Our Ratings, Checked.
A rating system you can’t check is a black box with confidence. This page puts ours next to the most widely cited public rating in college football and the final AP poll, season by season, and shows where they agree, where they don’t, and the design differences behind the gaps. Everything here is descriptive — a check on how credibly our numbers summarize the past, not a claim about anyone’s future.
What our rating is
Our team rating is built from one model: a ridge regression over every drive of every FBS season since 2017, solving all teams’ offensive and defensive values simultaneously, so strength of schedule is part of the math rather than an adjustment after it. The per-drive components (Drive Rating) are scaled to a per-game expected-points rating, and an opponent-adjusted special-teams component is added on top — the Final Rating is expressed in points per game against an average opponent. It is descriptive: a summary of what happened on the field, not a prediction of what happens next.
How it lines up with SP+
SP+ is Bill Connelly’s tempo- and opponent-adjusted efficiency rating, published at ESPN — the most widely cited public rating system in college football. We compare our final season ratings against his publicly posted final SP+ ratings for the eight seasons they’re publicly available: 2017–2022, 2024, and 2025. The final 2023 FBS table was never publicly posted, so 2023 sits this comparison out rather than being filled from a secondhand source.
Across those eight seasons, the two systems’ final FBS rankings agree with a Spearman rank correlation between 0.83 and 0.96, median 0.92 — two independently built systems, looking at the same games, seeing largely the same hierarchy.
| Season | Teams compared | Rank correlation (Spearman) | Rating correlation (Pearson) |
|---|
| 2017 | 129 | 0.953 | 0.958 |
| 2018 | 130 | 0.911 | 0.928 |
| 2019 | 130 | 0.931 | 0.943 |
| 2020 | 127 | 0.827 | 0.851 |
| 2021 | 130 | 0.901 | 0.919 |
| 2022 | 131 | 0.892 | 0.910 |
| 2024 | 134 | 0.934 | 0.940 |
| 2025 | 136 | 0.958 | 0.960 |
The lowest agreement is 2020, and that’s the season where it should be: short, uneven, regionally siloed schedules are the hardest opponent-adjustment environment for any rating system, so independent systems naturally agree least there.
SP+ ratings © Bill Connelly / ESPN, from his publicly posted ratings sheets. This page publishes relationships between the systems — correlations and rank gaps alongside our own ratings — and does not reproduce SP+ ratings tables.
Where we disagree, and the texture of why
Disagreements are design choices showing up in the numbers, not errors to adjudicate: SP+ works at the play level with tempo adjustments and preseason priors that fade as the season goes; we work at the drive level with no priors at all. The clearest pattern in the gaps is pace: teams at the extremes of tempo — in either direction — are where drive-grain and play-grain designs separate most. Below are the largest rank gaps in each season, with the characteristics of each team that the two designs weigh differently. None of this says which system is right — that would be a prediction, and we don’t make those.
2017
Virginiaours No. 43 · SP+ No. 76we rank it higher by 33
No single design difference stands out in our diagnostics — the systems simply disagree here.
Boston Collegeours No. 13 · SP+ No. 42we rank it higher by 29
Where the designs differ on this team: one of FBS's fastest teams (No. 11 of 129 in pace); its garbage-time-filtered EPA differs notably from its raw EPA (-0.13/play).
Coloradoours No. 89 · SP+ No. 60SP+ ranks it higher by 29
No single design difference stands out in our diagnostics — the systems simply disagree here.
Florida Atlanticours No. 25 · SP+ No. 53we rank it higher by 28
Where the designs differ on this team: special teams adds 1.8 pts/game in our rating.
Marshallours No. 60 · SP+ No. 86we rank it higher by 26
Where the designs differ on this team: special teams adds 1.6 pts/game in our rating.
2018
Armyours No. 29 · SP+ No. 70we rank it higher by 41
Where the designs differ on this team: one of FBS's slowest teams (No. 130 of 130 in pace).
Florida Internationalours No. 43 · SP+ No. 81we rank it higher by 38
Where the designs differ on this team: one of FBS's slowest teams (No. 128 of 130 in pace); its garbage-time-filtered EPA differs notably from its raw EPA (+0.17/play).
Georgia Southernours No. 35 · SP+ No. 72we rank it higher by 37
Where the designs differ on this team: one of FBS's slowest teams (No. 127 of 130 in pace).
Buffaloours No. 44 · SP+ No. 78we rank it higher by 34
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (-0.16/play).
Eastern Michiganours No. 54 · SP+ No. 88we rank it higher by 34
No single design difference stands out in our diagnostics — the systems simply disagree here.
2019
Missouriours No. 67 · SP+ No. 25SP+ ranks it higher by 42
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (-0.12/play).
UABours No. 110 · SP+ No. 72SP+ ranks it higher by 38
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (+0.10/play).
Kentuckyours No. 54 · SP+ No. 18SP+ ranks it higher by 36
Where the designs differ on this team: one of FBS's slowest teams (No. 122 of 130 in pace); its garbage-time-filtered EPA differs notably from its raw EPA (-0.11/play).
Illinoisours No. 79 · SP+ No. 45SP+ ranks it higher by 34
Where the designs differ on this team: special teams adds 2.1 pts/game in our rating.
Miami (OH)ours No. 73 · SP+ No. 107we rank it higher by 34
No single design difference stands out in our diagnostics — the systems simply disagree here.
2020
Riceours No. 47 · SP+ No. 104we rank it higher by 57
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (-0.19/play).
LSUours No. 81 · SP+ No. 29SP+ ranks it higher by 52
Where the designs differ on this team: one of FBS's fastest teams (No. 4 of 127 in pace).
Michiganours No. 77 · SP+ No. 25SP+ ranks it higher by 52
Where the designs differ on this team: special teams adds 1.9 pts/game in our rating; its garbage-time-filtered EPA differs notably from its raw EPA (+0.11/play).
Utahours No. 67 · SP+ No. 15SP+ ranks it higher by 52
No single design difference stands out in our diagnostics — the systems simply disagree here.
Kent Stateours No. 48 · SP+ No. 96we rank it higher by 48
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (-0.15/play).
2021
Marshallours No. 21 · SP+ No. 70we rank it higher by 49
Where the designs differ on this team: one of FBS's fastest teams (No. 10 of 130 in pace).
Ball Stateours No. 61 · SP+ No. 105we rank it higher by 44
Where the designs differ on this team: special teams adds 3.6 pts/game in our rating; its garbage-time-filtered EPA differs notably from its raw EPA (-0.15/play).
Old Dominionours No. 68 · SP+ No. 107we rank it higher by 39
Where the designs differ on this team: special teams adds 1.9 pts/game in our rating.
Toledoours No. 39 · SP+ No. 77we rank it higher by 38
No single design difference stands out in our diagnostics — the systems simply disagree here.
Central Michiganours No. 48 · SP+ No. 85we rank it higher by 37
Where the designs differ on this team: one of FBS's fastest teams (No. 14 of 130 in pace); special teams adds 2.2 pts/game in our rating.
2022
Ole Missours No. 62 · SP+ No. 14SP+ ranks it higher by 48
Where the designs differ on this team: special teams subtracts 1.9 pts/game in our rating.
Vanderbiltours No. 116 · SP+ No. 73SP+ ranks it higher by 43
No single design difference stands out in our diagnostics — the systems simply disagree here.
UTSAours No. 19 · SP+ No. 58we rank it higher by 39
No single design difference stands out in our diagnostics — the systems simply disagree here.
North Texasours No. 46 · SP+ No. 84we rank it higher by 38
No single design difference stands out in our diagnostics — the systems simply disagree here.
Fresno Stateours No. 20 · SP+ No. 57we rank it higher by 37
Where the designs differ on this team: special teams adds 1.6 pts/game in our rating.
2024
Miami (OH)ours No. 33 · SP+ No. 68we rank it higher by 35
No single design difference stands out in our diagnostics — the systems simply disagree here.
Navyours No. 25 · SP+ No. 57we rank it higher by 32
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (+0.14/play).
North Carolinaours No. 81 · SP+ No. 49SP+ ranks it higher by 32
No single design difference stands out in our diagnostics — the systems simply disagree here.
Missouriours No. 50 · SP+ No. 19SP+ ranks it higher by 31
No single design difference stands out in our diagnostics — the systems simply disagree here.
Bowling Greenours No. 47 · SP+ No. 77we rank it higher by 30
Where the designs differ on this team: its garbage-time-filtered EPA differs notably from its raw EPA (-0.12/play).
2025
Syracuseours No. 81 · SP+ No. 115we rank it higher by 34
No single design difference stands out in our diagnostics — the systems simply disagree here.
Old Dominionours No. 16 · SP+ No. 49we rank it higher by 33
Where the designs differ on this team: one of FBS's fastest teams (No. 6 of 136 in pace).
UConnours No. 82 · SP+ No. 54SP+ ranks it higher by 28
No single design difference stands out in our diagnostics — the systems simply disagree here.
UCLAours No. 72 · SP+ No. 99we rank it higher by 27
Where the designs differ on this team: one of FBS's slowest teams (No. 132 of 136 in pace).
UTSAours No. 88 · SP+ No. 62SP+ ranks it higher by 26
No single design difference stands out in our diagnostics — the systems simply disagree here.
The AP sanity anchor
The AP poll is a human ranking, so we treat it as a sanity anchor, not a target: our final top 25 should mostly overlap with the voters’, and where it doesn’t there should be a visible statistical reason. Across 2017–2025 our final top 25 shares an average of 18.9 of 25 teams with the final AP poll.
| Season | Top-25 overlap | Largest gap vs the poll |
|---|
| 2017 | 18 / 25 | USC — AP No. 12, ours No. 33 |
| 2018 | 20 / 25 | Northwestern — AP No. 21, ours No. 50 |
| 2019 | 22 / 25 | Minnesota — AP No. 10, ours No. 20 |
| 2020 | 21 / 25 | Texas A&M — AP No. 4, ours No. 22 |
| 2021 | 17 / 25 | Oregon — AP No. 22, ours No. 65 |
| 2022 | 18 / 25 | UCLA — AP No. 21, ours No. 52 |
| 2023 | 16 / 25 | Mississippi State — AP No. 9, ours No. 89 |
| 2024 | 19 / 25 | Missouri — AP No. 22, ours No. 50 |
| 2025 | 19 / 25 | Tulane — AP No. 18, ours No. 53 |
What we don't claim
This page checks that our ratings describe the past credibly. It makes no claim that they predict the future, beat a market, or should change what anyone does with their money. Not picks. Not hype. Just data.
Methods: for each season, both systems are re-ranked over the identical set of matched FBS teams (the “teams compared” column) before any correlation or gap is computed, so neither system is penalized for rating teams the other doesn’t cover; the only FBS team-season missing from the comparison is 2017 Idaho, which our dataset does not rate. Spearman is computed on ranks, Pearson on the rating values. Every number on this page comes from a single committed data file, generated by a script run against Bill Connelly’s publicly posted SP+ sheets and our own ratings — re-running it reproduces this page exactly.