below lists the factors and their weights; the sections after it explain how they work together, with a worked example.
How the five factors fit together
Each factor produces its own score on a 0 to 100 scale. The scores are then combined using the weights in the table above (they are read straight from the ranking system's configuration file, so the table is always current). Roughly speaking, the composite is two-thirds about what the team did on the field - its winning percentage and how its latest game went - and one-third about the circumstances of those results: where the games were played and how good the opponents were.
- Win/loss record is simply winning percentage. An undefeated team scores 100 here, a team that has lost every game scores 0. It carries by far the most weight because wins and losses are the whole point of the sport.
- Recency looks only at the most recent completed game: 100 for a win, 0 for a loss. A bye week carries the previous result forward. This is why a team can fall several spots after a single loss, and why a team that has just won tends to hold its position. At the end of the season this bites hardest: a team's last game, the conference championship or the final regular-season game, is the result the final rankings carry, and there is no next week to make up for a loss.
- Location bonus rewards winning away from home. An away win is worth +1.5, a neutral-site win +1.25, and a home win +1.0. Losses are scored the other way round: a home loss costs -1.0, a neutral-site loss -0.75, and an away loss -0.5. The average across all games is then scaled onto 0 to 100.
- Opponent record (at the time) asks how good the opponent looked when the game was played. A win over a team with a strong record counts for more than a win over a team with a weak one, and a loss to a strong team costs less than a loss to a weak one. If the opponent was on a winning streak, the game's value is boosted by 5% per streak win, up to 50%.
- Opponent record (current) asks the same question with hindsight: every past opponent is re-valued using their record as of today. A win over a team that has since gone on a run becomes more valuable each week, and a win over a team that has since collapsed loses value.
A worked example
Take a team that is 4-0 with three home wins and one road win, and whose opponents have gone a combined 14-5 since. Its factor scores might look like this:
| Factor | Score | Weight | Contribution |
|---|---|---|---|
| Win/loss record | 100.0 | 55% | 55.0 |
| Recency | 100.0 | 15% | 15.0 |
| Location bonus | 87.5 | 15% | 13.1 |
| Opponent record (at the time) | 81.7 | 9% | 7.4 |
| Opponent record (current) | 85.0 | 6% | 5.1 |
| Composite | 95.6 |
The location score comes from the four games' values (+1.0, +1.0, +1.0 and +1.5), which average +1.125 and map to 87.5 on the 0 to 100 scale. The same calculation is shown for every real team on its page: open any team from the rankings and use "Show calculation" to see each game that went into each factor.
Why the weights look the way they do
The weights are a deliberate statement about what matters. Winning is weighted far above anything else, and the two opponent-quality factors together carry only 15% - enough to separate a team that beat good opponents from one that beat poor ones, not enough to let a loss be explained away as a "quality loss." Conference membership, preseason expectations, brand, and prior seasons do not appear anywhere in the formula. The weights are a starting point, not a claim of perfection, and may be revised in future versions of the system; if they are, the change will be reflected in this page.
Ties
Teams with the same score (shown to one decimal) share a rank, marked "T-" in the tables: two teams tied for 10th are both T-10, and the next team is 12th. Tied teams are listed alphabetically, which is only a display order, not a tiebreaker. Early in the season, when results are just wins and losses, many teams have the same score and ties are large.
Tiers
Besides its rank, every team is placed in a tier. The tier is just a label for where a team sits in the order; it never changes a score. From week 3 on, a tier is a band of 25 ranks. Because teams that tie share a rank (and so a tier), a tier can hold a few more or fewer than 25 teams. The last tier has no upper limit, so it simply holds every team ranked below 125 (the number of FBS teams varies a little from year to year).
| Tier | Rank range |
|---|---|
| Winning | ranks 1-25 |
| Feeling Good | ranks 26-50 |
| Character Building | ranks 51-75 |
| Do We Need to Talk? | ranks 76-100 |
| Sick of It | ranks 101-125 |
| Oh Crap | ranks 126 and below |
In the first two weeks there are too few results to split teams into bands of 25, so the tiers follow the record instead: unbeaten teams are Winning, teams with a win and a loss are Character Building, and winless teams are Oh Crap. The other tiers appear in week 3. Tier changes are only shown from week 4, once two weeks in a row use the same bands.
Tiers appear next to the rank in the rankings table (with a divider between tiers when the table is in rank order), and in each team's page header. See every team by tier.
The Davis Line (TDL)
The Davis Line, or TDL, is the league average for the week: the mean TDR score of every ranked FBS team. It works as a quick benchmark, a way to see at a glance whether a team is on the good or bad side of average and by how much. The number of teams above it changes a little from week to week, which is the point: it follows the real spread of scores, not a fixed half. It is only a display and never changes a score or a rank.
In the rankings table, a dashed row marks the line between the last team at or above it and the first team below it (when the table is in rank order), and a small +/- beside each score shows the distance from it: a team scoring 68.5 when the line is 50.4 shows +18.1. On the Standings page, a dashed line across the conference comparison chart marks the same point. It is the same average the conference comparison table uses, so a conference's average score can be read directly against it. FCS opponents are not ranked and are not part of it.
Like tiers, the line is hidden in the first two weeks, when too few games have been played for an average to mean much. It appears in week 3 and is recomputed every week, so it moves a little as results come in.
When and how the rankings update
Rankings are recomputed automatically several times a week during the season, from late August through the conference championships. Sunday morning picks up Saturday's games. Sunday afternoon adds the AP and Coaches polls. Monday to Wednesday catch Sunday and Monday night games, the CFP committee rankings (once they start; they are released Tuesday evening, so they show on Wednesday) and corrections to the underlying data. Thursday, Friday and Saturday morning refresh kickoff times and TV listings for games not yet played, and add the result of any weekday game that has finished to each team's schedule. They do not change a ranking: a Tuesday to Friday night game counts in the rankings with the rest of its week on Sunday. Runs can start late, so a number can lag by a few hours. A week becomes the "current" week once its main slate of games (normally Saturday) is over and at least half of its games have been played, so a single midweek game does not reshuffle the board.
During the season you can browse any past week using the Season and Week selectors on the home page, and the Projected Final table shows a hypothetical end-of-season order that assumes the currently favored team wins every remaining game. It is an illustration of the formula, not a forecast.
Data sources and limitations
Game results, schedules, team information, and the AP, Coaches, and CFP polls come from the College Football Data API. The polls are shown next to our ranking for comparison only and are never used in the score. Because the formula only sees results, it has real limits: it cannot know about injuries, margin of victory, or how a game was played, and early in the season, when teams have played only a few games, scores are volatile and a team's rank says more about a small sample than about its true strength. Those are known trade-offs of keeping the method simple and fully transparent.
Think something looks wrong, or have an idea for the formula? See the contact page.