Michigan’s third-down conversion rate vs a ranked opponent and vs an unranked opponent
, ,
Against what the season’s ratings expected, Michigan was 9.0 percentage points under vs a ranked opponent and 9.4 percentage points over vs an unranked opponent over its last 25 games, 9 games against 16. The plain averages are 29.6% and 51.3%, a difference of 21.7%. All 4 games after a bye were vs an unranked opponent. Both bowl and playoff games were vs a ranked opponent.
Michigan’s −18.4% vs a ranked opponent and vs an unranked opponent over expected ranks 76th of the 77 teams with 4 games on each side over their last 25 games.
third-down conversion rate a game over expected
A −18.4% difference in third-down conversion rate over expected vs a ranked opponent and vs an unranked opponent is unusual: shuffle which games were vs a ranked opponent and it comes up 2 times in 100.
A −21.7% difference in third-down conversion rate vs a ranked opponent and vs an unranked opponent is unusual: shuffle which games were vs a ranked opponent and it comes up once in 100.
The games
| Side | Date | Game | Result | 3rd downs | Expected |
|---|---|---|---|---|---|
| Nov 9, 2024 | @ Indiana | L | 33.3% | 39.6% | |
| Nov 23, 2024 | vs Northwestern | W | 68.8% | 40.6% | |
| Nov 30, 2024 | @ Ohio State | W | 50.0% | 37.3% | |
| Dec 31, 2024 | vs Alabama (neutral) | W | 29.4% | 37.5% | |
| Aug 30, 2025 | vs New Mexico | W | 46.2% | 42.9% | |
| Sep 6, 2025 | @ Oklahoma | L | 21.4% | 37.6% | |
| Sep 13, 2025 | vs Central Michigan | W | 72.7% | 48.1% | |
| Sep 20, 2025 | @ Nebraska | W | 33.3% | 43.6% | |
| Oct 4, 2025 | vs Wisconsin | W | 41.7% | 38.4% | |
| Oct 11, 2025 | @ USC | L | 22.2% | 43.4% | |
| Oct 18, 2025 | vs Washington | W | 46.2% | 41.2% | |
| Oct 25, 2025 | @ Michigan State | W | 40.0% | 41.5% | |
| Nov 1, 2025 | vs Purdue | W | 54.5% | 41.0% | |
| Nov 15, 2025 | vsNorthwestern (neutral) | W | 64.3% | 41.4% | |
| Nov 22, 2025 | @ Maryland | W | 85.7% | 41.6% | |
| Nov 29, 2025 | vsOhio State | L | 11.1% | 37.5% | |
| Dec 31, 2025 | vs Texas (neutral) | L | 22.2% | 39.7% | |
| Sep 5, 2026 | vs Western Michigan | W | 30.0% | 45.1% | |
| Sep 12, 2026 | vsOklahoma | W | 30.8% | 37.0% | |
| Sep 19, 2026 | vs UTEP | W | 69.2% | 46.5% | |
| Sep 26, 2026 | vs Iowa | L | 38.5% | 41.6% | |
| Oct 3, 2026 | @ Minnesota | L | 30.0% | 37.8% | |
| Oct 17, 2026 | vs Penn State | W | 41.7% | 35.5% | |
| Oct 24, 2026 | vsIndiana | L | 30.0% | 40.0% | |
| Oct 31, 2026 | @ Rutgers | W | 75.0% | 42.3% |
Method
- Expected is an offense and a defense a team, fit over every game of its season between two teams in the table with the game being read left out, plus 0.6% in 2024, plus 0.7% in 2025, plus 1.4% in 2026 for the home team (the edge the fit finds once strength is in), nothing at a neutral site.
- Each expected number is then fit to the season’s games: a game bears out about 50% in 2024, 56% in 2025, 40% in 2026 of the difference the model predicts.
- The shuffle keeps the season’s shape: the games are cut in two where one date leaves the least variance, each game is read against its stretch, and the sides are shuffled only inside each stretch.