Ranking Lab History Check
Ad hoc historical check
This page reruns the current model across recent completed seasons and compares the result with the final reference rankings for those seasons.
This model is directionally aligned with recent final rankings, but it still takes a few meaningful swings in order.
What this page is doing
It reruns this model across recent completed seasons, then compares the model's final top 25 with the real final reference rankings for each season.
What this page is not
It is a fit check, not a prediction test. A model can be useful even if it disagrees sharply with the final poll.
Avg Rank Gap
6.4
Average spots each ranked team is away from the final reference poll. Lower is better.
Final Top 4 Match
2.7
Average number of teams your top 4 shares with the final reference top 4. Higher is better.
Final Top 10 Match
7.7
Average number of teams your top 10 shares with the final reference top 10. Higher is better.
Order Match
0.652
Solid ordering match. Closer to 1.000 is better; negative means the order often runs against the final poll.
How To Read The Scores
- Avg Rank Gap: lower means the model is landing teams closer to their final poll spots.
- Top 4 / Top 10 Match: higher means the model is catching more of the same teams near the top.
- Order Match: closer to 1.000 means the ordering is similar, around 0 is weak, and negative means the order is often upside down versus the final poll.
Season Table
Each row shows one full-season replay. Use it to spot whether the model is steady year to year or whether one season is dragging the averages down.
6.4 lower is better
7.7 higher is better
Solid ordering match
| Season | Avg Rank Gap | Top 4 Match | Top 10 Match | Order Match |
|---|---|---|---|---|
| 2025 | 5.2 | 3 | 8 | 0.767 |
| 2024 | 7.8 | 2 | 7 | 0.652 |
| 2023 | 6.2 | 3 | 8 | 0.411 |