Overview
The Split Row poll rates every NCAA football team (FBS, FCS, Division II and Division III) from completed games. Each Sunday the poll is re-run, a new Top 25 and full boards are published for each division, and next week's games are projected from the same ratings.
This is an independent computer poll. It isn't the AP poll, the Coaches Poll or the College Football Playoff rankings, and it doesn't try to predict what those polls will do.
What goes into the ratings
Each team's rating is a weighted blend of seven pieces:
| Component | Weight | What it means |
|---|---|---|
Game results Every game earns credit for the result, scaled by how good the opponent was. | 25% | Every game earns credit for the result, scaled by how good the opponent was. |
Strength of schedule The average rating of everyone you've played. | 20% | The average rating of everyone you've played. |
Quality wins Wins over teams in the top 40 of the all-division poll (maxes out at three). | 15% | Wins over teams in the top 40 of the all-division poll (maxes out at three). |
Offense Opponent-adjusted offensive stats, ranked within your division. | 12.5% | Opponent-adjusted offensive stats, ranked within your division. |
Defense Opponent-adjusted defensive stats, ranked within your division. | 12.5% | Opponent-adjusted defensive stats, ranked within your division. |
Recency Recent games count a little more than September games. | 7.5% | Recent games count a little more than September games. |
Bad losses A penalty for losing to teams rated well below you. | −7.5% | A penalty for losing to teams rated well below you. |
About three quarters of the rating comes from results (who you played, who you beat, and how) and about a quarter from stats. That keeps an unbeaten team with a tough schedule ahead of a team that has great numbers but more losses.
Stats: offense and defense use measures like yards per play, EPA per play, success rate, explosiveness, havoc, red-zone finishing, turnover margin and scoring. Each one is adjusted for the quality of the opposing unit, then ranked within the division. Division II and III offense/defense use scoring only, because detailed stats aren't consistently available at those levels.
Conference strength: each conference gets a strength weight that starts from a preseason prior and shifts toward how its teams actually perform in non-conference games (65% prior, 35% performance), within set limits.
How each game counts
- Result credit: a win is worth 1.00, a tie 0.50 and an overtime loss 0.35. A regulation loss is worth up to 0.25, shrinking as the margin grows (a 28+ point loss earns nothing).
- Opponent adjustment: that credit is multiplied by a factor between 0.75 and 1.25 based on the opponent's rating, so beating a strong team is worth much more than beating a weak one.
- No hindsight: each game is valued with the opponent's rating from the week before it was played, the same information anyone had at the time.
- Finished weeks only: a week isn't counted until its games are done, so a midweek game never produces a half-finished poll.
- Home field and margin of victory beyond the loss scale aren't part of the poll itself.
Early-season ratings
In Week 1 nobody has a 2026 rating yet, so opponents are valued using each team's own final 2025 rating from the same model. If a team has no 2025 rating, its division average is used instead. These starting values only affect how opponents are valued early on; by around Week 6 the 2026 results drive almost everything.
Opponents outside the NCAA (NAIA, junior college and others) aren't rated by the poll. When an NCAA team plays one, that game uses a conservative estimated rating for the opponent.
Weekly archive
Every week's poll is saved. On the rankings page, use the week picker to see exactly where the poll stood after any week of the season, for every division. Links include the week (for example ratings.splitrow.com/2026?week=3), so you can share a past poll.
Note: if the model itself is ever changed, past weeks are re-run with the updated model so every week in the archive is calculated the same way.
Predictions
- Projected margin = 0.75 × the rating difference between the two teams, plus 3 points for the home team (0 at neutral sites).
- Win probability comes from that margin on a smooth curve: a 10-point favorite wins about 69% of the time.
- Projected scores start from a typical 54-point game total and adjust it using each offense against the opposing defense, then split it to match the projected margin.
- These are computer projections. They don't account for injuries, weather, rest or travel, and they're not betting advice.
Production Value
Production Value rates FBS players within their position group: quarterbacks, running backs, receivers and tight ends, defensive linemen, linebackers, defensive backs, kickers, punters and returners.
- Every stat is turned into a 0–100 percentile among qualified players at that position (for example, quarterbacks need 50 pass attempts).
- Volume blends counting stats (yards, touchdowns, tackles, sacks). Efficiency blends rate stats (yards per attempt, yards per carry, field-goal percentage and so on).
- The two are combined, mostly about 55% volume and 45% efficiency, with small differences by position.
- A team strength multiplier (0.85 to 1.20) rewards production against tougher schedules and on stronger teams, using the poll's strength of schedule and rating. That's why the best players can score above 100.
Offensive linemen aren't rated (public box scores have no blocking stats), defense is box score only, and early-season samples are small. See the leaders →
Coverage & limits
- Rated: all NCAA FBS, FCS, Division II and Division III teams that have played.
- Not rated: NAIA, junior college, U Sports (Canada) and other non-NCAA programs.
- Some Division II and III games are missing detailed line scores, so overtime can't always be detected; those games are treated as regulation.
- Game and stat data come from CollegeFootballData.com.