The September 9 Model Snapshot: Top of the Table, Schedule Strength, and Where the Model Missed

One archived snapshot read three ways, because the interesting questions about a rating system are not answered by the order of the list.

On 9 September 2026 the rankings pipeline wrote a file covering 148 Alabama teams and 207 rated games. This page takes that one archived snapshot and asks three different things of it: who was rated highest, who had actually been tested, and where the model was most wrong. A short follow-up at the end compares it against the current file, because the season has kept going.

Everything below was recomputed from the archived September 9 snapshot rather than copied forward, which matters: one figure that circulated in the original version of this analysis turned out to be mislabelled, and it is corrected in place below.

Who was rated highest

Central High School, Phenix City led at 38.5 rating points, just ahead of Moody High School at 37.3. The rating is denominated in expected points against an average Alabama team, so the gap between those two was worth about a point in a single-game prediction. That is a real edge and nowhere near a certainty.

TeamDivisionRecordRating
Central High School, Phenix City6A4-038.5
Moody High School1A-4A5-037.3
Spain Park High School6A3-028.7
Smiths Station High School6A5-127.9
Montgomery Catholic Preparatory SchoolPrivate2-122.4
Brewbaker Tech Magnet High School1A-4A4-222.1

Where the season was moving the model fastest

The published rating blends discounted prior-season information with this year's results. The pipeline also computes a cold-start rating that ignores last season entirely. Subtracting one from the other shows which teams the model was still arguing with itself about.

TeamDivisionRecordBlendedCold startDifference
Dothan High School6A0-16.0-3.2+9.2
G.W. Carver-Birmingham High School1A-4A0-1-27.7-18.8-8.8
Central High School, Phenix City6A4-038.529.7+8.8
Tallassee High School1A-4A1-111.22.8+8.3
i3 Academy High School1A-4A1-2-26.9-18.7-8.3
Montgomery Catholic Preparatory SchoolPrivate2-122.414.6+7.9
Greene County High School1A-4A1-05.012.7-7.7
Opelika High School6A1-213.76.4+7.3
Smiths Station High School6A5-127.920.6+7.3
Leeds High School1A-4A0-1-8.5-1.8-6.8

This table covers only teams that had played by September 9. Teams still at 0-0 were excluded, because their cold-start rating is 0 by construction and the difference would just be their prior season restated, which tells you nothing about 2026.

The last column is not a grade. A large positive number usually means a team is still carrying a strong prior that this season has not yet confirmed; a large negative number means the early scores are pulling against last year's resume. The coaching question is whether the opponent on next week's schedule looks like their record or like their underlying rating.

Who had actually been tested

Records alone do not say who has played anybody. Schedule strength is the model's estimate of the average opponent faced so far, put on the same point scale as the rating. A team can be 0-3 against a brutal opening slate and be more dangerous than a 3-0 team whose wins came against provisional opponents.

TeamDivisionRecordSchedule strengthRating
Dothan High School6A0-127.96.0
Johnson Abernathy Graetz (JAG) High School6A0-327.30.5
Booker T. Washington Magnet High School1A-4A0-122.1-2.5
Montgomery Catholic Preparatory SchoolPrivate2-121.922.4
Oak Mountain High School6A2-117.714.4
Moody High School1A-4A5-017.637.3
Mae Jemison High School5A0-217.4-11.3
Opelika High School6A1-216.613.7
Hale County High School1A-4A0-115.3-13.9
James Clemens High School6A0-315.1-2.1

Read the first three rows together: three teams with one win between them, all of whom had faced the hardest schedules in the state. Nothing about their records tells you that, and a coach preparing for any of them would be badly served by the standings.

Two different division averages, and a correction

An earlier version of this analysis printed a set of per-division averages and described them in schedule-strength language. They were in fact averages of team rating, which is a different quantity and in two divisions carries the opposite sign. Both are worth seeing, so both are below, correctly labelled this time.

DivisionTeamsAverage ratingAverage schedule strength
6A31+7.6+7.2
5A45+1.1+1.3
1A-4A55-3.4-2.5
Private14-3.3+3.4
Unclassified3-8.8+2.9

The Private row is the one that makes the distinction concrete. Those fourteen schools rated below average as teams and above average on schedule faced, which is what you would expect from programs that schedule across division lines rather than inside a region. Collapsing the two numbers into one would have hidden exactly the thing worth knowing. Schedule-strength averages are computed over teams with a rated opponent so far, so their counts are smaller than the division rosters.

In any case the averages matter less than the spread inside each division. The strongest team in a small division is a harder opponent than the middle of a large one.

Where the model was most wrong

A rating system that is never surprised is not measuring anything. These were the rated games through September 9 where the model's direction and the actual margin disagreed most sharply.

DateResultModel home edgeActual home margin
2026-09-08John Carroll Catholic 20, Springville 6-7.2+14
2026-08-25B.C. Rain 0, McGill-Toolen Catholic 13+4.1-13
2026-08-25Satsuma 19, St. Michael Catholic 12-5.9+7
2026-08-25Spanish Fort 13, Fairhope 19+5.2-6
2026-09-08Northside Methodist Academy 18, Charles Henderson 12-4.9+6
2026-08-25J.B. Pennington 6, Springville 0-3.5+6
2026-08-22Auburn 6, Albertville 12+3.2-6
2026-09-01Brewbaker Tech Magnet 7, Vestavia Hills 13+3.1-6

The model home edge is the home team's rating, plus the home-field estimate in force at the time (2.12 points), minus the opponent's rating. The actual home margin is the scoreboard. Opposite signs put a game on this list.

One upset erases very little prior information, especially in September. Repeated misses in the same direction are how a team actually climbs or falls. The useful follow-up is to turn each miss into a question that later games can answer: was the margin driven by explosive plays, short fields, fourth-down decisions, or a matchup the model cannot see? If the same pattern shows up again, it is probably a real team attribute. If it vanishes, it was noise.

What changed by September 17

The file has moved from 207 rated games to 305. The top of the table firmed up rather than reshuffling:

TeamRecordRating on Sep 17Rating on Sep 9
Central High School, Phenix City5-140.238.5
Moody High School7-036.037.3
Spain Park High School5-032.428.7
Smiths Station High School5-229.227.9
Prattville High School5-226.119.7

Two things in that table are worth more than the order. Moody went 7-0 and its rating fell slightly, which is what happens when a team keeps winning against opposition the model does not rate highly. Prattville gained more than six points while going 5-2, because who you played is doing as much work as whether you won.

The quietest change is in the model itself. The fitted home-field advantage dropped from 2.12 points on September 9 to 0.87 on September 17. That term is estimated from the games in hand, and with a hundred more of them the apparent value of hosting shrank by more than half. Any prediction made from the September 9 file was carrying a home-field assumption more than twice the current one.

Use the live table for today's order

This page is a dated snapshot kept for comparison, not a current ranking. Its figures are frozen at September 9 except where a row says otherwise. For today's numbers, go to the current rankings table, which rebuilds nightly.

Data and methods

The September 9 figures were recomputed from that day's archived public/data/rankings.json, recovered from this repository's version history rather than transcribed from the earlier write-ups. That recomputation is what caught the mislabelled division averages described above. The September 17 figures come from the current published snapshot. Ratings, cold-start ratings, schedule strength and the home-field term are all produced by the public Python pipeline in this repository; no rating is hand-edited. Teams with fewer than four rated games are provisional throughout. Score entry by schools runs late, so game counts on any given day are a floor rather than a full accounting. The rating method itself is described in how the rankings model works, and corrections policy is in the editorial policy.